Building and Environment 126 (2017) 355–372

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

CFD and wind-tunnel analysis of outdoor ventilation in a real compact T

heterogeneous urban area: Evaluation using “air delay”
Nestoras Antonioua,b,∗, Hamid Montazerib,c, Hans Wigod, Marina K.-A. Neophytoua,
Bert Blockenb,c, Mats Sandbergd
a
Department of Civil and Environmental Engineering, University of Cyprus, Nicosia, Cyprus
b
Department of the Built Environment, Eindhoven University of Technology, Eindhoven, The Netherlands
c
Department of Civil Engineering, KU Leuven, Leuven, Belgium
d
Department of Building, Energy and Environmental Engineering University of Gävle, Gavle, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: Outdoor urban ventilation in a real complex urban area is investigated by introducing a new ventilation in-
Computational Fluid Dynamics (CFD) dicator – the “air delay”. Computational Fluid Dynamics (CFD) simulations are performed using the 3D steady
Urban microclimate Reynolds-Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES) approaches. The up-to-date litera-
Large Eddy Simulations (LES) ture shows the lack of detailed evaluations of the two approaches for real compact urban areas. This study
Steady RANS
further presents a systematic evaluation of steady RANS and LES for the assessment of the ventilation conditions
City breathability
Air delay
in a dense district in Nicosia, Cyprus. The ventilation conditions within the urban area are investigated by
calculating the distribution of the age of air. To better assess the outdoor ventilation, a new indicator, the “air
delay” is introduced as the difference between the local mean age of air at an urban area and that in an empty
domain with the same computational settings, allowing the comparison of the results in different parts of the
domain, without impact of the boundary conditions. CFD results are validated using wind-tunnel measurements
of mean wind speed and turbulence intensity performed for the same urban area. The results show that LES can
accurately predict the mean wind speed and turbulence intensity with the average deviations of about 6% and
14%, respectively, from the wind-tunnel measurements while for the steady RANS, these are 8% and 31%,
respectively. The steady RANS simulations overestimate the local mean air delay. The deviation between the two
approaches is 52% at pedestrian level (2 m).

1. Introduction
More than 80% of the population living in urban areas is exposed to
poor air quality levels [1]. In addition, the urban population has
reached 54% of the world's population, while it is expecting to increase
up to 66% by 2050 [2]. Consequently, without proper mitigation
measures, outdoor air pollution is expected to become the main reason
for premature mortality worldwide by 2050 [3]. In order to improve the
air quality in urban areas, strategies for improving urban ventilation
need to be evaluated and implemented.
Urban ventilation, also alternatively termed as urban breathability,
is defined as the ability of an urban area to dilute pollutants, heat and
moisture by exchanging air between inside and above of the urban
canopy [4], while earlier research has shown that it can play a sig-
nificant role in urban air quality [5]. Urban ventilation depends on the
combination of a wide variety of urban geometry parameters such as
frontal area density (λf), plan area density (λp) [6,7] as well as the
aspect ratio of urban canyons [8–12]. For example, it has been shown in
recent research work that building height variation can be beneficial in
terms of breathability levels [7,11,13,14], while larger aspect ratios of
urban canyons can lead to higher pollutant concentrations inside the
street canyons [8,11]. The impact of meteorological parameters, such as
wind speed, wind direction, temperature and humidity is also extremely
important for urban ventilation. Furthermore, the actual pollutant
concentration distribution in canyons was also found to be strongly
linked to the source release characteristics [12] while compact cities
appear to have less well ventilated conditions [6].
While the contribution of laboratory and field experiments in the
progress of our understanding of urban fluid dynamics is well-estab-
lished, in the past decades, Computational Wind Engineering (CWE) has
become an increasingly established research method and field [15],
covering a wide range of topics such as pedestrian-level wind comfort
around buildings [7,16,17], pollutant dispersion [6,18,19], outdoor
thermal environment [5,20–22], convective heat transfer [23–25], etc.

∗
Corresponding author. Department of Civil and Environmental Engineering, University of Cyprus, Nicosia, Cyprus.
E-mail address: antoniou.nestoras@ucy.ac.cy (N. Antoniou).

http://dx.doi.org/10.1016/j.buildenv.2017.10.013
Received 25 June 2017; Received in revised form 7 October 2017; Accepted 9 October 2017
Available online 13 October 2017
0360-1323/ © 2017 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
N. Antoniou et al. Building and Environment 126 (2017) 355–372

Table 1
Overview of CFD studies on urban wind flow and outdoor ventilation conditions in urban areas.

Study Ref. Scale Geometry Turbulence model Validation Focus

Juan et al. (2017) [26] Neighb. Real Steady RANS (RKE) Field a, c, d
Wen et al. (2017) [11] Neighb. Generic Steady RANS (SKE) WT a,c
Nazarian and Kleissl (2016) [27] Neighb. Generic LES (WMLES) WT b, c, d
Hang et al. (2015) [28] Neighb. Generic Steady RANS (SKE, RNG) WT a, b, d
Buccolieri et al. (2015) [29] Neighb. Generic Steady RANS (SKE) WT a, c, d
Ramponi et al. (2015) [10] Neighb. Generic Steady RANS (SKE) WT a, c, d
Lin et al. (2014) [30] Neighb. Generic Steady RANS (SKE) WT a, d
Panagiotou et al. (2013) [5] Neighb. Real Steady RANS (RSM) – a
Hang et al. (2013) [31] Neighb. Generic Steady RANS (SKE, RNG, RKE, RSM) WT a, d
Hang et al. (2012b) [32] Neighb. Generic Steady RANS (SKE, RNG) WT a, c
Hang et al. (2012a) [13] Neighb. Generic Steady RANS (SKE, RNG) WT a
Tominaga and Stathopoulos (2011) [18] Street Generic Steady RANS (RNG), LES (St.) WT –
Salim et al. (2011) [33] Street Generic Steady RANS (RSM, SKE) LES (Dyn.) WT –
Gousseau et al. (2011) [19] Neighb. Real Steady RANS (SKE), LES (Dyn.) WT c, d
Moonen et al. (2011) [34] Street Generic Steady RANS (RKE), LES (Dyn.) – a, d
Kim et al. (2010) [35] Street Real Steady RANS (SKE) WT a, d
Buccolieri et al. (2010) [6] Neighb. Generic Steady RANS (SKE) WT a, c
Hang and Li. (2010b) [36] Neighb. Generic Steady RANS (SKE) WT a, c
Hang and Li. (2010a) [7] Neighb. Generic Steady RANS (SKE) WT a, c, d
Hang et al. (2010b) [38] Neighb. Generic Steady RANS (SKE, RNG) WT a
Hang et al. (2010a) [37] Neighb. Generic Steady RANS (SKE) WT a
Hang et al. (2009b) [39] City Generic Steady RANS (SKE, RNG) WT a, d
Hang et al. (2009a) [40] City Generic Steady RANS (SKE, RNG) WT a, c, d
Yim et al. (2009) [41] Neighb. Generic Steady RANS (RKE) WT a, c, d
Bu et al. (2009) [9] Neighb. Generic Steady RANS (SKE) – a, d
Cheng et al. (2008) [42] Neighb. Generic Steady RANS (RNG) WT a, c
Blocken et al. (2008) [43] Street Generic Steady RANS (RKE) WT a, c, d
Bady et al. (2008) [44] Street Generic Steady RANS (SKE) – a, d
Yoshie et al. (2007) [45] Neighb. Real Steady RANS (SKE) WT, Field c, d
Blocken et al. (2007a) [46] Street Generic Steady RANS (RKE) WT a, c
Xie and Castro (2006) [47] Neighb. Generic Steady RANS (SKE, RSM, MKE) LES (Dyn.) WT –
Li et al. (2005) [48] Street Generic Steady RANS (RNG) WT a
Liu et al. (2005) [8] Street Generic LES (Dyn.) WT a
Skote et al. (2005) [49] City Generic Steady RANS (SSTKO) – a, c, d

Neighb. = Neighborhood scale, Real. = Realistic, WT = Wind-tunnel measurements, Field = Field measurements RANS = Reynolds-averaged Navier-Stokes, RKE = Realizable k-ε
model, MKE = Modified k-ε model, SKE = Standard k-ε model, RNG = Renormalization Group k-ε model, LES = Large Eddy Simulation, Dyn. = Dynamic Smagorinsky-Lilly SGS model,
St. = Standard Smagorinsky-Lilly SGS model, WMLES = Wall-Modeled SGS model, SSTKO = SST k-ω model. The entry “Focus” refers to different aspects that have been investigated in
each study, which are (a) geometrical characteristics (building height, street width, plan area density (λp), etc.), (b) heated surfaces, (c) grid resolution, (d) wind direction.

