Building and Environment 70 (2013) 266e276

Contents lists available at ScienceDirect

Building and Environment

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

Assessing practical measures to reduce urban heat: Green and cool

roofs
Andrew M. Coutts a, c, *, Edoardo Daly b, c, Jason Beringer a, c, Nigel J. Tapper a, c
a

School of Geography and Environmental Science, Monash University, Melbourne, VIC 3800, Australia
Department of Civil Engineering, Monash University, Melbourne, VIC 3800, Australia
c
CRC for Water Sensitive Cities, Australia
b

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 10 June 2013
Received in revised form
2 August 2013
Accepted 4 August 2013

As cities continue to grow and develop under climate change, identifying and assessing practical approaches to mitigate high urban temperatures is critical to help provide thermally comfortable, attractive
and sustainable urban environments. Green and cool roofs are commonly reported to provide urban heat
mitigation potential; however, their performance is highly dependent upon their design, particularly
green roofs that vary in substrate depth, vegetation species, and watering regime. This study compares
the insulating properties, the radiation budget and surface energy balance of four experimental rooftops,
including a green roof (extensive green roof planted with Sedum) and a cool roof (uninsulated rooftop
coated with white elastomeric paint), over the summer of 2011e12 in Melbourne, Australia. For the roof
treatments explored here, results suggest that cool roofs, combined with insulation, provide the greatest
overall benet in terms of urban heat mitigation and energy transfer into buildings. The high albedo of
the cool roof substantially reduced net radiation, leaving less energy available at the surface for sensible
heating during the day. Under warm and sunny conditions, when soil moisture was limited, evapotranspiration from the green roof was low, leading to high sensible heat uxes during the day. Irrigation
improved the performance of the green roof by increasing evapotranspiration. Daytime Bowen ratios
decreased from above four during dry conditions, to less than one after irrigation, yet sensible heat uxes
were still higher than for the cool roof. These results demonstrate that rooftops must be designed
accordingly to target specic performance objectives, such as heat mitigation.
2013 Elsevier Ltd. All rights reserved.

Keywords:
Green roofs
Cool roofs
Surface energy balance
Evapotranspiration
Irrigation

1. Introduction
Governments, city managers and urban residents are seeking
practical approaches to improve urban heat mitigation at least cost
as adaptation to global warming, extreme heat events and urban
heat effects. Green roofs are commonly purported as a key
approach for mitigating heat in urban areas [1e4] because of their
thermal benets, including the insulating effect of the soil substrate
and vegetation, the shading from the plant canopy and transpirational cooling [5]. In a review of green roof studies, Chen and Wong
[6] found that green roofs could greatly reduce rooftop surface
temperatures and create energy savings for buildings, while also
reducing ambient air temperatures. Likewise, cool roofs (white
and/or reective roofs) may also provide efcient mitigation of
* Corresponding author. School of Geography and Environmental Science, Monash University, Room 205, Building 28, Wellington Rd., Clayton, VIC 3800, Australia.
Tel.: 61 03 9905 8284.
E-mail address: Andrew.Coutts@monash.edu (A.M. Coutts).
0360-1323/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.buildenv.2013.08.021

atmospheric heating and building energy savings through an increase in surface albedo [7,8].
Several studies have been undertaken comparing the benets of
green and cool roofs in terms of building energy efciency and
rooftop microclimates. Takebayashi and Moriyama [4] used both
experimental and modelling approaches to compare green and
white roofs in Kobe, Japan. They found that, during the day, peak
sensible heat uxes (QH) were small for the white roof (153 W m
2)
due to the low net radiation (Q*) achieved by high solar reectance.
QH was also relatively small (361 W m
2) on the green roofs
because of the large evapotranspiration (QE), which peaked between 400 and 600 W m
2. However, despite the energy spent for
evapotranspiration, QH of the green roof was still twice as high as
the white roof. Scherba et al. [9] modelled the performance of green
and white roofs, nding that peak daytime QH was similar, but the
total daily QH was higher for green roofs because the thermal mass
of the green roof maintained positive QH at night. Scherba et al. [9]
acknowledged that only one green roof conguration was
modelled, and noted that factors such as irrigation specications

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

could impact green roof QH. In a review of green and cool roof
effectiveness, Santamouris [10] deemed that, when the albedo of
reective roofs is 0.7 or higher, cool roofs present a higher heat
island mitigation potential compared to green roofs. Santamouris
[10] also argued that green roofs could deliver similar cooling potential during peak temperature periods if QE exceeds 400 W m
2;
this would be possible for very well irrigated vegetated roofs (in
studies reviewed, solar radiation varied between 500 and
1000 W m
2).
Accordingly, of critical importance in green roof design is the
hydrological performance, which inuences the rooftop surface
energy balance and, hence, rooftop microclimate. For instance, the
choice of substrate type and depth affects QE, due to different water
retention [11], as well as the insulating effects of green roofs [12].
Vegetation type affects the rate of water loss from green roofs
through transpiration and the subsequent cooling effect; additionally, plants can have different shading effects on roof surface
cooling [13]. Many green roofs are constructed using a combination
of shallow soils and drought tolerant plant species that can survive
harsh rooftop environments and low water availability [14,15]. Such
environments may compromise plant transpiration due to stomatal
closure [16] and limit green roof cooling potential via QE. If the
specic goal of green roofs is for urban heat mitigation, then they
need to be designed carefully to maximize the benets, as green
roof performance varies widely [17].
A key function of green roofs is to capture and retain rainfall on
the roof and they are often implemented to help manage urban
stormwater. While a number of studies have documented a
reduction in stormwater runoff volumes from green roofs [18,19],
few have directly quantied rates of evapotranspiration [20],
although agreeing that green roofs mitigate high rooftop heating
partly through an increase in QE [21,22]. In greenhouse trials of
green roof systems planted with Sedum mexicanum and Disphyma
austral, Voyde et al. [20] observed that the rapid water loss via QE in
the days following watering gradually reduced as water supplies
became limited, until plants stopped transpiring to conserve water.
Similarly, using a weighing platform as part of an environmental
chamber laboratory setup, Tabares-Velasco and Srebric [23] found
that QE decreased as substrate water content decreased. Evapotranspiration occurred both during the day and night time periods,
and Tabares-Velasco and Srebric [23] determined that substrate

267

water content was the most important factor in determining QE.

