Beruflich Dokumente
Kultur Dokumente
Abstract: Roof surfaces represent a significant portion of the impervious area associated with urban development. Storm-water runoff from
those surfaces causes stream degradation in receiving waters attributable to excess volume of water runoff. This paper investigates the
influence of roof surface type on storm-water runoff and specifically considers the benefits of a vegetated roof, or green roof, as a
storm-water best management practice (BMP). Runoff data were collected over a 6-month period from three full-scale roofs, which were
retrofitted with flow meters and automated water-quality samplers. The roof surfaces included an asphalt roof (for control purposes), a
vegetated extensive green roof, and a stone ballasted roof. Both the green roof and stone roof were effective at reducing runoff volume
and attenuating peak discharge, with the green roof being more efficient for rainfall events less than 2.54 cm. Overall, the green roof retained
68.25% of rainfall volume and reduced peak discharge by an average of 88.86%. Water-quality results were inconclusive, but did provide
some indication that green roof systems could reduce nutrient loadings. DOI: 10.1061/(ASCE)IR.1943-4774.0000185. © 2011 American
Society of Civil Engineers.
CE Database subject headings: Storm-water management; Best Management Practice; Hydrology; Water quality; Runoff; Urban
development; Subtainable development; Roofs.
Author keywords: Storm-water management; Best management practice; Hydrology; Water quality; Runoff; Urban development;
Low-impact development; Green roof.
Introduction by the vegetation. Excess precipitation drains from the green roof
through roof drains similar to those of traditional roofs.
The U.S. Environmental Protection Agency (EPA) defines a best A key feature of vegetated roofing systems is their ability to
management practice (BMP) as a “measure, technique or a struc- mitigate storm-water runoff in urban environments where imper-
tural control which manages the quantity and improves water qual- vious surfaces are prolific (Liptan 2003). Research has shown that
ity of storm-water runoff, for a given set of conditions, in the most the quantitative benefits associated with green roofs include a
cost-effective manner” (U.S. EPA 2003). BMPs such as rain gar- decrease of runoff peak discharge (Carter and Rasmussen 2006;
dens, storm-water treatment wetlands, and detention ponds are in Moran et al. 2004; Kurtz 2008), a delay in runoff occurrence
wide practice, but require significant surface areas for construction. (Carter and Rasmussen 2006; Moran et al. 2004; VanWoert et al.
When ground surface area is limited or unavailable, such as in a 2005), and a reduction in runoff volume (Kolb 2004; Moran et al.
dense urban environment, a vegetated roof has proved to be a good 2004; Carter and Rasmussen 2006; Teemusk and Mander 2007;
alternative roof covering because of the many benefits associated VanWoert et al. 2005; Kurtz 2008; Hutchinson et al. 2003). How-
with vegetated roofs including storm-water management. ever, the performance of green roofs has been shown to be based on
A vegetated roof, or green roof, is a rooftop planted with veg- factors such as roof slope, media depth, and roof surface (Kolb
etation over layers of soil media and drainage systems. There are 2004; VanWoert et al. 2005).
two general categories of green roofs—extensive and intensive. Green roofs can improve water quality by retaining and filtering
Extensive green roofs tend to use thinner soil media, weigh less, rain water through the soil medium and root uptake zone (Dramstad
are less expensive and have lower maintenance requirements. In et al. 1996). However, the literature does not clearly demonstrate
contrast, intensive green roofs typically have deeper soil layers, the influence of vegetated roofing systems on water quality. Studies
a wider variety of plants, are heavier, and require higher levels have found that green roofs can cause an increase in nutrient con-
of maintenance. In both roof types, precipitation is retained by centrations in runoff (Hutchinson et al. 2003; Moran et al. 2005;
the green roof and either evaporated into the atmosphere or used Teemusk and Mander 2007; Van Seters et al. 2007) but reduce
the concentration of other common storm-water pollutants includ-
1
Associate Professor, Civil Engineering Dept., Lawrence Technological ing metals and polycyclic aromatic hydrocarbons (Van Seters et al.
Univ., 21000 W. Ten Mile Rd., Southfield, MI 48075. E-mail: 2007; Berhage et al. 2007; Buccola et al. 2008). Even the reported
carpenter@ltu.edu nutrient data is mixed with a reduction in nitrogen reported in some
2
Research Assistant, Civil Engineering Dept., Lawrence Technological investigations (Berhage et al. 2007; Van Seters et al. 2007).
