Keywords: Urban Ventilation,CFO Simulation,Urban 1988 [1]). The major inaccuracies of the model are its
Canopy. overestimation if (of) the value of the turbulence energy
around the frontal corner and its underestimation of the
Abstract value of the lateral component of normal stress in the re
circulation region behind the model, and the results of
The past half century has seen an unprecedented growth of LES (Large Eddy Simulation) agree fairly well with the
the world's urban population. Increasing urbanization and experimental data (SHUZO MURAKAMI et at., 1990 [2]).
concern about sustainability and quality of life issues have But Y. Gao pointed out that the standard turbulence model
produced considerable interest in air quality in urban areas. can predict well the separated and reattachment flow on
Urban ventilation implies that rural areas may supply rela the top surface of the cube and rib if the wall boundary is
tively clean air into urban canopies and distribute rural air treated properly (Y. Gao et at., 2003 [3]). The influence of
within them to help air exchange and pollutant dilution. the street geometrical aspect ratio leads to a refinement of
This paper numerically studied rural air flow through square the flow separation in three regimes, popularized by Oke
building arrays at different wind angles. The building area (1988) [4], taking into account not only the influence of
the buildings on the flow of the surface layer just above
density (Ap' i. e. the rario between the plan area of build
the roofs but also the structure of the recirculating flow
ings viewed from above and the total underlying surface within the street. Results show that street aspect ratio (Oke,
area) is restricted to 0.25, while the building numbers in 1988 [4]),area densities (Buccolieri et at., 2010 [5]),build
the square region change from 9 to 81. A twodimensional ing arrangements (Kanda, 2006 [6]), and street orienta
(20) k8 turbulence model is employed to solve the Rey tions (Kim and Baik, 2004 [7]) are the key urban parame
noldsaveraged NavierStokes equations.The model is em ters. Kanda (2006b) found that staggered arrays tend to
ployed to predict the influence of the building numbers on produce stronger drag force to the approach wind than
flow capacity in wind angle altering from 00  4Y. 3D mo aligned arrays. Hang et at. (2009) [8] studied pollutant dis
del is employed to solve this problem as well. It's conflrmed persion in idealized round city model with ambient wind
that fewer buildings get better flow capacity. Smaller wind directions of 00, 15°, 30°, 45° . They found that small in
angle improves this process as well. °
cident wind angles of 00 and 15 are better for the pollut
ant removal and air exchange than large incident angle of
1 Introduction °
30 . This method is effectively and widely used in urban
Frequent hazy weather in Chinese cities has aroused much ventilation research, but there is a lack of research on the
concerns, and inspired many studies on urban ventilation. influence of different building scales. Understanding the
The wind from rural areas is relatively clean and cool,which influence of different building scales on urban ventilation
purifles the urban canopy environment when flowing through is quite important for improving urban air quality and op
it. The buildings in urban areas play a vital role in this proc timal design. The present work aims to calculate the flow
ess. Understanding the impact caused by it can help to de capacity in building arrays with different building scales
velop guidelines for better urban planning [9] [16]. and same area density with CFO code ANSYS FLUENT.
An acceptable pedestrian level wind environment in out Ambient wind angle is changed as well, and 2D simula
door public areas has become a major design considera tion with ks model is used.
tion for new urban planning. Well designed building ar
rays can provide flne wind environment, which is impor 2 Description of Numerical Simulations
tant for human comfort and energy conservation.
Wind tunnel experiment is the main method for urban 2.1 Governing Equations
ventilation research, which is expensive and timeconsum
The Reynoldaveraged NavierStokes (RANS) equations
ing. Numerical simulation is provided to be a much cheap
er and timesaving way, and proved to be effective when with standard ks turbulence model are widely used to
compared with wind tunnel experimental data. model urban turbulence flow. The momentum conser
Numerical simulation using the model accurately re vation equations and two transport equations for the turbu
produce the velocity and pressure field with a finemesh lent kinetic energy (k) and its dissipation rate (s) are
resolution around the model (SHUZO MURAKAMI et at., solved to get timeaverage flow variables.
'Corresponding author. The dimensionless governing equations for a steady, in
compressible, and neutrallystratified fluid flow are as
follows:
au av
ax
+
ay
=0 (1)
���
( ) a (uv ) a
ax +ay =ax
a uu ( ) ( )au
+
a au

ap
���
( ) ( )
Veffax ay Veff ay ax
(2)
+
a au
ax Veffax
+
a av
ay Veffax
���
( ) +a (vv ) = �
a uv ( ) ( )
v
av
+
�
v
av
_
ap Figure 1. The distribution ofbuildings in building area.