These results highlight the risk that green roofs may not be able to
provide a strong benet to rooftop microclimates during extreme
heat events when it is most needed because of water shortage in
the substrate. It therefore becomes fundamental to dene the
desired performance characteristics of urban roofs and the appropriate design to meet those criteria. As such, if heat mitigation is a
primary goal, then irrigation may be necessary to support vegetation health and promote QE.
The design of rooftops has a strong inuence on the partitioning
of energy at the roof surface, and hence the adjacent rooftop
microclimate and internal building temperatures. To help identify
and compare the effectiveness of rooftop treatments to mitigate
urban heat, this study examines the thermal performance of
different roofs, focussing on green and cool roofs. The comparison is
based on a) energy storage and heat transfer through each rooftop,
and b) the surface energy balance of each rooftop. The emphasis
here was on retrotting existing roofs, so an extensive green roof
with shallow soil substrate was used in the experiments. Due to the
importance of design and water availability on the performance of
green roofs for cooling by evapotranspiration, we also assessed the
role of irrigation on the surface energy balance. Because green roofs
are considered a form of water sensitive urban design (WSUD) and
given our exploration of irrigation effects, our study also touches
upon issues of urban stormwater management. As such, the study
concludes by assessing the performance of different roofs within
the context of urban stormwater management, discussing strategies to best promote urban cooling and improved human thermal
comfort [24].
2. Methodology
2.1. Experimental setup
Four experimental roofs were compared: a conventional steel
sheet roof (STEEL), a steel sheet roof covered with white, high albedo paint (WHITE), a vegetated roof (VEG), and a roof with just the
soil layer (no vegetation) (SOIL). These experimental roofs were
2.4 m  2.4 m wooden platforms erected on stilts at a height of 1 m
(Fig. 1). They were in close proximity (<2 m) to each other and were
inclined at a slope of 15 . STEEL had a base of 20 mm plywood, with

Fig. 1. Construction of the vegetated roof (VEG) and experimental approach. VEG consisted of a plywood base, steel sheet, black poly membrane, plastic egg cup sheet, geotextile
layer and a scoria-soil mixture. The vegetation type was Sedum rubrotinctum. The measured daytime energy balance is represented schematically (left) along with the measured soil
and cavity variables (right).

268

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

a sheet of green steel overlaid; this rooftop acted as a control.

WHITE was constructed as STEEL, but it was painted with Thermoshield, a highly reective elastomeric coating containing hollow
ceramic particles. This coating is designed to reect radiation across
the entire shortwave radiation spectrum. VEG and SOIL were constructed as per typical extensive green roofs, one with vegetation
and the other with just the soil layer. Fig. 1 shows a schematic of
VEG. As with STEEL and WHITE, the 20 mm plywood was overlaid
with a sheet of steel. This was covered with a black poly membrane,
and then overlaid with a sheet of plastic egg-cups, which act as
small reservoirs. Over this was laid a geo textile layer and then a
coarse scoria-soil mixture that was 0.15 m deep (SOIL) with a
measured bulk density of 1.31 g.cm
3. VEG was planted with the
succulent vegetation type Sedum rubrotinctum (Sedum is commonly
used on extensive green roofs) which exhibits the crassulacean acid
metabolism (CAM) photosynthetic pathway during water stressed
conditions, making it drought tolerant. CAM plants have the ability
to take CO2 up at night for later use in photosynthesis in order to
limit water loss during the day, allowing survival in long drought
periods [13,25,26]. VEG and SOIL were not irrigated but were
exposed to ambient rainfall and weather conditions. Drainage holes
were located on the downward slope to allow runoff from the roofs
(Fig. 1).
Pyranometers (SP110, Apogee) at a height of 0.9 m above each
roof measured incoming shortwave radiation (KY); downward
facing pyranometers at a height of 0.3 m above each roof surface
measured reected shortwave radiation (K[). One pyrgeometer
(CGR3, Kipp and Zonen) measured incoming longwave radiation
(LY). Infrared temperature sensors (SI-121, Apogee), mounted at a
height of 0.9 m above each roof, provided outgoing longwave radiation (L[) and a derived surface temperature. This set of measurements was used to estimate the net radiation (Q*) at the surface
of each roof. VEG and SOIL were also instrumented with two soil
heat ux plates (HFP01, Hukseux) at a depth of 0.08 m, soil temperature (T107, Campbell Scientic) at three depths (0.04, 0.05 and
0.08 m), and one soil moisture probe (CS616, Campbell Scientic) at
a depth of 0.08 m. These additional measurements allowed the
calculation of ground heat uxes (QG) at the surface. For STEEL and
WHITE, thermocouples (Type T) were secured under the steel sheet
and a small amount of thermal paste was used to improve contact.
These data were used as boundary conditions in a heat transfer
equation to estimate QG. In the heat transfer equation, the thermal
conductivity of plywood was assumed to be 0.12 W m
1 K
1, the
specic heat capacity 1210 J kg
1 K
1, and the plywood density
540 kg m
3. The temperature was assumed to be initially linear
between the measured values at the two sides of the plywood. The
time step to solve the equation was 15 s and the depth of the
plywood was discretised using 7 intervals of 2.86 mm depth. Air
temperatures above the rooftops were not measured as there was
insufcient upwind fetch.
An enclosed air space was tted below each roof to replicate a
roof cavity air space. Thermocouples (Type T) were mounted in the
centre and the upper side of the roof cavities to observe the

insulating effects of the different roof treatments (Fig. 1). Shade

cloth was installed on the north and west sides of the experimental
roofs to minimise direct heating of the roof cavity spaces from the
sides. STEEL, VEG, and SOIL were instrumented from 1 October 2011
to 31 March 2012, while WHITE was instrumented from 1
December 2012. The four roofs experienced slightly different degrees of shading in the morning due to a building and trees to the
east of the experimental setup (mean daily KY in January 2012 was
893.7 MJ day
1 for STEEL, 917.7 MJ day
1 for WHITE,
1051.1 MJ day
1 for VEG, and 1003.3 MJ day
1 for SOIL).

