Building and Environment 45 (2010) 1880e1889

Contents lists available at ScienceDirect

Building and Environment

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

A simple method for designation of urban ventilation corridors and its

application to urban heat island analysis
Man Sing Wong, Janet E. Nichol*, Pui Hang To, Jingzhi Wang
Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Kowloon, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: This paper describes urban wind ventilation mapping, using the concept of “building frontal area index”,
Received 7 December 2009 and uses the Kowloon peninsula of Hong Kong as an example of a dense, sub-tropical urban environment
Received in revised form where ventilation is critical for human health. The frontal area index is calculated for uniform 100 m grid
22 February 2010
cells, based on three dimensional buildings in each cell, for eight different wind directions. The frontal
Accepted 24 February 2010
area index is then correlated with a land use map, and the results indicate that high density commercial
and industrial areas with large building footprints had higher values than other urban land use types.
Keywords:
Using the map of frontal area index, the main ventilation pathways across the urban area are located
Frontal area index
Geographic information systems
using least cost path analysis in a raster GIS. Field measurements of urban winds confirmed the signif-
Remote sensing icance and functionality of these modelled ventilation paths. Comparison of the pathways with a map of
Urban heat island the urban heat island suggests that ventilation is a key parameter in mitigating heat island formation in
Ventilation path the study area. Planning and environmental authorities may use the derived frontal area index and
ventilation maps as objective measures of environmental quality within a city, especially when
temperatures in the inner city are a major concern.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction Mfula et al. [13] tested a very large building model at 1:100 scale to
identify pollution sources affecting buildings, based on wind and
The urban heat island (UHI) is defined as the temperature pollutant patterns at the surface. Although wind tunnel studies of
difference between urban and rural areas. As urban populations urban ventilation can provide accurate wind models measured
increase, many cities in both temperate [1e3] and tropical regions under constrained conditions, the small area coverage, high
[4e9] are reporting significant heat island effects resulting from computer processing requirements and high operational cost often
high building densities. Air flow between rural and urban areas is prohibit their usage. In recent years, a variety of numerical models
one of the parameters governing urban heat island formation and have been developed for modelling air ventilation, such as the PSU/
the build-up of pollution [3,10]. Low horizontal wind speeds are NCAR mesoscale model (known as MM5) and the computational
usually associated with high surface roughness, where energy is fluid dynamics (CFD) model. The MM5 model works for mesoscale
lost by vertical instability due to a high density of built structures phenomena such as sea breezes and mountainevalley flows [14]
[11]. The pressure differences along temperature gradients can also with large area coverage at coarse resolution, while the CFD model
induce low-level breezes across the urban-rural boundary. simulates urban wind flows over smaller areas in greater detail. The
Most of the data included in wind and air quality studies are from CFD model is being widely used in engineering flow analysis,
ground level instruments. The gathering of data over large regions building and structural design, urban wind flow predictions [15],
such as a city therefore, is a major challenge to these studies. Wind and air pollution dispersal modelling [16e18]. It comprises a set of
tunnel models provide another method for visualising the local physical models which attempt to closely match the real urban
wind direction and pollutant dispersion at large scales over a district. geometry and thus simulate the air flow around buildings and along
For example, Duijm [12] used the wind tunnel model in Lantau streets. The CFD model is thus highly computer-intensive and
island, Hong Kong at a large scale (1:4000) over a small area, and generally inapplicable to large areas or whole cities. The only known
exception to this is the use of a CFD model running on a supercom-
puter, which models temperature and air flow over a 5
5 km area of
* Corresponding author. Tel.: þ852 27665952; fax: þ852 23302994.
Tokyo [19]. Therefore, wind ventilation modelling at city scale,
E-mail addresses: [email protected] (M.S. Wong), [email protected]. especially over densely urbanised regions with complex street and
hk (J.E. Nichol). building structures is challenging.

