Urban Forestry & Urban Greening 11 (2012) 73–85

Contents lists available at SciVerse ScienceDirect

Urban Forestry & Urban Greening

journal homepage: www.elsevier.de/ufug

Weather effect on thermal and energy performance of an extensive tropical

green roof
C.Y. Jim ∗ , Lilliana L.H. Peng
Department of Geography, The University of Hong Kong, Pokfulam Road, Hong Kong

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

Keywords: This study investigated the weather effect on thermal performance of a retrofitted extensive green roof on
Cooling load a railway station in humid-subtropical Hong Kong. Absolute and relative (reduction magnitude) ambient
Energy conservation and surface temperatures recorded for two years were compared amongst antecedent bare roof, green
Extensive green roof
roof, and control bare roof. The impacts of solar radiation, relative humidity, soil moisture and wind
Heat flux
speed were explored. The holistic green-roof effect reduced daily maximum tile surface temperature
Passive cooling
Thermal performance by 5.2 ◦ C and air temperature at 10 cm height by 0.7 ◦ C, with no significant effect at 160 cm. Green-roof
passive cooling was enhanced by high solar radiation and low relative humidity typical of sunny summer
days. High soil moisture supplemented by irrigation lowered air and vegetation surface temperature, and
dampened diurnal temperature fluctuations. High wind speed increased evapotranspiration cooling of
green roof, but concurrently cooled bare roof. Heat flux through green roof was also weather-dependent,
with less heat gain and more heat loss on sunny days, but notable decline in both attributes on cloudy
days. On rainy days, green roof assumed the energy conservation role with slight increase instead of
reduction in cooling load. Daily cooling load was 0.9 kWh m−2 and 0.57 kWh m−2 , respectively for sunny
and cloudy summer days, with negligible effect on rainy days. The 484 m2 green roof brought potential
air-conditioning energy saving of 2.80 × 104 kWh each summer, equivalent to electricity tariff saving of
HK$2.56 × 104 and upstream avoidance of CO2 emission of 27.02 t at the power plant. The long-term
environmental and energy benefits could justify the cost of green roof installation on public buildings.
© 2011 Elsevier GmbH. All rights reserved.

Introduction
Green roofs can serve as surrogate green spaces in urban
areas dominated by artificial structures and surfaces (Arnold and
Gibbons, 1996). Many cities, especially the compact ones, have
adopted this innovative way of extending urban greening into oth-
erwise bare rooftops. Green roofs are well known for their wide
range of environmental, ecological, social and economic benefits
(Emilsson, 2006; Mentens et al., 2006; Oberndorfer et al., 2007).
With rapid urbanization and aggravation of the urban heat island
(UHI) effect in recent years, the thermal and energy performance of
green roofs has attracted the attention of researchers, professionals,
civil servants and citizens.
Many empirical and modelling studies of green roofs have
been conducted in different climatic zones and cities of different
sizes and development modes. The findings demonstrate notable
green-roof contribution to lowering roof surface and ambient
air temperature, reducing transmission of solar heat through the
roof slab into indoor space, and trimming building cooling load
∗ Corresponding author. Tel.: +852 2859 7020; fax: +852 2559 8994.
E-mail address: hragjcy@hku.hk (C.Y. Jim).
(Takakura et al., 2000; Papadakis et al., 2001; Akbari, 2002; Wong
et al., 2003a,b,c; Sonne, 2006; Santamouris et al., 2007; Wong et al.,
2007a,b; Spala et al., 2008; Eumorfopoulou and Kontoleon, 2009;
Teemusk and Mander, 2009; He and Jim, 2010; Jim and He, 2010;
Williams et al., 2010). Cooling is attributed to the combined effect of
evaporation from soil, transpiration from plants, shading by plants
and substrate, and insulation by the soil and drainage layers with
constituent moisture (Harazono, 1990/91). Summer cooling could
range from 15 to 45 ◦ C in daily peak surface temperature and 2–5 ◦ C
in peak air temperatures. Energy saving in air-conditioning elec-
tricity in summer could amount to 10–80% for individual buildings
(Getter and Rowe, 2006).
Besides cooling, green roofs can bring environmental and social
benefits. The living vegetation can filter or absorb air pollutants
to improve air quality. The soft vegetation cushion, in conjunc-
tion with soil, can serve as a noise barrier. The soil and vegetation
layers can clean and reduce stormwater discharged from build-
ings to improve water quality and abate flood hazard (DeNardo
et al., 2003; Villarreal and Bengtsson, 2004; VanWoert et al., 2005).
The vegetated roofs attract wildlife and flora to the above-ground
habitats and stepping stone sites to enhance urban biodiversity
(Brenneisen, 2004; Burgess, 2004). Buildings with well-designed
and maintained green roofs can fetch higher rental and sale values.