Univ., 21000 W. Ten Mile Rd., Southfield, MI 48075. E-mail: There exists a need to investigate full-scale vegetated roof sys-
pk000547767@ltu.edu tems over a range of climatic conditions to better quantify the
Note. This manuscript was submitted on September 25, 2009; approved
hydrologic and water-quality response of green roofs on storm-
on June 18, 2010; published online on June 22, 2010. Discussion period
open until August 1, 2011; separate discussions must be submitted for water management. Monitoring full-scale and retrofitted roofs al-
individual papers. This paper is part of the Journal of Irrigation and Drai- lows investigators to determine how vegetated roofs will respond in
nage Engineering, Vol. 137, No. 3, March 1, 2011. ©ASCE, ISSN 0733- actual as-built conditions that may vary from reported monitoring
9437/2011/3-161–169/$25.00. project results on smaller pilot scale roofs. The opportunity to
Sedum
Planting Mix
Filter Fabric
The discharge through a flume, in this case storm-water runoff, was Attenuation of the peak discharge can be analyzed by consid-
calculated based on the water level readings in the flumes using a ering both the reduction in magnitude and the lag time associated
calibrated weir equation with peak discharge. To compare the three roofs, which were differ-
ent in size, the discharge rates were normalized per 100 m2 of the
Q ¼ 1:68 × H 1:9 ð1Þ corresponding roof areas and represented as hydrographs. The re-
duction in peak discharge can be determined by comparing the
where Q = discharge in ft3 =s and H = flume water level (ft) as re- green roof and stone roof peaks with the asphalt roof peak dis-
corded by the flowmeter. Values were then converted into metric charge serving as the control.
units for analysis and reporting. The discharge rate from Eq. (1) Two methods were utilized to determine peak discharge time
was integrated over the length of time that runoff was measured delay. The first compared the time from which rainfall was first
to obtain the volume of the runoff recorded to the time in which the highest discharge was recorded.
Z t However, during most events, multiple peaks were recorded, mak-
V¼ QðtÞdt ð2Þ ing this type of analysis suspect. The second method determined
0 the time to peak discharge using a precipitation centroid method
where V = volume of runoff in m3 ; Q = discharge in m3 =s; and t = (Carter and Rasmussen 2006) that calculates the depth-weighted
time in seconds. These calculations were repeated for each roof time
system. For comparison between roofs, volume was converted into
1X n
a runoff depth by dividing by roof surface area. Rainfall was mea- tp ¼ Pt ð5Þ
sured in 5-min increments, with the total rainfall depth for a given P i¼1 i i
storm event being the summation of rain depths from each incre-
ment. The amount of water retained by each roof is the total rainfall where P = total storm precipitation depth (cm); n = total number of
depth for a storm event minus the runoff depth. Finally, a percent time intervals; Pi = depth of precipitation in each interval (cm); and
retained was determined ti = observation time of each interval (s). This method was used to
calculate a time to peak discharge centroid for the asphalt roof, the
rainfall depth ðcmÞ runoff depth ðcmÞ green roof, and the stone roof.
% retained ¼ × 100%
rainfall depth ðcmÞ
Water Quality
ð3Þ
The Avalanche Sampler has a refrigerated compartment with a 14-
The volume of rainfall for each event and the corresponding bottle configuration for sequential collection of water samples. The
runoff volume for the three roofs have been utilized to calculate sampler was programmed to draw a sample when volumetric runoff
the volumetric runoff coefficient. A volumetric runoff coefficient reached the equivalent of 0.15 cm of rainfall depth over the roof
Fig. 4. Normalized flow hydrographs for each roof for the May 30, 2008, rainfall event
surface. Once the volumetric threshold was reached, the sampling exhibit a reduced volume and delayed discharge response. The fol-
protocol was initiated and water was drawn from the flume and lowing sections provide quantitative analysis of these trends.
deposited into a bottle every 15 min over the duration of the storm
event. Once the storm was finished, 50 cm3 from each of the 14 Volume
bottles were mixed to form a composite sample for analysis.
A total of 21 storm events with measureable runoff were observed
The composite sample was then utilized to determine an event
during a six-month period from April 1, 2008, to September 30,
mean concentration (EMC) for total phosphate, nitrate (NO3 -N),
2008 (Table 2), with the cumulative volume of runoff depicted
and total solids for each storm event using standardized laboratory
in Fig. 5. Unfortunately, equipment malfunction for the stone roof
testing protocols (Johnson et al. 2000). The total mass for each
made the water-quantity monitoring results for the final two storm
water-quality parameter was then determined by multiplying total
event runoff volume with concentration over the event duration.