However, there is still need for further investigation of the performance
of CFD for different applications in wind engineering and urban phy-
sics. Table 1 provides a non-exhaustive overview of a number of CFD
studies focusing on urban ventilation. The following observations can
be made:
• Based on the spatial classification by Britter and Hanna [50], these
studies focus on either street scale (∼100 m), neighborhood scale
(∼1 km) or city scale (∼10 km). It should be noted that at the city
scale, the focus is on the general shape of cities, while individual
buildings are not modeled explicitly in these studies [39,40].
• The vast majority of the studies have been performed for generic
urban configurations, with idealized geometry, while there is a clear
gap on information of ventilation conditions in real urban en-
vironments where the geometry is much more complex.
• In addition, 3D steady RANS is the most commonly used approach
for such CFD studies. It should be noted that in almost all the studies
using transient simulations, the simulations were performed for
generic configurations, while for the study by Gousseau et al. [19],
in which transient simulations were used in a real urban area, the
focus was not on the ventilation conditions.
The comparison of RANS and LES has been a subject of many studies
[19,51–53]. However, based on the literature review that presents a
sample of relevant studies, most of the formal cross-comparative in-
vestigations deal with idealized urban geometries and the few dealing
with real complex urban geometries do not address outdoor ventilation
as measured through specific quantitative metrics; in fact, to the best of
our knowledge, the detailed evaluation of the two approaches for the
outdoor ventilation in complex urban areas has not yet been in-
vestigated. Therefore, the new contribution of this paper is to provide
an investigation of the outdoor ventilation conditions for real complex
urban areas and provide a quantitative evaluation of the accuracy of
LES and RANS (through the wind-tunnel study validation) for relevant
metrics for such outdoor ventilation.
It is well known that steady RANS is incapable of capturing the
inherently transient behavior of separation and recirculation down-
stream of buildings and of von Karman vortex shedding in the wake
[51,52,54,55]. Transition simulations with Large-Eddy Simulation
(LES), on the other hand can provide accurate descriptions of the mean
and instantaneous flow field around bluff bodies, at the expense of
much larger computational requirements. However, steady RANS is still
the most frequently used method for urban breathability studies based
on the information provided in Table 1, since it can be very effective in
terms of time and computational cost. It should be noted that earlier
studies on urban ventilation have shown the significant contribution of
turbulence diffusion in the air mixing between inside and above of
urban canopies [5,30,39,53,56], while turbulence can act as the main
mechanism controlling pollutant removal from an urban canopy layer
[12,57]. Given the important influence of turbulence on outdoor ven-
tilation, a detailed evaluation of steady RANS and LES for outdoor
ventilation of complex real urban areas is in order. Furthermore, using
the information provided in this paper in terms of accuracy and com-
putational time, engineers, urban planners and researchers can easily
decide which of these turbulence modeling approaches would be more
beneficial.

