Final report OpenFOAM CFD simulation of pollutant dispersion in street canyons: Validation and annual impact of trees Action 9.2/Part II Stijn Vranckx, Peter Vos VITO www.vmm.be www.vito.be Lisa Blyth Email: lisa.blyth@vito.be Tel: (+32-14) 33 67 57 Fax: (+32-14) 32 27 95 www.life-atmosys.be http://ec.europa.eu/environment/life
Table of contents Table of contents 2 1. Summary 3 2. Introduction 4 3. Modeling strategy 5 3.1 Wind tunnel validation data 5 3.2 Model Setup 6 4. Results Validation Study 10 4.1 Flow and kinetic energy profiles 10 4.2 Flow field around the buildings and inside the street canyon 11 4.3 W/H =1 14 4.4 W/H = 2.0 20 5. Annual average effects using meteo statistics 22 6. Conclusions 29 7. References 30 Page 2
1. Summary During the last decade, a variety of studies has been presented on the effects of urban vegetation on the urban air quality. Vegetation affects the air quality in street canyons in several ways, with the main effects the reduction of pollutant concentration through deposition and oppositely an increase in concentrations caused by decreased ventilation reduces the air flow and wind speeds in a street canyon. In literature, several street canyon orientations have been investigated for fixed wind directions for single vegetation characteristics, with the overall conclusion trees increase pollutant concentrations. Our aim is to investigate the annual average effects of trees on air quality, including deposition and air flow influences, for a range of vegetation settings. Firstly, the pollutant concentrations have been calculated for all wind directions at a 15 interval. This has been completed for 10 different vegetation settings, ranging from no vegetation and significant deposition speeds and momentum loss terms. All model simulations have been performed using a Computational Fluid Dynamics (CFD) model. The choice for an open source CFD approach using OpenFOAM has been made, applying a k-epsilon model which has been enlarged to capture vegetation and pollutant dispersion. The CFD results are validated against the CODACS wind tunnel database. Fair agreement between the CFD results and these data is obtained, confirming the validity of the simulation approach. This was a first goal of this study. Next, meteo statistics have been applied to use the individual results for single wind directions to calculate annual effects. The statistical averaging, using the meteo data of the Belgian meteo station Luchtbal, take into account the seasonal effects of trees and the influences of the street canyon orientation. Further, the background concentration and pollutant emissions for a busy urban road in Antwerp have been applied for both PM 10 and EC. As can be expected, the urban vegetation effects are significantly larger for EC, as the influence of the background on local concentrations is much smaller as for PM 10. Finally, we obtain an annual increase of concentrations due to the presence of trees in street canyons ranging from 0.2% to 2.6% for PM 10 and 1% to 13% for EC, depending on the type of vegetation. Page 3
2. Introduction Within the framework of the ATMOSYS Life + project, we have started a study on the thorough validation of the CFD model in use. Over the last decade, OpenFOAM has emerged as an interesting and promising CFD software package. The OpenFOAM (Open Field Operation and Manipulation) CFD Toolbox is a free, open source CFD software package produced by OpenCFD Ltd. It has a large user base across most areas of engineering and science, from both commercial and academic organizations. OpenFOAM has an extensive range of features to solve anything from complex fluid flows involving chemical reactions, turbulence and heat transfer, to solid dynamics and electromagnetics (see OpenFOAM.com). These advantages made us choose to set up OpenFOAM for air quality simulations on micro-scales; and as a first step to validate the OpenFOAM model against wind tunnel data for the turbulent dispersion and diffusion of pollutants in a street canyon and the influence of trees on the concentration field. In this study we aim at validation of an airquality modeling using CFD to obtain a practical, useful tool for urban air quality studies at microscales. Further, the validated model is applied to study the annual average effect of trees on the air quality in urban street canyons. Page 4
3. Modeling strategy 3.1 Wind tunnel validation data Several sets of wind tunnel studies have been published in the past that could be used to validate CFD models. Given our interest in air quality, we have selected the CODACS data set (see CODASC) which gives a wealth of data on the dispersion of traffic exhausts in urban street canyons with and without tree plantings. The results have been obtained for narrow street canyons with a ratio between the building height (H) and canyon width (W) of 1.0 and broader street canyons where this ratio equals 2.0. The street canyon setup for W/H =1.0 is depicted in Figure 1, the dimensions are a building height of 18 m, building length of 180 m, building width of 18 m and a canyon width of 18 m (Gromke et al., 2008). Two line sources represent the traffic in the canyon, extended by 10% out of the canyon on both sides. Tree plantings could be placed inside the canyon with varying porosities and degrees of disturbance in the flow pattern inside the canyon. The experimental data have been obtained in an atmospheric boundary layer wind tunnel at scale 1:150. In this test section, a boundary layer flow was developed with the mean velocity (u(z)) and turbulence intensity (I u ) profiles described by the following power law equations: ( ) (1) ( ) (2) The reference height is in this case the building height of 18 m with u z equal to 4.65 m/s. The concentrations have been determined close to both building walls inside the canyon. The mean concentrations per location inside the canyon are reported as normalized concentrations (3) With c the measured concentrations, u H the wind speed at building height and Q/l the line sources strength per unit length. Full description of the wind tunnel setup and a detailed discussion of all experimental result together with validation of simulations using the commercial CFD package FLUENT can be found in (Gromke et al., 2008; Buccolieri et al., 2009, 2011; Gromke and Ruck, 2008a, 2012; Gromke, 2011; Gromke and Ruck, 2007, 2008b). Results have been reported for three different wind direction; one with an angle of 90 between the street canyon axis (y in Figure 1) and the flow direction at the inlet (case depicted in Figure 1), one with an angle of 45 and with the flow parallel to the street canyon, an angle of 0. Page 5