( ) ( )
ax ay ax effax ay eff ay ax
(3) buildings in lateral and longitudinal direction are equal,
a au a av
+ v + 
which means there is N x N buildings in each array.
ax eff ax ay Veffax
( ) ( )
The building array is placed in a numerical wind tunnel,
a (uK) a (vK)
which provides reasonable boundary conditions for the
ax+aY = ax
a aK a aK
rK + rK numerical calculation. The computational domain in Fig
ax ay ay (4)
ure 2 is used for all wind directions ( 0", IS" ,300 ,4S" ) . Mass
+GK E
( J �( J
flux is calculated in four faces (see Figure 3). The distance
a uE ( ) + a (vE) = � r
aE
+ r
aE
between building domain and outlet boundary are large
enough for outflow boundary condition, where the distur
ax ay ax r. ax ay r. ay
(S) bance of the building areas are negligible.
E E2
+G1£GK  2.3 FLUENT Flow Setup
C £K
K 2
The second order upwind differencing scheme was used to dis
K2 V K2 I . cretize the advection terms. All transport equations were dis
vt = C,
,  = 
+C  = +Vt IS the turbu
E r
Veff UoL JI E Re cretized by the second order upwind scheme. The discretized
differential equations were solved by the SIMPLE algorithm.
lence viscosity;
Velocity inlet is used in both south and west face of the
rK = 
v CII K2
+
1 vt
= + 'IS the d'IsslpatlOn rate 0f
. . computational domain, and at north and east face outflow
uoL O'k E Re O'k boundary condition is used. The governing equation of
k; outflow is as follow:
v CJl K2 1 vt • . . ap ap au av
lle" = 
UoL
+
0'" E
= + IS the d'IsslpatlOn rate 0f
O'e

_ 
_ 
_ 
_ 0 (8)
Re ax ay ax ay
&;
The velocity at inlet is defmed as:
( a)
To evaluate GK in a manner consistent with Boussinesq
{ [( J ( J] ( J}
assumptions,it can be further written as, U = urefsin
v = vr cos ( a )
(9)
2 ef
au av 2 au aV 2
GK = vt 2 
+  + + 
(6)
ax ay ax ay where the reference velocity Uf f = .fim/s ;
e
Noslip wall and wall function is used at solid boundary.
where Cel' Co2' C , O'k' O'e are closure constants, the The wall function is described as follows, and the law of a
Jl
value of which are as follows: logarithm is used;
y+ (11)
( a ) are set equal to length of
buildings are treated as aligned square obstacles. The dis
v
tance between buildings
�
'#��
The result shows that when the number of buildings in
ia
creases, the flow capacity of the building arrays gets
worse. Less rural wind will flow through the urban street D8� 1 (� _
(see Figure 4). While there exists an unexpected rise
when the number of buildings increase from 8 x8 to 9 x 9  �
1
Computational domain
)0
N
20l
W+E
S
 � 5L
�
20L
building arrays
5L
D'6
�=� =::: �
N !
: W�lq � [iI�
)0 i Il I .
LE
f �.
�'f!!/�=
 S / ��
_.0
_ JI
��= �(
0� ��l�==c r ;� , I , ,
Figure 3. The diagram ofthe flow monitoring. 0 0.2 0.4 0.6 0.8
2.1
1.9
1.7
1.5
1.3
1.1
Appendix 1
2 3 4
"f
12
� "f 'f� ,..
0.8
0.6
>
OA
0.2
0.2
 . 04
I I t I [ I !
0.5 0.5 0.5 .5
0 0.5 0.5
x
0.5
X X
5 6 7 8
1.'
1.2
0.8
D.'
>
OA
0.2
0.2
U.4
[ [
0.' D.' 0.' 0.'
t
D.'
I I
0.'
X x
0.'
X X
9 10 11 12
,
"
,2
0.8
0.'
>
0..
0.2
·0,2
OA
I I
0.' 0
[ [
0.' 0.'
X x
13 14 15 16
> >
17 18 19 20
14
,
1.2
0.'
06
>
0.4
0.'
0,2
r r
0,4
.o.
, , ,
0 . 5
,
0.' 0.' .o.• 0.' 0.'
X X X
21 22 23 24
25 26 27 28