w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

Available online at www.sciencedirect.com

ScienceDirect

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

Water quality and quantity investigation of green

roofs in a dry climate

S. Beecham, M. Razzaghmanesh*
Centre for Water Management and Reuse, School of Natural and Built Environments, University of South Australia,
Adelaide, Australia

article info abstract

Article history: Low-energy pollutant removal strategies are now being sought for water sensitive urban
Received 25 July 2014 design. This paper describes investigations into the water quality and quantity of sixteen,
Received in revised form low-maintenance and unfertilized intensive and extensive green roof beds. The factors of
4 December 2014 Slope (1
and 25
), Depth (100 mm and 300 mm), Growing media (type A, type B and type C)
Accepted 8 December 2014 and Species (P1, P2 and P3) were randomized according to a splitesplit plot design. This
Available online 17 December 2014 consisted of twelve vegetated green roof beds and four non-vegetated beds as controls.
Stormwater runoff was collected from drainage points that were installed in each area.
Keywords: Samples of run-off were collected for five rainfall events and analysed for water retention
Stormwater management capacity and the water quality parameters of NO2, NO3, NH4, PO4, pH, EC, TDS, Turbidity,
Water sensitive urban design Na, Ca, Mg and K. The results indicated significant differences in terms of stormwater
Green roof runoff water quality and quantity between the outflows of vegetated and non-vegetated systems.
Dry climate The water retention was between 51% and 96% and this range was attributed to the green
roof configurations in the experiment. Comparing the quality of rainfall as inflow, and the
quality of runoff from the systems showed that green roofs generally acted as a source of
pollutants in this study. In the vegetated beds, the intensive green roofs performed better
than the extensive beds with regard to outflow quality while in the non-vegetated beds, the
extensive beds performed better than intensive systems. This highlights the importance of
vegetation in improving water retention capacity as well as the role of vegetation in
enhancing pollutant removal in green roof systems. In addition growing media with less
organic matter had better water quality performance. Comparison of these results with
national and international standards for water reuse confirmed that the green roof outflow
was suitable for non-potable uses such as landscape irrigation and toilet flushing.
© 2014 Elsevier Ltd. All rights reserved.

This report also anticipated that most of the world's popu-

1. Introduction lation growth will take place in the urban areas in the next
four decades. This will result in urban areas becoming larger
It has been reported by the United Nations (2011) that in with increased impervious areas, which in turn will lead to
2010, for the first time in history more than half of the many environmental problems. These impacts are in the
world's population lived in urban rather than rural areas. form of changing hydrology, climate change, water and food

* Corresponding author.
E-mail addresses: [email protected], [email protected] (M. Razzaghmanesh).
http://dx.doi.org/10.1016/j.watres.2014.12.015
0043-1354/© 2014 Elsevier Ltd. All rights reserved.
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 371

scarcity and urban heat island effects (Gill et al., 2007). Also, the runoff time of concentration have been critical research
increased flooding and deteriorating water quality in issues.
receiving waters are particularly associated with these One of the solutions to these problems is adopting new
ongoing changes. Climate change consequences have stormwater management strategies such as Low Impact
recently appeared more in the form of extreme rainfall Development (Voyde et al., 2010), Sustainable Urban Drainage
events in the UK, New Zealand and in many parts of Europe Systems (SUDS) (Stovin, 2010), Low Impact Urban Design and
(Carter, 2011) and North America. Generally in areas with Development (LIUDD) (Van Roon, 2005) and Water Sensitive
dry climates, such as Australia, increasing urban tempera- Urban Design (WSUD) (Beecham and Chowdhury, 2012). In
tures and urban heat island effects are one of the most particular, introducing green infrastructure through WSUD is
important issues. In the high rainfall regions in the North- one of the possible solutions to reduce the harmful impacts of
ern Hemisphere as well as in New Zealand, because of urbanization while providing additional amenity and water
flooding problems, reducing peak flow rates and delaying quality benefits for communities and the environment

Fig. 1 e (A) Green roof bed configurations, (B) Statistical experimental layout, PC (platform columns), Area (area), P
(platforms), S (slope), Pl (plots), and D (media depth) and P (plant species).
372 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

(Beecham, 2003; Beecham et al., 2012). For this reason, green low-maintenance green roof beds regardless of drainage layer
roofs have become more widely used in recent years function.
(Emilsson et al., 2006).
One of the important strategies in sustainable urban
drainage systems (SUDS) and Water Sensitive Urban Design 2. Materials and methods
(WSUD) systems is source control of runoff to improve its
quality and to reduce its quantity (Alsup et al., 2010; Voyde 2.1. Experimental setup
et al., 2010). Installing green roofs is viewed as a best man-
agement practice to attenuate peak runoff flows in urban The study area was in a northern suburb of Adelaide, South
areas (Palla et al., 2010) and also to improve stormwater Australia (34.55
S, 138.35
E), which has a hot Mediterranean
quality (Razzaghmanesh et al., 2014a). Therefore, one of the climate based on the Ko € ppeneGeiger climate classification.
most important objectives in green roof studies is determining The experiment was conducted in sixteen small-scale green
how green roofs can affect stormwater quality and quantity. roof systems (Fig. 1-A) constructed at the University of South
This requires an understanding of the hydrological perfor- Australia, Mawson Lakes campus. This experiment was
mance of green roofs. Hydrological studies of green roofs, monitored from early November 2012 until the end of
particularly studies about their water retention capacity, November 2013. It involved multiple green roof design pa-
began in Germany several decades ago (Mentens et al., 2006). rameters and particularly their corresponding effects on
In addition to the rapid progress of the green roof industry in runoff water quality. The statistical design was based on a
Germany, North America has also shown significant progress splitesplit plot experiment in which four factors were
in this area. While much attention has focused on hydrology, considered. The first was Slope for which the selected values
several researchers have also studied changes in the water were 1
(representing a mild slope) and 25
(representing a
quality of outflows (Berndtsson et al., 2009) from green roof steep slope) to cover the range of slopes described in the
systems. In particular, in Europe, most green roof water literature by Villarreal and Bengtsson (2005) and Getter et al.
quality studies have been conducted in cool climate countries (2007). The second factor was Depth for which media depths
(Steusloff, 1998; Ko € hler et al., 2002; Berndtsson et al., 2006; of 100 mm and 300 mm were selected to represent extensive
Emilsson et al., 2006; Teemusk and Mander, 2007; and intensive profiles, respectively (Dunnett and Kingsbury,
Berndtsson et al., 2009; Berndtsson, 2010). Water quality and 2004; Williams et al., 2010b). The first two factors of Slope (1