Figure 1: Street canyon setup (reproduced from (Gromke et al., 2008)) 3.2 Model Setup CFD model All 3D simulations have been completed using the OpenFOAM CFD package. A solver has been developed starting from the standard OpenFOAM steady-state solver for incompressible, turbulent flow, SimpleFoam, coupled to support for the dispersion of pollutants and the influence of vegetation. This Reynolds-Averaging Navier-Stokes (RANS) equations solver s vegetation module captures the effect of trees on the momentum equations (sink), the turbulence equations (source/sink) and the dispersion (deposition). All simulations have been completed applying a k- epsilon model, which is a RANS-based turbulence, linear eddy viscosity model with two extra transport equations included to describe the turbulent properties of the flow. This type of model is chosen as it can be considered an industry standard. Our use of a k-epsilon model including vegetation terms is here considered a sufficiently accurate model given the large uncertainties present on the experimental data. The goal of this study is not to model this problem at the highest level detail; we aim at validating the k-epsilon approach to have a validated tool for urban air quality results at micro-scales. The influence of the vegetation on the turbulence has been added to the k-epsilon model in use. All concentrations reported troughout this report are normalized using equation 3. Page 6
Mesh The computational domain has been built using Gmsh (Geuzaine and Remacle, 2009) a threedimensional finite element mesh generator with built-in pre- and post-processing facilities. Gmsh supports both regular hexahedral grids and irregular grids. For the present application, we have combined both grid types into one. Inside and around the canyon, the highest resolution is requested and here a constant grid resolution of z/h = 0.028 and x/h = y/h = 0.05 is chosen with regular hexahedral cells. For the case with the flow perpendicular to the canyon, the vertical resolution is further increased to z/h = 0.014 to better capture the formation of the vortex above the building facing the atmospheric boundary flow. This domain resolution is used in the region within a distance of 1 H of the buildings. It is ensured the computational results are mesh resolution independent. Upon increasing the resolution beyond those given above, hardly any effects on the modeled flow patterns and concentration profiles are observed. To keep the amount of cells in the domain manageable the cell size increases by an expansion ratio of maximum 1.2 between connected cells outside this central region and the grid becomes irregular. The domain is vertically extended by 5H, where H is the building height, laterally by 5H and the extension in the flow direction is 8H and 15 H respectively in front (approach flow) and behind (the wake) the buildings. The final number of cells is in the range 2.800.000 cells to 5.000.000 cells depending on the vertical resolution. The computational domain is developed taking into account the best practice guidelines of the COST Action 732, a sketch is given in Figure 2. Figure 2: CFD computational grid for W/H = 1 Boundary conditions The boundary conditions of a CFD simulation are used to represent the influence of the environment outside the computational domain. They have a strong influence on the final results and have to be accurately determined. The inflow boundary conditions are defined as follows: Mean velocity profile u z given by the power law equation of eq. 1 Page 7
The turbulent kinetic energy, k, inlet profile: The turbulent dissipation rate, epsilon, profile: (5) The kinematic turbulence viscosity, nut, uniform 0 The pressure, p, has a zero gradient boundary The equations have been set using the groovybc tool for OpenFOAM, where δ is the boundary layer depth of 108 m (6H) of the computational domain, κ the Von Kàrmàn constant (0.40), u * = 0.52 m/s the friction velocity and C µ = 0.09 (Buccolieri et al., 2009). Symmetry boundary conditions have been specified to ensure a parallel flow near the lateral sides of the domain. The values for the velocities and the turbulence quantities of the inflow profile at the height of the top boundary have been applied to over the entire top boundary while the outlet boundary aims at ensuring a fully developed flow. The simulations are iteratively solved and all residuals dropped below 10-6, indicating full convergence has been reached. Second order upwind discretization schemes have been applied for k, ε, pressure and momentum. As described in (Blocken et al., 2007) it is important to correctly specify the wall roughness of the ground surface to avoid stream wise gradients in the atmospheric boundary layer flow. The equivalent sand-grain roughness height, K S, and the roughness constant, C S, are thus optimized to avoid these stream wise gradients. The value of K S has been set to 0.12 m combined with a C S value of 8. Using the following equation this combination represents an aerodynamic roughness length of 0.1 m. This represents rough, open terrain like agricultural land with scattered buildings. This seems to agree quite well to the wind tunnel set up where roughness elements have been introduced. (6) (4) The model buildings in the wind tunnel have a very smooth surface of Perspex; in the simulations their surfaces thus have low K S and C S values of only 0.0033 and 0.5. The settings of all simulations thus meet the basic requirements for ABL flow simulations listed by (Blocken et al., 2007) quite well: A sufficiently high mesh resolution in the vertical direction close to the bottom of the computational domain (e.g. height of the first cell < 1 m) A horizontally homogenous ABL flow in the upstream and downstream region of the domain A distance z P from the center point P of the wall adjacent cell to the bottom of the domain that is larger than the physical roughness height k S of the terrain Knowing the relationship between the equivalent sand-grain roughness height k S and the corresponding aerodynamic roughness length y 0. All four requirements should be satisfied in the upstream and downstream region of the computational domain and the first and third also in the central region. The cell heights are either 0.25 or 0.5 m depending on the case; both are in the range where the resolution is sufficiently high (< 1 m) and the distance from the center of the bottom cell to the ground surface, z P, is larger than k S (0.12 m). Page 8