356
N. Antoniou et al.
To gain insights into the performance of steady RANS compared to
LES for predicting urban ventilation in a real compact heterogeneous
urban area, this paper presents steady RANS and LES CFD simulations
of the wind flow and the local mean age of air (τp ) distribution in a
dense district in the old city of Nicosia, Cyprus. Age of air represents the
time it takes for the external air to reach a location after entering the
urban canopy layer [58]. The evaluation is based on validation with
wind-tunnel measurements of mean wind speed and turbulence in-
tensity for a reduced-scaled urban configuration representing this dis-
trict. The wind-tunnel measurements, CFD simulations and ventilation
analysis are performed for a reduced-scale urban model.
The concept of age of air was firstly introduced as a ventilation
indicator for indoor environment, where it was used to characterize the
freshness of the air, representing the time that has passed since the air
entered the room [59]. This concept can be interpreted as the link
between a concentration level and a time scale [6]. For indoor en-
vironments, the inlets and outlets are easily defined, according to the
position of the openings. However, in an open environment such as an
urban area, this is not straightforward. In order to address this, Hang
et al. [40] used the effective age of air, where the emission source is set
in the entire urban canopy. This was also applied in other studies with
generalized geometries [10]. For real complex urban environments
such the case of this study, where the urban canopy cannot be easily
defined, this method is less applicable. Other currently available me-
trics such as exchange velocity, pollutant exchange velocity [4,5,12],
air exchange rate [48] and pollutant exchange rate [8] characterize
ventilation using a process occurring over the rooftop level of canyons,
not the process of dilution, removal and recirculation of contaminants
which occur inside the urban areas. This motivation arises from the
interest that within large long canyons in highly complex real urban
areas local ventilation processes are bound to occur in a non-uniform
357
Building and Environment 126 (2017) 355–372
Fig. 1. Case study area and surroundings: (a) Top
view, (b) Aerial view taken from point 1 on
Fig. 1a and (c) taken from the Shakolas tower
indicated with point 2 on Fig. 1a. (d) Street view
at Ledras street. Black line on Fig. 1a indicates
borders of the case study area.
way. To cover this void, and to better assess the outdoor ventilation for
the urban area, a new indicator called air delay (τd) is introduced,
which is actually the difference between the local mean age of air inside
the area of interest, and the local mean age of air in an empty domain
(when the buildings are not present) with the same computational
settings and parameters. The air delay is independent of the boundary
conditions as it presents the amount of time delay caused by the pre-
sence of the urban geometry in each point inside the domain. In addi-
tion, the air delay parameter can be used to easily assess and compare
ventilation conditions in different regions of an urban area, but also in
different urban areas.
In Section 2 the selected urban area and its surroundings are pre-
sented. Section 3 describes the wind-tunnel measurements. Section 4
describes the computational settings and parameters for the CFD si-
mulations. In Section 5, the validation of the CFD results with the wind-
tunnel measurements is presented, the performance of LES versus
steady RANS is evaluated, and the limitations and further work are
discussed. The main conclusions are presented in section 6.
2. Description of urban area and surroundings
The complex heterogeneous urban area under study covers
0.247 km2 and is composed of low-rise and medium-rise buildings
(Fig. 1), corresponding to Local Climate Zone 3 (compact low-rise
buildings) based on the classification made by Steward and Oke [60]
and with hot-summer Mediterranean climate (Csa) based on the
Köppen-Geiger classification [61]. The average and maximum building
height is about 8.1 m and 45.5 m (Shakolas Tower), respectively. The
plan area density λp is 0.40, as derived from European Research Project
TOPEUM [20]. Most of the buildings have flat terrace-type roofs and
internal courtyards typical of urban textures in Mediterranean cities.
N. Antoniou et al.
The area of interest is composed of narrow streets with widths ranging
from 6 m to 10 m, where the aspect ratio of the street canyons (width to
height) ranges from 0.5 to 1. All streets are single direction for vehicles,
while the main street crossing the area in the south-north direction is
used exclusively by pedestrians (Ledras Street shown in Fig. 1a and d).
The lack of green areas in combination with the use of concrete and
asphalt, as main building materials, lead to relatively high air tem-
peratures inside the street canyons during the summer period.
In this study, we also investigate the association of urban ventilation
with morphological and aerodynamic characteristics of the urban area.
To do so, a scale-adaptive description of the morphological and aero-
dynamic parameters is used, based on the Multi-Resolution Analysis
(MRA) method proposed by Mouzourides et al. [62]. The area is divided
into cells of horizontal surface area of 64 × 64 m2 and produces a
spatial varying map of the parameters. Fig. 2 presents the building
height (H) (Fig. 2a), aerodynamic roughness length (z0) (Fig. 2b), plan
area density (λp) (Fig. 2c) and the zero plane displacement height dis-
tribution (zd) (Fig. 2d) of the area of interest. The calculation of z0 and
zd has been done using the relations proposed by Kastner-Klein and
Rotach [63] derived after a series of wind-tunnel measurements in
358
Building and Environment 126 (2017) 355–372
Fig. 2. (a) Top view of the case study area with
buildings colored by building height (zHmax =
45.5 m). Results of Multi-Resolution Analysis: (b)
Aerodynamic surface roughness length z0 (z0max
= 0.5 m), (c) Plan area density λp (λpmax =
0.8 m), (d) Zero plane displacement height zd
(zdmax = 7 m).
which they related the aerodynamic roughness parameters with planar
density and the average building height of the investigated neighbor-
hood. Then the MRA method by Mouzourides et al. [62] was used to
provide scale-adaptive values of the z0 and zd at different resolutions
relevant and appropriate to our calculations. Since z0 and zd are both
defined based on geometrical characteristics of the urban area such as
average building height and planar density, it would be reasonable to
investigate their relationship with the breathability levels. Moreover,
the analysis of the aerodynamic characteristics is considered extremely
important in order to predict and describe the urban wind flow and
pollutant dispersion at different scales [50].
3. Wind-tunnel measurements
The wind-tunnel measurements are conducted in the atmospheric
boundary layer wind tunnel at the University of Gävle, Sweden. The
closed-circuit wind-tunnel is 11 m long and has a test section of
3 × 1.5 m2 (width × height). A reduced-scale model (1:250) of the
urban area is manufactured (Fig. 3). The blockage ratio is 2.7%. The
buildings are modeled with details of 2 m (full scale). The
N. Antoniou et al.
measurements are performed for wind direction 290° from North, de-
duced as prevailing wind direction from analyzing relevant associated
field measurements in the broader area [64]. The atmospheric
boundary layer is generated by roughness elements and spires. The
mean wind speed and turbulence intensity are measured using hot-wire
anemometers at 1261 points along 38 vertical lines (Fig. 4). The mea-
surement lines are located in different parts of the domain as shown in
Fig. 4: (i) 3 lines in the upstream part of the area of interest to capture
the approach-flow profiles (Fig. 4a); (ii) 3 lines downstream of the area
(Fig. 4b); (iii) 16 lines on top of the buildings (Fig. 4c) and (iv) 16 lines
inside the urban canyons (Fig. 4d). The heights of the measurement
points are distributed in a way to have more information near the
ground plane, with intervals as shown in Fig. 5, where Hmax is the
359
Building and Environment 126 (2017) 355–372
Fig. 3. (a) Top view of reduced-scale model used in wind-
tunnel measurements. (b) Close up of the model around the
Shakolas Tower.
height of the highest building (45.5 m in full scale). The resulting
vertical profile of mean wind speed at the middle upstream measuring
location “2” (Fig. 4a) is represented by a log law with an aerodynamic
roughness length z0 = 0.008 m (2 m in full scale) and a friction velocity
∗
uABL = 0.79 m/s. The results of the wind-tunnel measurements will be
presented along with the CFD results in Section 5. In this study, due to
lack of high-resolution on-site and wind-tunnel data of age of air for
complex urban areas, the CFD validation study is performed based on
wind-tunnel measurements of mean wind speed and turbulence in-
tensity.
Fig. 4. Top view of the case study area with the
position of measurement lines: (a) Upstream of
the urban area (b) Downstream of the urban area
(c) at roof top levels (d) at street levels.
N. Antoniou et al.
4. CFD simulations
4.1. Computational geometry, domain and grid
The computational geometry of the urban area reproduces the
geometry in wind-tunnel measurements at scale 1:250. The dimensions
of the computational domain are chosen based on the best practice
guidelines by Refs. [65–67] (Fig. 6). The height of the domain is 6Hmax
where Hmax is the maximum building height. The upstream and
downwind domain size is 5Hmax and 15Hmax, respectively. The width of
the domain is set the same as the wind-tunnel width resulting in a
blockage ratio of 2.7%, which is below the maximum value of 3.0%
recommended by the aforementioned guidelines. The computational
grid was created using the surface-grid extrusion technique by van
Hooff and Blocken [68]. The procedure was executed with the aid of the
pre-processor Gambit 2.4.6, resulting in a grid with 13,408,998 hex-
ahedral cells. The computational grid is shown in Fig. 7.
4.2. Boundary conditions
The boundary condition types are depicted in Fig. 6. The inlet
boundary conditions (mean wind speed U, turbulence kinetic energy k
and turbulence dissipation rate ε) are based on the measured incident
profiles of mean wind speed and longitudinal turbulence intensity Iu
along the middle line upstream of the model (line with label “2” in
Fig. 4a). The aerodynamic roughness length for the surroundings is
determined based on the measured incident profiles of mean wind
speed, as it was described in Section 3. For the inner part of the domain,
Fig. 6. Perspective view of computational domain with indication of main dimensions
and boundary conditions.
360
Building and Environment 126 (2017) 355–372
Fig. 5. Height of measuring points corresponding to vertical
lines presented in Fig. 4 (Hmax = 0.182 m).
where the buildings are modeled explicitly, i.e. with their main shape,
the ground plane and building walls are modeled as smooth walls with
z0 = 0 m. At the outlet of the domain, zero static gauge pressure is
imposed.
Note that in the measurements, the lateral homogeneity of the ap-
proach-flow profile is investigated measuring the mean wind speed and
turbulence intensity along the three lines (Fig. 4a), upstream of the area
of interest, as presented in Fig. 8. The average deviation between the
line in the middle and the other two lines is within 4% for mean wind
speed and 6% for turbulence intensity. The turbulence kinetic energy k
is calculated from the measured U and Iu using Eq. (1) where “ɑ” is a
constant parameter ranging between 0.5 and 1.5. In this study, ɑ = 1 is
taken, as recommended by Tominaga [66]. The turbulence dissipation
rate ε at height z from the ground is given by Eq. (2) where uABL
∗
is the
ABL friction velocity and κ the von Karman constant (0.42).
k (z ) = ɑ (Iu (z ) u (z ))2 (1)
∗3
uABL
ε (z ) =
κ (z + z 0) (2)
4.3. Computational settings
The commercial CFD code ANSYS/Fluent 15.0 based on the finite
volume method is used to perform the simulations. The simulations are
performed on the HPC cluster (×86_64 architecture) at the Department
of the Built Environment of Eindhoven University of Technology, the
Netherlands. For each simulation, a 22-core node (Intel(R) Xeon(R)