quantity issues associated with green roofs have been inves- and 25
) and Depth (100 mm and 300 mm) were randomized to
tigated by several researchers (Monterusso et al., 2004; beds (main plots) so that each combination occurred on 3
Hathaway et al., 2008; Bliss et al., 2009; Van Seters et al., beds. The third factor was Species, which was based on three
2009; Alsup et al., 2010; Carpenter and Kaluvakolanu, 2011; plant species, namely P1 e Brachyscome multifida (Cut-leaved
Gregoire and Clausen, 2011). Moreover, there have been Daisy), P2 e Chrysocephalum apiculatum (Everlasting Yellow
studies in high rainfall areas in Asia, especially in Singapore Buttons), and P3 e Disphyma crassifolium (Round-leaved Pig-
and Japan (Berndtsson et al., 2009; Vijayaraghavana et al., face). These were selected by reviewing the recommended
2012). native Australian plants for green roofs described by Hopkins
Green roof technologies in Australia are still very much and Goodwin (2007) together with consideration of the avail-
in their infancy and there are several barriers to the wide- ability, sizes, health, maintenance, fire risk and adaptability to
spread adoption of this technology. Although the environ- the local climate. The fourth factor was Media for which types
mental benefits of green roofs are well understood in some A, B and C were chosen. Also, each area was split into 3 plots
countries the scientific understanding of this WSUD tech- and 3 Species (P1, P2, P3) randomized to the plots within an
nology is still in the earliest stages in Australia (Williams area. In addition to the twelve vegetated beds, there were four
et al., 2010a). Many studies indicate that it is the depth non-vegetated beds which were used as controls. Further-
and type of substrate, vegetation and climate that are the more, a drainage point was placed at the bottom of the each
major influences on green roof water retention capacity area of a main bed and this was connected through a pipe to
while others have opposing views and have considered an individual container so that the outflow water could be
rainfall properties and slope of the green roof to be more collected for measuring volumes and also for taking water
important than these other factors on water retention samples for analysis. Each green roof bed was divided into
(Steusloff, 1998; Monterusso et al., 2004; VanWoert et al., three sections using aluminium sheet dividers and also each
2005; Teemusk and Mander, 2007; Dunnett et al., 2008). part was sealed from the other two parts using aquarium safe
However, Berndtsson (2010) concluded that the factors silicon adhesive covered with high strength tape. The design
which are effective for improving the water quality and of the system was based on a non-weighing lysimeter utilizing
quantity of green roof outflows include the type of substrate the principles of free drainage. Each of the three sub-plots in
material (composition of soil), the depth of the growing each bed were 0.15 m2 in area. Of the 16 beds, 8 were deep or
media and the type of vegetation and the physicochemical intensive (I), with depth ¼ 300 mm. These comprised 6 vege-
properties of pollutants. tated and 2 non-vegetated (or bare) green roofs that only
The purpose of this paper is to investigate the effects of contained growing media but no plants. The remaining 8 beds
these factors on the water quality and quantity of green roof were shallow or extensive (E), with depth ¼ 100 mm. These also
outflows in the dry climate of South Australia. In particular, comprised 6 vegetated and 2 non-vegetated green roofs.
this paper describes the results of a research project investi- Each bed consisted of a bottom insulation layer, a drainage
gating the water quality and quantity of sixteen unfertilized, layer, a geo-textile root barrier, the growing media and finally
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 373

Table 1 e Rainfall characteristics.

Magnesium
Event date Rainfall Rainfall ADWPa Maximum

s
s
s
duration depth (mm) (days) rainfall
(hh:mm) intensity
(mm/hr)
21/04/2013 10:40 22.4 21 8.4

Calcium
15/05/2013 2:20 13.8 2 16.8

s
s
s
03/06/2013 13:10 56.0 7 32.4
15/07/2013 8:40 17.2 6 4.8
23/08/2013 27:00 11.5 1 6.4

Sodium
a
ADWP ¼ Antecedent dry weather period.

s
s
s
the plants, where applicable. Each bed contained a different
Growing media in each of the three sub-plots. These were: type

Potassium
A (denoted Brick mix in this study), which contained red

s
s
s
crushed brick, 20 mm scoria, coir fibre and composted or-
ganics; type B (denoted Scoria mix in this study), which
comprised 20 mm scoria, composted pine bark and hydro-cell
flakes; and type C (denoted Organic mix in this study), which

Orthophosphate
was a combination of 50% media type B with the addition of
50% organic compost. Scoria is a commonly used landscape

s
s
s
material and is a dark coloured, extrusive igneous rock with a
vesicular texture. The detailed composition and physico-
chemical properties of the green roof growing media are
described in Razzaghmanesh et al. (2014b). Three species of
plants used, namely P1 e B. multifida (Cut-leaved Daisy), P2 e C.