It is well known that the turbulent Schmidt number, Sc T, has a significant influence on the simulation of pollutant concentrations using a k-ε model. The optimum values for Sc t vary with the flow characteristics and are distributed in the range 0.2 1.3 (Tominaga and Stathopoulos, 2007). For this study we have optimized the Sc t number for each wind direction and W/H ratio independently, the range of used values is 0.3-1.0. Modeling of trees By including the vegetation module, the effect of trees and other vegetation types can be incorporated in the simulation. Vegetation zones do affect the momentum equation (sink), the turbulence equations (source/sink) and the dispersion (deposition). For the momentum sink term, Cx, we have assumed the values 0 0.24 0.53 1.33 m -1, while for the deposition term LADvd the values range from 0 0.008 0.022 0.08 0.22 s -1. The momentum sink (same as pressure loss coefficient) and deposition term is assigned to the cells in the computational domain occupied by the tree crowns. The wind tunnel study does not consider deposition and the porosity is varied representing full scale momentum sink values of 0, 0.53 and 1.33 m -1. In the present study we have selected 1 extra momentum sink value and have studied the effect of deposition as well to obtain a broader view on the effects of trees in street canyons. Cx equals the product of the leaf area density, LAD, and the drag coefficient, C d ; while LADvd equals the product of LAD and the deposition velocity, v d. Several studies are available on these 3 parameters exist, yielding a range of values and agreeing the parameters depend on the kind of tree under study, and for the deposition on the pollutant under study. Litschke and Kuttler give a good overview of the deposition velocity for trees, summarizing several studies available which report a range of deposition velocities as a function of type of tree, particle size and atmospheric conditions (Litschke and Kuttler, 2008). The range of velocities is 0.0002 m/s to 0.3 m/s, with about 0.02 m/s an average value for PM 10. The leaf area index of several types of trees is discussed in (Lalic and Mihailovic, 2004). The range for LAD seems to be from 0.2 to 2.2 m 2 /m 3, with an average value through the canopy of full grown deciduous trees of about 0.8 m 2 /m 3. The effect of seasons leads to variation of this parameter throughout the year for deciduous trees. The third parameter which determines the effect of trees is the drag coefficient C d. Here the range for Cd is given as 0.1 0.5 (even 0.1 0.3 for most vegetation) by (Katul et al., 2003) and (Endalew et al., 2009). In this study, Page 9
4. Results Validation Study The OpenFOAM CFD model is set up to simulate most of the cases studied in the CODACS wind tunnel study. We validate against the results for both W/H = 1.0 and 2.0, for the three different wind directions with an angle between the inlet flow direction and the street canyon axis of 90, 45 and 0, for the cases without trees and with porous trees with varying pressure loss coefficients (wind tunnel λ = 80 and 200 m -1, corresponding to full scale values of 0.53 and 1.33 m -1 ). All together this forms 18 validation cases. For each case the concentrations inside the canyon at planes close to both buildings are analysed and discussed. We only investigated the cases with continuous trees, meaning tree crowns are adjactent, as the differences with lower tree densities are small. For W/H = 1.0, only the cases with densest trees have been investigated. 4.1 Flow and kinetic energy profiles The first step on the way to validation is inspection of the simulated flow profile. As discussed in the previous section, the inlet flow profile and boundary conditions have been selected for an isolated street canyon and atmospheric boundary layer flow. These settings reflect the conditions in the wind tunnel. A simulation has been completed without buildings to inspect to what degree the flow profile is retained over the domain. Figure 3 shows the flow profile is retained throughout the domain and agrees well with the experimental data. Changes in the profile between the inlet boundary and the first building are less than about 10% at any height, which is more than satisfactory. Figure 3: Vertical profile of the flow velocity in the CFD simulations at the inlet of the domain and the location of the flow facing building s front wall and the wind tunnel data of (Gromke and Ruck, 2007). Page 10
Height (m) CFD windtunnel validation Besides the flow profile, wind tunnel measurements have been completed for the turbulence kinetic energy of the approaching flow. These data are compared in Figure 4 with the simulated profile of k, both the inlet boundary and the location of the front face of the first building, of the simulation using the computational domain without buildings. At increased heights the profiles meet each other fairly well; near the surface the simulated k values do not meet the maximum observed in the wind tunnel data. This can indicate that the turbulence near the ground surface is somewhat underestimated with k-epsilon turbulence closure scheme in use. The under estimation of the turbulence is partly countered by optimizing the turbulent Schmidt number, Sc t, as described in literature (Tominaga and Stathopoulos, 2007). 120 100 80 60 40 20 0 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 turbulent kinetic energy (m 2 s -2 ) inlet frontfacea Gromke et al. Figure 4: Vertical profile of the turbulent kinetic energy in the CFD simulations at the inlet of the domain and the location of the front face of the first building and the measured wind tunnel data of the approaching flow. 4.2 Flow field around the buildings and inside the street canyon The next step after ensuring the flow and turbulence kinetic energy profiles are retained over the computational domain, is the study of the flow in the empty, tree-free street canyon. To start, the case with a flow perpendicular to the canyon, angle of 90, is further inspected. A sketch of the flow field in and around the canyon is shown in Figure 5, which explains the dominant effects that are observed for this configuration, a canyon vortex in the central region of the canyon and corner eddies that dominate closer to the corners of the canyon. For this configuration, flow fields have been measured using laser Doppler velocimetry (LDV). The vertical flow field components have been measured and can be compared with the CFD results, as shown in Figure 6. The contour plots Page 11