2.2. Evapotranspiration observations

A portable closed-chamber was used to measure evapotranspiration rates from VEG and SOIL. The collection of QE provided for
the calculation of QH as a residual of the energy balance equation
(Q* QH QE QG). Evapotranspiration measurements were
conducted on four clear sunny days representing conditions when
the cooling effects of green roofs are most desirable (Table 1).
Measurements were taken hourly during these experimental periods. Chambers restrict the volume of air available for gas exchange between the surface and the atmosphere and the net
emission or uptake of gases can be measured as a concentration
change [27]. Three sample locations were established on VEG and
two locations on SOIL. Several weeks prior to the chamber experiment beginning, 300 mm diameter plastic collars were inserted
into the soil of the two roofs. A circular perspex chamber with a
diameter of 300 mm and a height of 415 mm was used (Fig. 2). The
chamber operated by drawing air from the top of the chamber,
along short (2 m) bevaline tubing to an LI-840 infrared gas analyser
(Li-Cor Biosciences), to measure the concentration of water vapour
(mmol/mol), before being recirculated back to the chamber in a
closed system. Two fans were built into the chamber to help
circulate the air. The chamber was placed on each measurement
point for a period of 120 s, and data were sampled every second and
the 5 s average logged on a CR800 control datalogger (Campbell
Scientic). Prior to each measurement, the chamber and tubes were
purged with air for 90 s. QE was determined from the change in the
mass concentration of water over time using a rectangular hyperbola as a saturation function [28] to determine the initial slope
(ux). This approach determined the slope at t 5 s using the rst
derivatives of the least square t. The combination of short tube
lengths and sampling times, taking the initial slope, and the manual
approach of placing and removing the chamber for each measurement, minimised inuences of adsorption of water molecules
on the internal wall of the sampling tubes [29] and the creation of
internal chamber microclimates [28,30]. Evapotranspiration from
STEEL and WHITE was assumed to be zero when chamber measurements were undertaken; this is deemed to be accurate for days
without rainfall or dewfall. As QH was calculated as a residual, any
errors in the radiation and soil observations, or latent heat uxes,
will accumulate in this term.

Table 1
Dates and times when chamber observations were conducted, along with maximum daily air temperature and volumetric water content (at 09:00). Times are Eastern Standard
Time. Plan vegetation coverage (lv) is for each plot on VEG where chamber measures were conducted.
Date

Time

Tmax ( C)

VEG (%WFP)

SOIL (%WFP)

lv1 (cm2)

lv2 (cm2)

lv3 (cm2)

19 Oct 2011
24 Nov 2011
22 Dec 2011
17 Jan 2012
1 Feb 2012
2 Feb 2012
3 Feb 2012

05:30e21:30
07:30e19:30
07:30e19:30
07:30e19:30
07:30e19:30
07:30e19:30
07:30e19:30

29.0
24.9
29.2
34.0
24.4
27.0
27.8

11.1
12.3
10.9
10.7
21.2
17.4
15.0

10.0
12.1
10.9
10.9
23.7
18.6
15.6

41.75
71.35
e
150.53
115.32

39.64
48.68
e
94.36
85.66

57.99
139.52
e
220.84
174.86

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

269

Fig. 2. Photograph of the experimental vegetated roof with the chamber system in December 2011; the vegetation coverage in each of the plots is from January 2012.

Soil moisture is also presented in Table 1 for each of the days of

evapotranspiration measurements. Data suggested that soil
moisture was higher for SOIL than VEG. However, we suggest that
these differences were probably largely instrumental in nature.
The CS616 volumetric water content (VWC) reectometer has an
accuracy of 2.5%, using the standard calibration provided by the
manufacturer, and probe-to-probe variability is 0.5% VWC in dry
soil and 1.5% VWC in typical saturated soil [31]. Inaccuracies can
also arise from improper installation. Hand-held soil moisture
measurements (CS620 Hydrosense, Campbell Scientic) indicated
that soil moisture was higher on the downward slope of the
rooftops; therefore, differences between VEG and SOIL could also
derive from slightly different positioning of the sensors. Observations of VWC during wet and dry conditions from the hand-held
soil moisture sensor also showed that soil moisture was similar
between VEG and SOIL, and these values were more in line with
the CS616 values observed on VEG. Soil moisture from VEG was
thus considered most representative. Accordingly, VWC data for
SOIL was rescaled to the VWC observed on VEG by determining a
linear relationship between the maximum and minimum VWC on
SOIL and the maximum and minimum VWC observed on VEG.
VWC was then converted to per cent water lled pore space (%
WFP).
The vegetation on VEG was rst established in July 2008 and
over time the vegetation coverage diminished as the experimental roof was un-irrigated and un-fertilized. Before the study
was undertaken, VEG was fertilized and the coverage and health
of the vegetation improved before the start of the experiment
and as the experiment progressed. The plan area vegetation

cover (lv) for each plot used to measure evapotranspiration is

given in Table 1; lv was determined from digital photographs
taken on days of chamber measurements. Each image was loaded
into the leaf area analysis program WinFOLIA [32]. Cover
increased from October through to January before declining
slightly in February as a result of summertime stress; this
reduction could also result from shading in the digital image. So,
contrary to many previous studies, this experiment investigates
what was effectively a long established un-irrigated green roof,
rather than a newly established roof, and the overall coverage of
S. rubrotinctum on VEG was only around 50% (Fig. 2). While VEG
was a long-established roof treatment, WHITE was constructed
specically for this project, so the white paint was new, but over
time, cool roofs can become weathered and lose some reective
capacity [10].