CPU - ×5650 @ 2.67 GHz), with 64 GB of system memory is used. The
computational settings and parameters for the steady RANS and LES
simulations are presented in Sections 4.3.1 and 4.3.2, respectively.
4.3.1. Steady RANS
The 3D steady RANS equations are solved in combination with
realizable k-ε turbulence model by Shih et al. [69] for closure for in-
compressible isothermal turbulent flow. The standard wall functions by
Launder and Spalding [70] are used with roughness modification by
Cebeci and Bradshaw [71]. The values of the roughness parameters, i.e.
the sand-grain roughness height ks and the roughness constant Cs, are
determined using their consistency relationship with the aerodynamic
roughness length z0 introduced by Blocken et al. [46,72]:
9.793z 0
ks =
Cs (3)
Note that up to this version of ANSYS/Fluent, the ks value cannot be
larger than yp (distance between the center point of the wall-adjacent
cell and the wall). Therefore, in this study, ks is set equal to yp and Cs is
chosen to satisfy Eq. (3). The SIMPLE algorithm is chosen for pressure-
velocity coupling and second-order discretization schemes are used for
both the viscous and convection terms of the governing equations.
N. Antoniou et al.
Fig. 7. (a) Aerial view of the case study area and surrounding buildings, (b and c)
Corresponding view of the computational grid (13,408,998 cells).
Convergence is assumed to have been achieved when the scaled nor-
malized residuals stabilize at a minimum of 10−6 for continuity, x, y
and z momentum, 10−8 for k and 10−11 for ε.
4.3.2. LES
The three-dimensional time dependent Navier-Stokes equations are
solved with LES as a turbulence modeling approach, for incompressible
isothermal turbulent flow. The closure of the equations for the filtered
361
Building and Environment 126 (2017) 355–372
velocity is obtained using the Smagorinsky-Lilly sub-grid scale model
[73,74]. Time discretization is second-order implicit and for pressure
interpolation second order is employed. For time advancement, the
iterative time step method is adopted [75–77], where all the equations
are solved iteratively for a given time-step and for a number of outer
iterations per time step [78]. The specification of the number of outer
iterations depends on the problem complexity [79]. The results of a
sensitivity analysis, performed in the present study, reveal that the use
of 10 outer iterations per time step is sufficient.
The Vortex Method (VM) [80] is used as turbulent inflow generator
technique with 190 2D vortices. The satisfactory performance of this
technique was already shown in several occasions in the past
[19,34,80].
The time-step value (Δt) was 0.0003 s, which corresponds to an
average CFL number ( = u × Δt/hcell), where u is the local velocity
value and hcell is the local cell size) of 0.2. The maximum CFL number is
4.8, which occurs at the top of the domain, far away from the area of
interest. Two initialization steps are performed. First, the simulations
are initialized with the solution from the steady RANS simulation on
which random noise is superimposed [78]. This initialization does not
affect the results, but it reduces the computational time. Second, an LES
initialization run is performed with an initialization period (Tinit) of 4 s,
corresponding to approximately 4 flow-through times (time needed for
wind flow to cover the distance from inlet to outlet with a reference
wind speed):
Lx
Tft =
Uh (4)
where Uh is the wind speed at a the height of the highest building in the
domain and Lx is the length of the computational domain [19,81]. After
initialization, LES simulations and sampling are performed for 20 s,
corresponding to 20 flow-through times.
4.3.3. Age of air and air delay
The ventilation is assessed using the local mean age of air (τp ), as-
suming that the air originating from outside the urban area is cleaner.
The homogeneous emission source method was used in which the local
mean age of air can be calculated from the local concentration C (kg/
m3) of each cell, based on the uniform release rate ṁ (kg/(m3s)) of the
source where the emission is released homogeneously in the entire
domain:
C
τp =
ṁ (5)
The turbulent Schmidt number (Sct = νt / Dt ) is assumed to be con-
stant in the whole domain (Sct = 0.7 where νt is the turbulent viscosity
and Dt is the turbulent diffusivity of the species), in line with similar
Fig. 8. Measured profiles of (a) ratio of normal-
ized streamwise mean wind speed, U, to reference
wind speed at Shiakolas Tower height, Uref, and
(b) streamwise turbulence intensity in the wind-
tunnel measurements.
N. Antoniou et al.
previous studies with the passive homogeneous emission source method
[40,44,82].
For the steady RANS simulations, the local mean age of air is cal-
culated using the mean concentration of the passive homogeneous
emission source. For LES, it is derived from the instantaneous local age
of air values for each time step using Eq. (6):
n along the vertical lines positioned at the roof top levels (points with
∑i = 0 τ pinst
τ pLES =
n (6)
where τ pLES is the local mean age of air, τ pinst is the instantaneous local
mean age of air and n is the number of time steps during the averaging
period.
5. CFD results and discussion
5.1. Comparison with wind-tunnel experiments
In order to check the horizontal homogeneity of the approaching
flow, the appearance of streamwise gradients of mean wind speed and
turbulence intensity between the inlet and the upstream measuring
locations is analyzed by comparing the CFD results of the normalized
streamwise velocity, U/Uref, and streamwise turbulence intensity, Iu,
with the measured data along the line indicated by the point with label
“2” upstream of the urban area (Fig. 4a). The results are shown in
Fig. 9. Note that Hmax = 0.182 m (45.5 m in real scale) is the height of
the highest building in the domain and Uref = 6 m/s is the mean wind
speed at z = Hmax at the inlet. According to this figure, the average
deviation of U/Uref from the measured data is about 5% and 2% for
steady RANS and LES, respectively. For Iu, however, this deviation is
more pronounced along the line and it reaches its maximum value of
35% and 69%, respectively. The roughness modeling is considered the
main reason to cause this decay. For LES a smooth ground surface is
assumed, which leads to reduction of turbulence levels as the flow is
developing. This is less intense in RANS where wall functions are used
and roughness of wall boundaries is modeled using the roughness
parameters (details in section 4.3.1). This decay will be a factor con-
tributing to possible discrepancies between the CFD results and the
wind-tunnel measurements inside the area of interest, as also reported
in previous studies [66,83]. As suggested by Blocken et al. [46], to
reduce the streamwise decay of turbulence intensity, the standard de-
viation of the measured stream-wise wind speed is artificially increased
by a constant factor for all measuring heights, leading to higher values
of turbulence intensity along the line. The results show that a factor of
0.3 is needed to achieve a satisfactory match between the LES results
and the measured data. For the steady RANS simulations, this factor is
0.1. No adjustment is applied for the inlet mean velocity profiles. It can
362
Building and Environment 126 (2017) 355–372
Fig. 9. Comparison of wind-tunnel measurements
and CFD results of (a) normalized streamwise
mean wind speed and (b) streamwise turbulence
intensity along line with label “2” upstream of the
urban area (see Fig. 4a) for two cases: with and
without adjusted approach-flow profile.
be seen that this adjustment significantly reduces the appearance of the
streamwise decay. In this case the decay of Iu drops to 25% and 19% for
steady RANS and LES, respectively. The deviations are about 6% and
4% for U/Uref.
Figs. 10 and 11 compare the wind-tunnel results and CFD results of
normalized streamwise mean wind speed and turbulence intensity
labels 2, 7 and 16 in Fig. 4c) and inside the street canyons (points with
labels 6, 8 and 14 in Fig. 4d), respectively. For LES, a good agreement is
observed for both the mean wind speed and the turbulence intensity.
This is especially the case for the mean wind speed where the deviation
is about 4% for the six lines. For the turbulence intensity, higher de-
viations can be clearly observed in Figs. 10f and 11f, where LES tends to
overestimate the results, though it can still capture the trends. Note that
the line indicated by the point with label “16” in Fig. 10 is located in the
wake of the Shakolas tower, while the line indicated by the point with
label “14” in Fig. 11 is located close to the ground inside a canyon.
Steady RANS also provides a satisfactory agreement for the mean wind
speed for most of the lines, though its deficiency in reproducing the
flow pattern around the tower can be clearly observed in Fig. 10e–f
where the flow is complex. This can be explained by the incapability of
RANS in predicting the periodic motions of the flow [51]. In this case,
the average deviation for the six lines presented in Figs. 10 and 11 is
about 20%. The average deviation increases to about 37% for the tur-
bulence intensity for these lines. This is in line with the results of earlier
studies in which the deficiency of steady RANS in capturing inherently
transient behavior of separation and recirculation downstream of bluff
bodies and of von Karman vortex shedding in the wake has been shown
[18,19,34]. Note that the higher accuracy by LES is obtained at the
expense of much larger computational costs (factor 15 compared to
steady RANS).
Table 2 presents the overall average deviation of mean wind speed
and turbulence intensity for all 1261 points along the 38 lines. The
overall deviations for the case if the inflow adjustment at the inlet
would not have been implemented is also reported in this table. The
average deviation of mean wind speed is 8% and 6% for steady RANS
and LES, respectively. This is about 31% and 14% for turbulence in-
tensity. It can be seen that using the adjusted profiles at the inlet can
significantly improve the agreement for the LES results. For steady
RANS, however, this adjustment has less impact on the accuracy of the
results.
Fig. 12 presents the distribution of the normalized mean wind speed
at pedestrian level (2 m from the ground), and the normalized turbu-
lence kinetic energy at the average canyon roof height (6.25 m from the
ground) obtained by steady RANS and LES, given the high importance
of turbulence fluctuations on top of urban canyons [32,38]. Considering
N. Antoniou et al.
that LES outperforms RANS as shown in Figs. 10 and 11, Fig. 12 in-
dicates that RANS generally underestimates the local mean wind speed,
in line with previous studies on pedestrian-level wind conditions [84].
The overall average deviation between steady RANS and LES is about
45%. Nevertheless, the channeling effect between parallel buildings
[85], lower wind velocities inside street canyons and courtyards [21]
and wake regions are well predicted by both steady RANS and LES. On
the contrary, a significant difference in the distribution of the normal-
ized turbulence kinetic energy can be clearly observed in Fig. 12c and d
where values are underestimated by RANS with about 70% deviation
from LES. Similar findings were also reported in previous CFD studies
for wind flow around isolated buildings [51,54,56] and inside complex
urban areas [19,34].
363
Building and Environment 126 (2017) 355–372
Fig. 10. Comparison of wind-tunnel measure-
ments and CFD results of (a, c, e) normalized
streamwise mean wind speed and (b, d, f)
streamwise turbulence intensity along three lines
(with labels “2”, “7” and “16” in Fig. 4c) located
on top of buildings.
5.2. Outdoor ventilation
In this paper, the results will be presented as the difference between
the local mean age of air inside the area of interest τp , and the local
mean age of air in an empty domain (when the buildings are not pre-
sent) with the same computational settings and parameters, τ pEMP . This
difference, which is called “air delay” in the remainder of this paper, is
given by Eq. (7):
τd = τp − τ pEMP (7)
where τd is the local mean air delay, τp is the local mean age of air in
the domain when the urban geometry is present, and τ pEMP is the local
mean age of air in an empty domain. Note that in CFD simulations of
N. Antoniou et al. Building and Environment 126 (2017) 355–372