show similar profiles with reasonable quantitative agreement. The CFD k-epsilon turbulence model simulates lower vertical flow velocities inside the street canyon. Above the street canyon the canyon vortex has less influence on the flow, while the wind tunnel data show significant vertical velocities above the canyon. A scatter plot of the simulations against the wind tunnel 1-D LDV data has been made, depicted in Figure 7. The vertical flow near the canyon center is for the simulation only 50% of the measured values. The high coefficient of determination confirms the good agreement in the shape of the profiles. The data points outside the canyon, the blue diamonds which are not overlapped by the green triangles, differ however significantly from the experimental data. All together, the ventilation of the canyon in the simulation will be less, leading to possible increased pollutant concentrations. Figure 5: Sketch of the flow field around the isolated street canyon with perpendicular flow (90 ), reproduced from (Gromke and Ruck, 2007). Page 12
Figure 6: Normalized vertical velocities w + at y/h = 0.5 in a tree free street canyon with W/H =1.0 and 90 wind angle, on the left wind tunnel LDV measurements reproduced from (Gromke et al., 2008), on the right the CFD simulations. Figure 7: Scatter plot of the CFD simulated normalized vertical flow component vs. the LDV measured wind tunnel data (Gromke et al., 2008). The dashed black line represents the x=y line. The Page 13
full line is a linear fit to the data within the street canyon. The blue diamonds which are not overlapped by the green triangles are the data points directly above the canyon. Besides the effect of the canyon vortex protruding into the air flow above the canyon, a second phenomenon affects the flow in this region. Above the roof top of the windward building A, a small vortex is formed influencing the flow field and causing vertical components in the flow. To capture this effect well, the resolution in the vertical direction has been increased to z/h = 0.014. Still, the simulations show less flow separation than the wind tunnel observations. Additionally, the increased resolution leads to a better description of the street canyon vortex protruding into the upper street canyon flow. The limitation in modeling the roof top vortex is however inherent to the use of a k-epsilon model. The detailed comparison of the simulated flow field with the wind tunnel is only made for the W/H = 1.0 case at 90. For all other cases no experimental data are available and the flow field is discussed together with the presentation of the concentration profiles. Figure 8: A vector plot of the CFD simulated flow field around the windward facing building in the plane at y/h = 0.5. 4.3 W/H =1 90 : no trees This case without trees is the reference case for which the flow field has been discussed in the previous section. When comparing the flow fields, it is obvious that the ventilation of the street canyon is less sicne the central canyon vortex extends less to above the canyon. One can thus expect higher concentrations inside the canyon. To counter this reduced turbulence in the simulations, the Sc t number has been optimized to 0.3, bringing the concentrations into better Page 14
agreement with the data. Overall, the shapes of the concentration profiles near both walls agree fairly well, as shown in Figure 9. All average and maximum concentrations per wall are summarized at the end of this section, scatter plots of the simulations against the experimental data have been completed and the coefficient of determination has been reported in this table. Figure 9: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 90, no trees. Top two profiles are from the CFD simulations, bottom two reproduced from (CODACS). 90 : porous trees (λ = 200 m -1 -> C x = 1.33) The introduction of trees in the street canyon reduces the flow velocities. Reduced flow velocities means reduced ventilation of the pollutants and both the wind tunnel (WT) data and the CFD simulations show an increase in concentrations. The CFD and WT profiles agree well on average concentrations for Wall A. Increased maxima are however observed in the simulations, as the strength of the central canyon vortex is reduced while the strength of the corner eddies is over predicted compared to the experimental data. The profiles for Wall B agree fairly well in shape, but significant over predictions are present. Page 15
Figure 10: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 90, porous trees with C x = 1.33. Top two profiles are from the CFD simulations, bottom two reproduced from (CODACS). 45 : no trees The Sc t number has been kept at 0.3 for the cases with W/H = 1.0 and 45. Good agreement has been obtained for both cases. Page 16
Figure 11: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 45, no trees. Top two profiles are from the CFD simulations, bottom two reproduced from (CODACS). 45 : porous trees (λ = 200 m -1 -> C x = 1.33) The effect of trees is well simulated in this case near wall A. The agreement is much less for wall B only the effect on the maximum concentration near Wall B proves to be over predicted. Figure 12: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 45, porous trees with C x = 1.33. Top two profiles are from the CFD simulations, bottom two reproduced from (CODACS). Page 17