Concentration and mass of total phosphate, nitrate (NO3 -N), and
total solids were determined for five sampled rainfall events that
exhibited rainfall greater than 1.27 cm in a 24 h period. To deter-
mine whether there was a statistically significant difference be-
tween the roof surfaces, a Mann-Whitney nonparametric test for
independent paired samples was performed on each pairwise
comparison.
The normalized hydrographs for a May 30, 2008, rain event are
presented as an example of the influence of roof type on hydro-
graph response (Fig. 4). It is clear that the asphalt roof responds
almost instantaneously to the rainfall and exhibits multiple peaks.
Fig. 5. Accumulative runoff from each roof (April–September 2008)
This is not the case for the green roof or the stone roof, which
Depth (cm)
Stone Roof %
6 Retained
50 Green Roof %
Retained
40
4
30
20
2
10
0 0
Apr(1.727) May(5.385) Jun(8.940) July(6.477) Aug(0.737) Sep(10.770)
Fig. 6. Monthly retention for each roofing system (total monthly rainfall is in parentheses for each month)
events questionable. Those values were removed from the analysis, roof is 0.883 with a standard deviation of 0.063. These values
with stone roof results being reported for 19 events. This is increas- are 0.399 and 0.192 for the stone roof (April to August) and
ingly unfortunate because those were the two largest rainfall events 0.155 and 0.215 for the green roof. To demonstrate the influence
recorded. that rainfall volume has on the volumetric runoff coefficient, rain-
For the measured events, the asphalt roof recorded 30.48 cm of fall events have been categorized into three sizes: rainfall volumes
runoff, for a retention rate of 10.45%. The green roof recorded that varied between 0.40 cm to 1.27 cm were categorized as small
10.81 cm of runoff, for an overall retention rate of 68.25%. Finally, rainfall events, 1.27 cm to 2.54 cm were considered medium events,
the stone roof recorded 9.83 cm of runoff from the 23.27 cm of rain and greater than 2.54 cm were considered large events (Table 3).
experienced from April to August (retention rate of 57.75%). As a The green roof is highly efficient at retaining smaller rainfall
direct comparison, the green roof retained 83.79% of the rainfall events, with the mean C v being nearly zero for events under
during the same 5-month period. 1.27 cm, but the green roof C v increases significantly as storm size
Fig. 6 depicts the monthly depth of runoff for each of the three increases. The stone roof had a mean C v of 0.40 for small and
roofs as bars with percent retained as a line plot. The asphalt roof medium-size storm events, but also had a higher value for the
discharge is relatively constant around 10% except for the month of one large storm event. Finally, the asphalt roof retained similar vol-
April, when 20% of the precipitation was retained. The first two umes of water for all storm events.
precipitation events in April were a wet snow/rain mixture, which The variability in roof response to a rainfall event is attributable
probably influenced performance. Runoff from the stone roof was to various factors that affect the rainfall retention, including rainfall
more variable than that from the asphalt roof, retaining between intensity/duration/frequency (i-d-f ), evapotranspiration rates, and
53.2% (April and May) and 67.9% (July). Finally, the green roof in the case of the green roof, initial soil medium moisture. When
exhibited significantly more variability with regard to rainfall re- the soil media on the green roof is already saturated, additional rain-
tention, retaining as much as 99.45% of rainfall in August and fall was not absorbed before runoff was recorded from the green
as little as 34.66% for September. The month of September re- roof. When a longer time period between storm events exists,
corded the highest rainfall of any month in the investigation, with evapotranspiration removed water temporarily stored in the soil
10.77 cm of rainfall recorded in two events. media and water storage layer.
The variability of response to individual storm events for the As an example, May had a rainfall of 5.38 cm over four events
different roof surfaces is depicted as volumetric runoff coefficient that were at least 4 days apart, whereas in June (Fig. 7), the total
(C v ) in Table 2. The mean volumetric coefficient for the asphalt rainfall was 8.91 cm over seven storm events, with several events
Table 3. Mean and Standard Deviation (SD) of Cv for All Three Roofs Based on Storm Size
Asphalt roof Green roof Stone roof
Size category Storm size (cm) Number of observations Mean SD Mean SD Mean SD
Small 0.40–1.27 10 0.866 0.075 0.044 0.092 0.409 0.191
Medium 1.27–2.54 8 0.891 0.263 0.131 0.123 0.37 0.191
Large > 2:54 3 0.917 0.923 0.591 0.669 0.559a NA
a
Stone roof volumetric runoff coefficients for large storm size category is based on a single storm as compared to three storms for the small and medium sized
storms.