0360-1323/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2010.02.019
 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889 1881

Now, geographic information systems (GIS) and remote sensing urban form, with population densities in the Kowloon peninsula
techniques can provide alternative solutions by adopting simplified exceeding 52,000 km2 in some areas [9]. With temperature
assumptions and numerical approximations. Wind modelling for differences up to 9e10  C between rural areas and the urban core
near surface conditions can be simplified mathematically by [6,9], the Hong Kong population has recently shown great concern
estimating roughness parameters from building structures. Several about the heat island effect. This is exacerbated by planning policies
studies have modelled surface roughness using GIS and remote which allow building developers to maximise profits by blocking
sensing techniques and several parameters have been suggested for sea views resulting in the so-called “wall effect” around the
calculation of surface roughness. These are zero-plane displace- coastline, and depriving inner areas of ventilation. The study area is
ment height (zd) and the roughness length (z0) [20,21], plan area the Kowloon peninsula in Hong Kong, which is 160 km2 in extent
density (lp), frontal area index (lf) [22,23], average height weighted and has a population over 2 million. It comprises mainly high
with frontal area (zh), depth of the roughness sub-layer (zr) [24,22] density residential and commercial districts, with one large park
and the effective height (heff) [25] etc. (Kowloon Park) and a few small urban parks of less than 1 ha each.
Among these urban morphological parameters, the “frontal area The topography is mainly flat, but at the northern edge, elevation
index” has been suggested as a good indicator of the roughness of rises to 300 m. Media reports suggest a strong belief by residents
the urban surface for mesoscale meteorological and urban disper- that the wall effect is a major cause of the urban heat island, by
sion models [22,23]. Frontal area index is the measurement of preventing cool sea breezes from reaching the inner city, and since
building walls facing the wind flow in a particular direction (frontal Kowloon is a peninsula, this is a reasonable assumption. In planning
area per unit horizontal area) (Fig. 1). It has a strong relationship for future urban renewal, data are needed to confirm and analyse
with surface roughness z0, and is a function of the flow regime the influence of ventilation on urban temperatures.
within urban street canyons [23]. Gál and Unger [26] calculated the
frontal area index from lot area polygons in Szeged, Hungary for
depicting the potential ventilation paths over the city. They sug- 3. Methods
gested lot area polygons as the unit for frontal area calculation since
the buildings in Hungary are individually separable and the density 3.1. Calculation of frontal area index
of built area in Szeged is far less (11%) than in the Kowloon
peninsula of Hong Kong (71%). Although Gál and Unger [26] The frontal area index (lf) is calculated as the total area of
provided a thorough method for the depiction of ventilation building facets projected to plane normal facing the particular wind
paths using frontal area index and other parameters, the pathways direction (and independent of the angle of the building facets),
were located mainly by visual inspection, and the results were not divided by the plane area (equation (1)) [22,23].
practically validated. More details of frontal area index will be given
in Section 3.
lf ¼ Afacets =Aplane (1)
The aims of this paper are (i) to demonstrate a simple automated
where lf is the frontal area index, Afacets is the total area of building
method for deriving the frontal area index from three dimensional
facets facing the wind direction, and Aplane is the plane area.
GIS building data, (ii) to demonstrate the use of the frontal area
Therefore, a lf of 1.0 means that wind is mostly blocked by build-
index map in least cost path (LCP) analysis to derive the frequency
ings within a selected plane region (e.g. Afacets  Aplane), and a lf of ca.
of occurrence of ventilation paths over the study area, and (iii) to
0.5 means that wind is half blocked (e.g. 2
Afactors zAplane ).
identify the locations of major ventilation corridors and their
relationships to urban heat island formation in the study area. Burian et al. [23] used a similar approach for estimating the lf in
Los Angeles. Digital data of building polygons at 1:5000 scale were
obtained from the Hong Kong Lands Department. A program was
2. Study area written in ESRIÒ ArcGISÔ 9.2 software to estimate the total frontal
area in the projected plane normal to the specific wind direction. In
Hong Kong, a sub-tropical city with hot humid summers, suffers this study we modified Grimmond and Oke and Burian et al.'s
from the urban heat island effect caused by a high-rise, high density [22,23] algorithm by eliminating the areas blocked by buildings
upwind, from the blocked area on leeward buildings (Fig. 1). This
program is first set up for a particular wind direction and it
generates projected lines in the wind direction with a 5 m hori-
zontal increment. If the projected lines hit the first facet and do not
reach the second facet, only the frontal area of the first facet is
calculated. This modification of the original method is important
for irregular building groups, and can reduce the number of facets
being calculated in computer memory. The calculated frontal areas
are then re-grouped based on horizontal plane polygons (e.g. grid
cell of 100 m
100 m, which is the resolution of the study).
We used grid cells of 100 m
100 m size to calculate the lf
because the Hong Kong Planning Department [27] in Hong Kong
found that a grid resolution of 100 m was compatible with all vari-
ables used for determining dynamic potential and thermal load
contributions in an urban climatic study. Also, since Nichol and Wong
[28] show that a resolution of 200 m corresponds to intra-urban
differences in air temperature between different land cover types,
then a resolution of 100 m is more than adequate. Thus, our study
calculated lf at 100 m grid resolution over the Kowloon peninsula
(approx. 11 km by 7 km) for eight different wind directions (north,
Fig. 1. Example of frontal area calculation. northeast, east, southeast, south, southwest, west, and northwest).
1882 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889