Fig. 11. Comparison of wind-tunnel measure-

ments and CFD results of (a, c, e) normalized
streamwise mean wind speed and (b, d, f)
streamwise turbulence intensity along three lines
(with labels “6”, “8” and “14” in Fig. 4c) located
inside street canyons.

Table 2 largely dependent on the definition of the “inlet” based on which the
Average relative error from wind-tunnel measurements for all measuring points. calculations are performed. This can therefore be more problematic for
complex urban geometries, like that in the present study, as the local
Without inflow adjustment With inflow adjustment
mean age of air distribution inside the urban area is more sensitive to
eAVG(%) RANS LES RANS LES the definition of the initial values. The air delay is however in-
U 9 9 8 6 dependent of the initial values as it presents the amount of time delay
Iu 41 48 31 14 caused by the presence of the urban geometry in each point inside the
domain.
Fig. 13 shows contours of the normalized air delay at pedestrian
outdoor ventilation, the age of air distribution inside an urban geo-
height (2 m from the ground) obtained by steady RANS and LES. To
metry is normally assessed as the time that it takes for the external air
normalize the results, the maximum local mean air delay in the LES
to reach a certain location after entering the computational domain
simulations, τdmax = 0.5 s, is used. Fig. 14 presents the distribution of
[6,10,40]. Therefore, the age of air inside an area of interest can be
the difference between the local mean air delay obtained by the two