0 : no trees As for the previous cases, the Sc t number has been kept 0.3. Good agreement in both shape and absolute concentrations, including the maxima, is obtained. Figure 13: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 0 without trees. The top profile shows the CFD simulations, the bottom profile is reproduced from (CODACS). 0 : porous trees (λ = 200 m-1 -> Cx = 1.33) The introduction of trees in the canyon, leads to increased concentrations inside the canyon. The simulations give slightly higher average concentrations near the canyon walls with a lower maximum. Figure 14: Normalized concentration profiles inside the street canyon near the walls of both buildings for the case W/H = 1.0, 0, porous trees with C x = 1.33. The top profile shows the CFD simulations, the bottom profile is reproduced from (CODACS). Summary W/H = 1.0 The results from the simulations of all W/H = 1.0 cases are summarized in Table 1 and Table 2. For each individual case a scatter plot has been made and its coefficient of determination has been included in the Table 1. An example of this kind of plots is given in Figure 15. The first table compares the wall average concentrations and the effects of the introduction of trees. All trends Page 18
observed in the wind tunnel experiments are reproduced by our simulation. The introduction of trees in a street canyon leads to an increase in the concentration of pollutants emitted in the canyon. An important remark is that positive environmental effect trees, the deposition of pollutants on the leafs, has been left out for this validation. Overall, the agreement with the experimental data near Wall A proves to better than near Wall B. Only for the case 45 with trees, near wall B significant difference in the shape of the profile is observed (see Figure 12), as indicated by the low coefficient of determination. Further the trend in maximum value near wall B is not accurately reproduced either. For the wall maximal concentrations, larger differences in absolute values are simulated. Some differences in the flow profile can lead to a local maximum, while the average concentration is overall quite well simulated. Table 1: Summary of the wall average normalized concentrations near the buildings inside the idealized street canyon W/H = 1.0, presented CFD results vs. the wind tunnel (WT) results (CODACS). For each case the coefficient of determination of a scatter plot of the simulations against the experimental data and the relative difference (diff.) between the simulation and the exp. value is included. All trees have λ = 200 m-1 -> Cx = 1.33. Angle Wall No trees Trees Effect Trees WT CFD diff. R 2 WT CFD diff. R 2 WT CFD 90 A 19.7 23.0 17% 0.78 32.7 35.0 7% 0.77 66% 52% 90 B 5.34 8.03 50% 0.95 2.69 7.10 169% 0.83-50% -12% 45 A 18.4 14.3-22% 0.81 31.0 21.6-30% 0.75 69% 51% 45 B 3.70 2.41-35% 0.80 5.27 11.6 120% 0.27 42% 380% 0 A/B 7.10 7.22 2% 0.88 9.73 12.2 26% 0.88 37% 70% Table 2: Summary of the wall maximal normalized concentrations near the buildings inside the idealized street canyon W/H = 1.0, presented CFD results vs. the wind tunnel (WT) results (CODACS). For each case the relative difference between the simulation and the experimental value is included. Angle Wall No trees Trees Effect Trees WT CFD difference WT CFD difference WT CFD 90 A 42.6 68.0 60% 65.4 128.9 97% 54% 90% 90 B 11.8 20.2 71% 5.37 24.4 354% -55% 21% 45 A 46.5 45.0-3% 75.8 81.3 7% 63% 81% 45 B 10.8 5.89-45% 14.7 37.8 157% 37% 541% 0 A/B 32 25.2-21% 46.3 33.7-27% 45% 34% The validation against the W/H = 1.0 wind tunnel proves to be of the same level as the earlier valdation of CFD simulations using a commercial CFD software package such as Fluent, (Gromke et al., 2008). Page 19
Figure 15: Scatter plot of the normalized concentration near Wall A for the case 45 without trees. The dashed line represents the y=x line; the full line is a linear fit for this plot with equation as given in the graph. 4.4 W/H = 2.0 The validation against the (CODACS) wind tunnel data set has been continued on the cases with W/H = 2.0, with a comparison of the simulations without trees and those with trees (Cx = 1.33). The results of the CFD simulations using OpenFOAM have been summarized in the the tables below. All trends observed in the wind tunnel are in general well captured by the simulations. For each wind direction the Sc t number has been optimized, best results are optained applying Sc t = 1.0 for 90 and 45 and Sc t = 0.3 for 0. The absolute values agree even better than for the W/H = 1.0 cases, although lower coefficients of determinations are obtained. The maximal concentration agree fairly well with the experimental data as well. The only exception is the maximum concentration near wall B for the 90 cases. Due to the symmetry, this prves the hardest case to obtain a converged solution with all residuals meeting the criteria. The flow pattern leads to a isolation of the central region in the canyon where relatively high concentrations are simulated, yielding the higher maximum values. The individual contour plots of each case have not been included at the moment, but will be made available for the final report. Page 20
Table 3: Summary of the wall average normalized concentrations near the buildings inside the idealized street canyon W/H = 2.0, presented CFD results vs. the wind tunnel (WT) results (CODACS). For each case the coefficient of determination of a scatter plot of the simulations against the experimental data and the relative difference (diff.) between the simulation and the exp. value is included. All trees have λ = 200 m-1 -> Cx = 1.33. Angle Wall No trees Trees Effect Trees WT CFD diff. R 2 WT CFD diff. R 2 WT CFD 90 A 15.0 10.8-28% 0.68 20.9 19.5-7% 0.73 40% 81% 90 B 5.14 8.23 60% 0.73 3.46 5.88 70% 0.93-33% -29% 45 A 9.84 9.83 0% 0.45 18.4 13.2-29% 0.80 87% 34% 45 B 0.87 0.89 2% 0.55 3.77 3.28 18% 0.59 218% 269% 0 A/B 1.46 1.16-21% 0.83 2.10 2.42 15% 0.82 44% 109% Table 4: Summary of the wall maximal normalized concentrations near the buildings inside the idealized street canyon W/H = 2.0, presented CFD results vs. the wind tunnel (WT) results (CODACS). For each case the relative difference between the simulation and the experimental value is included. Angle Wall No trees Trees Effect Trees WT CFD difference WT CFD difference WT CFD 90 A 26.1 18.6-29% 39.5 35.1-11% 52% 89% 90 B 12.2 40.7 234% 8.97 21.4 138% -26% -47% 45 A 15.2 19.6 26% 43.4 43.1-1% 186% 120% 45 B 2.36 3.00 27% 7.47 6.89-8% 217% 130% 0 A/B 4.63 3.56-23% 9.98 8.11-19% 116% 128% Page 21