Since the most densely built areas are devoid of vegetation, and pixel, e.g. the higher lf, the higher the friction value. The friction
elsewhere street planting is severely restricted by lack of space [29], values represent the percentage of obstruction of wind ventilation
urban vegetation is small and fragmented. Since urban vegetation or air flow for that 100 m pixel. By varying the lf classes and/or the
has a small frontal area compared to buildings, trees are not friction weightings, it is possible to designate a few major pathways
considered in the lf calculation. In addition, since the urban or many minor pathways. For example, this study adopted variable
topography is mainly flat, terrain is also not considered. weightings in order to generate many different paths.
Firstly, the lf map was imported to IDRISI v.14.02 (Clark Labs.,
3.2. Frontal area index in different land use types Worcester, MA, USA). The lf pixel values were reclassified into five
classes and each class was given a friction value according to the
In Hong Kong, as in most cities, different land use types support degree of wind obstruction, with higher friction values given to large
different building structures and characteristics. For example high- frontal area values. The friction values were set to increase gradually
rise buildings in residential areas have over 50 floors compared with higher lf values (Table 2). This assumes that air is most
with only 20e30 floors in commercial districts, and industrial obstructed in areas with high lf values, but is not completely
buildings have larger footprint, and 20e50 storeys. Therefore, it obstructed if the cost-distance of using an alternative path
was expected that these areas would have different lf properties. outweighs the cost of a cell's lf value. Secondly, starting points for the
This study used a digital land use map at 10 m resolution from the air flows were designated. For example, fifty points were spaced
Hong Kong Planning Department for analysing the relationship evenly along the east coast of the Kowloon peninsula representing
between lf and different land uses. starting points for the easterly wind travelling across the peninsula
from east to west. The friction surface was then created by the IDRISI
3.3. Relationship between frontal area index and environmental COST module which computed the cost surfaces for the fifty starting
parameters points. Thirdly, fifty ending points located along the west coast of
Kowloon peninsula were input to the PATHWAY module [30] for
Five environmental parameters, namely heat island intensity generating LCPs from the starting points to these ending points.
(HII), aerosol optical thickness (AOT) air quality derived from A total of 2500 and 5186 LCPs represent eastward and northeasterly
a satellite image, vegetation cover (NDVI) derived from a Landsat winds respectively, from the combinations of multiple starting and
image, building density (BD), and building height (BH) were mapped ending points. Due to the large number of starting and ending points,
over the study area. HII represents the air temperature difference the LCPs overlapped each other in many places, i.e. many grid cells
between the urban pixels and a rural climatic station at the image have many LCPs running through them. The occurrence frequencies
time, derived from a previous study [9], and the Normalized were calculated by counting the overlaid LCP pathways for each cell.
Difference Vegetation Index (NDVI) is a measure of vegetation Thus, grid cells with high occurrence frequencies are associated with
amount. Table 1 describes the sources of the parameters, their units, low frontal area index values represented by low friction values.
and their resolutions. The parameters HII and AOT represent typical These cells are likely to have stronger air flow and better ventilation
distributions of aerosols and temperature over the urban area. than those with few, or no paths passing through them. All these
Although the data were derived from a single image time, both HII processes were implemented by customised scripts in IDRISI, and
and AOT are closely related to the urban structure and activities, and the results are shown in Section 4.4.
are observed by the researchers to be spatially stable at different
seasons of the year. All maps were then regressed against the derived 4. Results
lf map and the correlations were tabulated (see Section 4.2).
4.1. Map of lf
3.4. Least cost path analysis
Fig. 2 shows the distribution of lf which represents the average
In order to predict the likely air flow corridors based on the frontal area index values calculated for all eight directions. An
premise that wind across a city follows the path of least resistance, average value of lf ¼ 0.25 is observed over the whole study area,
least cost path (LCP) analysis was undertaken to designate these with significantly high lf values near the west coast of the penin-
pathways across the city. LCP analysis identifies the path of least sula, caused by newly developed high-rise buildings (>300 m high)
resistance across a cost surface from a starting point to an ending creating the so-called “wall effect” described in Section 2. These
point. The pathways represent routes with a high probability of high-rise buildings are also shown by our study to have a significant
strong ventilation and a high degree of connectivity between relationship with the UHI (Section 2).
starting and ending points. This study adopted the approach of
allocating variable weightings to the frontal area index value of each 4.2. Relationship between lf and different land use types