70
Downloaded from ascelibrary.org by Dr S & SS Ghandhy Government Engineering College on 10/09/19. Copyright ASCE. For personal use only; all rights reserved.
Percent Retained
1.5 60
Depth (cm)
50
1 40
30
0.5 20
10
0 0
7-Jun 10-Jun 13-Jun 14-Jun 23-Jun 25-Jun 28-Jun
Date
Fig. 7. Green roof retention for individual rainfall events for June 2008
occurring only 1–3 days apart. For June, the green roof retained If the best-fit relationship for the stone roof is used to predict the
81% of the total precipitation, compared with 94% for May. response to the larger events, the stone roof would be expected to
The effect of temporal spacing and magnitude of rain events on retain a total 48.58% of rainfall instead of the 57.75% of the rainfall
retention rate is clearly depicted in Fig. 7. it retained from April to August.
A plot of runoff versus rainfall (Fig. 8) depicts the correlation of
storm size with runoff. Statistical best-fit trend lines indicated a lin- Discharge
ear relationship for the asphalt roof (r 2 ¼ 0:99) and the stone roof For green roofs, the soil media and the drainage layer absorb rain
(r2 ¼ 0:60), whereas the green roof exhibited a best-fit power func- water, and runoff occurs after the soil is saturated and the drainage
tion relationship based on coefficient of correlation (r 2 ¼ 0:71). layer is full. This led to a delay for the runoff to occur, and a time
One use of these equations is to determine how the stone roof might delay is observed between when rainfall begins and when the green
have performed for the two larger events measured in September. roof exhibits measureable runoff. This difference varied from
Asphalt Roof Stone Roof Green Roof Linear (Asphalt Roof) Linear (Stone Roof) Power (Green Roof)
2.5
y = 0.9208x - 0.0374
R2 = 0.9886
2
Roof Run off (cm)
1.5
y = 0.47x - 0.0764
R2 = 0.6005
0.5
y = 0.0243x3.7455
R2 = 0.7104
0
0 0.5 1 1.5 2 2.5
Rainfall (cm)
Fig. 8. Runoff comparison for the three roofs for storms with less than 2.54 cm of total rainfall
several minutes to several hours depending on rainfall characteris- creased, as evidenced by higher standard deviation values for larger
tics and media saturation. For the stone-ballasted roof, the storm events. Similar trends were not evidenced by the stone roof,
rock layers provided roughness as well as storage area for the which reduced peak discharges by approximately 70% for all storm
storm water and exhibited response characteristics similar to event categories.
those of the green roof. In addition to reducing peak discharge, the green roof and the
For comparison purposes, the peak runoff discharge values stone roof also delay hydrologic response. Time to peak discharge
for each roof surface normalized to 100 m2 (Fig. 5) to account was computed on the basis of depth-weighted time for the asphalt
for the fact the roofs have different surface areas. The mean peak roof, green roof, and stone roof (Table 6). The mean peak delay
runoff discharge rate was 0:65 L=s=100 m2 for the asphalt roof, response time for the asphalt roof is 10.8 min (0.18 h) and varied
0:10 L=s=100 m2 for the green roof, and 0:14 L=s=100 m2 for from less than a minute to 24.6 min. The stone roof had a mean
the stone roof (Table 4). Using the asphalt roof peak discharge response time of 1.25 h and varied from 8.3 min to 3.36 h. Finally,
as the control roof, the green roof reduced peak discharges by the green roof had a mean response time of 2.16 h and exhibited
84.86% and the stone roof by 69.28%. However, it is worth noting variability similar to that of the stone roof, with response times
that the largest two storm events were not captured by the stone ranging from 10.2 min to 4.8 h. The variability of response time
roof. Similar to volumetric retention, the rainfall distribution also for the green roof and stone roof appeared to depend on rainfall
influenced the peak discharge reduction and the stone roof exhib- volume, with mean peak delay being less for large storm events
ited much higher variability than the green roof. when compared with small and medium-sized storm events
To determine the influence of storm size on peak discharge re- (Table 7).