Ammonia
apiculatum (Everlasting), P3 e D. crassifolium (Round-leaved Pig-

n.s
face). Half of the beds were set up with Slope ¼ 1
(mild) and

s
s
half with Slope ¼ 25
(steep). Water quality samples were
collected immediately after each of the five rainfall events.
The characteristic of the sampled rainfall events are shown in
Nitrite

n.s
n.s
Table 1.
s

2.2. Water quality test procedures and methods

Table 2 e Significance of the effects of growing media and roof groups.
Nitrate

*s: significant, p < 0.050, **n.s: not significant, p > 0.050; df ¼ degrees of freedom.
n.s
s
s

The first set of water samples was collected 6 months after

plant establishment. The sample collection depended on col-
lecting sufficient volume of runoff from the drainage points
EC
s
s
s

for chemical analysis purposes. In this experiment, five sets of

water quality samples were collected. For each of the five
Turbidity

events, samples were collected from sampling points that

n.s

drained each of the three areas in each bed and two additional
s
s

rainwater samples were also collected. Therefore a total of 50

samples were collected per event. These were refrigerated
after collection and then analysed for pH, Turbidity, Electrical
n.s**
pH

Conductivity (EC), TDS, Nitrate (NO 

3 ), Nitrite (NO2 ), Ammonia
s*
s

þ
(NH34 ), Orthophosphate (PO3
4 ), Potassium (K ), Sodium (Naþ),
2þ 2þ
df
2
7
14

Magnesium (Mg ) and Calcium (Ca ). A multi-probe real-

time model 90 FLT Field Lab Analyser-TPS instrument was
used for measurement of pH, turbidity and conductivity. Po-
Growing media  Roof groups

tassium was measured using an Atomic Absorption Spec-

trometer according to EPA 258.1 standard (EPA, 1983).
Orthophosphate was measured by a Helius Gamma (UVeViS)
spectrophotometer and Automated Flow Colorimetry using
Growing media

the APHA 4500-P standard (APHA, 1998). Nitrate and nitrite

Roof groups

were both measured using a UV-Spectrophotometer auto

Factors

sampler and the data were checked by Automated Flow

Colorimetry using the APHA 4500-NO3-I standard method
(APHA, 1998). Ammonia was measured using Microdistillation
374 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

Fig. 2 e Mean retention of the vegetated and non-vegetated green roofs. The same capital letters (A, B, C and D) show no
statistically significant differences (n ¼ 5, one-way ANOVA, p < 0.05), while the same lower case letters (a, b, c, d & e) show
no statistically significant differences (n ¼ 5, two-way ANOVA, p < 0.05).

Flow Analysis (Davey et al., 2011). An Analyst 400 atomic ab- where in Equation (1), Naþ, Ca2þ and Mg2þ are in meq/l and
sorption spectrophotometer was used to determine the con- SAR is dimensionless. In Equation (2), Naþ, Ca2þ, Mg2þ, Kþ are
centrations of Naþ, Ca2þ and Mg2þ (APHA, 1998). The in meq/100 g soil for the calculation of ESP as a %.
analytical detection limits are discussed in (Razzaghmanesh
et al., 2014a). Also the sodium adsorption ratio (SAR) and
2.3. Statistical analysis
exchangeable sodium percentage (ESP) were calculated ac-
cording to Equations (1) and (2) (Evangelou, 1998):
For conducting statistical analyses, eight groups of vegetated
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u and non-vegetated green roof systems were created. The
u Naþ
S:A:R ¼ t 2þ 2þ (1) vegetated groups were VS1E (Vegetated, Slope ¼ 1
, Extensive),
Ca þMg
2 VS25E (Vegetated, Slope ¼ 25
, Extensive), VS1I (Vegetated,
   Slope ¼ 1
, Intensive) and VS25I (Vegetated, Slope ¼ 25
, Inten-
ESP ¼ Naþ Naþ þ Ca2þ þ Mg2þ þ Kþ  100 (2)
sive). Similarly, the non-vegetated groups were NS1E (Non-

Fig. 3 e Mean pH values of the vegetated and non-vegetated roofs and inflow stormwater. The same capital letters (A, B, C
and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while the same
lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA, p < 0.05),
Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 375

vegetated, Slope ¼ 1
, Extensive), NS25E (Non-vegetated, Growing media, the results of the 2-way ANOVA are shown in
Slope ¼ 25
, Extensive), NS1I (Non-vegetated, Slope ¼ 1
, Table 2 and these indicate that there were no significant
Intensive) and NS25I (Non-vegetated, Slope ¼ 25
, Intensive). (p < 0.05) statistical differences between retention in the
Statistical analyses were applied using the SPSS software vegetated and non-vegetated areas containing Scoria mix and
package version 21. First, the data were examined for Organic mix, but there were statistically significant differences
normality and homogeneity of variance using a Kolmogor- with areas that contained Brick mix. In addition, the results
oveSmirnov test. In the case of non-normality base 10 loga- show that in vegetated beds, the range of retention was be-
rithms of the original data were used in the analyses. A one- tween 52% and 95% while in the non-vegetated beds, the
way ANOVA was used for finding significant differences be- corresponding range was between 31% and 65%. As described
tween the means of the inflow stormwater quality and the by Berretta et al. (2014), evapotranspiration is a key parameter
quality of the green roof outflows from the eight statistical that influences stormwater retention capacity, and thus the
groups. Also a two-way ANOVA (Growing media  Roof groups) hydrological performance of green roofs. The results of this
was conducted to study the main effects of either Growing study also mean that for shorter ADWP periods, the retention
media or Roof groups or interactions of Growing media  Roof mostly depends on the growing media properties and for
groups, as reported in Table 2. Whenever the ANOVA showed a longer ADWP periods it depends more on the presence of
significant difference among the means, a post hoc analysis plants. This is perhaps due to plant water uptake and
was used to investigate where the differences were located. increased evapotranspiration increasing the retention capac-
ity of the green roof beds. Similarly, under sub-tropical con-
ditions in Auckland, New Zealand, Voyde et al. (2010) found
3. Results and discussion that for a green roof at the University of Auckland there was
66% retention of rainfall over a one year period. Their results
3.1. Stormwater retention showed that in some individual events, green roofs could
retain an average of 82% of the rainfall and can reduce up to
Comparing the two main groups of vegetated and non- 93% of the peak flow.
vegetated roofs, it can be seen from Fig. 2, that vegetated
beds generally showed higher retention performance. Among 3.2. pH
the vegetated groups, VS1I (89.66 ± 2.12%) displayed the
highest mean retention and it also showed a statistically sig- In a biological system similar to a green roof, the pH level of
nificant difference (p < 0.05) to the groups of VS25I the growing media is mainly dependent on the concentration
(82.96 ± 2.14%), VS1E (79.62 ± 2.46%) and VS25E (78.13 ± 2.12%). of free hydrogen ions in the media. Plant root activity can
Also, the results indicate that there were no statistically sig- make hydrogen ions available to the soil media (Gymer, 1973).
nificant differences (p > 0.05) between the non-vegetated Generally, at a pH lower than 5.5, many nutrients become very
groups although the intensive non-vegetated beds of NS1I soluble and are readily leached from the soil profile while at
(67.66 ± 4.28%) and NS25I (67.55 ± 3.55%) showed higher higher pH levels, nutrients become insoluble and plants
retention coefficients than the non-vegetated extensive roofs cannot readily extract them. In this study the results showed
of NS1E (63.74 ± 5.07%) and NS25E (65.12 ± 4.62%). Regarding significant statistical differences between mean pH values of