2.3. Irrigation experiment

The four individual days of chamber observations (Table 1) were
representative summer time periods when the soils of VEG and
SOIL were relatively dry. To investigate the role of irrigation in
enhancing QE and to observe patterns in QE as soil moisture
decreased, an irrigation experiment was conducted over a threeday period from 1 to 3 Feb 2012. At 17:00 on 31 Jan 2012, VEG
and SOIL were heavily irrigated until free drainage was observed.
On the following days, observations were conducted hourly from
08:00 to 20:00. The irrigation event increased the %WFP from 11.1%
to 29.9% for VEG and from 11.9% to 30.1% for SOIL.

270

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

3. Results and discussions

3.1. Radiation budget and energy transfer
The mean diurnal incoming and outgoing short and longwave
radiations for each experimental roof during January 2012 are
presented in Fig. 3. Incoming solar radiation for each roof varied
markedly until 11:00 as a result of the different degrees of shade in
the morning. The most striking difference in the radiation budget
between each roof was in K[ that was related to the albedo of the
roofs. The albedo of WHITE was very high, measuring 0.71 at solar
noon (approx. 13:30 EST) during January 2012. In comparison, the
albedo was low on SOIL (0.10) due to the dark colour of the scoriasoil mix, while the presence of the lighter coloured vegetation
increased the albedo slightly on VEG (0.15). The albedo of STEEL
was 0.21, resulting in a value of K[ higher than both VEG and SOIL.
With regard to outgoing longwave radiation (Fig. 3), WHITE
showed the lowest levels, largely due to the high albedo of the
surface, which, reecting radiation across the shortwave spectrum,
did not heat as intensely as the other roofs. Despite the higher albedo of STEEL relative to VEG and SOIL, L[ was higher as the STEEL
surface heated more easily than VEG and SOIL. The thin layer of
steel and underlying plywood had a low thermal mass and the steel
had a high thermal conductivity, meaning that less energy was
needed to heat the materials; consequently, the roof surface heated
intensely. The inuence of the vegetation layer can be also seen to
reduce L[, where the vegetation itself and shading of the soil surface serves to reduce surface temperatures and hence emitted
longwave radiation. During the night, L[ of STEEL and WHITE were
similar and lower than VEG and SOIL.
The overall effect of each experimental roof type on Q* is presented in Fig. 4 for January 2012. As the effect of the vegetation
served to both increase K[, but decrease L[ by similar amounts,
this meant that Q* was similar between VEG and SOIL. Blanusa et al.
[16] showed high Q* over bare soil treatments (665.1 W m
2)
compared to Sedum mix treatments (552.7 W m
2), but they had a
higher vegetation coverage. The higher albedo of STEEL, and to a
much greater extent WHITE, resulted in reduced Q* at the surface.
Mean net radiation at midday for WHITE was around 78% lower
than VEG. Fig. 4 also presents energy transferred into the roofs at
the roof surface (QG) for January 2012. The highest peak of QG was

calculated for STEEL as energy was conducted through the steel and
plywood easily due to their high thermal conductivity. In contrast,
the high albedo of WHITE meant that there was little net radiative
energy available at the surface for conduction through the roof,
though the bulk of available energy was partitioned into QG. The
presence of vegetation on VEG resulted in energy absorption and
shading of the soil surface, resulting in reduced QG in comparison
with SOIL. While the differential shading of the experimental roofs
earlier in the day makes comparisons less distinct, the vegetated
roof clearly reduced heat transfer compared to the non-insulated
STEEL control; WHITE reduced heat transfer through the reduction in available energy at the surface, rather than any changes to
conductive properties.
Previous studies stated that green roofs reduce heat ow
through buildings, particularly through the insulating effects of the
soil layer [33]. Fig. 5 presents the mean diurnal air temperatures
observed in the cavity airspace (Tcav) of each experimental roof for
January 2012. The insulating effects of VEG and SOIL were observed
with a large reduction in Tcav. VEG appeared to slightly reduce heat
transfer through the roof compared to SOIL, but the soil layer was
providing the bulk of the insulating benet. VEG and SOIL not only
generated lower Tcav, but they also caused a lag between the peak of
Tcav and the peak of atmospheric temperature. Results also showed
reduced substrate temperatures for VEG compared to SOIL (not
shown), as also reported by Blanusa et al. [16]; this was due to a
higher albedo and shading of the surface from the vegetation. A
denser plant coverage or alternative species may have improved
the insulating capacity of VEG. For WHITE and STEEL, because there
was no form of insulation below the roof, a large proportion of the
energy in QG at the surface was transmitted into the cavity airspace
below. Energy is conducted into the rooftop materials raising the
temperature of the plywood (Tbase) and then the roof cavity
airspace is heated via convective and radiative heating from the
underside of the rooftop surface, raising Tcav. This highlights the
importance of roof insulation in building design. The lower Q* of
WHITE meant that less energy was available for transfer through
the roof; Tcav in WHITE was lower than STEEL during the day, while
during the night they were comparable and lower than VEG and
SOIL (Fig. 5). While VEG produced the coolest roof cavity, Tcav,
during the day, it remained warmer at night, offsetting some of the
daytime benets in terms of human thermal comfort and building

Fig. 3. Ensemble diurnal average shortwave (left panel) and longwave (right panel) radiation for January 2012. Bars denote standard error.