Table 3 shows lf as a function of different land use types. The

Table 1 industrial and commercial areas have significantly higher lf (0.324
Data sources.
and 0.305) than other areas. The residential class has moderate to
Parameters Unit Source Source resolution high lf (w0.254) values, and public transportation and warehouses

HII C ASTER image 10 m after emissivity
on 31 Jan 2007 correction [8]
AOT e CHRIS/PROBA 90 m Table 2
satellite images Friction values allocated to different classes of lf.
on 18 Dec 2005 and
06 Feb 2007
Allocated friction value lf
NDVI e Landsat image 30 m 20 <0.2
on 17 Sep 2001 40 0.2e0.4
BD e LIC (GIS) 1:1000 data Tertiary Planning Unit 60 0.4e0.6
BH Meter LIC (GIS) 1:1000 data Rasterised to 10 m, 80 0.6e0.8
Tertiary Planning Unit 100 >0.8
 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889 1883

Fig. 2. Map of frontal area index.

have lower lf (w0.15) values. Our findings indicate a direct 1% confidence level. Of the individual parameters, HII obtained the
relationship between the frontal area index and the building highest correlation (r ¼ 0.757), and both BD and BH also have strong
characteristics, with higher lf always associated with wider and correlations with lf, since lf represents the plane and vertical
taller buildings (e.g. industrial and commercial districts). Burian aspects of building geometry. Vegetation cover (NDVI) shows
et al. [23] found that the lf in residential, commercial, industrial and strong negative correlation with lf. Only moderate correlation was
public transportation districts in Los Angeles, were 0.176, 0.246, observed with AOT (r ¼ 0.315, r ¼ 0.251 at regional and pixel levels
0.095, 0.011 respectively based on the irregular grid cells respectively), but this may be due to the difficulty of mapping
(>10000 m2) of the street network. Grimmond and Oke [22] aerosols accurately at high resolution from satellite images.
studied the lf over small areas immediately surrounding meteo-
rological stations in North America, finding the highest lf in the city 4.4. Ventilation paths
center in Vancouver, Canada (0.3) and in suburban residential areas
in Arcadia, United States (0.33). Since their methods of calculating lf The eastward and northeastward maps of lf gave somewhat
included the frontal areas for every facet, their values would be similar results (Fig. 3a and b), but they differ in the distribution of
expected to be much higher than ours. The fact that our values are path frequencies (Fig. 3c and d; the cross and triangle symbols
higher, testifies to the extremely high density built environment of represent the starting and ending points of the LCPs). A total of
the Kowloon peninsula, and the problems of human comfort from 2500 and 5186 pathways were created from easterly and north-
which it suffers, compared with these temperate zone cities. easterly directions respectively. The distribution of the occurrence
frequency of LCPs shows that the areas with high probability of
4.3. Correlations between lf and environmental parameters easterly wind ventilation paths are mostly located on