Fig. 4 e Mean Turbidity (NTU) values of the vegetated and non-vegetated roofs and inflow stormwater. The same capital
letters (A, B, C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while
the same lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA,
p < 0.05), Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
376 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

the inflow and the outflow from the vegetated and non- general, for the non-vegetated beds, NS25I (30.90 ± 9.44 NTU)
vegetated roofs (one-way ANOVA, p < 0.05, Fig. 3) but there displayed the highest turbidity values, followed by NS1E
was no significant statistical differences among these sub- (23.01 ± 4.76 NTU), NS1I (22.72 ± 5.30 NTU) and NS25E
groups. Stormwater inflow displayed the highest pH values (13.04 ± 3.03 NTU). For the vegetated groups, VS1E (12.84 ± 2.31
(7.50 ± 0.01) and this was followed by VS1I (6.98 ± 0.07), VS1E NTU) had the highest turbidity value and this was followed by
(6.90 ± 0.11), VS25E (6.80 ± 0.10) and VS25I (6.70 ± 0.13) among VS25E (11.64 ± 2.88 NTU), VS25I (7.17 ± 1.27 NTU) and VS1I
the vegetated groups. Among the non-vegetated treatments, (7.16 ± 1.75 NTU). Moreover, not surprisingly, inflow storm-
NS25E had the highest pH value (6.20 ± 0.21), followed by NS1E water displayed the lowest turbidity value (1.30 ± 0.008 NTU).
(6.0 ± 0.10), NS25I (5.94 ± 0.18) and NS1I (5.65 ± 0.16). In this These results also showed that areas containing Organic mix
study, the pH values in the soil media were between 5.5 and generated the highest turbidities, followed by Brick mix areas
6.5 in the non-vegetated beds, which indicates that these and then Scoria mix areas (Fig. 4). Also there was a correlation
would provide a suitable media for plant growth and activity between turbidity and colour (in web version) of the collected
(Evangelou, 1998). Moreover, the results showed that both samples, with the darkest coloured outflow water draining
vegetated and non-vegetated roofs have significantly changed from the Organic mix areas. This indicates that leaching of
the pH of the stormwater and also that vegetated systems organic matter in the growing media may have been the main
reduce the pH values less than non-vegetated systems, indi- reason for increased turbidity in the outflow water.
cating a potentially important role of vegetation. Also, the
results show that Growing media type changed the pH values 3.4. Electrical Conductivity (EC) and total dissolved
significantly (Table 2). However, the results show that except solids (TDS)
for areas of the vegetated beds which contained Scoria mix,
green roofs can maintain the pH above 7.0, which should A one-way ANOVA was employed to compare the effects of
protect downstream receiving waters from acidification. This vegetated and non-vegetated green roof configurations and
is consistent with the findings of earlier research studies type on outflow water EC values. The results showed that
(Teemusk and Mander, 2007; Berndtsson et al., 2009; Bliss there were significant statistical differences among vegetated
et al., 2009; Berndtsson, 2010; Vijayaraghavana et al., 2012; beds, non-vegetated beds and inflow stormwater EC content
Razzaghmanesh et al., 2014a). (one-way ANOVA, p < 0.05, Fig. 5). Among the non-vegetated
groups, NS1I (with a Log10 value of 3.56 ± 0.21 ms/cm) dis-
3.3. Turbidity played the highest value and this was followed by NS25I
(3.34 ± 0.37 ms/cm), NS25E (2.97 ± 0.38 ms/cm) and NS1E
Often in the period soon after a green roof installation, many (2.94 ± 0.50 ms/cm). Of the vegetated beds, VS1E showed the
soil particles are washed through the beds, and so turbidity highest value (Log10 value ¼ 2.99 ± 0.24 ms/cm) followed by
measurement was of particular interest in this study. The VS25E (2.95 ± 0.33 ms/cm), VS25I (2.93 ± 0.25 ms/cm) and VS1I
results show that there were no significant statistical differ- (2.80 ± 0.24 ms/cm). Moreover, the lowest measured value was
ences between the outflow turbidities in all the non-vegetated again for inflow stormwater, with a Log10 value of
roofs, and vegetated extensive beds from the vegetated 0.81 ± 0.13 ms/cm. In terms of the growing media that there
groups. However, these groups showed significant statistical was no significant statistical difference between those areas
differences with inflow stormwater and the vegetated inten- containing Scoria mix and Organic mix while these two groups
sive beds (VS1I and VS25I) (one-way ANOVA, p < 0.05, Fig. 4). In displayed significant differences with Brick mix areas that had

Fig. 5 e Mean Log10 EC (Log mS/cm) values for vegetated and non-vegetated roofs and inflow stormwater. The same capital
letters (A, B, C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while
the same lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA,
p < 0.05), Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 377

Fig. 6 e (I) Mean Log10 nitrate, (II) Mean nitrite, (III) Mean Log10 ammonia values of the vegetated and non-vegetated roofs
and inflow stormwater. The same capital letters (A, B, C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5,
nIs ¼ 10, one-way ANOVA, p < 0.05), while the same lower case letters (a, b, c, d & e) show no statistically significant
difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA, p < 0.05), Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow
stormwater).
378 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

Fig. 7 e Mean Orthophosphate values for vegetated and non-vegetated roofs and inflow stormwater. The same capital
letters (A, B, C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while
the same lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA,
p < 0.05), Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).