1618-8667/$ – see front matter © 2011 Elsevier GmbH. All rights reserved.
doi:10.1016/j.ufug.2011.10.001
74 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85
Fig. 1. Locations of the green roof experimental plot (A) and the bare control plot (Z) on the top of the Tai Po Railway Station.
The elevated green spaces provide recreational and amenity venues
in semi-private or private grounds. Home-owners can enjoy out-
door life in a safe, clean, secluded and serene setting. Offices and
factories with pleasant green roofs can attract and retain workers,
who tend to be happy, healthy, productive and loyal. Due to multi-
ple environmental, economic and social benefits, green roofs offer a
promising, welcomed and cost-effective contemporary technology
for buildings.
A modern green roof is composed of multiple layers of syn-
thetic and natural materials installed in sequence from the bottom
upwards: root barrier, drainage, filter, substrate, and vegetation.
The first three layers are synthetic, manufactured and durable
materials; the last two are natural. Two major types of green roofs,
extensive and intensive, are distinguished by vegetation height and
biomass structural complexity, which require different soil depth.
The more common extensive green roof is usually covered by herbs,
grasses or drought-tolerant succulent plants such as Sedum species.
The intensive green roof embraces trees and shrubs often planted
closely. The extensive green roof usually has 2–20 cm of substrate,
and the intensive, over 20 cm (Oberndorfer et al., 2007).
Characterized by an exceptionally compact urban morphology,
Hong Kong suffers from severe shortage of ground-level green
spaces and intense UHI effect. The green-space deficit has degraded
the quality of the urban environment and associated quality of
life (Jim, 2004, 2008). Most of the 40,000 odd high-rise buildings
mainly with flat reinforced-concrete rooftops have adequate load-
bearing capacity to receive green roofs. Thus urban Hong Kong has
a huge potential to embrace the green-roof innovation (Tsang and
Jim, 2010). Empirical studies of thermal performance could throw
light on underlying factors and processes, furnish hints to optimize
design, and offer an objective basis to promote green roofs (Niachou
et al., 2001; Köhler et al., 2002; Köhler, 2004). More local or regional
studies in the humid-subtropical context could be attempted, as the
results from other climatic zones may not be directly applicable.
The railway station at Tai Po covers a large area with significant
demand on air-conditioning energy to cool the indoor space in sum-
mer. Installing a green roof could bring thermal and energy benefits,
reduce cooling load and improve landscape and amenity values.
This study focuses on the weather effect on green roof thermal and
energy benefits with three specific objectives:
(1) Explore the impacts of the green roof on summer daily surface
and air temperature;
(2) Evaluate the influence of meteorological variables on green-
roof thermal behaviour; and
(3) Analyse green-roof thermal performance on the days represen-
tative of typical summer weather conditions.
Comparing with the literature, this study is different in two
aspects:
(1) Evaluate green-roof thermal effect with reference to a compre-
hensive range of summer weather conditions covering sunny,
cloudy and rainy days; in contrast, most studies concentrate on
sunny days; and
(2) Acquire antecedent environmental monitoring of the experi-
mental site for one year before green roof installation to permit
longitudinal comparisons between before and after scenarios.
Study area and method
Study area
Hong Kong is situated at the south coast of China, at 22◦ N
latitude and 114◦ E longitude with a typical humid-subtropical cli-
mate influenced by the dominating Asian monsoon climatic system.
The summer is hot and humid with frequent showers and thun-
derstorms and occasional typhoons. It extends from late April to
September, with August the hottest month often exceeding 33 ◦ C.
The rainy season largely covers summer with annual rainfall over
2000 mm. The winter is relatively short and mild, running from
January to February with average temperature above 10 ◦ C. The
long-hot period vis-a-vis the relatively short-cool period indicate
summer cooling as a priority in sustainable urban design to usher
environment and human health.
An extensive green roof was retrofitted in July 2009 on the Tai Po
railway station situated in a suburban new town in Hong Kong. The
station is a one-to-two storey low-rise structure facing the platform
and tracks. The large flat rooftop is composed of several parcels with
different elevation and area (Figs. 1 and 2). This study enlists the
largest plot (the experimental site with green roof) which is square
in shape with 484 m2 . A nearby plot (the control site with original
bare roof), with 106 m2 , provides a baseline for comparison.
Experimental design and environmental monitoring
The modern extensive green roof (Figs. 3 and 4) was installed
on the reinforced-concrete flat-roof protected by a waterproofing
membrane, screed and cement tiles. A proprietary multiple-layer
green-roof system (Nophadrain, Kirkrade, the Netherlands) was
laid directly on the tile surface with a 2% gradient to shed drainage
water. From bottom to top, it contains five layers: plastic (polyethy-
lene) root barrier, plastic (high impact polystyrene) drainage
(2.5 cm), geotextile filter, rockwool water retention (5 cm), and
growing medium (5 cm). The soil mix is composed of mineral soil
(locally derived completely decomposed granite) mixed with 20%
(v/v) mature compost. It has a sandy loam texture (with around 78%
sand, 10% silt and 12% clay) and granular structure (Soil Survey Staff,
1999) to facilitate free drainage and aeration, with a water holding
capacity of about 20% (v/v), and pH range of 5.5–5.8. It has been
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85
Fig. 2. Location of the green roof study site on the rooftop of the taller structure
(indicated by red arrow) of Tai Po Railway Station. (For interpretation of the refer-
ences to color in this figure legend, the reader is referred to the web version of the
article.)
Fig. 3. The experimental green roof well covered by vigorous herbaceous ground-
cover vegetation Perennial Peanut (Arachis pintoi).
Fig. 4. Position and height of environmental monitoring sensors at green roof experimental plot (site A) and bare roof control plot (site Z).
75
formulated to emulate the topsoil of the local red earth (Oxisol soil
order) commonly found in the humid-tropical area. This soil com-
position is widely used as a growing medium for landscape planting
in Hong Kong and the south China region (Jim, 1996). The soil mix
has been tested at other green roof sites prior to our experimental
study, and has been found to be a suitable material that is readily
available from local sources, relatively inexpensive, and rather free
from inherited weed seeds.
A perennial, herbaceous and evergreen vine, Arachis pintoi
(Perennial Peanut), was chosen for its vigorous growth and forma-
tion of a tight and complete green cover. It is a tropical low-growing
(about 20 cm thick), ground-hugging, non-twining and nitrogen-
fixing legume. The species has high-temperature tolerance, and
rather good resistance to weed, pest and disease invasion (CTAHR,
2010). Its ornamental value is attributed to the continuously
verdant green foliage and many dainty golden-yellow flowers pre-
sented for seven warm months in a year. Vigorous stem cuttings
were spread at a high density on the prepared soil surface. The
vigorous plant was able to achieve 100% site coverage in three
months of growth in the warm season. Thereafter, a continu-
ous green mantle has been maintained with gradual addition of
new stems and foliage to enhance the thermal insulation effect.
An automatic sprinkler irrigation system provided supplementary
water supply at 5 L/m2 /day, sustaining an average soil moisture
content of about 0.3 m3 /m3 . Watering was regulated by a rain-
fall detector to stop the pump when antecedent rainfall exceeded
10 mm.
Environmental sensors were placed at the centre of both sites
(Figs. 4 and 5) to minimize the edge effect for long-term monitoring
of meteorological and soil parameters. The sensors were located at
a spot free from the influence of shadows or reflected radiation
from adjacent buildings or trees (Fig. 3). Readings were taken at
15-min interval and stored in Hobo (Onset, Cape Cod, MA, USA)
data loggers. The types and positions of sensors are given in Table 1
and Fig. 4, and the measured parameters in Table 2. The following
notations are used to denote the site terminology:
• Site B: Experimental plot before green-roof installation;
• Site A: Experimental plot after green-roof installation;
• Site Z: Control plot (bare baseline roof).
76 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85

Table 1
Environmental sensors and installation positions on the experimental and control roof sites.

Site Sensor Measured environmental parameter and position Brand and model

Experimental Infrared temperature sensor
(B or A) Air temperature sensor
Relative humidity sensor
Tile temperature sensor
Soil moisture sensor
Pyranometer
Anemometer
Control Infrared temperature sensor
(Z) Air temperature sensor
Relative humidity sensor
Surface temperature of ceiling and vegetation canopy
Air temperature at 10 cm and 160 cm above the ground
Relative humidity at 10 cm and 160 cm above the ground
Surface temperature of concrete slab (below green roof)
Soil moisture at 35 mm depth
Intensity of solar radiation
Wind speed and wind direction at 160 cm above the ground
Surface temperature of ceiling and bare concrete slab
Air temperature at 10 cm and 160 cm above the ground
Relative humidity at 10 cm and 160 cm above the ground
Apogee SI-111 (Logan, UT, USA)
Onset Hobo S-THB-M002 (Bourne, MA, USA)
Onset Hobo S-THB-M002 (Bourne, MA, USA)
Onset Hobo S-THB-M017 (Bourne, MA, USA)
Onset Hobo S-SMC-M005 (Bourne, MA, USA)
Onset Hobo S-LIB-M003 (Bourne, MA, USA)
Onset Hobo S-WCA-M003 (Bourne, MA, USA)
Apogee SI-111 (Logan, UT, USA)
Onset Hobo S-THB-M002 (Bourne, MA, USA)
Onset Hobo S-THB-M002 (Bourne, MA, USA)

Data analysis Table 2

Symbols and measurement units used in this study.