duction, the analysis was repeated using the same storm size cat-
Water Quality
egories (Table 5). Using the asphalt roof as the control, the percent
reduction in peak discharge varied from 98.60% for small events to Water-quality concentrations were computed for runoff samples
52.70% for large events. The green roof also exhibited increased from five rainfall events (May 30, June 25, July 2, July 12, and
variability in peak discharge reduction as storm event size in- September 5) that exhibited rainfall greater than 1.27 cm in a
Table 5. Mean and Standard Deviation (SD) of Peak Discharge Reduction for All Three Roofs Based on Storm Size
Asphalt roof Green roof Stone roof
Peak (L=s=100 m2 ) Peak (L=s=100 m2 ) Percent reduction Peak (L=s=100 m2 ) Percent reduction
Size Category Mean SD Mean SD Mean SD Mean SD Mean SD
Small 0.42 0.28 0.01 0.01 98.60 1.03 0.10 0.06 68.24 16.95
Medium 0.77 0.28 0.07 0.08 90.24 12.18 0.19 0.11 70.64 20.59
Large 1.07 0.68 0.51 0.34 52.70 21.82 0.19 NAa 68.83 NAa
a
Stone roof values for large storm size category are based on a single storm as compared to three storms for the small- and medium-sized storms.
24 h period. The green roof exhibited a significantly higher total concentrations for the green roof were lower than those for the as-
solids concentration ( p ¼ 0:045) than the asphalt roof (Table 8), phalt roof and stone roof, but not at the p ¼ 0:05 significance level.
with the stone roof having the second-highest measured solids con- Finally, the stone roof exhibited the highest measured nitrate con-
centration for all five sampled runoff events. The total phosphate centration for all five events with the green roof runoff having
Table 7. Mean and Standard Deviation (SD) of Peak Discharge Delay for All Three Roofs Based on Storm Size
Asphalt roof Green roof Stone roof
Peak delay (h) Peak delay (h) Peak delay (h)
Size category Mean SD Mean SD Mean SD
Small 0.16 0.14 2.15 1.29 1.33 0.83
Medium 0.22 0.29 2.50 1.50 1.29 0.81
Large 0.15 0.15 1.27 1.23 0.20 NAa
a
Stone roof values for large storm size category are based on a single storm as compared to three storms for the small and medium sized storms.
Table 8. Mean and Standard Deviation (SD) of Concentration for Water-Quality Parameters
Asphalt roof Green roof Stone roof
Concentrations (mg=L) Mean SD Mean SD Mean SD
Total solids 108 55.52 190 51.84 152 60.17
Total phosphate 0.19 0.09 0.13 0.05 0.10 0.08
Nitrate 0.37 0.55 0.41 0.28 0.81 0.64
Table 9. Mean and Standard Deviation (SD) of Mass for Water-Quality Parameters
Asphalt roof Green roof Stone roof
Mass (g) Mean SD Mean SD Mean SD
Total solids 503.7 498.24 946.72 1,490.85 101.94 97.49
Total phosphate 0.88 0.83 0.63 0.99 0.05 0.03
Nitrate 0.87 0.96 0.69 1.11 0.31 0.30
and mass of total solids from the green roof are from the green roof
growing media. With regard to mass, the only other statistically
References
significant parameter was phosphate, with the asphalt roof and
green roof both exhibiting higher levels of phosphate mass than Berhage, R., Beattie, D., Jarrett, A., and O’Connor, T. (2007). “Greenroof
the stone roof. It is important to reiterate that the roofs were differ- runoff water quality.” Proc., 2007 Greening Rooftops for Sustainable
ent in area so a comparison of total mass can be misleading, but it is Communities Conf., Minneapolis.
included to demonstrate the influence of volumetric runoff on Buccola, N., Spolek, C., and Johnson, G. (2008). “A laboratory comparison
water-quality performance of roof surfaces. of green roof runoff water quality.” Proc., 2008 Low Impact Develop-
Sampling was limited to five storm events, which is not ment Conf., Seattle.
Carter, T. L., and Rasmussen, T. C. (2006). “Hydrologic behavior of
adequate to generalize the water-quality results on the whole,
vegetated roof.” J. Am. Water Resour. Assoc., 42(5), 1261–1274.
but the data demonstrates the potential of a green roof to reduce Dramstad, W. E., Olson, J. D., and Forman, R. T. T. (1996). Landscape
nutrients. Observed nitrate concentrations were highest from the ecology principles in landscape architecture and land-use planning,
stone roof, with the green roof and asphalt roof releasing similar Harvard Univ. Graduate School of Design, Island Press, Washington,
concentrations of nitrate. Mean mass values of total phosphate DC, 39.
and nitrate were lower for the green roof than for the asphalt roof. Huff, F., and Angel, J. (1992). Rainfall frequency atlas of the Midwest,
It might be possible that the green roof would show a decrease in Midwestern Climate Center and Illinois State Water Survey,
the nutrient concentrations if additional data were collected over a Washington, DC.