The relationships between lf and environmental parameters i. Boundary Street (A in Fig. 3c). This route traverses across low-
were analysed at regional scale, within 68 Tertiary Planning Units rise low-density residential areas in which high-rise building
(TPU), which sub-divide the city into a number of street blocks and was restricted due to the airport flight path. This street marks
village clusters, and are regarded as basic socioeconomic units the boundary between Kowloon in the south which was
(Table 4). All parameters at regional scale have higher correlations ceded to the United Kingdom in 1860, and the New Territories
than at pixel scale and most of the parameters are significant at the which were leased to the United Kingdom in 1898. The
occurrence frequency of LCPs along this route is greater than
Table 3
Frontal area index in different land use types.
Table 4
Land use type Averaged lf with Correlations between lf and six environmental parameters at TPU district and pixel
eight directions levels, and the number of samples.

Private residential 0.267 Parameter Correlation (r)

Public residential 0.241
Commercial/business & offices 0.305 Regional level Pixel level
Industrial 0.324 HII 0.757 0.452
Warehouse & storage 0.155 AOT 0.315 0.251
Rural settlements 0.056 NDVI 0.674 0.449
Vegetation 0.075 BD 0.603 0.522
Public transportation 0.15 BH 0.527 0.383
Vacant development land 0.191 n (Number of samples) 68 6504
1884 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889

30%, i.e. 750 out of the 2500 eastewest paths are generated southern end. The occurrence frequency of LCPs along this
along Boundary Street. route is ca. 16% (i.e. 400/2500) which is the lowest frequency
ii. Argyle Street and Cherry Street (B in Fig. 3c). This is the among the three main paths.
shortest route, but is a wide four-lane dual-carriageway
which connects the old airport in the east to the west coast. These three routes have occurrence frequencies greater than
This route provides a strategic air corridor in the eastewest 16%, and flow directions are from the east toward reclaimed land
direction since it bisects the most densely built part of areas on the west coast. For the northeasterly paths generated, the
MongKok. The occurrence frequency of LCPs along this route corridor with the greatest frequency of paths (28%) was located on
is greater than 18% (i.e. 450/2500).
iii. Ho Man Tin Hill road, King's Park road and Gascoigne road iv. Princess Margaret road, Chatham road, Hong Chong road and
(C in Fig. 3c). This route traverses a large patch of urban Salisbury road (D in Fig. 3d). This route is located along the
vegetation, low-density residential development with direction of the roads as they join end to end, and passes
fragmented trees, and dense commercial areas at the through low-rise residential areas, urban parks, a university

Fig. 3. a. Frontal area index map in eastewest direction; b. Frontal area index map in northeastesouthwest direction; c. Occurrence frequency of ventilation paths in eastewest
direction (total number of paths is 2500); d. Occurrence frequency of ventilation paths in northeastesouthwest direction (total number of paths is 5186).
M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889 1885

Fig. 3. (continued).
1886 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889

campus, a commercial district and harbor walkway. This route all the winds in Hong Kong, only these two directions were
is represented by occurrence frequency greater than 28% calculated in this study.
(1500 out of 5186 paths). These maps of ventilation paths facilitate the visualisation of
wind ventilation and show the specific locations in the city e.g. those
These four pathways generated were considered to have a high in red and purple colours, which have the highest frequency of
probability of stronger and more frequent air movement than other occurrence of paths are potential air flow corridors. (For interpre-
areas. Since the easterly and northeasterly winds account for 66% of tation of the references to colour in text, the reader is referred to the