The monitoring data in August 7–September 17, 2008 before Symbol Unit Explanation
green-roof installation represent the “before session” (site B). The Study site:
data acquired after roof greening in August 7–September 17, 2009 A na Experimental plot after green roof installation
represent the “after” session (site A). Thus the study collected B na Experimental plot before green roof installation
antecedent data of the experimental plot before green-roof instal- Z na Control (bare roof) plot
Surface temperature:
lation to serve as an in situ baseline. Table 3 presents the summary ◦
Tc C Ceiling surface temperature
weather statistics of a nearby station in the study period. One- Tt ◦
C Concrete tile surface temperature
way ANOVA test confirmed no significant difference in background ◦
Ttb C Concrete tile surface temperature of site B
◦
weather condition between the before and after sessions to render Tta C Concrete tile surface temperature of site A
◦
Ttz C Concrete tile surface temperature of site Z
the comparison valid. ◦
Tv C Vegetation canopy surface temperature
Four temperature variables were derived as indicators of green- Air temperature:
roof thermal effect: U10 ◦
C Air temperature at 10 cm height
◦
U160 C Air temperature at 160 cm height
◦
U10b C Air temperature at 10 cm height of site B
• Daily maximum temperature (DM) of Tv , Tt , U10 , U160 ; U160b ◦
C Air temperature at 160 cm height of site B
• Average daily maximum (ADM) of Tv , Tt , U10 and U160 ; U10a ◦
C Air temperature at 10 cm height of site A
• Daily heat flux; U160a ◦
C Air temperature at 160 cm height of site A
• Cooling load. Derived temperature attribute:
◦
DM C Daily maximum temperature
◦
DA C Daily average temperature
◦
ADM C Average daily maximum temperature (monthly
basis)
◦
DMR C Daily maximum temperature reduction
◦
DAR C Daily average temperature reduction
Thermal attribute:
k W m−1 K−1 Thermal conductivity
V W m−2 Heat flux
Other weather parameter:
RH10 % Relative humidity at 10 cm height
RH160 % Relative humidity at 160 cm height
SR W m−2 Solar radiation
WS m s−1 Wind speed
SM m3 /m3 Soil moisture

Heat flux through the concrete slab was calculated based on the
temperature difference between roof tile surface (Tt ) and ceiling
surface (Tc ):

k(Tt − Tc )
Heat flux through the roof slab, V = (1)
d

where k is the thermal conductivity of the roof slab (W m−1 K−1 );

Tt is the roof tile surface temperature (K); Tc is the ceiling surface
temperature (K); and d (m) is the thickness of the roof deck. A
positive V value indicates incoming heat flux to the interior space
(heat gain), whereas a negative value denotes outgoing heat flux
(heat loss).
Daily cooling load change of the roof slab was calculated by
accumulating heat fluxes over 24 h in a day:


n=96
Cooling load = 0.25 × Vi × S (2)
Fig. 5. The instrument stand situated at the centre of the experimental green roof
supporting the environmental sensors and data loggers. i=1
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85 77

Table 3
Comparison of background weather parameters between the before and after sessions.

Air temperature (◦ C) Relative humidity (%)

Mean Stan. error F Sig. Mean Stan. error F Sig.

Before 31.20 0.35 3.77 0.06 82.20 1.49 1.22 0.27

After 32.20 0.31 84.30 1.16

Fig. 6. Daily maximum (DM) surface temperature in August and September for: (a) before session in 2008 between Ttb and Ttz ; and (b) after session in 2009 amongst Tta , Tv
and Ttz .

where Vi is the ith heat flux over a day, with V1 meaning heat flux
at 0000 h, V2 heat flux at 0015 h, and so on; and S is the roof area
(m2 ).
Data analysis was implemented in three stages:
(1) Compare surface and air temperature between site A and sites
B and Z to detect green-roof thermal effects in summer;
(2) Explore the effects of key meteorological variables on green-
roof thermal performance;
(3) Analyse green-roof thermal performance on selected summer
days with typical weather conditions.
Green-roof thermal performance
Surface temperature
Roof surface temperature critically affects sensible heat flux and
longwave radiation to the air, and downward heat flux to the inte-
rior space. Fig. 6 compares the daily maximum DM of tile surface
(Tt ) at sites B, A and Z. In the before session, site Z has consistently
higher DM than site B (Fig. 6a). The temperature peaks and troughs
of the two sites are synchronous, but the amplitude at site Z exceeds
site B.
In the after session, DM at Ttz remains high, with 58% (19 out of
33) of the days exceeding 40 ◦ C (Fig. 6b). In contrast, DM at Tta drops
notably with only three days exceeding 30 ◦ C. Such drastic temper-
ature reduction at the green-roof bottom is attributed to passive
shading and insulation in tandem with active evapotranspiration.
The difference in DM between sites Z and A (Ttz − Tta ) ranges from
0.48 ◦ C (on August 12) to 12.5 ◦ C (on August 23), indicating notable
daily variations in green-roof thermal performance. The weather
influence is evident. On relatively cool days, the thermal reduction
is subdued; on hot days, the contrast widens.
The control tile temperature Ttz and green-roof vegetation
canopy temperature Tv are compared (Fig. 6b). Contrary to expecta-
tion, DM at Tv is higher than Ttz on most days, with a few exceptions
in late August and early September. This vegetation canopy anomaly
might be explained by: (1) higher net solar radiation on green roof
due to lower albedo of vegetation leaves; (2) trapping of heat in the
canopy when evapotranspiration rate declines in early afternoon
due to exhaustion of stored soil moisture; and (3) the incomplete
coverage of the groundcover herb resulting in a notable proportion
of exposed soils.
Table 4 presents the average daily maximum (ADM) surface
temperature of sites B, A and Z. Comparing sites A and B, green
roof reduces the tile ADM by 5.2 ◦ C (Tta − Ttb ), which denotes the
holistic green-roof effect. The vegetation canopy ADM is raised by
3.4 ◦ C (Tv − Ttb ), indicating the vegetation canopy anomaly. Com-
paring sites A and Z, green roof lowers tile ADM by a notable 8.1 ◦ C
(Tta − Ttz) . The vegetation canopy suppresses the tile ADM by 0.5 ◦ C
(Ttz – Tv ).
Air temperature
In the before session, summer DM air temperature at site Z is
warmer than site A at U10 by 2.4 ◦ C, and at U160 by 2.5 ◦ C (Fig. 7a and
b). At site Z, DM at the low level U10 is warmer than the high level
U160 by about 1 ◦ C, but less so at site A. After roof greening, site A is
notably cooler than site Z at U10 by 4.4 ◦ C (September 5) (Fig. 7c).
At U160 , site A is cooler than Z by a maximum of 2.3 ◦ C (Fig. 7d),

Table 4
Comparison of average daily maximum (ADM) surface temperature (◦ C) in the before and after sessions and between experimental and control sites.

Baseline to assess green roof effect

Site A before or after green roof installation Control roof Site Z before or after green roof installation

Holistic green roof effect Vegetation canopy effect Holistic green roof effect Vegetation canopy effect

ADM timing Tta − Ttb Tv − Ttb Ttz Tta − Ttz Ttz − Tv

After state, Y 30.20 38.80 38.3 30.20 38.30
Before state, X 35.40 35.40 40.5 38.30 38.80
Y−X −5.20 3.40 na −8.10 −0.50
78 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85

Fig. 7. Daily maximum (DM) air temperature in August and September between: (a) sites B and Z in 2008 at U10 ; (b) sites B and Z in 2008 at U160 ; (c) sites A and Z in 2009 at
U10 ; and (d) sites A and Z in 2009 at U160 .