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

271

Fig. 4. Ensemble diurnal average net radiation (left panel) and ground heat ux (right panel) of each experimental roof for January 2012. Bars denote standard error.

energy use for air conditioning. The thermal performance of a green

roof will undoubtedly depend on what type of roof it is being
compared to. If STEEL and WHITE had been insulated (e.g. ceiling
batts), the heat transfer to the cavity airspace would have been
reduced, possibly leading to temperatures similar to those
measured for VEG and SOIL. This was also discussed by Jaffal et al.
[34], who noted that green roofs only show strong insulating effects
for uninsulated or moderately insulated buildings.
3.2. Surface energy balance and atmospheric heating
Modifying the roof surface through implementation of green or
white roofs alters surfaceeatmosphere interactions and the
resulting roof microclimate. The aim is to reduce urban air temperatures by limiting available energy partitioning into sensible
heat. Fig. 6 presents the surface energy balance for STEEL and
WHITE for January 2012. These two roofs received approximately
the same incoming solar radiation despite the differential shading

Fig. 5. Ensemble diurnal average air temperature (Ta), and cavity air temperatures
(Tcav) observed in the roof space of each experimental roof for January 2012. Bars
denote standard error.

in the morning (Fig. 3). As highlighted above, the high albedo of

WHITE dramatically reduced Q*, and, as a result, there was only a
small amount of energy available at the surface to support sensible
heat uxes. In contrast, QH for STEEL was high, particularly in the
late afternoon and evening. Slightly positive QH was also observed
in the evening for WHITE. During the evening, energy stored in the
roofs during the day is lost from the surface and supports a positive
sensible heat ux. These energetics are a key reason for the
development of the urban heat island (UHI) at the city scale; urban
areas remain warm into the evening and night as stored energy is
lost, supporting atmospheric warming, while rural areas cool
rapidly. In this study, QE was assumed to be zero for STEEL and
WHITE; but, it is recognized that QE can be the dominant ux at
times following rain or in the early morning when dewfall is
present.
For VEG and SOIL, chamber measurements were conducted to
quantify QE during the day. Chosen monitoring days (Table 1) were
warm and sunny and mostly towards the end of dry spells (Fig. 7).
Previous research on these experimental rooftops found that over
26 July 2008 to 3 May 2009 runoff from VEG was 65% less than that
observed for STEEL [35]. Soil moisture on VEG and SOIL on these
monitoring days was low (Fig. 7), suggesting that during summer,
when green roofs are required to promote cooling through
enhanced QE, there may in fact be little soil moisture available to
support evapotranspiration. Therefore, active management
through irrigation strategies may be required to make green roofs
effective. While the green roof promotes inltration, soil moisture
also declines rapidly to allow for water capture during the next rain
event [36]. Despite our green roof design containing the plastic
egg-cup sheet for storing water below the soil, during January and
February, soil moisture was low for extended periods of time
(Fig. 7). Fig. 7 presents QE for each individual plot on VEG and SOIL
for each of the monitoring days. Generally, QE was low across both
VEG and SOIL. Plot 3 on VEG showed higher QE due to its location on
the downward edge of the sloping roof, so the upward edge of the
roof dried out quicker than the downward edge. Plot 3 also had the
highest vegetation coverage (Table 1). On 24 November 2011, soil
moisture was slightly higher; consequently, QE was greater across
all the plots and particularly SOIL 2 (Fig. 7), which was also on the
more downward portion of the roof. The higher QE on SOIL 2 could
also suggest that the wet soil freely evaporates, while for VEG the
lower surface temperatures and water uptake by vegetation limits
QE along with constrained transpiration by the Sedum. This will be
discussed further in section 3.3. Overall, QE was low during dry
periods, and was variable across VEG.

272

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

Fig. 6. Mean surface energy balance for January 2012 for STEEL (left panel) and WHITE (right panel). QE was assumed to be zero on these dry impervious surfaces. Bars denote
standard error.

The surface energy balance for VEG and SOIL over the four days
of un-irrigated chamber observations (Fig. 8) shows that generally
only a small portion of Q* was partitioned into QE (mean mid-day
[10:30e16:30] evaporative fraction [QE/Q*] of 0.15 for VEG and
0.13 for SOIL). As such, under these conditions, the green roof
provided little rooftop cooling through evapotranspiration, except
for 24 November 2012 when soil moisture was slightly higher. In
comparison with STEEL, the VEG and SOIL roofs actually showed a

slightly higher peak daytime QH. During the day, QG for VEG and
SOIL was reduced compared to STEEL due to lower thermal conductivity and albedo, meaning more energy was available at the
surface for partitioning into sensible and latent heat. Since the
contribution of QE under these conditions was small (Fig. 8), QH was
consequently slightly higher. In the evening and night, the positive
QH was not present on VEG and SOIL due to reduced daytime heat
storage in the roofs. Further, once Q* became negative on VEG and

Fig. 7. Half hourly soil moisture for VEG and precipitation (upper panel) over the study period (1 Oct 2011e31 Mar 2012). Circles represent days when chamber observations were
conducted. The rectangle represents days when the irrigation experiment was conducted. Chamber observations of evapotranspiration on VEG and SOIL experimental roofs (lower
panels). Observations were conducted on warm, sunny days.

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

273

Fig. 8. Ensemble average diurnal surface energy balance over the four days for which un-irrigated chamber observations were conducted for VEG and SOIL and the corresponding
period for STEEL. The WHITE experimental roof was not included as observations began in December 2011. Bars denote standard error and dotted error whiskers denote mean,
maximum and minimum.