Fig. 4. Occurrence frequency of ventilation paths overlaid with field measurements in a. eastewest direction, b. northeastesouthwest direction.
 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889 1887

web version of this article.) This enables better understanding of air were chosen because of the high occurrence frequencies of ventila-
ventilation at local scale, as well as over the whole urban area, tion paths. Fig. 4 shows the field measurements overlaid onto the
whereas previous studies using the concept of lf have been ventilation paths, and the dot sizes represent the wind speeds along
conducted at local scale, for small city districts. When the whole the route. Along route A, average and maximum wind speeds of
urban area is considered, identification of the ventilation corridors 9.3 ms1 and 17.8 ms1 were observed respectively. However, wind
by visual inspection from a building plan and/or a map of frontal area speeds off the pathway were much lower (ca. 2.3 ms1), and more
index alone would be impossible due to the complexity of the similar to the background wind speed of 2 ms1 for Kowloon
building and street network. For example, in our study, because the recorded by the Hong Kong Observatory at the time. About 55% of
four pathways traverse the whole urban area, each path is influenced the field readings with wind speed above 9.1 ms1 fall in route A
by a large number of buildings (average of 436) and comprises many (which has 24% of all paths generated). This identifies route A as
changes of direction (average of 24) along a route, and furthermore, a key connecting corridor along which air flows in the dominant
LCP analysis considers distance values as well as lf in the compu- wind direction, from the eastern to the western parts of the Kowloon
tation. On the other hand, the study by Gál and Unger ([26], op cit) peninsula. Along route D, the average and maximum wind speeds
describe a visual (non-automated) method for identifying the were lower than for route A, namely 3.5 ms1 and 12.4 ms1
ventilation paths, which appear unable to traverse the central urban respectively. But again the speeds along the route were much higher
areas, and only general ventilation paths are drawn. than that off the route (ca. 1.1 ms1) as well as higher than the
background wind speed of 0.65 ms1. This validation study
4.5. Validation strategy confirmed that analysis of the frequency of occurrence of ventilation
paths does indeed indicate relevant ventilation corridors over the
In order to investigate the significance and functionality of the city, since the highest field measured wind speeds corresponded to
ventilation paths, fieldwork was undertaken on 09 Oct 2009, 13 Oct the corridor with the highest frequency of paths from the LCP
2009, 15 Oct 2009, when winds were at 45, 90, and 45 , which are analysis i.e. route A with 30% of all paths generated, and wind speeds
representative of the dominant east and northeasterly winds off the paths were much lower.
throughout the year. A slow walk was undertaken along the differing Fig. 5 shows the designated ventilation paths overlaid onto the
sections of the routes A and D, recording the wind speed and GPS image of HII which has a 10 m resolution. Red and purple areas
locations along the routes. The wind speeds were measured four indicate the core regions of the urban heat island, having large
times for each section, resulting in 22 and 48 readings from routes A temperature differences from the rural areas. All four paths cross
and D respectively, and the results were averaged. These two routes these core areas at the shortest distance, especially route B, which

Fig. 5. Designated ventilation paths overlaid on heat island intensity image at 10 m resolution.
1888 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889

Fig. 6. Ventilation paths located using offshore starting and ending points to test the “wall effect” along coastlines. Red arrows indicate main locations of “wall effects” due to high-
rise building parallel to the coast, referred to in media reports. Colour legend as for Fig. 3c. (For interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)