Table 5 with bare roof, respectively DM reduction (DMR) and DA reduction

Comparison of average daily maximum air temperature (ADM, in ◦ C) of the control
(Z) and experimental (B or A) sites in the before and after sessions.
Sensor Before session (Z – B) After session (Z − A)
Mean Std. error Mean
U10 1.37 0.1 2.1
U160 1.28 0.1 1.1
indicating more effective vegetative cooling near the ground and
its notable reduction with height. However, within each site (A and
Z), U10 remains warmer than U160 .
In the before session, summer ADM air temperature at U10 at
site A is cooler than site Z by 1.37 ◦ C; after roof greening, it increases
to 2.10 ◦ C, indicating vegetative cooling by around 0.7 ◦ C (Table 5).
At U160 , the corresponding temperature reductions are 1.28 ◦ C and
1.1 ◦ C for the before and after sessions. Before roof greening, the
inter-site temperature differences at the two levels are similar. The
green-roof cooling effect, however, is notably subdued (0.1 ◦ C) at
the higher level.
Effect of weather parameters on green-roof cooling
General observations
This section explores underlying weather factors on thermal
effects of green roof. Pearson correlation analyses identified the
key meteorological variables, including daily average values of solar
radiation (SR), relative humidity (RH), soil moisture (SM), and wind
speed (WS) (Table 6). At the four sensor positions of the green
roof, namely Tta , Tv , U10 and U160 , four thermal-effect indicators
were computed: daily maximum (DM) and daily average (DA) tem-
perature, and temperature reduction of green roof in comparison
(DAR).
Overall, the weather parameters exert more influence on two
sensors that are contiguous (Tta ) or close to the green-roof stra-
Std. error tum (U10 ) than the somewhat elevated (Tv ) and the significantly
elevated (U160 ). Of the four sensor positions, Tv has the smallest
0.25
0.12 number of significant correlations and relatively weaker corre-
lation coefficients. For weather parameters, SR and RH have the
largest number of significant correlations and relatively stronger
coefficients. WS has a smaller number of significant correlations
and weaker coefficients. Within a given weather parameter, the
sign of significant correlation coefficients is consistently either pos-
itive or negative, except WS which has both.
Solar radiation
SR is positively correlated (p < 0.01) with all thermal-effect indi-
cators at all sensor positions except DA at Tv (Table 6). High SR
can induce high temperature at vegetation surface Tv and at green-
roof bottom Tta , and high air temperature at U10 and U160 . More
SR can sustain evapotranspiration to bring cooling by latent heat at
green roof in comparison with bare roof. The cooling effect is more
pronounced on hot days with high SR.
Comparing the four sensor positions, more DMR and DAR at Tta
as indicated by stronger correlation coefficients, curtail downward
heat flux to the indoor space. For air temperature, the SR effect is
stronger on DM and DA but less so for DMR and DAR. SR is not
correlated with DA at Tv , suggesting that living foliage has different
response to SR in comparison with abiotic materials.
Relative humidity
Similar to SR, RH is significantly correlated with all thermal-
effect indicators (p < 0.05 or p < 0.01) again except DA at Tv , but
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85 79

Table 6
Pearson correlation coefficients between weather parameters and thermal-effect indicators of the green roof.

Sensor position Thermal-effect indicatora Weather parameter

Solar radiation Relative humidity Soil moisture Wind speed

SR RH SM WS

Surface temperature of vegetation canopy, Tv DM .523** −.473** −0.33 0.00

DA 0.00 0.22 0.04 −0.31
DMR .529** −.517** −0.19 0.29
DAR .498** −.644** −.382* .711**
Surface temperature of roof tile, Tta DM .597** −.700** −.478** .440**
DA .465** −.663** −.561** .493**
DMR .701** −.644** −0.32 0.11
DAR .855** −.916** −569** .594**
Air temperature at 10 cm height, U10 DM .702** −.452** −.425** −.683**
DA .860** −.814** −.755** −.453**
DMR .571** −.611* −.505** 0.19
DAR .599** −.553* −.502** 0.01
Air temperature at 160 cm height, U160 DM .788** −.581** −.515** −.578**
DA .704** −.816** −.726** −.411**
DMR .608** −.555** −0.36 −0.06
DAR .535** −.426** −.420* −\0.06
a
DM: daily maximum, DA: daily average, DMR: daily maximum reduction, DAR: daily average reduction; reduction refers to the lowering of green roof temperature in
comparison with bare roof.
*
Correlation is significant at the 0.05 level (2-tailed).
**
Correlation is significant at the 0.01 level (2-tailed).

the relationships are all negative (Table 6). High RH induces lower
temperature and limited temperature reduction. This is related to
the frequent association between high RH and low SR in the local
summer weather to bring lower surface and air temperature. The
restrained green-roof cooling could be attributed to elevated RH
incurring a reduced vapour pressure gradient between the sur-
face and air, and leading to suppression of evapotranspiration. The
dampening effect of RH is expressed by its strong negative cor-
relation with DAR at Tta (r = −0.916), indicating the notably stifled
cooling effect during high RH periods often associated with overcast
or rainy days.
Soil moisture
On sunny or cloudy days, the soil moisture could reach
0.35 m3 /m3 after the irrigation, and drop to around 0.25 m3 /m3
later in the afternoon. On rainy days, it can rise to 0.45 m3 /m3
depending on the intensity and duration of the rain event. SM effect
is similar to RH, with negative significant (p < 0.05 or p < 0.01) cor-
relations with all except five thermal-effect indicators associated
with Tv , Tta and U160 (Table 6). Tv is particularly detached from
SM effect, except DAR. The SM correlation coefficients are weaker
than SR and RH. The correlations with DM and DA at U10 and U160
demonstrate the enhancement of evapotranspiration cooling by
SM. SM also has negative effect on DM and DA at Tta , reflecting
enhancement of thermal insulation and reduction of heat flux into
the indoor space. SM plays a dual and complex role in green-roof
heat budget and thermal performance. The high thermal capac-
ity of soil water can subdue temperature rise, but the stored heat
can retard temperature drop and maintain the elevated temper-
ature for a longer period. The overall result is a lower maximum
temperature coupled with a higher minimum temperature, and a
contracted diurnal temperature range. The supply of SM is required
to sustain the evapotranspiration cooling effect.
The correlations between SM and DMR and DAR are negative. Of
the four thermal-effect indicators, DMR is only statistically signifi-
cant at U10 . DAR, however, is significant for all four sensor positions.
This is related to the superposition of artificial (irrigation) on the
natural (rainfall) regulation of SM. The green roof was irrigated
daily to nourish the newly planted vegetation. On days without
rainfall, the daily water supply matches the SM pattern with limited
amplitude of fluctuation. The main variations in SM are attributed
to rainfall and its attendant influence on energy dynamics on both
green and bare roofs. Rainfall triggers almost immediate temper-
ature drop on bare roof, whereas green roof shows a delayed and
subdued drop due to stored heat mainly in soil water, and inter-
ception and shielding effect of plants. Thus higher SM is associated
with less green-roof cooling.
Wind speed
WS is correlated with only half of the thermal-effect indicators
(Table 6). Unlike the previous three weather attributes which have
consistent negative or positive significant coefficients, WS has four
positive contrasting with four negative coefficients. Tv is the least
correlated, whereas Tta the most correlated. High WS incurs lower
DM and DA at U10 and U160 on green roof. DMR and DAR are not cor-
related with WS at U10 and U160 , but DAR is significantly correlated
with WS at Tv and Tta . Wind can augment evapotranspiration cool-
ing on green roof, and carry away heat from both bare and green
roofs to cool surface and air temperature. WS does not bring DMR
or DAR of air temperature at both heights, but it can enhance DAR at
Tv and Tta . Thus WS effect on green roof is largely exercised through
intermediary effect of evapotranspiration.
Overall assessment
SR and RH present the key weather determinants of green-roof
cooling effect. SM exerts as a multiple and interactive factor, with
important influence on cooling by evapotranspiration, heat stor-
age and associated dampening of temperature fluctuations, and
thermal insulation to accentuate green-roof thermal performance.
WS can reduce air temperature, but its effect on green roof is not
notably better than bare roof. The cooling effect of weather parame-
ters on green roof at Tta and Tv is partly offset by concurrent cooling
of bare roof.
Comparing the sensor positions, Tv has the weakest correla-
tion with weather parameters. Tv is somewhat elevated above the
soil surface by a three-dimensional lattice of interwoven stems of
Perennial Peanut. The foliage surface temperature is largely regu-
lated internally by plant thermal properties and operational state of
physiological functions. The external weather influence has limited
influence on the four thermal-effect indicators of living foliage with
an estimated leaf area index of about 1.2. Other plant growth forms
80 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85

Table 7
Key weather conditions of three sets of typical summer days in 2008 and 2009.