SOIL, energy loss through QG appeared to support positive QE into

the evening (and possibly throughout the night) rather than supporting positive QH into the evening as occurred with STEEL.
Evapotranspiration from a green roof will vary with plant species,
vegetation coverage, and irrigation regimes. These results show
that this long-established green roof, where vegetation coverage
was not 100%, succulent vegetation was present and irrigation was
absent, provides only a marginal benet for rooftop microclimates
through evapotranspiration, and is concentrated towards the early
evening. Getter et al. [5] suggest that increasing substrate depth or
irrigating green roofs would allow the use of plants with greater
biomass and leaf area, leading to higher QE. Our ndings also suggest that there was little difference between QE for VEG and SOIL.
The evapotranspiration from VEG may be mostly soil evaporation,
with little transpiration from the Sedum rubrotincum, which could
be utilising CAM under these dry soil moisture conditions.
Accordingly, the support of vegetation health and evapotranspiration from VEG and SOIL could be done using a strategic watering
regime. This could be achieved by sourcing t-for-purpose water
for irrigation of green roofs from sustainable sources, such as
stormwater, as for example promoted by WSUD principles.
3.3. Irrigation experiment
Given the low QE observed for VEG under dry conditions, an
irrigation experiment was conducted to examine changes in QE
from higher soil moisture levels under clear, sunny conditions.
Fig. 9 presents the surface energy balance for the three day period
(1e3 Feb) following the irrigation event. After irrigation, there was
a substantial increase in QE for both VEG and SOIL, before evapotranspiration began to decline over progressive days. Maximum
rates of QE increased on VEG (SOIL) from a mean of around
100 W m
2 (90 W m
2) observed for the individual days (Fig. 9), to
a mean of around 260 W m
2 (480 W m
2) on VEG (SOIL) the day
after irrigation. As seen prior to the irrigation event, there was a
large amount of variability in QE on VEG and SOIL due to the slope of
the roofs (see maximum and minimum range bars for QE in Fig. 9).
The day after irrigation, the entire roof area showed high QE, but in
days following, the upper portion of the sloped roof dried out
quicker. Evapotranspiration rarely reached the 400 W m
2 suggested by Santamouris [10] to make green roofs comparable to cool

roofs (with albedos >0.7) in terms of UHI mitigation potential

during the peak period, except for the day after irrigation where
peak QE for plot 2 on SOIL reached as high as 600 W m
2. However,
in contrast with the results of Scherba et al. [9] VEG did eliminate
the late afternoon and evening sensible heat uxes seen on STEEL
(Fig. 10) by promoting evapotranspiration into the evening instead.
Generally, QE was higher for the SOIL plots after irrigation than
the VEG plots, similar to the second day of chamber observations
(24 November 2011). Moisture freely evaporated from the soil
surface, while water uptake and resistance to water loss by the
succulent vegetation limited QE. Post-irrigation, the mean daytime
evaporative fraction increased from 0.13 for observation days prior
to irrigation, to 0.41 for VEG immediately after irrigation, and 0.13
to 0.51 for SOIL. By the third day following irrigation, the mean
evaporative fraction had reduced to 0.26 for VEG and 0.38 for SOIL.
Results suggest that having soil alone may deliver a greater effect
on QE, as the resistance to water loss from vegetation is not present,
nor is the surface cooling effect of vegetation. However, by
removing the vegetation, the roof would lose much of its amenity
and biodiversity values. In trials of 34 different green roof treatments in Halifax, Canada, MacIvor et al. [37] found that treatments
with little or no cover (controls) were amongst the best performers
in terms of water loss (indicative of evapotranspiration). In
contrast, in green house trials in New Zealand, Voyde et al. [20]
found that planted treatments of S. mexicanum and Disphyma
australe showed higher QE than bare substrate when plants were
not water stressed. Shading of the substrate surface, a higher albedo
and reduced substrate temperatures on VEG may also have
contributed to reduced QE, as suggested by Lundholm et al. [36].
VanWoert et al. [26] also suggested that shading from vegetation
cover lowered substrate moisture evaporation, leading to higher
substrate soil moisture levels in vegetated treatments compared to
substrate only treatments. In this study, substrate temperatures
were lower for VEG (mean January substrate temperatures at 8 cm
at 17:00 were 31.2  C and 35.0  C for VEG and SOIL respectively) and
may have contributed to lower QE.
In our study, while QE was higher on SOIL than VEG post irrigation, when the soil was dry, QE was lower on SOIL (Fig. 8), and in
combination with the low albedo resulted in high QH. This is well
presented in the mid-day (10:30e16:30) Bowen ratios (b) for VEG
and SOIL (Fig. 10). Prior to irrigation, b for both VEG and SOIL were

274

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

Fig. 9. Surface energy balance of each experimental roof over 1e3 Feb 2012 following irrigation. Whiskers denote maximum and minimum observed.

high (b > 4), representing conditions similar to semi-arid deserts,

and b for VEG was lower than SOIL. After irrigation, SOIL displayed
lower b values (b < 1 for both rooftops on Day 1). This highlighted
the regulating effect of vegetation on water loss and hence surface
energy partitioning. While peak QE post irrigation was not as high
on VEG as it was on SOIL, water was likely retained on VEG in the
substrate and vegetation over a longer period. Still, QE from VEG
when irrigated was >200 W m
2, suggesting that despite the
presence of drought tolerant S. rubrotinctum, irrigation did support
increased evapotranspiration when water was available. Increasing
water availability on green roofs could support a wider range of
species and Blanusa et al. [16] suggested that additional expenses
for supplementary irrigation could allow species selection that
provide greater eco-system service potential. Yet, any irrigation
regime would need to be carefully designed to ensure sufcient
water capture during precipitation events is not compromised.
Finally, in order to compare the different rooftop treatments,
particularly VEG and WHITE, Fig. 10 presents the mean afternoon
(12:30-19:30) sensible heat uxes for the days following irrigation
(at times when chamber observations were made and there were
no shading effects). Clearly, the white elastomeric paint delivered
the greatest reduction in QH during the day. Of the small amount of
energy available at the surface because of the high albedo, most of
this was conducted into the uninsulated rooftop and demonstrates

that reducing Q* is an effective mechanism for limiting daytime QH.