364
N. Antoniou et al.
approaches. For easier comparison and generalization of the conclu-
sions, the results of mean air delay are normalized with the maximum
value in the area of study. The distributions presented in Fig. 13a and b
are fairly similar, though steady RANS tends to overestimate the air
delay at pedestrian height. In this case, the overall average deviation
between steady RANS and LES is about 52%. In addition, the influence
of building height variations on the local mean air delay values is well
predicted by the two approaches. A closer look at this figure reveals
that steady RANS and LES predict a relatively low level of air delay in
parts with high building height variations. For example, parts B6, B7,
C4, C5 and C6. The opposite trend is predicted by both approaches in
parts with relatively low building height variations, where the local air
delays are relatively higher. This is especially the case in parts E5, E6,
E7, F5, F6, F7, G5, G6 and G7 (Fig. 2a). This is in agreement with findings
of previous studies performed for generic urban configurations in which
the positive influence of building height variation for urban ventilation
has been reported [7,13].
In order to reveal more clearly differences and better understand the
performance of steady RANS and LES, two vertical planes at different
locations inside the area of interest are selected, where contours of the
normalized mean wind speed, turbulence kinetic energy and local mean
365
Building and Environment 126 (2017) 355–372
Fig. 12. Contours of (a) normalized mean wind
speed at pedestrian level (2 m from the ground)
and (c) turbulence kinetic energy at canyon top
level (6.25 m from the ground) obtained by
steady RANS. (b, d) Same for LES (Umax = 6 m/s,
Uref = 6 m/s).
air delay are provided. Fig. 15 shows the location of the two planes,
Plane A and Plane B. Fig. 16 presents the results across plane A. It can
be seen that the approaching wind diverges over and around the tower,
resulting in a substantial amount of fresh air to flow downwards in front
(windward side) of the tower and to enter the Ledras Street. This leads
to a lower level of air delay in this region that is well predicted by both
steady RANS and LES. It should be noted that the turbulence levels
across the roof level and inside the street canyon (with label “1” in
Fig. 16) are well predicted for both steady RANS and LES, though the
turbulence kinetic energy is overestimated by the steady RANS at the
stagnation point in front of the tower and especially near the leading
edge; this observation is in accordance with similar findings of previous
studies [51,54,56]. Behind the tower, a relatively large separation re-
gion is generated. The extent of the separation region is larger for
steady RANS than for LES. As shown in Fig. 17, this leads to different
flow fields inside canyons 2 and 3 downstream of the tower. In addi-
tion, less turbulence is swept around the leading edge and over the roof
of the tower into the downstream region for steady RANS, leading to
lower levels of turbulence on top of street canyons with labels “2” and
“3” with deviation from LES of 65% and 85%, respectively. Figs. 16e
and f clearly indicate a significant difference in the local mean air delay
N. Antoniou et al.
Fig. 14. Distribution of ratio of difference between local mean air delay at pedestrian
level (2 m) obtained by RANS and LES along with buildings colored by building height
(τdmax = 0.5 s and Hmax = 45.5 m).
distributions obtained by the two approaches especially in the street
canyon with label “3”. The average deviation between steady RANS and
LES inside the street canyons with labels “2” and “3” is about 20% and
80%, respectively. Note that earlier research have shown that the ver-
tical mean flows and turbulence fluctuations across the canyon roof
level contribute significantly to air exchange between the street can-
yons and the external flow above them [7,38].
Fig. 18 compares the steady RANS and LES results at a vertical plane
in the region with a relatively low building height variation. It can be
seen that the wind flow is generally well predicted by both steady RANS
and LES, showing relatively low wind velocities inside street canyons.
However RANS, again underestimates the turbulence kinetic energy
366
Building and Environment 126 (2017) 355–372
Fig. 13. Contours of ratio of local mean air delay
to maximum mean air delay (τdmax = 0.5 s) at
pedestrian height (2 m) obtained by (a) steady
RANS and (b) LES.
Fig. 15. 3D map of buildings geometry with streamwise vertical bounded planes A and B.
across the roof level of street canyons with average deviation of 80%.
This overestimation can be considered the main reason for the generally
higher air delay values predicted by steady RANS, and it is consistent
with the results presented by previous studies [18,34]. In a more de-
tailed analysis inside the urban canyons with label “4” and 5 (Fig. 19),
the wind flow predicted by both RANS and LES presents skimming flow
characteristics as described by Oke [86]. Inside the urban canon with
label “4”, a region with underestimated mean wind speed by RANS is
observed with 85% deviation from LES. At the same location, air delay
predicted by RANS is reaching up to 3 times higher than LES while in
canyon with label 5 this is around 1.5 times higher. Since turbulence is
overestimated by RANS at the top of canyons with label “4” and “5”
within the same range of 80%, the under prediction of wind speed by
RANS in canyon with label “4” seem to enhance the overestimation of
mean air delay at the same location.
Similar morphological characteristics of the area under study can be
found in several examples of historical towns mainly in the
Mediterranean region. Hence the conclusions and any possible re-
commendations for improving the breathability of the area under in-
vestigation are considered important, and can be useful for urban
planners and decision makers of these regions.
The comparison of Figs. 13 and 2a shows a clear impact of building
height variability on the breathability at the pedestrian level. Areas that
are poorly ventilated appear to be associated with lower building-
N. Antoniou et al.
height variation. Therefore, an increase in the variability of building
height in a given neighborhood or urban area can improve the
breathability in complex urban areas. This in line with the results of
previous studies for generic urban areas [11,13,14,41,58].
The planar density of an urban area has been proven to significantly
affect the breathability levels [14], where more compact cities appear
367
Building and Environment 126 (2017) 355–372
Fig. 16. Distribution of normalized (a) mean velocity, (c)
turbulence kinetic energy and (e) local mean air delay in a
vertical plane obtained by steady RANS. (b, d and f) same
for LES (Umax = 6 m/s, Uref = 6 m/s, τdmax = 0.8 s).
to have less well ventilated conditions due to the stronger recirculation
zones [6]. According to Fig. 2a, regions with the highest planar density
can be identified in and around square C5 and regions with the lowest
planar density in squares B2, B6, D6 and F2. However, based on the
results presented in Fig. 13, a clear impact of planar density on air delay
levels cannot be identified. The reason for this is considered to be the
N. Antoniou et al.
complexity of the urban area, which creates difficulties in identifying
these relations, while previous findings were based on generic config-
urations. The same conclusions can be made for the relations of aero-
dynamic characteristics z0 and zd where their impact on breathability
levels cannot be identified clearly. Consequently, the creation of formal
connections between the morphological parameters and breathability
in a real urban geometry can be the scope and objective of further re-
search.
5.3. Limitations and future work
Some important points of this study can be further discussed:
• Based on the complexity of the specific problem, and the high
computational cost, only one wind direction has been considered.
Hence, conclusions made for analysis of local mean age of air, are
valid for the specific wind direction. However as it has been stated
in Section 2, the wind direction was chosen after analysis of the
prevailing wind conditions of the same area, derived from a field
measurement campaign [87]. In a next step, other wind directions
can be considered in order to investigate in more depth the
368
Building and Environment 126 (2017) 355–372
Fig. 17. Distribution of normalized (a) mean ve-
locity and velocity streamlines and (c) turbulence
kinetic energy obtained by steady RANS. (b and
d) Same for LES (Umax = 6 m/s, Uref = 6 m/s).
ventilation conditions in the selected area.
• In order to reduce computational cost, buildings were modeled with
simplified geometry. However geometrical details of buildings have
proven to affect significantly the flow pattern around them [88],
and a further analysis using a more detailed model could provide
more information on the wind flow conditions of the case study
area.
• The temperature distribution inside the urban canopy layer can play
an important role in air mass transfer and flow exchange. In addi-
tion, compact urban cities with lack of open spaces and green areas
are expected to have more chance in experiencing excessive heat
stress periods and urban heat island effect. Hence it would be of high