Typical weather Date Sunshine Maximum air Minimum air Cloud cover Mean relative Rainfall
duration (h) temperature (◦ C) temperature (◦ C) (%) humidity (%) (mm)

Sunny summer August 14, 2008 11.2 33.5 26.0 21.0 74.0 0.0
August 21, 2009 11.3 33.4 26.1 21.0 70.0 0.0
Cloudy summer August 19, 2008 5.8 31.2 27.9 47.0 78.0 0.0
August 25, 2009 6.7 31.1 27.6 80.0 78.0 0.0
Rainy summer August 08, 2008 0.1 27.1 24.8 90.0 93.0 108.0
August 13, 2009 0.2 27.4 24.8 90.0 96.0 70.0

such as shrubs and trees with taller and more complex biomass
structure could provide better thermal insulation and transpiration
cooling.
Temperature and heat flux of typical weather scenarios
This section explores the green-roof thermal effects under three
typical summer weather conditions, namely sunny, cloudy (over-
cast) and rainy days. Green roof is compared with bare roof in terms
of diurnal variations in surface and air temperature and heat flux.
The marked differences in weather conditions of the sampled sum-
mer days, for the before and after sessions, are listed in Table 7. The
surface temperature at Ttb , Tta and Tv are plotted in Fig. 8.
Surface temperature
On sunny summer day, green roof at Tta demonstrates drastic
suppression of surface temperature and fluctuation amplitude than
the antecedent bare roof at Ttb due to the holistic thermal effect
of the vegetated roof system (Fig. 8a). Tta experiences a limited
range of 28.2–30.3 ◦ C, whereas Ttb a wide range of 25.3–39.4 ◦ C.
The largest difference (11 ◦ C) between Ttb and Tta occurs at 1200 h
when Ttb temperature peaks. In the 24-h cycle, Ttb exceeds Tta dur-
ing the day, but drops below Tta after sunset at 1800 h until sunrise
at 0600 h. This nighttime reversed temperature difference, how-
ever, is limited to 1–2 ◦ C only. Thus green roof displays a nocturnal
inhibition of heat release. It affords a heat sink to store solar energy
acquired in daytime, largely due to thermal capacity of soil mois-
ture, which is released slowly in nighttime to sustain the elevated
nocturnal temperature.
Still on sunny day, the diurnal temperature trend of Tv is simi-
lar to Ttb (Fig. 8a). Tv is slightly cooler in the afternoon from 1100
to 1800 h due to high evapotranspiration rate. However, Tv has
slightly warmer nighttime and morning temperature than Ttb due
to retained heat in the foliage water and rapid nocturnal radiative
cooling at bare roof.
The green-roof cooling effect on cloudy summer day is notably
suppressed (Fig. 8b) in comparison with sunny day, mainly due to
cooler Ttb than warmer Tta . Ttb ranges from 27.1 to 36.1 ◦ C, whereas
Tta is similar to sunny day and fluctuates within merely 2 ◦ C. The
green roof can reduce peak Tta by 6.3 ◦ C, and the widest difference
between Tta and Ttb of 7.2 ◦ C occurs at around 1300 h. The nighttime
scenarios display a reversed pattern similar to sunny day, with Tta
slightly warmer than Ttb .
The cloudy day Tv trend deviates from sunny day (Fig. 8b). Tv
remains warmer than Ttb in both daytime and especially in night-
time. This situation could be attributed to the combined effect of
evapotranspiration suppression at green roof under overcast con-
dition, and reduced radiative cooling at night.
On rainy day, the surface temperature contrasts with sunny and
cloudy days (Fig. 8c). The energy conservation role of green roof is
mobilized, with Tv notably warmer than Ttb by >5 ◦ C throughout
daytime and nighttime. The widest difference between Tv and Ttb of
5.7 ◦ C occurs in the early afternoon. Tta is warmer than Ttb through-
out the 24-h cycle except a short period from 1600 to 1800 h, and
with the widest difference of 3.8 ◦ C occurring at around midnight.
Rainfall could cause a rapid and dramatic drop in surface temper-
ature of bare roof, whereas the effect on green roof is subdued in
terms of both speed and magnitude. This could be explained by
vegetation interception of rainfall, and soil heat reservoir of soil
moisture. The moisture in the living foliage can also trap heat and
keep the green surface relatively warm.
Air temperature
On sunny summer day, the green-roof diurnal air temperature
pattern could be demarcated into two episodes (Fig. 9a). At U10 , the
short first episode from 0700 to 1130 h records a slightly warmer
green roof. The long second episode from 1130 h to 0700 h entails
prominent and continuous cooling of the vegetated roof. At U10 ,
the peak temperature of 36 ◦ C is attained at around 1400 h, which
is 1.4 ◦ C lower than bare roof. The widest difference of 4.5 ◦ C occurs
at 1545 h when U10 reaches its maximum at bare roof. At U160 ,
the duration and magnitude of green-roof cooling are restrained

Fig. 8. Diurnal surface temperature variations between the green and the bare roof: (a) on a typical sunny summer day; (b) cloudy summer day; and (c) rainy summer day.
Ttb denotes surface temperature of the bare roof tile at site B before green roof installation; Tta denotes surface temperature of roof tile beneath the green roof at site A; and
Tv denotes surface temperature of vegetation canopy on the green roof at site A.
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85
Fig. 9. Diurnal air temperature variations on typical sunny summer days: (a) at U10 for before and after sessions; (b) at U160 for before and after sessions; (c) at U10 and U160
for before session; and (d) at U10 and U160 for after session. The before session was recorded on August 14, 2008 and after session on August 21, 2009.
in comparison with U10 (Fig. 9b). The cooling episode runs from
1200 to 1800 h, with maximum cooling of 3.4 ◦ C occurring around
1330 h. Overall, the cooling effect is more effective at U10 than U160 ,
indicating a height-decay function.
U10 and U160 on sunny day are plotted for before and after ses-
sions (Figs. 9c and 6d). They display a similar diurnal temperature
pattern in daytime, with U10 slightly warmer than U160 . In daytime,
the warmed ground surface, with or without green roof, serves as a
heat source to warm the near-ground air by conduction and convec-
tion. In nighttime, the two levels diverge, with U10 slightly warmer
than U160 on bare roof, but slightly cooler than U160 on green roof.
In nighttime, bare roof surface continues to warm the air by releas-
ing longwave radiation, whereas green roof cools the air due to
subdued but continuous evaporation. Thus green-roof cooling of
ambient air on sunny summer day is characterized by warming in
daytime and cooling in nighttime.
On cloudy day, the green-roof cooling effect is smaller than
sunny day due to solar screening by clouds. No decrease in peak
U10 is observed, and some cooling continues from 1500 to 0600 h,
with maximum difference of 2.3 ◦ C occurring at 1730 h (Fig. 10a). At
U160 , the limited difference between green and bare roofs through-
out the day reflects less effective cooling (Fig. 10b). The vertical
comparisons between U10 and U160 also indicate green-roof cooling
in nighttime on cloudy day (Fig. 10c and d).
More air temperature variations occur on rainy summer day and
they deviate notably from sunny and cloudy days. At U10 , green
roof is warmer in daytime but cooler in nighttime than bare roof
(Fig. 11a). The antecedent heat retained by green roof and the solar
radiation absorption in daytime, albeit subdued, provide sufficient
energy to continue warming the air in daytime. The green-roof
nocturnal cooling effect has not been suppressed by rainfall. At
U160 , green roof is consistently warmer than bare roof except for a
brief period before dawn, showing limited influence of nighttime
cooling (Fig. 11b). The bare roof continually warms the surrounding
81
air and contributes to a slightly warmer U10 compared with U160
throughout the day (Fig. 11c). The green roof shows little differ-
ence between U10 and U160 in daytime, but slightly cooler U10 in
nighttime, corroborating the nocturnal near-ground cooling effect
(Fig. 11d).
Daily heat flux pattern
On sunny summer day, the heat flux of the antecedent bare roof
increases from zero from around 0600 h to peak value of 84 W m−2
at 1130 h (Fig. 12a). It then drops to zero at 1800 h, and continues
to slide into negative values until 0600 h. The heat flux profile indi-
cates pronounced heat gain by the building in daytime and small
heat loss in nighttime.
In contrast, green roof shows consistent heat loss throughout
the day, being higher in daytime and approaching zero in night-
time (Fig. 12a). Even on hot summer days, the solar heat received
by green roof cannot penetrate into the building indoor space. Neg-
ative heat flux begins to intensify from 0600 h in the morning,
peaking at 50.5 W m−2 around 1200 h, which corresponds with the
maximum heat gain at bare roof. It then declines to near-zero at
around 2100 h, and remains nearly constant from 2100 to 0600 h.
Compared with daytime data, the nocturnal heat loss from green
roof is marginal, and it is even lower than bare roof. This is due to
the insulation effect of soil and vegetation which can retard heat
transmission both downwards and upwards. Thus the cooling effect
of green roof on hot sunny summer days occurs mainly in day-
time, and that it is propelled mainly by evapotranspiration cooling
fuelled by solar energy.
Heat flux on cloudy summer day follows a similar pattern as
the sunny situation, but it demonstrates markedly reduced magni-
tude and amplitude on both green and bare roofs (Fig. 12b). The
curves have been compressed in the y-axis orientation towards
the zero heat flux line. On bare roof, the maximum positive heat
82 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85