slope ¼ 25
. Regarding EC values, it is concluded that both 3.5. Nitrogen

vegetated and non-vegetated green roof test beds have acted
as a source of EC. Also in vegetated beds the EC values were 3.5.1. Nitrate
less than in the non-vegetated systems and intensive sys- Nitrate was one of three forms of nitrogen that were
tems produced higher EC values than extensive systems. This investigated in this study, the others being nitrite and
indicates that salt leaching occurred from the green roof ammonia. Log10 (NS25E) (2.22 ± 0.27 mg/l) displayed the
media profiles. TDS values were also estimated using the highest value and this was followed by Log10 (NS1I)
empirical relationship TDS (ppm) ¼ 0.64  EC (ms/cm) (Sparks, (2.16 ± 0.40 mg/l), Log10 (NS1E) (2.05 ± 0.22 mg/l) and Log10
1995), and this allowed comparison of outflow water quality (NS25I) (1.88 ± 0.46 mg/l) among non-vegetated categories. In
with recommended maximum values in water quality addition, among vegetated configurations, Log10 (VS1E)
guidelines. (1.05 ± 0.62 mg/l) displayed the maximum value and this

Fig. 8 e Mean potassium values for vegetated and non-vegetated roofs and inflow stormwater. The same capital letters (A, B,
C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while the same
lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA, p < 0.05),
Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 379

was followed by Log10 (VS25I) (0.85 ± 0.75 mg/l), Log10 (VS25E) in outflows from intensive beds were higher in non-vegetated
(0.77 ± 0.71 mg/l) and Log10 (VS1I) (0.49 ± 0.59 mg/l) and the systems while in the vegetated beds the concentrations were
inflow stormwater nitrate value was around zero. The re- all very similar. In this study no fertilizer was added during the
sults showed no statistically significant difference between project duration and therefore the majority of the nitrite was
any of the vegetated areas of VS1E, VS25E, VS1I, VS25I and the sourced from the growing media constituents.
inlet stormwater nitrate values but these had a significant
statistical difference with non-vegetated beds (one-way 3.5.3. Ammonia
ANOVA, p < 0.050, Fig. 6-I). In general, the results showed For determining the statistical differences between Roof
(Table 2) that the outflow water that drained from areas groups, a one way ANOVA was conducted. The results showed
containing Organic mix showed the highest nitrate content, that there were no statistically significant differences among
and this was followed by areas containing Scoria mix and vegetated treatments (p > 0.05) in terms of outflow ammonia
then Brick mix. level. The concentrations were significantly higher than in the
inflow stormwater, for three of the vegetated groups but the
3.5.2. Nitrite difference was not statistically significant for VS25I (Fig. 6-III).
Nitrite was also investigated to understand how nitrogen is For the non-vegetated groups, extensive roofs recorded higher
transformed in vegetated and non-vegetated green roof sys- ammonia concentrations than intensive roofs which could be
tems. A one-way ANOVA was conducted to examine any due to the lower volume of growing media and probably less
statistical differences between the groups. The results reaction time for nitrification to nitrite and nitrate.
revealed that there were no significant statistical differences
between nitrite levels in the outflows of all the vegetated 3.6. Orthophosphate
groups. Of the vegetated groups, only VS25I showed any sig-
nificant difference with the inflow stormwater nitrite level. Phosphorus is one of the three nutrients generally added to
However, all the vegetated groups showed statistically sig- soils in fertilizers. The main role of phosphorus is to transfer
nificant differences from both non-vegetated extensive and energy in living organisms. Although phosphorus is one of the
intensive. In this study, of the non-vegetated beds NS25I necessary elements for plant growth, poor management of
(11.87 ± 2.51 mg/l) displayed the highest value and this was this element can decrease water quality. Inorganic phosphate
followed by NS1I (11.11 ± 3.53 mg/l), NS1E (3.69 ± 1.14 mg/l) and (PO3
4 ) is the only form available to living organisms (Hsieh
NS25E (3.52 ± 1.02 mg/l). Among the vegetated beds, the et al., 2007). In this study results of effect of Roof groups on
highest nitrite concentration was for VS25I (0.99 ± 0.98 mg/l), the outflow water quality showed that there were no statis-
followed by VS1E (0.69 ± 0.68 mg/l), VS25E (0.50 ± 0.38 mg/l) and tically significant differences among the vegetated beds or
VS1I (0.48 ± 0.43 mg/l). The lowest concentration was recorded between these and the inflow stormwater orthophosphate
in the inflow stormwater (0.08 ± 0.01 mg/l) (Fig. 6-II). Further content. However, these all showed statistically significant
investigation using a post hoc multiple comparison test and p difference with the non-vegetated beds (one-way ANOVA,
value interpretation (Table 2) showed that there were no sig- p < 0.050, Fig. 7), which generated much higher orthophos-
nificant statistical differences among Growing media in all of phate concentrations. This indicates that plant uptake of
the vegetated and non-vegetated groups except for the area of orthophosphate occurred in this study. Also a post hoc mul-
NS25I that contained Scoria mix. Moreover, the concentrations tiple comparison test and p value results (Table 2) showed that

Fig. 9 e Mean sodium values for vegetated and non-vegetated roofs and inflow stormwater. The same capital letters (A, B, C
and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while the same
lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA, p < 0.05),
Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
380 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

Fig. 10 e Mean calcium values for vegetated and non-vegetated roofs and inflow stormwater. The same capital letters (A, B,
C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while the same
lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA, p < 0.05),
Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).

the areas that contained Organic mix produced higher ortho- beds, the phosphorous concentration was reduced by 60%e
phosphate values and this was followed by Scoria mix and then 80%.
those sections with Brick mix. This indicates that leaching of
organics in the growing media results in higher orthophos- 3.7. Cation analysis
phate concentrations. Overall, a comparison of inflow
stormwater orthophosphate content with the outflow quality 3.7.1. Potassium
from both vegetated and non-vegetated beds shows that the Potassium was one of four nutrient cations investigated in
beds acted as a source of phosphorous but in the vegetated this study, the others being sodium, calcium and magnesium.