5. Annual average effects using meteo statistics The validated OpenFOAM CFD model has been used to study the year round effect of trees. This study focuses on the broader street canyons W/H = 2.0 as this type of canyon is more relevant for Belgian cities. Besides the three wind directions studied for validation, 90, 45 and 0, cases with 15, 30, 60 and 75 have been added. For each wind direction the following vegetation parameters have been varied: momentum sink C X deposition term LADv d From the variety of reported leaf area densities, LAD, deposition speeds and drag coefficients, we have selected ten combinations to cover the range of values. These three parameters depend on the type of tree, time of the year and location, type of pollutant. The combinations we have selected are: - C X = 0 m -1 and LADv d = 0 s -1 (absence of trees) - C X = 0.24 m -1 and LADv d = 0 s -1 - C X = 0.24 m -1 and LADv d = 0.008 s -1 - C X = 0.24 m -1 and LADv d = 0.08 s -1 - C X = 0.53 m -1 and LADv d = 0 s -1 - C X = 0.53 m -1 and LADv d = 0.0088 s -1 - C X = 0.53 m -1 and LADv d = 0.088 s -1 - C X = 1.33 m -1 and LADv d = 0 s -1 - C X = 1.33 m -1 and LADv d = 0.022 s -1 - C X = 1.33 m -1 and LADv d = 0.22 s -1 All seventy simulations, combination of seven wind directions and ten vegetation settings, have been completed using a k-ε model in OpenFOAM. A summary of all results is presented in Table 5, listing the wall average normalized concentrations near both buildings for the tree-free case and change in these concentrations when introducing trees. Several trends can be observed in this overview of CFD results. Significant changes in wall average concentrations are observed when including trees for all wind directions. The momentum sink effect of trees hinders the inflow of fresh air, lowering the wind velocities in the canyon and its ventilation, leading to increased concentrations. Even for the high deposition velocities, an overall increase is observed. For the 30 and 15 inflow, the presence of the trees changes the flow profile through the canyon. In the absence of trees, the inflow efficiently ventilates the pedestrian zone close to wall B, where the highest wind speeds are found. The inclusion of trees alters this flow pattern as it shields wall B partly from the inflow of fresh air. Near wall A, the concentrations however drop as the trees partly shield the pedestrian zone close to wall A from the emissions in the center of the canyon. As can be expected, including the deposition effects lowers the pollutant concentrations in all cases. Page 22
Table 5: Summary of the wall average normalized concentrations near the buildings inside the idealized street canyon W/H = 2.0 for all wind directions with a 15 interval. The presented CFD results are given as normalized concentrations for the tree-free case and as the change in concentration compared to the tree-free case for the nine different types of vegetation studied. Units: C X in m -1 LADv d in s -1. 90 75 60 45 30 15 0 Cx LADvd A B A B A B A B A B A B A/B 0 0 10.8 8.23 9.37 2.35 9.26 1.12 9.83 0.89 7.62 0.65 3.77 0.7 1.16 0.24 0 36% 2% 65% 30% 53% 53% 40% 72% 15% 202% -42% 240% 98% 0.24 0.008 34% -15% 64% 11% 52% 36% 39% 58% 11% 186% -46% 209% 90% 0.24 0.08 24% -62% 60% -51% 48% -28% 34% 2% -9% 114% -64% 126% 37% 0.53 0 56% -16% 89% 32% 67% 73% 48% 269% 6% 331% -60% 301% 106% 0.53 0.0088 52% -34% 88% 11% 66% 51% 47% 113% 1% 311% -64% 279% 95% 0.53 0.088 38% -79% 82% -50% 61% -12% 38% 62% -25% 220% -80% 163% 32% 1.33 0 81% -29% 103% 29% 71% 127% 34% 269% -34% 558% -79% 403% 109% 1.33 0.022 70% -56% 100% -10% 67% 82% 29% 239% -46% 502% -85% 339% 78% 1.33 0.22 46% -89% 91% -56% 56% 32% 15% 198% -68% 343% -94% 169% -10% The next step is to use annual statistics to consider the year round effect of trees. We are considering the influence of the change of the leaf area density throughout the year for deciduous trees, different deposition speeds and several orientations of the street canyon with aspect ratio W/H = 2.0. The wind directions can be split up in 24 sectors of 15. Since an inflow angle between 90 and 180 is the mirror image along the 90 plane and angles between 180 and 360 are the mirror images along the street canyon axis, the seven cases summarized in Table 5 give the normalized concentrations for all 24 wind sectors. The hourly meteo data for the year 2012 from the meteo station Luchtbal near Antwerp in Belgium have been used. For each hour the normalized concentrations are used from the respective wind sector to come to a year average concentration. As wind directions are unevenly distributed, the orientation of the street canyon influences the year average result. We have therefore considered four different orientations along the cardinal and intercardinal directions. Looking at the influence of deciduous trees in a moderate climate, the effect of the trees varies throughout the year. The growth of leafs and the leaf loss vary from year to year. We have approximated here this seasonal effect by using the results with trees for the period from the 15 th of April to the 1 st of November (Wilson and Baldocchi, 2000). For the winter period the results without trees are used, as no deposition is possible and the disturbance of the flow is significantly reduced. Following this procedure, the annual average effect near both buildings is given in Table 6. For each orientation, the annual concentration averaged close to both buildings is calculated starting from the data in Table 5 and further an average of these two values. Depending on the deposition and flow disturbance characteristics the effect of trees on the concentration of pollutants emitted in a street canyon ranges from 5% to 30% increase. The four sets of results which exclude effects of deposition are in Figure 16 applied to show the influence of the momentum loss parameter, C X. This figure plots the percentage increase in annual wall average concentration as a function of C X. The curve shows a strong increase for the introduction of very porous trees reaching a plateau for dense trees with a range of 19% to 40% increase. The spread between the results per canyon wall is caused by the different orientations and the prevailing winds, reflecting the different effects per wind sector summarized in Table 5. Page 23