Fig. 10. Diurnal air temperature variations on typical cloudy summer days: (a) at U10 for before and after sessions; (b) at U160 for before and after sessions; (c) at U10 and
U160 for before session; and (d) at U10 and U160 for after session. The before session was recorded on August 19, 2008 and after session on August 25, 2009.

Fig. 11. Diurnal air temperature variations on typical rainy summer days: (a) at U10 for before and after sessions; (b) at U160 for before and after sessions; (c) at U10 and U160
for before session; and (d) at U10 and U160 for after session. The before session was recorded on August 08, 2008 and after session on August 13, 2009.
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85 83

Fig. 12. Diurnal heat flux profile under typical weather conditions for before and after sessions: (a) sunny summer day; (b) cloudy summer day; and (c) rainy summer day.

Table 8
Daily heat flux pattern, cooling load reduction and estimated electricity energy cost of the green-roof cooling effect.

Typical weather Roof type Heat flux Cooling load reduction Electricity cost savinga
condition

Heat gain Heat loss Unit area Whole roof b Whole Unit area Whole roofb Whole
(kJ m−2 day−1 ) (kJ m−2 day−1 ) (kWh m−2 day−1 ) (kWh day−1 ) summerc (HK$ m−2 day−1 ) (HK$ day−1 ) summerc
(kWh) (HK$)

Sunny summer Bare roof 2055.00 608.00

Green roof 0.00 1789.00 0.90 435.60 17,424.00 0.82 398.57 15,942.96
Cloudy summer Bare roof 899.00 565.00
Green roof 0.00 1714.00 0.57 275.88 11,035.20 0.52 252.43 10,097.21
Rainy summer Bare roof 0.00 1504.00
Green roof 0.00 1414.00 −0.03 −12.10 −484.00 −0.02 −11.07 −442.86
Total 699.38 27,975.20 639.93 25,597.31
a
The local electricity tariff is HK$0.915 per kWh in March 2011.
b
The green roof covers an area of 484 m2 .
c
The summer season in Hong Kong lasts about 120 days, and each typical weather condition is assigned 40 days.
The average CO2 emission from a coal-fire power plant is 966 g per kWh (Currie et al., 2002), hence the green roof could reduce 27.02 t of CO2 emission per summer based
on cooling load reduction alone.