prolonged period (Hunt et al. 2006; Moran et al. 2005). In addition, Hunt, W., Hathaway, A., Smith, J., and Calabria, J. (2006). “Choosing the
right green roof media for water quality.” Proc., 2006 Greening Roof-
intermediate nonsampled smaller rainfall events would have
tops for Sustainable Communities, Boston.
flushed nutrients and sediment off the asphalt roof and stone roof Hutchinson, D., Abrams, P., Retzlaff, R., and Liptan, T. (2003).
for cases in which the green roof did not exhibit measureable run- “Stormwater monitoring of two ecoroofs in Portland, Oregon, USA.”
off. The capture of those events might influence the total mass dis- Proc., 1st North American Greenroof Conf., Chicago.
charged from the roofs. Johnson, R., Redding, K., and Holmquist, D. (2000). Water quality with
computers, Vernier Software and Technology, Beaverton, OR.
Kurtz, T. (2008). “Flow monitoring of three ecoroofs in Portland, Oregon.”
Conclusions Proc., 2008 Low Impact Development Conf., Seattle.
Kolb, W. (2004). “Good reasons for roof planting—Green roofs and
The goal of this research was to determine the effect of roof surface rainwater.” Int. Conf. on Urban Horticulture, Acta Horticulturae,
type on runoff retention, peak discharge attenuation, and water Waedenswil, Switzerland, 643, 295–300.
Liptan, T. (2003). “Planning, zoning and financial incentives for ecoroofs in
quality. Of specific interest was the performance of a full-scale
Portland, Oregon.” Proc., 1st North American Green Roof Conf., Chi-
established nonirrigated extensive green roof. Results indicated that cago, 113–120.
the green roof was highly efficient in capturing small storm events Moran, A., Hunt, B., and Jennings, G. (2004). “A North Carolina field
and was able to retain 68.25% of rainfall volume. The ability of the study to evaluate green roof runoff quantity, runoff quality, and plant
green roof to capture and slowly release rainfall runoff was very growth.” Proc., 2nd North American Green Roof Conf., Portland,
pronounced for events less than 2.54 cm, which historically repre- OR, 446–460.
sents more than 90% of all rainfall events in the northern temperate Moran, A. C., Hunt, W. F., and Smith, J. T. (2005). “Green roof hydrologic
climate of Michigan. The green roof reduced peak discharge by and water quality performance from two field studies in North Caro-
more than 90% for 17 of the 21 measured storm events and on aver- lina.” Proc., 2005 Watershed Management Conf., ASCE, Reston, VA.
National Oceanic and Atmospheric Association (NOAA). (2004). “Clima-
age, delayed the centroid of runoff mass by more than 2 h. The
tography of the United States, 1971–2000.” 〈www.ncdc.noaa.gov/oa/
stone roof also demonstrated the ability to mitigate storm-water climate/normals/usnormalshist.html〉.
runoff, capturing approximately 50% of the rainfall volume, exhib- Teemusk, A., and Mander, U. (2007). “Rainwater runoff quantity and qual-
iting a significant reduction in peak discharge, and creating a time ity performance from a greenroof: The effects of short-term events.”
delay of 1.2 h. Ecol. Eng., 30(3), 271–277.
Volumetric runoff coefficients were developed for each roof sur- U.S. EPA. (2003). “Protecting water quality from urban runoff.” 〈http://
face and will aid in modeling and designing future storm-water www.epa.gov/npdes/pubs/nps_urban-facts_final.pdf〉.
management systems that incorporate green or stone roofs. Volu- Van Seters, T., Rocha, L., and MacMillan, G. (2007). “Evaluation of the
runoff quantity and quality performance of an extensive greenroof in
metric runoff coefficients varied significantly based on rainfall vol-
Toronto, Ontario.” Proc., 2007 Greening Rooftops for Sustainable
ume, which is important when predicting a roof system’s response Communities Conf., Minneapolis.
to a design rainfall event. With regard to water quality, there was VanWoert, N. D., Rowe, D. B., Andresen, J. A., Rugh, C. L.,
significant variability between individual rainfall events, and the Fernandez, R. T., and Xiao, L. (2005). “Green roof stormwater reten-
nutrient removal and water-quality performance portion of the tion: Effects of roof surface, slope, and media depth.” J. Environ. Qual.,
investigation is inconclusive. However, the data provide some 34(3), 1036–1044.