For each momentum sink term, the influence of deposition is calculated and plotted in Figure 17. The deposition shows a similar trend as the momentum sink, the influence of more deposition becomes less pronounced with increasing deposition strengths. Overall, both figures prove the applied range of momentum loss and deposition term values covers the full range of no effect to nearly maximal effect of both parameters on the effects of trees on pollutant dispersion in street canyons. Table 6: Annual average effect of nine types of vegetation on the concentration of pollutants emitted in a W/H = 2.0 street canyon. The annual effect is given as the percentage difference with the tree free street canyon. W-E orientation N-S orientation NE-SW orientation NW-SE orientation Cx Ladvd A B average A B average A B Average A B average 0.24 0 21% 24% 22% 19% 25% 22% 24% 19% 21% 24% 19% 21% 0.24 0.008 18% 22% 20% 17% 23% 20% 21% 16% 19% 22% 14% 18% 0.24 0.08 5% 15% 10% 9% 11% 10% 10% 5% 7% 15% -1% 7% 0.53 0 29% 31% 30% 23% 37% 30% 32% 22% 27% 32% 22% 27% 0.53 0.0088 22% 28% 25% 20% 30% 25% 27% 18% 23% 29% 16% 23% 0.53 0.088 9% 20% 14% 10% 17% 14% 14% 6% 10% 21% 0% 10% 1.33 0 29% 30% 30% 19% 40% 29% 34% 19% 26% 36% 22% 29% 1.33 0.022 21% 26% 23% 13% 32% 23% 27% 12% 19% 30% 12% 21% 1.33 0.22 8% 15% 11% 2% 17% 10% 11% -2% 5% 20% -3% 8% The analysis has so far only considered emissions from traffic inside the street canyon and the background concentration has been assumed to be zero, as has been assumed in wind tunnel studies (CODACS, 2008). In reality, the concentrations of pollutants emitted by traffic are significantly affected by sources outside of the immediate surroundings of an urban street. The background concentration in an urban street canyon is made up of the dispersion of all other pollution sources in the city of interest and the background concentration of longer distance transport of pollutants. Page 24
Change in wall average concentration CFD windtunnel validation 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% W-E North W-E South N-S West N-S East NE-SW NW NE-SW SE NW-SE SW NW-SE NE 0 0,2 0,4 0,6 0,8 1 1,2 1,4 Cx (m -1 ) Figure 16: The percentage increase in annual wall average pollutant concentrations as a function of the momentum loss term, C X, for both walls of different street canyon orientations W/H = 2.0. Page 25
Change in wall average concentration CFD windtunnel validation 0% W-E North W-E South N-S West N-S East NE-SW NW NE-SW SE NW-SE SW NW-SE NE -5% -10% -15% -20% -25% 0 0,05 0,1 0,15 0,2 0,25 LADvd (s -1 ) Figure 17: The percentage increase in annual wall average pollutant concentrations as a function of the deposition term, LADv d, for both walls of different street canyon orientations W/H = 2.0. To study the effect of trees on air quality in the presence of background emissions we have further investigated the dispersion of elementary carbon (EC) in urban street canyons, an air pollutant produced through combustion processes which can be considered a measure for the amount of soot particles in the air. Diesel soot forms a large fraction of the environmental EC concentration and it is therefore chosen as a pollutant linked to traffic. EC plays a crucial role in the adverse environmental effect of particulate matter (PM). Secondly, PM 10 (particulate matter with a diameter below 10 µm) is selected as a representative pollutant with a much smaller link to local traffic as a variety of sources and long scale transport strongly affect environmental concentrations. For this study, we have applied for EC and PM 10 the annual average urban background concentrations determined in the city of Antwerp, Belgium, for the year 2009, respectively 1.81 µg m - ³ and 29.32 µg m - ³ (ref. Atmosys hindcast). The emissions of both pollutants are the estimations of the MIMOSA4 model (Mensink et al., 2000; Vankerkom et al., 2009) for the busy urban street Frankrijklei in Antwerp, respectively 11.22 and 22.45 µg s -1 m -1. These values have been used together with the wind velocities of the 2012 meteo data of Luchtbal, Antwerp, to calculate Page 26
absolute EC and PM 10 concentrations using equation 5. For each hour, based on the wind direction, the correct wind sector is chosen and its normalized concentration multiplied with the emission strength and divided by the wind speed of the respective hour and the building height. The annual average EC and PM 10 concentrations are thus calculated using this meteo statistical approach, again assuming full grown leaves on trees form the 15 th of April to the 1 st of November and no effect of trees outside this period. The effects of trees on the annual average EC concentrations for the different combinations of vegetation parameters have been summarized in Table 7. The emission from a busy urban street canyon form 40% of simulated EC concentrations inside the canyon. Taking into account EC background values, the annual effect of trees ranges from 1% to 13% increase, depending on the orientation and type of vegetation. For PM 10 the contribution of the emissions inside the canyon is limited to about 7.5% of the total concentrations. The effect of trees on the PM 10 concentrations is therefore only 0.2% to 2.6% increase, as summarized in Table 8. The highest negative influences of 13% EC and 2.6% PM 10 concentration increases are for simulated vegetation types which combine a high momentum sink and negligible deposition characteristics. The combination of C x = 0.24 and LADv d = 0.008 seems the best estimate for an average Belgian street canyon tree, having an average leaf area density in its canopy of 1.0 m² m -3, a reasonable drag coefficient of 0.24 and a deposition velocity of 0.8 cm s -1. This combination yields an annual average EC increase of about 8% and PM 10 increase of about 1.4%. These conclusions for the annual average effect of trees on air quality in urban street canyon are based on the following approximations: Isolated street canyon Artificial trees Touching tree crowns Solutions for a single inflow profile No resuspension of pollutants No emissions from vegetation No deposition of background emissions in the street canyon No thermal effects The calculations are based on the isolated street canyon configuration used in wind tunnel studies to enable validation (CODACS, 2008). The flow field is therefore a free atmospheric boundary layer flow, and the calculations are performed for a single inflow. The flow pattern in the street canyon is largely independent of the wind speed. The wind velocities inside the canyon scale with the inflow wind speed at a reference height while the flow pattern remains the same. For nearly windless conditions the pattern of the flow inside the canyon is however expected to differ. Thermal effects and the meteorological stability of the atmosphere can further influence the ventilation of the street canyon. The trees have in the simulations been characterized with a single set of representative dimensions and deposition and momentum sink terms, while this can vary between trees and for a single tree as function of height. Only the case with continuous vegetation along the street canyon has been studied, implying touching tree crowns. Wind tunnel studies have concluded the difference between touching and non-touching tree crowns is however limited (CODACS, 2008). The seasonal effects have been roughly reproduced, excluding possible effects of trees during winter. The CFD simulations have been configured excluding background concentration, a chosen background value is later added during the analysis. The deposition of the Page 27