passes through the core heat island region of MongKok (the highest To evaluate ventilation corridors at city scale, LCP analysis was
populated district in the world). Route D also passes through green performed. However, whereas LCP analysis usually operates with
open spaces with low HII, avoiding the core built areas with high a single or few pathways, this study overlaid all LCP paths to derive
HII values. These observations suggest that adequate ventilation is maps of frequency of occurrence of LCPs. Four major ventilation
indeed a key requirement for controlling the temperatures in inner corridors travelling from easterly and northeasterly directions were
areas of densely built, high-rise cities such as Kowloon. Indeed it identified from these maps. The corridors, routes A, B, C and D
may also be suggested that without the observed ventilation accounted for 30%, 18%, 16% and 28% respectively, of all least cost
corridors, the existing core heat island areas would expand and paths, and they travel from easterly and northeasterly directions
amalgamate to form a larger and more compact entity. respectively. Wind speeds were measured in the field to evaluate
Fig. 6 shows ventilation corridors located by LCP analysis using the relevance of the models, and these compare well with the
starting and ending points offshore, in order to test the “wall effect” frequencies of the modelled pathways.
due to high-rise buildings along the coast, which has been a cause In densely urbanised Hong Kong, the ventilation corridors of
of concern to the Hong Kong public. It is obvious that the paths 100 m width used in this study may not appear to provide viable
avoid, and have to navigate around the two coastlines with major connecting corridors at street scale, but are sufficient for city scale
wall effect buildings (red arrowed areas). A comparison with Fig. 5 modelling, as demonstrated by the analysis of the urban heat
indicates high HII values immediately or more generally inland of island. However, the model of frontal area index can be adapted for
these highly built coastlines, comprising heat island core areas. any grid resolution depending on the morphology of any particular
city. This study offers a simple and low cost method to investigate
the ventilation conditions of any city at a highly detailed level. City
5. Conclusion planners and environmental authorities may use the derived maps
for pinpointing the key buildings restricting air flow to the urban
This paper investigates urban ventilation pathways using the core. Thus the living environment in future urban renewal projects
frontal area index model over a densely built urban area, the may easily be improved by a simple shift in location and orientation
Kowloon peninsula of Hong Kong. We calculated the frontal area of buildings, while maintaining the same plot ratios, and the
index based on three dimensional building data and expressed this creation of ventilation corridors will not decrease ventilation
as a function of land use type. Most of the frontal area indices as elsewhere. The lf values calculated for this study did not account
a function of land use type were found to be greater than those for terrain elevation, since the study area was mainly flat. This could
computed for other studies elsewhere, and commercial and easily be achieved by converting the grid to a DEM, and allocating
industrial areas were found to have significantly high values a DEM value weighting to the lfvalues in each cell. Furthermore, if
because these areas are high-rise in Hong Kong. The environmental the main objective of designating ventilation paths was to maxi-
parameters heat island intensity (HII), building density (BD), and mise human comfort rather than for heat island analysis, then
building height (BH) were found to have significant positive ground level is more important than the tops of tall buildings, and
correlations with frontal area index, while vegetation cover (NDVI) a lower weight should be allocated to high level building facets.
had a significant negative relationship. These suggest that the Other future improvements to the model may include consider-
frontal area index presented herein is a useful parameter in ation of the type of surface, since winds traversing a vegetated
mesoscale meteorological studies and can contribute to heat island surface have a greater cooling capacity than those traversing
analysis. impervious surfaces.
 M.S. Wong et al. / Building and Environment 45 (2010) 1880e1889 1889

Acknowledgments [14] Dudhia J, Gill D, Manning K, Wang W, Bruyere C. PSU/NCAR mesoscale