Fig. 11 e Mean magnesium values of the vegetated roofs, non-vegetated roofs and inflow stormwater. The same capital
letters (A, B, C and D) show no statistically significant difference (nv ¼ 15, nn-v ¼ 5, nIs ¼ 10, one-way ANOVA, p < 0.05), while
the same lower case letters (a, b, c, d & e) show no statistically significant difference (nv ¼ 15, nn-v ¼ 10, two-way ANOVA,
p < 0.05), Subscripts: v (vegetated), n-v (non-vegetated) and Is (Inflow stormwater).
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 381

The results of one-way ANOVA showed that there were no 3.7.2. Sodium
statistically significant differences (p > 0.05) between the Similar to potassium, the one-way ANOVA Sodium results
different vegetated green roof groups. However, these groups showed that there were no statistical significant differences
showed a statistically significant difference from the inlet (p > 0.05) between vegetated green roof groups while these
stormwater and also from all the non-vegetated groups groups did show statistically significant differences from
(Fig. 8). Among the non-vegetated beds, NS1I (4.32 ± 1.43 mg/l) the inflow stormwater and also from the non-vegetated
displayed the highest value and this was followed by the other beds. Among the non-vegetated beds, the two intensive
intensive bed, NS25I (3.76 ± 1.47 mg/l) and these were statis- groups displayed the highest sodium values, Fig. 9. The
tically significantly different to and higher than the two areas of the experimental beds that contained Organic mix
extensive beds. It is concluded that both vegetated and non- showed statistically significant differences from the other
vegetated green roof beds have acted as a source of potas- parts that contained either Scoria mix or Brick mix. It is
sium but that some potassium uptake by plants has occurred. concluded that both the vegetated and non-vegetated green
In addition, outflow from the non-vegetated intensive sys- roof beds acted as a source of sodium in this study but that
tems contained higher potassium concentrations than the the presence of plants reduced the sodium levels, indicating
outflow from extensive systems. plant uptake.

Fig. 12 e (A) SAR values from vegetated and non-vegetated green roof beds (B) ESP (%) values from vegetated and non-
vegetated green roof beds.
 382
Table 3 e Water quality of outflows from green roofs and impervious surfaces compared to Australian and international water quality standards.
Water Experimental data Experimental data from authors' Range recommended by Ranges reported in the literature [7]
quality previous studies [6] standards
parameter
Stormwater Green roofs (N-f) Non- Asphalt Aluminium Green roofs Potable Non-potable Urban Rainwater Roofs Trafficable Trafficable
vegetated roof roof (F) reuse reuse irrigation areas areas

w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4
roofs with low with
Brick or Organic Brick or
density high
scoria mix scoria
density
mixes mixes
pH 7.45e7.55 4.2e8.01 5.65e8.22 4.5e7.55 7.13 7.58 6.72e8.45 6.5e8.5 4.5e9.0 [4] 4.5e8.4 3.9e7.5 4.7e6.8 6.4e7.9 6.4e7.9
[1&2] [1&4]
Turbidity (NTU)* 1.3e1.35 1.51e67.5 1.54e104 1.8e100 2.98 1.26 4.0e300 <5 [3] <2 [4] to n.a. n.a. n.a. n.a. n.a.
2e5 [1]
EC (mS/cm) <0.5 1.82e59 1.27e86 0.85e75 75.00 23.00 <100- <200 [4] e 0e8100 [4] 28e223 25e269 n.a. 108e2436
Total dissolved 2.4e5.9 6.50e150 8.1e220 5.4e380 37.00 31.00 385.77 <600.00 [5] e e n.a. n.a. n.a. n.a.
solids (mg/l)
Nitrate (mg/l)* <1 1e40 1.01e100 20e350 2.62 1.90 2.20e39.20 <10 [4] <10 [4] <30 [4] 0.0e7.4 0.1e4.7 n.a. 0.0e16
Nitrite (mg/l) 0.07e0.10 0.02e3.5 0.04e4 2.2e16.5 n.a. n.a. n.a. <3 [2&3] <10 [3] <10 [3] n.a. n.a. n.a. n.a.
Ammonia (mg/l) 1.01e1.10 1e16.5 1e20 1e55 n.a. n.a. n.a. 0.50 [3] 20e30 [2] n.a. n.a. n.a. n.a.
Orthophosphate <0.01 0.03e2.37 0.04e4.39 0.46e7.5 0.14 0.16 0.20e2.20 0.1 [5] e 0.50 [4] <0.20 <0.50 n.a. 0.34
(mg/l)*
Potassium (mg/l) 0e0.46 0.03e3.0 0.52e6.45 0.05e7.03 9.60 3.00 38.37 10e20 [1] e e 0.46e0.65 n.a. n.a. 1.7e3.8
Sodium (mg/l) 0e1.84 0.10e10.4 1.26e15 0.16e19.46 n.a. n.a. n.a. <180 [3] 0.22e20 n.a. n.a. 5e474
Calcium (mg/l) 0 1e104 4e140 2e151 n.a. n.a. n.a. 60e200 [3] 200e500 [3] <400 [3] 1.1e67.13 1e1900 n.a. 13.7e57
SAR 0 0.15e0.21 0.18e0.22 0.2e0.42 n.a. n.a. n.a. n.a. n.a. <3 [1] n.a. n.a. n.a. n.a.

N-f: Non-fertilized, F: Fertilized, n.a.: data not available, * Pollutant levels exceed potable standards [1] (USEPA, 2012); [2] (EPA South Australia, 1999); [3] (EPA South Australia, 2003); [4] (Higgins et al.,
2007); [5] (NRMMC, 2011); [6] (Razzaghmanesh et al., 2014a); [7] (Go € bel et al., 2007).
 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4 383