interest to evaluate thermal comfort and microclimatic conditions in
the selected area.
• As shown in this study, there is a clear impact of morphological
characteristics like building height variability, on the ventilation
conditions inside real urban areas. However, more investigation on
different urban configurations must be performed, in order to arrive
at more generalized conclusions.
• In this study, the CFD validation is performed for mean wind speed
and turbulence intensity. Further research is required to (i) obtain
N. Antoniou et al.
high-resolution experimental data of age of air for real complex
urban areas and (ii) perform detailed CFD validation studies for air
delay in complex urban areas.
6. Conclusions
The main conclusions of this study are:
• LES can accurately predict the mean wind speed and turbulence
intensity in a complex urban area. Steady RANS, however, indicates
369
Building and Environment 126 (2017) 355–372
Fig. 18. Distribution of normalized (a) mean velocity, (c) turbu-
lence kinetic energy and (e) local mean air delay in a vertical plane
obtained by steady RANS. (b, d and f) Same for LES (Umax = 6 m/s,
Uref = 6 m/s, τdmax = 0.8 s).
a satisfactory agreement with the experiments only for mean wind
speed. For LES, the average deviations in wind speed and turbulence
intensity from the experiments are about 6% and 14%, respectively,
while for steady RANS they increase to 8% and 31%, respectively.
• In order to achieve satisfactory CFD results and fair agreement with
experimental data, it is often needed to adjust the inflow conditions.
• Based on the validation, the LES turbulence approach is suggested
for calculating local mean age of air in real complex urban en-
vironments.
• To better assess the outdoor ventilation for the urban area, a new
N. Antoniou et al.
indicator called “air delay” is introduced as the difference between
the local mean age of air at an urban area and that in an empty
domain with the same computational settings.
• RANS overestimates the local mean air delay with an average de-
viation from LES of 52% at pedestrian level. The main reason for this
is the overestimation of turbulence by RANS while maximum de-
viation is observed in regions where RANS underestimates both
wind speed and turbulence intensity.
•
Complex wind flow structures occurring in wake regions of tall
buildings and narrow span wise streets inside complex urban areas
cannot be captured well with RANS, resulting in overestimation of
local mean air delay at these locations.
Acknowledgements
Hamid Montazeri is currently a postdoctoral fellow of the Research
Foundation – Flanders (FWO) and is grateful for its financial support
(project FWO 12M5316N). The authors gratefully acknowledge the
partnership with ANSYS CFD. The TOPEUM project was funded by
ERA-NET (Urban-Net Call) through Cyprus Research Promotion
Foundation with contract number ΔΙΕΘΝΗ/URBAN-NET/0308/02.
370
Building and Environment 126 (2017) 355–372
Fig. 19. Distribution of normalized (a) mean ve-
locity and velocity streamlines and (c) turbulence
kinetic energy obtained by steady RANS. (b and
d) same for LES (Umax = 6 m/s, Uref = 6 m/s).
References
[1] WHO, Ambient Air Pollution: a Global Assessment of Exposure and Burden of
Disease, (2016).
[2] United Nations, World urbanization prospects: the 2014 revision, highlights,
Department of Economic and Social Affairs, 2014.
[3] OECD, Environmental Outlook to 2050: the Consequences of Inaction, (2012).
[4] M.K.-A. Neophytou, R.E. Britter, Modelling of atmospheric dispersion in complex
urban topographies: a computational fluid dynamics study of the central London
area, 5th GRACM Int. Congr. Comput. Mech. Limassol, 2 2005, pp. 967–974.
[5] I. Panagiotou, M.K.A. Neophytou, D. Hamlyn, R.E. Britter, City breathability as
quantified by the exchange velocity and its spatial variation in real inhomogeneous
urban geometries: an example from central London urban area, Sci. Total Environ.
442 (2013) 466–477.
[6] R. Buccolieri, M. Sandberg, S. Di Sabatino, City breathability and its link to pol-
lutant concentration distribution within urban-like geometries, Atmos. Environ. 44
(2010) 1894–1903.
[7] J. Hang, Y. Li, Ventilation strategy and air change rates in idealized high-rise
compact urban areas, Build. Environ. 45 (2010) 2754–2767.
[8] C. Liu, D.Y.C. Leung, M.C. Barth, On the prediction of air and pollutant exchange
rates in street canyons of different aspect ratios using large-Eddy simulation, Atmos.
Environ. 39 (2005) 1567–1574.
[9] Z. Bu, S. Kato, Y. Ishida, H. Huang, New criteria for assessing local wind environ-
ment at pedestrian level based on exceedance probability analysis, Build. Environ.
44 (2009) 1501–1508.
[10] R. Ramponi, B. Blocken, L.B. de Coo, W.D. Janssen, CFD simulation of outdoor
ventilation of generic urban configurations with different urban densities and equal
and unequal street widths, Build. Environ. 92 (2015) 152–166.
N. Antoniou et al.
[11] C. Wen, Y. Juan, A. Yang, Enhancement of city breathability with half open spaces
in ideal urban street canyons, Build. Environ. 112 (2017) 322–336.
[12] A. Kubilay, M.K. Neophytou, S. Matsentides, M. Loizou, J. Carmeliet, Urban Climate
the pollutant removal capacity of urban street canyons as quanti fi ed by the pol-
lutant exchange velocity, Urban Clim. 21 (2017) 136–153.
[13] J. Hang, Y. Li, M. Sandberg, R. Buccolieri, S. Di, The influence of building height
variability on pollutant dispersion and pedestrian ventilation in idealized high-rise
urban areas, Build. Environ. 56 (2012) 346–360.
[14] W. Wang, E. Ng, C. Yuan, S. Raasch, Urban climate large-Eddy simulations of
ventilation for thermal comfort — A parametric study of generic urban configura-
tions with perpendicular approaching winds, Urban Clim. 20 (2017) 202–227.
[15] B. Blocken, 50 years of computational wind engineering: past, present and future, J.
Wind Eng. Ind. Aerodyn. 129 (2014) 69–102.
[16] T. Stathopoulos, Pedestrian level winds and outdoor human comfort, J. Wind Eng.
Ind. Aerodyn. 94 (2006) 769–780.
[17] H. Montazeri, B. Blocken, W.D. Janssen, T. van Hooff, CFD evaluation of new
second-skin facade concept for wind comfort on building balconies: case study for
the Park Tower in Antwerp, Build. Environ. 68 (2013) 179–192.
[18] Y. Tominaga, T. Stathopoulos, CFD modeling of pollution dispersion in a street
canyon: comparison between LES and RANS, J. Wind Eng. Ind. Aerodyn. 99 (2011)
340–348.
[19] P. Gousseau, B. Blocken, T. Stathopoulos, G.J.F. van Heijst, CFD simulation of near-
field pollutant dispersion on a high-resolution grid: a case study by LES and RANS
for a building group in downtown Montreal, Atmos. Environ. 45 (2011) 428–438.
[20] M.K.-A. Neophytou, P. Fokaides, I. Panagiotou, I. Ioannou, M. Petrou, M. Sandberg,
H. Wigo, E. Linden, E. Batchvarova, P. Videnov, B. Dimitroff, A. Ivanov, Towards
optimization of urban planning and architectural parameters for energy use mini-
mization in mediterranean cities, World Renew. Energy Congr, 2011 Sweden.
[21] Y. Toparlar, B. Blocken, P. Vos, G.J.F. Van Heijst, W.D. Janssen, T. van Hooff,
H. Montazeri, H.J.P. Timmermans, CFD simulation and validation of urban mi-
croclimate: a case study for Bergpolder Zuid, Rotterdam, Build. Environ. 83 (2015)
79–90.
[22] H. Montazeri, Y. Toparlar, B. Blocken, J.L.M. Hensen, Simulating the cooling effects
of water spray systems in urban landscapes: a computational fluid dynamics study
in Rotterdam, The Netherlands, Landsc. Urban Plan. 159 (2017) 85–100.
[23] B. Blocken, T. Defraeye, D. Derome, J. Carmeliet, High-resolution CFD simulations
for forced convective heat transfer coefficients at the facade of a low-rise building,
Build. Environ. 44 (2009) 2396–2412.
[24] H. Montazeri, B. Blocken, D. Derome, J. Carmeliet, J.L.M. Hensen, Journal of Wind
Engineering CFD analysis of forced convective heat transfer coefficients at wind-
ward building facades: Influ. Build. geometry 146 (2015) 102–116.
[25] H. Montazeri, B. Blocken, New generalized expressions for forced convective heat
transfer coefficients at building facades and roofs, Build. Environ. 119 (2017)
153–168.
[26] Y. Juan, A. Yang, C. Wen, Y. Lee, P. Wang, Optimization procedures for enhance-
ment of city breathability using arcade design in a realistic high-rise urban area,
Build. Environ. 121 (2017) 247–261.
[27] N. Nazarian, J. Kleissl, Realistic solar heating in urban Areas: air exchange and
street-canyon ventilation, Build. Environ. 95 (2016) 75–93.
[28] J. Hang, Q. Wang, X. Chen, M. Sandberg, W. Zhu, R. Buccolieri, S. Di, City
breathability in medium density urban-like geometries evaluated through the pol-
lutant transport rate and the net escape velocity, Build. Environ. 94 (2015)
166–182.
[29] R. Buccolieri, P. Salizzoni, L. Soulhac, V. Garbero, S. Di, Urban Climate the
breathability of compact cities, Urban Clim. 13 (2015) 73–93.
[30] M. Lin, J. Hang, Y. Li, Z. Luo, M. Sandberg, Quantitative ventilation assessments of
idealized urban canopy layers with various urban layouts and the same building
packing density, Build. Environ. 79 (2014) 152–167.
[31] J. Hang, Z. Luo, M. Sandberg, J. Gong, Natural ventilation assessment in typical
open and semi-open urban environments under various wind directions, Build.
Environ. 70 (2013) 318–333.
[32] J. Hang, Y. Li, R. Buccolieri, M. Sandberg, S. Di Sabatino, On the contribution of
mean flow and turbulence to city breathability: the case of long streets with tall