modeling system tutorial class notes and user's guide: MM5 modeling system
version 3. NCAR; 2003.
The authors would like to acknowledge the Hong Kong Planning [15] Baik JJ, Kim JJ. A numerical study of flow and pollutant dispersion characteristics
Department for land use and Tertiary Planning Unit data and the in urban street canyons. Journal of Applied Meteorology 1999;38:1576e89.
Lands Department for building GIS databases. This project was [16] Blocken B, Carmeliet J, Stathopoulos T. CFD evaluation of the wind speed
conditions in passages between buildings e effect of wall-function roughness
funded by PolyU SURF project 1-ZV4V. modifications on the atmospheric boundary layer flow. Journal of Wind
Engineering and Industrial Aerodynamics 2007;95(9e11):941e62.
[17] Huber AH, Tang W, Flowe A, Bell B, Kuehlert K, Schwarz W. Development and
References applications of CFD simulations in support of air quality studies involving
buildings. In: Proceedings of 13th joint conference on the applications of air
[1] Baker L, Brazel A, Selover N, Martin C, McIntyre N, Steiner F, et al. Urbanization pollution meteorology with the Air & Waste Management Association. Van-
and warming of Phoenix (Arizona, USA): impacts, feedbacks and mitigation. couver, British Columbia, Canada; August 23e27 2004.
Urban Ecosystems 2003;6:183e203. [18] Kondo H, Asahi K, Tomizuka T, Suzuki M. Numerical analysis of diffusion
[2] Hawkins T, Brazel A, Stefanov W, Bigler W, Saffell E. The role of rural variability around a suspended expressway by a multi-scale CFD model. Atmospheric
in urban heat island determination for Phoenix, Arizona. Journal of Applied Environment 2006;40:2852e9.
Meteorology 2004;43(3):476e86. [19] Ashie Y, Tokairin T, Kono T, Takahashi K. Numerical simulation of urban heat
[3] Oke TR. The energetic basis of the urban heat island. Quarterly Journal of the island in a ten-km square area of central Tokyo. In: Annual report of the Earth
Royal Meteorological Society 1982;108:1e24. Simulator Center; 2007. p. 83e87.
[4] González JE, Luvall JC, Rickman DL, Comarazamy D, Picón AJ. Urban heat island [20] Lettau H. Note on aerodynamic roughness-parameter estimation on the basis
identification and climatological anlaysis in a coastal, tropical city: San Juan, of roughness-element description. Journal of Applied Meteorology
Puerto Rico. In: Weng Q, Quattrochi DA, editors. Urban remote sensing. Flor- 1969;8:828e32.
ida: CRC Press; 2007. p. 223e52. [21] Counihan J. Adiabatic atmospheric boundary layers: a review and analysis of
[5] Jusuf SK, Wong NH, Hagen E, Anggoro R, Hong Y. The influence of land use on data from the period 1880e1972. Atmospheric Environment 1975;9:871e905.
the urban heat island in Singapore. Habitat International 2007;31:232e42. [22] Grimmond CSB, Oke TR. Aerodynamic properties of urban areas derived from
[6] Fung WY, Lam KS, Nichol JE, Wong MS. Heat island study e satellite derived air analysis of surface form. Journal of Applied Meteorology 1999;34:1262e92.
temperature. Journal of Climate and Applied Meteorology 2009;48(4):863e72. [23] Burian SJ, Brown MJ, Linger SP. Morphological analysis using 3D building
[7] Nichol JE. High resolution surface temperature patterns related to urban databases. LA-UR-02-0781. Los Angeles, CA: Los Alamos National Laboratory;
morphology in a tropical city: a satellite-based study. Journal of Applied 2002. 36e42.
Meteorology 1996;35:135e46. [24] Bottema M. Urban roughness modelling in relation to pollutant dispersion.
[8] Nichol JE. Remote sensing of urban heat islands by day and night. Photo- Atmospheric Environment 1997;31:3059e75.
grammetric Engineering and Remote Sensing 2005;71(5):613e21. [25] Matzarakis A, Mayer H. Mapping of urban air paths for planning in Munchen,
[9] Nichol JE, Fung WY, Lam KS, Wong MS. Urban heat island diagnosis using 16. Wissenschaftlichte Berichte Institut for Meteorologie und Klimaforschung,
ASTER satellite images and ‘in situ’ air temperature. Atmospheric Research The University of Karlsruhe; 1992. 13e22.
2009;94(2):276e84. [26] Gál T, Unger J. Detection of ventilation paths using high-resolution roughness
[10] Haeger-Eugensson M, Holmer B. Advection caused by the urban heat island parameter mapping in a large urban area. Building and Environment
circulation as a regulating factor on the nocturnal urban heat island. Inter- 2009;44:198e206.
national Journal of Climatology 1999;19:975e88. [27] The Hong Kong Planning Department. Urban Climatic Map and Standards for
[11] Oke TR. Boundary layer climates. London and New York: Routledge; 1987. Wind Environment e Feasibility Study; 2009.
[12] Duijm NJ. Dispersion over complex terrain: wind-tunnel modelling and [28] Nichol JE, Wong MS. Spatial variability of air temperature over a city in
analysis techniques. Atmospheric Environment 1996;30(16):2839e52. a winter night. International Journal of Remote Sensing 2008;29(24):7213e23.
[13] Mfula AM, Kukadia V, Griffiths RF, Hall DJ. Wind tunnel modelling of urban [29] Jim CY. Impacts of intensive urbanisation on trees in Hong Kong. Environ-
building exposure to outdoor pollution. Atmospheric Environment 2005;39 mental Conservation 2004;25(2):146e59.
(15):2737e45. [30] Eastman R. IDRISI Andes tutorial. Worcester, USA: Clark Labs; 2006.