flux of 46.7 W m−2 occurs at 1130 h, and the maximum nega-
tive heat flux of 22.6 W m−2 occurs at 2115 h. Heat gain starts
from 0700 to 1845 h, and heat loss from 1845 to 0700 h. The
green roof exhibits all-day heat loss, with more in daytime and
less in nighttime. Compared with sunny day, cloudy day experi-
ences more heat loss in nighttime especially from sunset to around
midnight.
On rainy summer day, the diurnal heat flux pattern departs
prominently from sunny and cloudy days (Fig. 12c). Both bare and
green roofs show consistent heat loss throughout the day and with
subdued fluctuations. The heat flux of bare and green roofs dif-
fers by small margins in the whole day. Even on bare roof, no heat
is gained by the building indoor space throughout the day. From
about 1100 to 2300 h, slightly more heat is lost from green roof
than bare roof. From about 2300 to 1100 h, a reversed pattern is
observed. This is largely attributed to the cooler wet concrete sur-
face during a continuous rainfall event. In contrast, heat loss from
green roof on rainy day is considerably less than sunny and cloudy
days due to suppression of evapotranspiration.
Daily heat gain and loss were calculated to compare cooling load
reduction under different weather conditions (Table 8). On sunny
day, heat gain of bare roof is 2055 kJ m−2 and heat loss 608 kJ m−2 ,
whereas green roof has no heat gain but accompanied by heat loss
of 1789 kJ m−2 . Overall, green roof can reduce the building’s cool-
ing load by 0.90 kWh m−2 . On cloudy day, heat gain of bare roof is
reduced to 899 kJ m−2 and heat loss is 565 kJ m−2 . The green roof
has no heat gain and heat loss of 1714 kJ m−2 , resulting in a daily
cooling load reduction of 0.57 kWh m−2 . The rainy day has no heat
gain on both bare and green roofs. There is a tiny cooling load
reduction of −0.03 kJ m−2 on green roof.
Assuming that the summer weather conditions last on average
four months (120 days) in Hong Kong, which is equivalent to the
days that normally require air-conditioning, and the partitioning of
sunny, rainy, and cloudy days is 1:1:1, the potential energy saving in
summer due to installation of the 484 m2 green roof is estimated to
be equivalent to 2.80 × 104 kWh (Table 8). Translated into monetary
units, with electricity tariff at the site of HK$0.915 per kWh unit in
2010, the total energy cost saving is HK$2.56 × 104 for each summer
season. This is equivalent to an avoidance of 27.02 t per summer
of carbon dioxide emission from a coal-fired power plant (Currie
et al., 2002). The research findings could inform the building service
engineering team to adjust the cooling system on hot summer days
to realize the environmental benefits.
Conclusion
Based on monitoring data of the railway station extensive green
roof, we investigated green-roof impacts on summer temperature
and heat flux in the humid-subtropical context. Compared with
previous field experiments which focused on a few sunny summer
days, we evaluated thermal performance in relation to key meteo-
rological parameters and three main weather scenarios. The main
findings are summarized as follows:
84 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85
(1) The green roof demonstrates great variations in diurnal ther-
mal performance. On average, the holistic green-roof effect could
reduce daily maximum temperature at the bottom tile sur-
face (Tta ) by 5.2 ◦ C, air temperature at 10 cm height (U10 ) by
around 0.7 ◦ C, but little effect at 160 cm (U160 ). The vegetation
surface temperature (Tv ) is even higher than bare roof tile sur-
face (Ttz ) by 3.4 ◦ C, imparting the vegetation canopy anomaly due
to limited solar protection by plants that have not achieved a
complete coverage.
(2) Solar radiation and relative humidity are the key meteoro-
logical factors of the green-roof thermal effect. High solar
radiation combined with low relative humidity could opti-
mize green-roof thermal performance. Soil moisture enhances
green-roof passive cooling by storing heat, dampening diurnal
temperature fluctuations, and sustaining evapotranspiration
and associated cooling. Rainfall can offset the relative cooling
contribution of green roof due to rapid cooling of bare roof.
Wind can reduce surface and air temperature of green roof by
enhancing evapotranspiration, but it is not correlated with tem-
perature reduction due to effective dissipation of the stored
heat at bare roof.
(3) The green-roof thermal effect displays considerable variations
under different weather scenarios. It attains an optimal state
on sunny summer day, declines on cloudy day, and becomes
negligible on rainy day. On sunny summer day, the maxi-
mum temperature reduction by green roof at different heights,
namely tile surface (Tta ), vegetation surface (Tv ), and air at 10 cm
(U10 ) and 160 cm (U160 ), reaches respectively 11 ◦ C, 3.2 ◦ C,
4.5 ◦ C and 3.4 ◦ C. On cloudy day, the values decline to 7.2 ◦ C
at Tta , 2.3 ◦ C at U10 , with little reduction at U160 and an increase
at Tv . On rainy day, green roof assumes the energy conservation
role with minor temperature reduction except slight nocturnal
cooling.
(4) The green roof reduces cooling load of the railway station build-
ing by 0.9 kWh m−2 on sunny summer day, 0.57 kWh m−2 on
cloudy day, but adds a slight cooling load on rainy day. Over-
all, the 484 m2 green roof offers a potential energy saving for
summer air-conditioning equivalent to 2.80 × 104 kWh of elec-
tricity or HK$2.56 × 104 of electricity tariff in each summer (the
officially pegged exchange rate is US$1.00 = HK7.80). This can be
translated into an upstream avoidance of 27.02 t of CO2 emis-
sion at the power plant.
(5) Especially on fine sunny days and to a lesser extent on cloudy
days, the solar energy heats up bare roof considerably to incur
heat flux into the building indoor space, demanding a con-
siderable amount of air-conditioning energy. The same solar
energy works in the opposite direction on green roof to propel
evapotranspiration and to cool the roof and ambient air mainly
during daytime. The green roof can transform the otherwise
consumptive solar energy impact into productive and beneficial
ecosystem services in a clean and free-of-charge mode.
(6) Shading and insulation by the physical materials of the green
roof system accounts for most of the cooling load reduction
in the building interior space below the roof slab, whereas
evapotranspiration from vegetation and soil with sufficient
water supply (rainfall supplemented by irrigation) contributes
significantly to cooing of the rooftop near-surface ambient
air.
(7) The holistic economic benefits (Banting et al., 2005; Carter
and Keeler, 2008) of ameliorating the urban heat island effect,
avoidance of CO2 emission, improvement in air quality, and
reduction in stormwater discharge and pollutants, in conjunc-
tion with the direct monetary savings in electricity tariff due
to reduction of cooling load, could justify the capital and recur-
rent costs of establishing green roofs in public buildings in the
humid-tropical context.
Compared with findings of similar studies in tropical Singapore
(Wong et al., 2007a), the reduction in maximum surface temper-
ature at Tta on sunny summer day in our study is lower by 9 ◦ C,
indicating a subdued thermal insulation and cooling effect. This
could be attributed to the relatively moderate bare-roof surface
temperature (a relatively cooler baseline) in humid-subtropical
Hong Kong in comparison with equatorial Singapore with stronger
insolation and a longer-hotter summer. The cooling of ambient air
above the extensive green roof in our study is notable, whereas
the Singapore study found higher air temperature above green roof
than bare roof. This could be associated with the relatively moist
soil in our experiment.
The regular daily irrigation could maintain average soil moisture
content at about 0.3 m3 /m3 to facilitate and sustain effective evap-
otranspiration cooling. This finding substantiates the crucial role of
soil moisture in the passive cooling of tropical green roof. However,
this study does not aim at establishing a quantitative relationship
between soil moisture and green-roof cooling effect due to lim-
ited variations in moisture level and unpredictable precipitation
input. Future studies could investigate temperature responses to
different rainfall (heavy, medium and light) and moisture regimes
(wet, moist and dry) to identify an optimal irrigation strategy under
different weather scenarios.
The sunny summer days selected in our study represent the fine
weather scenario in Hong Kong with high solar radiation and low
relative humidity. The summer weather tends to oscillate amongst
sunny, cloudy and rainy episodes. A study of green-roof thermal
effect should encompass the main weather scenarios. The findings
of this study on cloudy and rainy summer days with moderate air
temperature and less insolation, in conjunction with sunny days,
offer a comprehensive coverage of weather conditions to assess the
weather-dependent thermal performance of humid-subtropical
green roofs.
Due to practical and resource limitations, we were not able
to replicate the sensors. The sensors are placed at the centre of
both the experimental and control sites to minimize the edge and