buildings, Sci. Total Environ. 416 (2012) 362–373.
[33] S.M. Salim, S.C. Cheah, A. Chan, Numerical simulation of dispersion in urban street
canyons with avenue-like tree plantings: comparison between RANS and LES, Build.
Environ. 46 (2011) 1735–1746.
[34] P. Moonen, V. Dorer, J. Carmeliet, Evaluation of the ventilation potential of
courtyards and urban street canyons using RANS and LES, Jnl. Wind Eng. Ind.
Aerodyn. 99 (2011) 414–423.
[35] T. Kim, K. Kim, B. Sean, A wind tunnel experiment and CFD analysis on airflow
performance of enclosed-arcade markets in Korea, Build. Environ. 45 (2010)
1329–1338.
[36] J. Hang, Y. Li, Wind conditions in idealized building clusters: macroscopic simu-
lations using a porous turbulence model, Bound. Layer. Meteorol. (2010) 129–159.
[37] J. Hang, M. Sandberg, Y. Li, L. Claesson, Flow Mechanisms and Flow Capacity in
Idealized Long-Street City Models 45 (2010), pp. 1042–1044.
[38] J. Hang, Y. Li, M. Sandberg, L. Claesson, Wind conditions and ventilation in high-
rise long street models, Build. Environ. 45 (2010) 1353–1365.
[39] J. Hang, M. Sandberg, Y. Li, Effect of urban morphology on wind condition in
idealized city models, Atmos. Environ. 43 (2009) 869–878.
[40] J. Hang, M. Sandberg, Y. Li, Age of air and air exchange efficiency in idealized city
models, Build. Environ. 44 (2009) 1714–1723.
[41] S.H.L. Yim, J.C.H. Fung, A.K.H. Lau, S.C. Kot, Air ventilation impacts of the wall
effect resulting from the alignment of high-rise buildings, Atmos. Environ. 43
(2009) 4982–4994.
371
Building and Environment 126 (2017) 355–372
[42] W.C. Cheng, C. Liu, D.Y.C. Leung, Computational formulation for the evaluation of
street canyon ventilation and pollutant removal performance, Atmos. Environ. 42
(2008) 9041–9051.
[43] B. Blocken, P. Moonen, T. Stathopoulos, F. Asce, J. Carmeliet, Numerical study on
the existence of the Venturi effect in passages between perpendicular buildings, J.
Eng. Mech. 134 (2008) 1021–1028.
[44] M. Bady, S. Kato, H. Huang, Towards the Application of Indoor Ventilation
Efficiency Indices to Evaluate the Air Quality of Urban Areas 43 (2008), pp.
1991–2004.
[45] R. Yoshie, A. Mochida, Y. Tominaga, H. Kataoka, Cooperative Project for CFD
Prediction of Pedestrian Wind Environment in the Architectural Institute of Jpn 95
(2007), pp. 1551–1578.
[46] B. Blocken, J. Carmeliet, T. Stathopoulos, CFD evaluation of wind speed conditions
in passages between parallel buildings-effect of wall-function roughness modifica-
tions for the atmospheric boundary layer flow, J. Wind Eng. Ind. Aerodyn. 95
(2007) 941–962.
[47] Z. Xie, I.P. Castro, LES and RANS for Turbulent Flow over Arrays of Wall- Mounted
Obstacles, (2006), pp. 1–30.
[48] X. Li, C. Liu, D.Y.C. Leung, Development of a k-ε Model for the Determination of Air
Exchange Rates for Street Canyons 39 (2005), pp. 7285–7296.
[49] M. Skote, M. Sandberg, U. Westerberg, L. Claesson, A.V. Johansson, Numerical and
experimental studies of wind environment in an urban morphology, Atmos.
Environ. 39 (2005) 6147–6158.
[50] R.E. Britter, S.R. Hanna, Flow and dispersion in urban areas, Annu. Rev. Fluid
Mech. 35 (2003) 469–496.
[51] W. Rodi, Comparison of LES and RANS calculations of the flow around bluff bodies,
J. Wind Eng. Ind. Aerodyn. 69–71 (1997) 55–75.
[52] K. Hanjalic, Will RANS survive LES? A view of perspectives, J. Fluids Eng. 127
(2005) 831–839.
[53] M. Neophytou, A. Gowardhan, M. Brown, An inter-comparison of three urban wind
models using Oklahoma City Joint Urban 2003 wind field measurements, Jnl. Wind
Eng. Ind. Aerodyn. 99 (2011) 357–368.
[54] S. Murakami, Comparison of various turbulence models applied to a bluff body, J.
Wind Eng. Ind. Aerodyn. 46–47 (1993) 21–36.
[55] Y. Tominaga, T. Stathopoulos, Numerical simulation of dispersion around an iso-
lated cubic building: model evaluation of RANS and LES, Build. Environ. 45 (2010)
2231–2239.
[56] T. Stathopoulos, Y. Tominaga, Numerical simulation of dispersion around an iso-
lated building, Atmos. Environ. (2007) 20–24.
[57] C. Liu, C.C.C. Wong, On the pollutant removal, dispersion, and entrainment over
two-dimensional idealized street canyons, Atmos. Res. 135–136 (2014) 128–142.
[58] J. Hang, Y. Li, Age of air and air exchange efficiency in high-rise urban areas and its
link to pollutant dilution, Atmos. Environ. 45 (2011) 5572–5585.
[59] M. Sandberg, What is ventilation efficiency? Build. Environ. 16 (1981) 123–135.
[60] D. Steward, T.R. Oke, Local climate zones for urban temperature studies, Bull. Am.
Meteorol. Soc. 93 (2012) 1879–1900.
[61] A.N. Srahler, A.H. Strahler, Elements of Physical Geography, John Wiley & Sons Ltd,
1984.
[62] P. Mouzourides, A. Kyprianou, M.J. Brown, B. Carissimo, R. Choudhary, M.K.-
A. Neophytou, Searching for the distinctive signature of a city in atmospheric
modelling: could the Multi-Resolution Analysis (MRA) provide the DNA of a city?
Urban Clim 10 (2014) 447–475.
[63] P. Kastner-klein, M.W. Rotach, Mean Flow and Turbulence Characteristics in an
Urban Roughness Sublayer, (2004), pp. 55–84.
[64] M.K.-A. Neophytou, E. Tryphonos, P. Fokaides, M. Sandberg, E. Batchvarova,
H.J.S. Fernando, J. Lelieveld, G. Zittis, Toward designing strategies for urban heat
island mitigation based on multi scale flow considerations, 34thInt. Conf. Energy
Conserv. Technol. Mitig. Adapt. Built Environ. Role Vent. Strateg. Smart Mater,
2013, pp. 1–10. Athens.
[65] J. Franke, A. Hellsten, H. Schlünzen, B. Carissimo, Best practice guideline for the
CFD simulation of flows in the urban environment, COST Action 732, 2007, pp.
1–52.
[66] Y. Tominaga, A. Mochida, R. Yoshie, H. Kataoka, T. Nozu, M. Yoshikawa,
T. Shirasawa, AIJ guidelines for practical applications of CFD to pedestrian wind
environment around buildings, J. Wind Eng. Ind. Aerodyn 96 (2008) 1749–1761.
[67] B. Blocken, Computational Fluid Dynamics for urban physics: importance, scales,
possibilities, limitations and ten tips and tricks towards accurate and reliable si-
mulations, Build. Environ 91 (2015) 219–245.
[68] T. van Hooff, B. Blocken, Coupled urban wind flow and indoor natural ventilation
modelling on a high-resolution grid: a case study for the Amsterdam ArenA stadium,
Environ. Model. Softw 25 (2010) 51–65.
[69] T.-H. Shih, W.W. Liou, A. Shabbir, Z. Yang, J. Zhu, A new k-ε eddy viscosity model
for high reynolds number turbulent flows, Comput. Fluids 24 (1995) 227–238.
[70] B.E. Launder, D.B. Spalding, The numerical computation of turbulent flows,
Comput. Methods Appl. Mech. Eng 3 (1974) 269–289.
[71] T. Cebeci, P. Bradshaw, Momentum Transfer in Boundary Layers, (1977)
Washington.
[72] B. Blocken, T. Stathopoulos, J. Carmeliet, CFD simulation of the atmospheric
boundary layer: wall function problems, Atmos. Environ 41 (2007) 238–252.
[73] J. Smagorinsky, General circulation experiments with the primitive equations, Mon.
Weather Rev. 91 (1963) 99–164.
[74] D.K. Lilly, A proposed modification of the Germano subgrid-scale closure method A
proposed modification of the Germano subgrid-scale closure method, Phys. Fluids
633 (1992) 633–635.
[75] J.B. Perot, An analysis of the fractional step method, J. Comput. Phys. 108 (1993)
51–58.
N. Antoniou et al.
[76] G.E.M. Karniadakis, S.A. Orszag, High-order splitting methods for the in-
compressible Navier-stokes equations, J. Comput. Phys. 97 (1991) 414–443.
[77] S. Armfield, R. Street, The Fractional-Step Method for the Navier – Stokes Equations
on Staggered Grids The Accuracy of Three Variations 665 (2006) 660–665.
[78] ANSYS FLUENT User's Guide, ANSYS Inc, 2013.
[79] F. Menter, Best Practice: Scale-resolving Simulations in ANSYS CFD, ANSYS
Germany GmbH, 2015.
[80] F. Mathey, D. Cokljat, K.-P. Bertoglio, E. Sergent, Assessment of the vortex method
for large eddy simulation inlet condition, Prog. Comput. Fluid Dyn 6 (2006) 58–67.
[81] S. Iousef, H. Montazeri, B. Blocken, P.J.V. Van Wesemael, On the use of non-con-
formal grids for economic LES of wind flow and convective heat transfer for a wall-
mounted cube 119 (2017) 44–61.
[82] S. Di Sabatino, R. Buccolieri, B. Pulvirenti, R. Britter, Simulations of pollutant
dispersion within idealised urban-type geometries with CFD and integral models,
Atmos. Environ 41 (2007) 8316–8329.
[83] R. Ramponi, B. Blocken, CFD simulation of cross-ventilation for a generic isolated
372
Building and Environment 126 (2017) 355–372
building: impact of computational parameters, Build. Environ 53 (2012) 34–48.
[84] B. Blocken, T. Stathopoulos, J.P.A.J. Van Beeck, Pedestrian-level wind conditions
around buildings: review of wind-tunnel and CFD techniques and their accuracy for
wind comfort assessment, Build. Environ 100 (2016) 50–81.
[85] M. Princevac, J. Baik, X. Li, H. Pan, S. Park, Lateral channeling within rectangular
arrays of cubical obstacles, J. Wind Eng. Ind. Aerodyn. 46–47 (2010) 21–36.
[86] T.R. Oke, Street design and urban canopy layer climate, Energy Build 11 (1988)
103–113.
[87] M.K.-A. Neophytou, H.J.S. Fernando, E. Batchvarova, M. Sandberg, J. Lelieveld,
E. Tryphonos, A scaling law for the urban heat island phenomenon: deductions from
field measurements and comparisons with existing results from laboratory experi-
ments, Proc. 4th US-European Fluids Eng. Div. Symp, Urban Fluid Mech, Chicago,
IL, USA, 2014.
[88] H. Montazeri, B. Blocken, CFD simulation of wind-induced pressure coefficients on
buildings with and without balconies: validation and sensitivity analysis, Build.
Environ 60 (2013) 137–149.