3.7.3. Calcium published data indicates that the addition of fertilizer is a

Again, the one-way ANOVA results showed that there were no major source of pollutants but so is leaching from the green
statistical significant differences (p > 0.05) between vegetated roof growing media.
green roof groups while these groups showed statistically
significant differences from the inflow stormwater. The
intensive non-vegetated beds displayed the highest calcium
4. Conclusion
levels and these were statistically significantly different from
the extensive beds, Fig. 10. In terms of soil media type, the
In this study, the quantity and quality of outflows from twelve
areas of the experimental beds that contained Brick mix and
vegetated and four non-vegetated green roofs were studied
Scoria mix showed statistically significant differences (Table 2)
over a twelve month period. In terms of water quantity, water
from areas that contained Organic mix. It is concluded that the
retention was higher in vegetated beds than in non-vegetated
soil media has acted as a source of calcium, but that the
beds and in both systems the intensive beds showed higher
presence of plants reduces the calcium levels, due to plant
retention capacity. In terms of water quality, both vegetated
uptake.
and non-vegetated systems acted as pollutant sources.
Generally the pollutant concentrations were higher in runoff
3.7.4. Magnesium
from non-vegetated beds. This was believed to be partly due to
There was a statistically significant difference between the
leaching from the growing media and partly due to plant up-
magnesium concentration in the inflow stormwater, and the
take in the vegetated beds. During the period of this study, no
outflow concentrations from the vegetated and non-
additional fertilizers were added to the green roof beds and
vegetated beds (one-way ANOVA, Fig. 11). The only other
three different Growing media were used. It can be concluded
statistically significant difference was between the non-
that the media containing higher organic content (media type
vegetated intensive roofs and other roof types. In terms of
C) generated higher nutrient concentrations. Generally, in the
the growing media, the only significant statistical difference
vegetated beds, intensive green roofs outperformed the
was between areas that contained Organic mix and areas that
extensive beds in terms of outflow water quality while in the
contained the other two media types, with higher values of
non-vegetated beds, extensive beds performed better than the
magnesium in the outflow water sourced from the Organic
intensive systems. In terms of the potential for recycling of
mix sections.
outflow water from green roofs, current state, national and
international standards indicate that such water can be
3.7.5. SAR and ESP
reused for urban landscape irrigation and for other non-
From the measured Na, Ca, Mg and K values, SAR (Fig. 12-A)
potable purposes such as toilet flushing. However, green
and ESP (Fig. 12-B) were calculated to investigate the green
roof outflow water should not be reused for potable con-
roof growing media properties. From Fig. 12, the range of ESP
sumption. Further study of the effects of ADWP and evapo-
indicates that the green roof media could be classified as
transpiration on water retention in dry climates is
moderately sodic. Sodic soils generally reduce the water
recommended as is further investigation of the effects of
infiltration rate which means that more elements can be
different growing media on water chemistry. A long-term
retained for longer periods in the soil profile allowing more
study of water quality from green roofs in a dry climate is
time for plant uptake. Also, all SAR values are less than 6
recommended to increase our understanding of their sus-
which means that the outflow water is suitable for urban
tainability, to quantify their environmental benefits such as
irrigation purposes (USEPA, 2012).
reduction of urban heat island effects and finally to under-
stand the important role that they can potentially play in
water sensitive urban design.
3.8. Feasibility of reusing green roof outflow water in
urban areas

There is not any specific standard for reusing runoff from references
green roofs (Berndtsson et al., 2006; Vijayaraghavana et al.,
2012; Razzaghmanesh et al., 2014a). Therefore in this study,
to investigate the potential for reusing runoff from the Alsup, S., Ebbs, S., Retzlaff, W., 2010. The exchangeability and
experimental green roofs, all available local, state, national leachability of metals from select green roof growth
substrates. Urban Ecosyst. 13, 91e111.
and international water quality guidelines were reviewed. The
APHA, 1998. Standard Methods for the Examination of Water and
qualities of the collected water samples were examined with Wastewater. American Public Health Association,
regard to alternative reuse scenarios such as potable con- Washington.
sumption and non-potable reuse, including as urban land- Beecham, S., 2003. Water sensitive urban design: a technological
scape irrigation and toilet flushing. From Table 3, it is assessment. J. Stormwater Ind. 17, 5e13.
acceptable to reuse green roof outflows for urban irrigation Beecham, S., Chowdhury, R.K., 2012. Effects of changing rainfall
patterns on WSUD in Australia. J. Water Manag. 165, 285e298.
and other non-potable purposes but not for potable con-
Beecham, S., Razzaghmanesh, M., Kazemi, F., 2012. Introducing
sumption. Also, the quality of outflow water from the green
green infrastructure into the built environment of Adelaide.
roofs is generally higher than from impervious surfaces such In: 6th International Conference and Workshop on the Built
as low and high density trafficable areas. Moreover, Environment in Developing Countries. Adelaide, South
comparing these experimental results with data from other Australia.
384 w a t e r r e s e a r c h 7 0 ( 2 0 1 5 ) 3 7 0 e3 8 4