shading effects, so as to obtain environmental data that are repre-
sentative of the green roof. The relatively large site with a square
shape helps to reduce the edge effect. The similar results obtained at
different times with comparable weather conditions could provide
a check on the validity of the sensor readings. The homogene-
ity of the groundcover vegetation in terms of species, vigour and
coverage throughout the green roof could eliminate internal spa-
tial variations in the measured parameters. Thus we believe that
the findings to a notable extent reflect the macroclimatic and soil
moisture conditions of an extensive green roof with humid-tropical
conditions.
Besides weather parameters, green-roof thermal effects are sen-
sitive to other factors, such as roof size, plant growth form and
biomass structure, soil composition and related thermal proper-
ties, building types and urban morphology in the environs. These
factors could be evaluated in other green roof studies to ascer-
tain their contributions to cooling performance. Our experimental
site represents a typical extensive green roof established on a low-
rise structure in a new town setting. The results could be applied
to other green roofs increasingly established on government and
school buildings which are mainly low-rise in Hong Kong and other
tropical cities. A deep understanding of green-roof thermal perfor-
mance in subtropical areas demands more empirical studies of a
wide variety of green roof systems.
Acknowledgements
We acknowledge with gratitude the research grant support
kindly provided by the railway company KCRC (recently merged
 C.Y. Jim, L.L.H. Peng / Urban Forestry & Urban Greening 11 (2012) 73–85
with MTRC). We are especially indebted to Mr. Yun Tai Li for
his bold vision and determination to install a green roof on
the Tai Po Railway Station and to fund our scientific study to
ascertain its environmental benefits. The kind advice and help
offered by Francis Goh, Brandon Chow, Grace Kwong, Jeannette
Liu, Raymond Wai Hung Law and Wing Yiu Wong are gratefully
appreciated.
References
Akbari, H., 2002. Shade trees reduce building energy use and CO2 emissions from
power plants. Environmental Pollution 116, S119–S126.
Arnold Jr., C.L., Gibbons, C.J., 1996. Impervious surface coverage: the emergence of a
key environmental indicator. Journal of the American Planning Association 62,
243–258.
Banting, D., Doshi, H., Li, J., Missios, P., 2005. Report on the Environmental Bene-
fits and Costs of Green Roof Technology for the City of Toronto. Department of
Architectural Science, Ryerson University, Toronto, Ontario.
Brenneisen, S., 2004. Green roofs: how nature returns to the city. Acta Horticulturae
643, 389–393.
Burgess, H., 2004. An Assessment of the Potential of Green Roofs for Bird Conserva-
tion in the UK. University of Sussex, UK.
Carter, T., Keeler, A., 2008. Life-cycle cost-benefit analysis of extensive vegetated
roof systems. Journal of Environmental Management 87, 350–363.
CTAHR, 2010. Cover Crops: Legume: Perennial Peanut. College of Tropical Agricul-
ture and Human Resources, University of Hawaii at Manoa, HI.
Currie, R., Elrick, B., Ioannidi, M., Nicolson, C., 2002. Renewables in Scot-
land. University of Strathclyde, Glasgow, UK, http://www.esru.strath.ac.uk/
EandE/Web sites/01-02/RE info/C02.htm (retrieved 01.01.2011).
DeNardo, J.C., Jarrett, A.R., Manbeck, H.B., Beattie, D.J., Berghage, R.D.,
2003. Stormwater detention and retention abilities of green roof
s. In: Proceedings World Water and Environmental Resources Congress,
June 23–26, 2003, Philadelphia, PA, pp. 1–7.
Emilsson, T., 2006. Extensive Vegetated Roofs in Sweden: Establishment, Develop-
ment and Environmental Quality. Department of Landscape Management and
Horticultural Technology, Faculty of Landscape Planning, Horticulture and Agri-
cultural Science, Swedish University of Agricultural Sciences, Uppsala, Sweden.
Eumorfopoulou, E.A., Kontoleon, K.J., 2009. Experimental approach to the contri-
bution of plant-covered walls to the thermal behaviour of building envelopes.
Building and Environment 44 (5), 1024–1038.
Getter, K.L., Rowe, D.B., 2006. The role of extensive green roofs in sustainable devel-
opment. Hortscience 41 (5), 1276–1285.
Harazono, Y., 1990/91. Effects of rooftop vegetation using artificial substrates on the
urban climate and the thermal load of buildings. Energy and Buildings 15/16,
435–442.
He, H.M., Jim, C.Y., 2010. Simulation of thermodynamic transmission in green roof
ecosystem. Ecological Modelling 221, 2949–2958.
Jim, C.Y., 1996. Edaphic properties and horticultural applications of some com-
mon growing media. Communications in Soil Science and Plant Analysis 27,
2049–2064.
Jim, C.Y., 2004. Green space preservation and allocation for sustainable greening in
compact cities. Cities 21, 311–320.
Jim, C.Y., 2008. Ecological design of sky woodland in compact urban Hong Kong. In:
Greening Rooftops for Sustainable Communities. Green Roofs for Healthy Cities,
Baltimore, pp. 1–15.
Jim, C.Y., He, H.M., 2010. Coupling heat flux dynamics with meteorological conditions
in the green roof ecosystem. Ecological Engineering 36, 1052–1063.
85
Köhler, M., 2004. Energetic effects of green roofs to the urban climate near to the
ground and to the building surfaces. In: International Green Roof Congress,
Nurtingen, Germany, 14–15 September, pp. 72–79.
Köhler, M., Schmidt, M., Grimme, F.W., Laar, M., de Assunção Paiva, V.L., Tavares,
S., 2002. Green roofs in temperate climates and in the hot-humid tropics: far
beyond the aesthetics. Environmental Management and Health 13, 382–391.
Mentens, J., Raes, D., Hermy, M., 2006. Green roofs as a tool for solving the rainwater
runoff problem in the urbanized 21st century? Landscape and Urban Planning
77, 217–226.
Niachou, A., Papakonstantinou, K., Santamouris, M., Tsangrassoulis, A., Mihalakakou,
G., 2001. Analysis of the green roof thermal properties and investigation of its
energy performance. Energy and Buildings 33, 719–729.
Oberndorfer, E., Lundholm, J., Bass, B., Coffman, R.R., Doshi, H., Dunnett, N., Gaffin,
S., Koehler, M., Liu, K.K.Y., Rowe, B., 2007. Green roofs as urban ecosystems:
ecological structures, functions, and services. Bioscience 57 (10), 823–833.
Papadakis, G., Tsamis, P., Kyritsis, S., 2001. An experimental investigation of the effect
of shading with plants for solar control of buildings. Energy and Buildings 33 (8),
831–836.
Santamouris, M., Pavlou, C., Doukas, P., Mihalakakou, G., Synnefa, A., Hatzibiros, A.,
Patargias, P., 2007. Investigating and analysing the energy and environmental
performance of an experimental green roof system installed in a nursery school
building in Athens, Greece. Energy 32, 1781–1788.
Soil Survey Staff, 1999. Soil taxonomy: A Basic System of Soil Classification for
Making and Interpreting Soil Surveys, 2nd ed. Department of Agriculture Soil
Conservation Service, Washington, DC.
Sonne, J., 2006. Evaluating green roof energy performance. ASHRAE Journal 48,
59–61.
Spala, A., Bagiorgas, H.S., Assimakopoulos, M.N., Kalavrouziotis, J., Matthopoulos,
D., Mihalakakou, G., 2008. On the green roof system. Selection, state of the art
and energy potential investigation of a system installed in an office building in
Athens, Greece. Renewable Energy: An International Journal 33, 173–177.
Takakura, T., Kitade, S., Goto, E., 2000. Cooling effect of greenery cover over a building.
Energy and Buildings 31, 1–6.
Teemusk, A., Mander, U., 2009. Greenroof potential to reduce temperature fluctua-
tions of a roof membrane: a case study from Estonia. Building and Environment
44, 643–650.
Tsang, S.W., Jim, C.Y., 2010. Game-theory approach for resident coalitions to allocate
green roof benefits. Environment and Planning A 43, 363–377.
VanWoert, N.D., Rowe, D.B., Andresen, J.A., Rugh, C.L., Fernandez, R.T., Xiao, L., 2005.
Green roof stormwater retention: effects of roof surface, slope and media depth.
Journal of Environmental Quality 34 (3), 1036–1044.
Villarreal, E.L., Bengtsson, L., 2004. Inner city stormwater control using a combina-
tion of best management practices. Ecological Engineering 22, 279–298.
Williams, N.S.G., Rayner, J.P., Raynor, K.J., 2010. Green roofs for a wide brown land:
opportunities and barriers for rooftop greening in Australia. Urban Forestry and
Urban Greening 9, 245–251.
Wong, N.H., Chen, Y., Ong, C.L., Sia, A., 2003a. Investigation of thermal benefits
of rooftop garden in the tropical environment. Building and Environment 38,
261–270.
Wong, N.H., Cheong, D.K.W., Yan, H., Soh, J., Ong, C.L., Sia, A., 2003b. The effects of
rooftop garden on energy consumption of a commercial building in Singapore.
Energy and Buildings 35, 353–364.
Wong, N.H., Tay, S.F., Wong, R., Ong, C.L., Sia, A., 2003c. Life cycle cost analysis of
rooftop gardens in Singapore. Building and Environment 38, 499–509.
Wong, N.H., Jusuf, S.K., Aung La Win, A., Kyaw Thu, H., Syatia Negara, T., Xuchao,
W., 2007a. Environmental study of the impact of greenery in an institutional
campus in the tropics. Building and Environment 42, 2949–2970.
Wong, N.H., Tan, P.Y., Chen, Y., 2007b. Study of thermal performance of extensive
rooftop greenery systems in the tropical climate. Building and Environment 42,
25–54.