Prior to the irrigation event (on 17 Jan 2012), QH on VEG and SOIL
was higher than for the control. The higher albedo of STEEL and
high heat transfer into the roof meant less energy was available for
QH. Again, it should be noted here that the performance of a green
roof will depend heavily on what conventional rooftop it is being
compared to [38]. After irrigation, QH from VEG and SOIL was
reduced to below that observed on STEEL, primarily through the
increase in evapotranspiration; QH was still at least double that
observed for WHITE. As seen by Takebayashi and Moriyama [4], the
soil dried out over successive days, QH increased for VEG and SOIL
but still remained lower than for STEEL. A longer period of study
would help determine how long these irrigation effects on QH last.
4. Conclusions
While experimental green and cool roof studies have been
conducted previously [17,39], this study directly compares the
diurnal radiation budget, the diurnal surface energy balance, and
insulating effects of green roofs and cool roofs simultaneously, in a
bid to assess the performance of practical measures to reduce the
contribution to urban heat from building rooftops. In this context,
roofs should be designed to target specic performance objectives
[17] that include heat reduction. This can be achieved in the urban

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

275

Fig. 10. Mid-day Bowen ratio (b) (10:30e16:30) for VEG and SOIL before and for the three days after irrigation (left panel) and afternoon/evening sensible heat uxes (QH) for the
period when chamber observations were taken (12:30e19:30 e excludes morning period when shading occurs) for each rooftop before (17 Jan 2012) and after irrigation (1e3 Feb
2012) (right panel).

landscape through enhanced evapotranspiration using irrigation

from t-for-purpose water, as suggested by WSUD practices. A
unique attribute of this study was the collection of diurnal observations of evapotranspiration from green roofs in the eld, that
helps address the lack of experimental data [11,20]. Furthermore,
little green roof research has been conducted under Australian
conditions [40].
With the roof treatments explored here, results suggest that the
high albedo rooftop, if combined with insulation, would probably
provide the greatest overall benet in terms of urban heat mitigation and heat transfer into buildings. Our results suggest that
irrigation is necessary for green roofs to provide a substantial
daytime microclimate benet, because, under dry conditions
(when intervention is most needed to alleviate heat stress), QE was
low and QH was high. Green roofs could provide as much benet as
cool roofs if they are regularly irrigated and planted with a dense
mix of actively transpiring vegetation. In our case, Sedum (a dryland
species) provided no signicant benet over a soil substrate roof
alone. This is due to the resistance to QE provided by the matt of
Sedum vegetation that does not transpire actively during the daytime. Although these species can provide benets, such as shallow
rooting depth and drought tolerance, and sustain stormwater
management objectives they are not suited to achieving a cooling
performance objective. Therefore, the selected species should
maximize transpirational losses, thereby enhancing evaporative
cooling effectiveness. Such species are likely to require more substantial water supply and management. Another alteration that
could improve the performance of the green roof is the inclusion of
a thin cover of white gravel or stones rather than the dark scoria
mix used here to increase the albedo.
An alternative framework to achieve cooling objectives may be
to harvest and store stormwater captured from cool roofs, which
here showed the larger rooftop microclimate benets; stormwater
could then be stored and successively used for irrigation at ground
level to support evaporative cooling and healthy vegetation (e.g.
street trees). This is likely to provide a greater benet to outdoor air
temperatures and human thermal comfort as its effects will be felt
where pedestrians reside, i.e. within the urban canopy layer, rather
than at the rooftop. Irrigation of harvested stormwater can also be
controlled, allowing for increased water use on warm, dry days
when evaporative cooling effects are most needed. Further research
and modelling is needed at the micro-scale that examines these

different urban congurations to quantify the best approaches to

mitigate urban heat and improve human thermal comfort.
Acknowledgements
Funding for this research was provided through the Cities as
Water Supply Catchments research program, the Cooperative
Research Centre (CRC) for Water Sensitive Cities, and the Engineering Small Grants Scheme within the Faculty of Engineering at
Monash University. Funding for construction of the experimental
roofs was provided by Melbourne Water. Thanks to Thermoshield
for providing the cool roof technology. Monash University provides
research into the CRC for Water Sensitive Cities through the Monash Water for Liveability Centre. Thanks to Ian McHugh for
providing the chamber system. Thanks to Richard Harris, Kerry
Nice, Ashley Broadbent, Jasmine Thom, Mahima Kalla and Rachel
Meinig for assisting with data collection and preparation.
References
[1] Ayata T, Tabares-Velasco PC, Srebric J. An investigation of sensible heat uxes
at a green roof in a laboratory setup. Building and Environment 2011;46:
1851e61.
[2] Niachou A, Papakonstantinou K, Santamouris M, Tsangrassoulis A,
Mihalakakou G. Analysis of the green roof thermal properties and investigation of its energy performance. Energy and Buildings 2001;33:719e29.
[3] Rosenzweig C, Solecki WD, Cox J, Hodges S, Parshall L, Lynn B, et al. Mitigating
New York citys heat Island: integrating stakeholder perspectives and scientic evaluation. Bulletin of the American Meteorological Society 2009;90:
1297e312.
[4] Takebayashi H, Moriyama M. Surface heat budget on green roof and high
reection roof for mitigation of urban heat island. Building and Environment
2007;42:2971e9.
[5] Getter KL, Rowe DB, Andresen JA, Wichman IS. Seasonal heat ux properties of
an extensive green roof in a Midwestern U.S. climate. Energy and Buildings
2011;43:3548e57.
[6] Chen Y, Wong NH. Thermal impact of strategic landscaping in cities: a review.
Advances in Building Energy Research 2009;3:237e60.
[7] Rosenfeld AH, Akbari H, Romm JJ, Pomerantz M. Cool communities: strategies
for heat island mitigation and smog reduction. Energy and Buildings 1998;28:
51e62.
[8] Synnefa A, Dandou A, Santamouris M, Tombrou M, Soulakellis N. On the use of
cool materials as a heat island mitigation strategy. Journal of Applied Meteorology and Climatology 2008;47:2846e56.
[9] Scherba A, Sailor DJ, Rosenstiel TN, Wamser CC. Modeling impacts of roof
reectivity, integrated photovoltaic panels and green roof systems on sensible
heat ux into the urban environment. Building and Environment 2011;46:
2542e51.