background pollutants had to be neglected. Suspension and resuspension of pollutants can further effect the local concentrations. Biogenic emissions have been considered negligible compared to the traffic emissions. Table 7: Annual average effect of nine types of vegetation on the EC concentration in a W/H = 2.0 street canyon. The annual effect is given as the percentage difference with the tree free street canyon. Based on an annual average EC background concentration of 1.81 µg m -3 and an EC emission strength of 11.22 µg s -1 m -1. W-E orientation N-S orientation NE-SW orientation NW-SE orientation Cx Ladvd A B average A B average A B average A B Average 0 0 2.87 3.18 3.02 3.12 2.86 2.99 2.83 2.92 2.88 3.22 3.02 3.12 0.24 0 8% 10% 9% 8% 9% 9% 8% 8% 8% 10% 8% 9% 0.24 0.008 7% 9% 8% 7% 8% 8% 7% 7% 7% 9% 7% 8% 0.24 0.08 3% 6% 4% 4% 4% 4% 4% 3% 3% 5% 2% 4% 0.53 0 11% 13% 12% 11% 12% 11% 11% 10% 10% 13% 10% 12% 0.53 0.0088 9% 12% 10% 7% 10% 9% 9% 8% 9% 12% 8% 10% 0.53 0.088 4% 8% 6% 4% 6% 5% 5% 3% 4% 7% 3% 5% 1.33 0 11% 13% 12% 9% 13% 11% 12% 9% 10% 14% 10% 12% 1.33 0.022 8% 11% 10% 7% 10% 8% 9% 6% 8% 12% 7% 9% 1.33 0.22 3% 6% 5% 2% 5% 3% 4% 1% 2% 7% 1% 4% Table 8: Annual average effect of nine types of vegetation on the PM 10 concentration in a W/H = 2.0 street canyon. The annual effect is given as the percentage difference with the tree free street canyon. Based on an annual average PM 10 background concentration of 29.32 µg m -3 and a PM 10 emission strength of 22.45 µg s -1 m -1. W-E orientation N-S orientation NE-SW orientation NW-SE orientation Cx Ladvd A B average A B average A B average A B Average 0 0 31.5 32.1 31.8 31.9 31.4 31.7 31.4 31.6 31.5 32.1 31.8 31.9 0.24 0 1.4% 2.0% 1.7% 1.7% 1.6% 1.6% 1.5% 1.4% 1.5% 2.0% 1.6% 1.8% 0.24 0.008 1.2% 1.8% 1.5% 1.5% 1.4% 1.4% 1.3% 1.2% 1.3% 1.8% 1.3% 1.5% 0.24 0.08 0.5% 1.2% 0.8% 0.7% 0.7% 0.7% 0.6% 0.5% 0.6% 1.0% 0.3% 0.7% 0.53 0 1.9% 2.6% 2.3% 2.1% 2.2% 2.1% 2.0% 1.8% 1.9% 2.6% 1.9% 2.3% 0.53 0.0088 1.6% 2.3% 1.9% 1.8% 1.8% 1.8% 1.7% 1.5% 1.6% 2.3% 1.6% 2.0% 0.53 0.088 0.7% 1.5% 1.1% 0.8% 1.0% 0.9% 0.9% 0.6% 0.8% 1.5% 0.5% 1.0% 1.33 0 1.9% 2.6% 2.2% 1.8% 2.3% 2.1% 2.1% 1.6% 1.9% 2.9% 1.9% 2.4% 1.33 0.022 1.4% 2.1% 1.8% 1.3% 1.8% 1.6% 1.6% 1.1% 1.4% 2.3% 1.3% 1.8% 1.33 0.22 0.5% 1.1% 0.8% 0.3% 0.9% 0.6% 0.7% 0.2% 0.4% 1.3% 0.3% 0.8% Page 28
6. Conclusions This report presents good validation against wind tunnel data for the dispersion of pollutants obtained with an OpenFOAM CFD model: Vegetation - dispersion module included in OpenFOAM Flow profile maintained over the computational domain Succesfull validation for different wind directions, street canyons, with and without trees Dispersion of traffic exhaust in an urban street canyon further investigated Model ready for application of micro-scale air-quality studies Detailed analysis of year-round effects of trees for a range of vegetation types The dispersion of traffic exhaust pollutants has been theoretically studied for street canyons using CFD. A vegetation-dispersion module has been included in the open-source CFD package OpenFOAM to simulate the effects of vegetation. Firstly, this model approach has been extensively compared against wind tunnel experiments of pollutant dispersion in street canyons (CODACS, 2008) to ensure a validated CFD approach for micro-scale air quality studies. The simulations for street canyons with different aspect ratios W/H = 1.0 and 2.0 reasonably agree with these wind tunnel studies. The CFD results agree that trees have an overall negative effect on the air quality in the street canyons, with smaller concentrations associated to broader street canyons and wind directions closer to a parallel flow. Similar trends are observed for both canyon aspect ratios. For most configurations the form of the concentration contour plots agree well for simulations and experiments, proving a similar flow pattern in the canyon is simulated. The validated CFD approach is further applied to study the annual average effect of trees on the air quality. Prior studies have mainly investigated the effect of trees on the dispersion of pollutants for a single wind direction, drawing conclusions for the share of the pollutant concentration emitted in the canyon only. The results for W/H = 2.0 street canyon configuration have been enlarged with extra wind directions to cover all directions with 15 sectors. A meteo statistical analysis of these results have been made using the measured meteo conditions in Antwerp (Belgium) for the year 2012, taking into account the seasonal effects of leaf loss for the vegetation. The positive influence of pollutant removal by trees through deposition, largely neglected in prior studies, has been simulated as well. As the deposition and flow disturbance characteristics of vegetation vary depending on the type of vegetation, pollutant and meteo conditions, nine different vegetation types have been studied to give a range for the air quality influence of vegetation in urban street canyons. Traffic pollutants are of course not only emitted inside a street canyon. For both EC and PM 10 the annual effects have been simulated for absolute concentrations, using the background concentration and emission strengths for a busy street canyon road in Antwerp. Based on the current CFD simulations, we find an annual average increase ranging from 0.2% to 2.6% for PM 10 and 1% to 13% for EC, depending on the type of vegetation. Page 29