Berndtsson, J., Emilsson, T., Bengtsson, L., 2006. The influence of Hsieh, C.-H., Davis, A.P., Needelman, B.A., 2007. Bioretention
vegetated roofs on runoff water quality. Sci. Total Environ. column studies of phosphorous removal from urban
355, 48e63. stormwater runoff. Water Environ. Res. 79, 177e184.
Berndtsson, J.C., 2010. Green roof performance towards € hler, M., Schmidt, M., Grimme, F.W., Laar, M., 2002. Green roofs
Ko
management of runoff water quantity and quality: a review. in temperate climates and in the hot-humid tropics e far
Ecol. Eng. 36, 351e360. beyond the aesthetics. Environ. Manag. Health 13, 382e391.
Berndtsson, J.C., Bengtsson, L., Jinno, K., 2009. Runoff water Mentens, J., Raes, D., Hermy, M., 2006. Green roofs as a tool for
quality from intensive and extensive vegetated roofs. Ecol. solving the rainwater runoff problem in the urbanized 21st
Eng. 35, 369e380. century? Landsc. Urban Plan. 77, 217e226.
Berretta, C., Poe €, S., Stovin, V., 2014. Moisture content behaviour Monterusso, M.A., Rowe, D.B., Rugh, C.L., Russell, D.K., 2004.
in extensive green roofs during dry periods: the influence of Runoff water quantity and quality from green roof systems.
vegetation and substrate characteristics. J. Hydrol. 511, Acta Hortic. 639, 369e376.
374e386. NRMMC, 2011. Australian Drinking Water Guidelines Paper 6
Bliss, D.J., Neufeld, R.D., Ries, R.J., 2009. Storm water runoff National Water Quality Management Strategy. National
mitigation using a green roof. Environ. Eng. Sci. 26, 407e417. Health and Medical Research Council, Canberra.
Carpenter, D.D., Kaluvakolanu, P., 2011. Effect of roof surface type Palla, A., Gnecco, I., Lanza, L., 2010. Hydrologic restoration in the
on stormwater run-off from full-scale roofs in a temperate urban environment using green roofs. Water 2, 140e154.
climate. J. Irrig. Drain. Eng. 137, 161e169. Razzaghmanesh, M., Beecham, S., Kazemi, F., 2014a. Impact of
Carter, J.G., 2011. Climate change adaptation in European cities. green roofs on stormwater quality in a South Australian urban
Curr. Opin. Environ. Sustain. 3, 193e198. environment. Sci. Total Environ. 470e471, 651e659.
Davey, D.E., McLeod, S., Chow, C.W., Ostrowski, J., Duker, P., Razzaghmanesh, M., Beecham, S., Brien, C.J., 2014b. Developing
Bustamante, H., et al., 2011. Development of an on-line resilient green roofs in a dry climate. Sci. Total Environ. 490,
nitrogen monitoring system using microdistillation flow 579e589.
analysis. In: ISSNIP 2011. Adelaide, Australia. Sparks, D.L., 1995. Environmental Soil Chemistry. Academic
Dunnett, N., Kingsbury, N., 2004. Planting Green Roofs and Living Press, INC, USA.
Walls. Timber Press, Portland (OR). Steusloff, S., 1998. Input and output of airborne aggressive
Dunnett, N., Nagase, A., Booth, R., Grime, P., 2008. Influence of substances on green roofs in Karlsruhe. In: Urban Ecology.
vegetation composition on runoff in two simulated green roof Springer-Verlag, Berlin,Germany.
experiments. Urban Ecosyst. 11, 385e398. Stovin, V., 2010. The potential of green roofs to manage urban
Emilsson, T., Berndtsson, J.C., Mattsson, J.E., Rolf, K., 2006. Effect stormwater. Water Environ. 24, 192e199.
of using conventional and controlled release fertiliser on Teemusk, A., Mander, U., 2007. Rainwater runoff quantity and
nutrient runoff from various vegetated roof systems. Ecol. quality performance from a greenroof: the effects of short-
Eng. 29, 260e271. term events. Ecol. Eng. 30, 271e277.
EPA, 1983. Methods for the Chemical Analysis of Water and United Nations, 2011. Population Distribution, Urbanization,
Wastes (MCAWW) 600/4-83/020. Internal Migration and Development: an International
EPA South Australia, 1999. South Australian Reclaimed Water Perspective. Department of economic and social affairs
Guidelines (Treated Effluent). Department of human services population division.
and environmental protection agency, Government of South USEPA, 2012. Guidelines for Water Reuse. EPA/600/R-12/618.
Australia. National Risk Management Research Laboratory Office of
EPA South Australia, 2003. Environment Protection (Water Research and Development, Cincinnati, Ohio, U.S. Agency for
Quality) Policy 2003. Environment Protection Authority, International Development.
Government of South Australia. Van Roon, M., 2005. Emerging approaches to urban ecosystem
Evangelou, V.P., 1998. Environmental Soil and Water Chemistry: management: the potential of low impact urban design and
Principles and Application. John Wiely & Sons, INC, USA. development principles. J. Environ. Assess. Policy Manag. 7,
Getter, K.L., Rowe, D.B., Andresen, J.A., 2007. Quantifying the 125e148.
effect of slope on extensive green roof stormwater retention. Van Seters, T., Rocha, L., Smith, D., MacMillan, G., 2009.
Ecol. Eng. 31, 225e231. Evaluation of green roofs for runoff retention, runoff quality,
Gill, S.E., Handley, J.F., Ennos, A.R., Pauleit, S., 2007. Adapting and leachability. Water Qual. Res. J. Can. 44, 33e47.
cities for climate change: the role of the green infrastructure. VanWoert, N.D., Rowe, D.B., Andresen, J.A., Rugh, C.L., Xiao, L.,
Built Environ. 33, 115e133. 2005. Watering regime and green roof substrate design affect
€ bel, P., Dierkes, C., Coldewey, W.G., 2007. Stormwater runoff
Go sedum plant growth. Hortscience 40, 659e664.
concentration matrix for urban areas. J. Contam. Hydrol. 91, Vijayaraghavana, K., Joshia, U.M., Balasubramanian, R., 2012. A
26e42. field study to evaluate runoff quality from green roofs. Water
Gregoire, B., Clausen, J., 2011. Effect of a modular extensive green Res. 46, 1337e1345.
roof on stormwater runoff and water quality. Ecol. Eng. 37, Villarreal, E., Bengtsson, L., 2005. Response of a sedum green-roof
963e969. to individual rain events. Ecol. Eng. 25, 1e7.
Gymer, R.G., 1973. Chemistry: an Ecological Approach. Harper & Voyde, E., Fassman, E., Simcock, R., 2010. Hydrology of an
Row. extensive living roof under sub-tropical climate conditions in
Hathaway, A.M., Hunt, W.F., Jennings, G.D., 2008. A field study of Auckland, New Zealand. J. Hdrol. 394, 384e395.
green roof hydrologic and water quality performance. Am. Williams, N.S.G., Rayner, J.P., Raynor, K.J., 2010a. Green roofs
Soc. Agric. Biol. Eng. 51, 37e44. for a wide brown land: opportunities and barriers for
Higgins, J., Warnken, J., Teasdale, P.R., 2007. A Review of Water rooftop greening in Australia. Urban For. Urban Green. 9,
Quality Criteria in Australian Reclaimed Water Guidelines and 245e251.
Sewage Effluent Discharge Licences. School of Environmental Williams, N.S.G., Hughes, R.E., Jones, N.M., Bradbury, D.A.,
and Applied Science, Griffith University. Rayner, J.P., 2010b. The performance of native and exotic
Hopkins, G., Goodwin, C., 2007. Living Architecture: Green Roofs species for extensive green roofs in Melbourne, Australia. Acta
and Walls. CSIRO. Hortic. 881, 689e696.