276

A.M. Coutts et al. / Building and Environment 70 (2013) 266e276

[10] Santamouris M. Cooling the cities e a review of reective and green roof
mitigation technologies to ght heat island and improve comfort in urban
environments. Solar Energy 2012 in press.
[11] Metselaar K. Water retention and evapotranspiration of green roofs and
possible natural vegetation types. Resources, Conservation and Recycling
2012;64:49e55.
[12] Castleton HF, Stovin V, Beck SBM, Davison JB. Green roofs; building energy
savings and the potential for retrot. Energy and Buildings 2010;42:1582e91.
[13] Wolf D, Lundholm JT. Water uptake in green roof microcosms: effects of plant
species and water availability. Ecological Engineering 2008;33:179e86.
[14] Nagase A, Dunnett N. Drought tolerance in different vegetation types for
extensive green roofs: effects of watering and diversity. Landscape and Urban
Planning 2010;97:318e27.
[15] Thuring CE, Berghage RD, Beattie DJ. Green roof plant responses to different
substrate types and depths under various drought conditions. HortTechnology
2010;20:395e401.
[16] Blanusa T, Vaz Monteiro MM, Fantozzi F, Vysini E, Li Y, Cameron RWF. Alternatives to Sedum on green roofs: can broad leaf perennial plants offer
better cooling service? Building and Environment 2012.
[17] Simmons M, Gardiner B, Windhager S, Tinsley J. Green roofs are not created
equal: the hydrologic and thermal performance of six different extensive
green roofs and reective and non-reective roofs in a sub-tropical climate.
Urban Ecosystems 2008;11:339e48.
[18] Czemiel Berndtsson J. Green roof performance towards management of
runoff water quantity and quality: a review. Ecological Engineering
2010;36:351e60.
[19] Mentens J, Raes D, Hermy M. Green roofs as a tool for solving the rainwater
runoff problem in the urbanized 21st century? Landscape and Urban Planning
2006;77:217e26.
[20] Voyde E, Fassman E, Simcock R, Wells J. Quantifying evapotranspiration rates
for New Zealand green roofs. Journal of Hydrologic Engineering 2010;15:
395e403.
[21] Del Barrio EP. Analysis of the green roofs cooling potential in buildings. Energy
and Buildings 1998;27:179e93.
[22] Theodosiou TG. Summer period analysis of the performance of a planted roof
as a passive cooling technique. Energy and Buildings 2003;35:909e17.
[23] Tabares-Velasco PC, Srebric J. Experimental quantication of heat and mass
transfer process through vegetated roof samples in a new laboratory setup.
International Journal of Heat and Mass Transfer 2011;54:5149e62.
[24] Coutts AM, Tapper NJ, Beringer J, Loughnan M, Demuzere M. Watering our
cities: the capacity for water sensitive urban design to support urban cooling
and improve human thermal comfort in the Australian context. Progress in
Physical Geography 2012;37:2e28.

[25] Terri JA, Turner M, Gurevitch J. The response of leaf water otential and crassulacean acid metabolism to prolonged drought in Sedum rubrotinctum. Plant
Physiology 1986;81:678e80.
[26] Vanwoert ND, Rowe DB, Andresen JA, Rugh CL, Xiao L. Watering regime and
green roof substrate design affect Sedum plant growth. HortScience 2005;40:
659e64.
[27] Livingston GP, Hutchinson GL. Enclosure-based measurements of trace gas
exchange: application and sources of error. In: Matson PA, Harriss RC, editors.
Biogenic trace gases: measuring emissions from soil and water. Great Britain:
Blackwell Science, Cambridge University Press; 1995.
[28] Langensiepen M, Kupisch M, Van Wijk MT, Ewert F. Analyzing transient closed
chamber effects on canopy gas exchange for optimizing ux calculation
timing. Agricultural and Forest Meteorology 2012;164:61e70.
[29] Ibrom A, Dellwik E, Flyvbjerg H, Jensen NO, Pilegaard K. Strong low-pass
ltering effects on water vapour ux measurements with closed-path eddy
correlation systems. Agricultural and Forest Meteorology 2007;147:140e56.
[30] Steduto P, etinkk , Albrizio R, Kanber R. Automated closed-system canopy-chamber for continuous eld-crop monitoring of CO2 and H2O uxes.
Agricultural and Forest Meteorology 2002;111:171e86.
[31] Campbell Scientic. CS616 and CS625 water content Reectometers: Instruction manual; 2012 [Logan, Utah].
[32] Regent Instruments. WinFOLIA for leaf analysis manual; 2009 [Canada].
[33] Liu KKY, Minor J. Performance evaluation of an extensive green roofIn
Greening rooftops for sustainable communities; 2005 [Washington DC].
[34] Jaffal I, Ouldboukhitine S, Belarbi R. A comprehensive study of the impact of
green roofs on building energy performance. Renewable Energy 2012;43:
157e64.
[35] Mitchell VG. Assessing the stormwater benets of thin vegetated roofs:
summary of stage 1 research ndings. Monash University; 2009.
[36] Lundholm J, MacIvor JS, MacDougall Z, Ranalli M. Plant species and functional
group combinations affect green roof ecosystem functions. PLoS ONE 2010;5:
e9677.
[37] MacIvor JS, Ranalli MA, Lundholm JT. Performance of dryland and wetland
plant species on extensive green roofs. Annals of Botany 2011;107:671e9.
[38] Hamdi R, Schayes G. Sensitivity study of the urban heat island intensity to
urban characteristics. International Journal of Climatology 2008;28:973e82.
[39] Wong NH, Jusuf SK, Win AAL, Thu HK, Negara TS, Xuchao W. Environmental
study of the impact of greenery in an institutional campus in the tropics.
Building and Environment 2007;42:2949e70.
[40] Williams NSG, Rayner JP, Raynor KJ. Green roofs for a wide brown land: opportunities and barriers for rooftop greening in Australia. Urban Forestry &
Urban Greening 2010;9:245e51.