Vous êtes sur la page 1sur 12

ENVIRONMENTAL ENGINEERING SCIENCE

Volume 26, Number 2, 2009


© Mary Ann Liebert, Inc.
DOI: 10.1089/ees.2007.0186

Storm Water Runoff Mitigation Using a Green Roof


Daniel J. Bliss,1 Ronald D. Neufeld,2 and Robert J. Ries3,*
1Department of Transportation, Commonwealth of Pennsylvania, Bridgeville, Pennsylvania.
2Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh, Pennsylvania.
3M.E. Rinker Sr. School of Building Construction, University of Florida, Gainesville, Florida.

Received: July 24, 2007 Accepted in revised form: May 8, 2008

Abstract

A prototype green roof was constructed and monitored in Pittsburgh, Pennsylvania, to demonstrate a way to
reduce storm water runoff and combined sewer overflow (CSO) events. The large impervious surface area cre-
ated by urban development in Pittsburgh creates a wet weather flow that the existing combined waste waster
and storm water sewer system cannot contend with. Green or vegetated roofs can reduce the amount of storm
water that reaches the sewer conveyance system by replacing an otherwise impervious roof with porous soil
that retains rainwater and plants that evapotranspire resulting in reduced storm water flows to the sewer sys-
tem. The prototype green roof reduced the runoff volume by up to 70% compared to a conventionally ballasted
roof covering a control portion of the same building. A reduction of at least 20% was found for rainfalls of 1.5
cm (0.60 inches) or less. The green roof also reduced the flow rate of runoff throughout storms. Peak values
from the green roof were between 5% and 70% lower than control roof flow rates. For some small storms, the
time of peak flow rate was delayed by up to several hours. In general, the green roof delayed the start of runoff
and extended the time period of low residual flows that existed at the end of a storm. Water quality tests in-
dicate that in most cases for the storms observed, a first-flush phenomenon was not evident in green roof runoff
samples. Levels of phosphorus and chemical oxygen demand (COD) were elevated by the green roof. Both the
control and green roofs produced runoff that demonstrated neutralization of slightly acidic rainfall.

Key words: storm water runoff mitigation, green roofs, combined sewer overflow, peak flow

Introduction depth. An intensive roof is similar to a roof-top park, usu-


ally containing large trees, shrubs and, at times, paved walk-

T HERE IS A NEED for appropriate technologies for urban


centers for managing storm water runoff that may lead
to sewer overflows. In particular, in cities such as Pittsburgh,
ways. Many intensive roofs are designed as usable green
space. For this configuration, a thick soil layer (usually a foot
or more) is required to support the large plants. The weight
Pennsylvania, that have a combined sewer system, precipi- of the soil and plants is substantial. Because people will fre-
tation events often result in combined storm and sewer wa- quently walk on the roof, a live load must be considered in
ter overflows containing diluted sewage to local rivers. design. Structural support can be an issue and plant main-
Green roofs are an emerging technology that provide tenance is often significant.
many positive environmental benefits, especially in attenu- Extensive roofs, on the other hand, are relatively light-
ation of storm water flows. The soil and plants on a green weight and low maintenance. The vegetation on an exten-
roof can reduce storm water runoff volume and peak runoff sive green roof is typically ground cover such as sedums,
flows and may delay the time to peak runoff flows by re- which lie low to the ground, require little maintenance, and
taining and partially attenuating the volume of water dur- are selected such that they can survive on the rainfall in a
ing a rainstorm event. given area. The soil layer is typically 3–5 inches thick, so the
There are two types of green roofs, intensive and exten- load on the structure is significantly reduced. An extensive
sive, which are differentiated by use, plant type, and soil green roof is built for aesthetics and environmental benefits,
and cannot be used for recreation.
A green roof is a complex layered structure. A water-
*Corresponding author: M. E. Rinker, Sr. School of Building Con- proofing membrane sits immediately on top of the structural
struction, University of Florida, 311 Rinker Hall, Gainesville, FL roof deck to prevent moisture from entering the building.
32611. Phone: 352-273-1155, Fax: 352-392-9606, E-mail: rries@ufl.edu Typically, above this membrane is a root barrier layer that

407
408 BLISS ET AL.

is designed to prevent roots from penetrating the water- stance, compared an extensive green roof to a bituminous
proofing membrane and the structural roof. A drainage layer roof and found that the green roof reduced runoff volume by
is next. The drainage layer typically has the appearance of a 54% and 23%, respectively, while also showing a reduced
perforated egg crate and is designed to carry excess runoff peak flow rate and a delay in green roof storm water runoff.
to roof drains, and to store water for the plants in dry peri- Research in Portland, Oregon (Liptan, 2000; Hutchinson et al.,
ods. Next, a filter fabric is installed to prevent soil from wash- 2003), Chicago (Green roof test Plot, 2004), North Carolina
ing away and compromising the drainage layer as water (Moran et al., 2004a, 2004b), Germany (Kohler et al., 2001) and
drains from the roof. Finally, the growing plants and asso- at Penn State University (DeNardo, 2003) showed similar re-
ciated substrate or growing medium completes the green sults, including up to an 87% reduction in peak flow rate and
roof. The substrate is often a lightweight synthetic soil that up to a 75% retention of rainwater by the green roofs.
is porous and inherently inert, with nutrients added for plant Water quality has also shown to be improved by green
growth. roofs in previous research. Kohler and Schmidt (2004) found
The storm water benefits of green roofs may be important that lead, cadmium, nitrate, phosphorous, and turbidity lev-
to the Pittsburgh region because of combined sewer over- els in green roof runoff were reduced by between 68% and
flows (CSOs). It has been reported that Pennsylvania has the 95%, respectively, compared to runoff from roofs made of
highest concentrations of sewer overflows of any state in the conventional materials and that green roofs neutralized acid
United States (Gilbert et al., 2005) and Allegheny County has rain. This study also found a first flush was evident in green
one of the highest concentrations in the United States. There roof runoff samples. First flush is defined as an initially high
are 265 CSOs and 52 sanitary sewer overflows (SSOs) in the pollutant discharge in runoff due to a build up on an ex-
county (3 Rivers Wet Weather Demonstration Program, posed surface that is washed away by rainfall runoff (New
2006). The Allegheny County Sanitary Authority (AL- South Wales Department of Environment and Climate
COSAN) is in the process of a 20-year, multibillion-dollar Change, 2007). A second study (Schueler, 2000) found that a
project to expand treatment facilities to make overflow first flush occurred in zinc, copper, and lead levels in runoff
events less likely (Gilbert et al., 2005). Green roofs offer an- from a conventional galvanized roof. However, a consensus
other technical approach to combat the sewer overflow cannot be drawn about phosphorus or nitrogen levels in
problem. The porous substrate absorbs water and replaces green roof runoff. Studies in North Carolina (Moran et al.,
an otherwise impervious roof. The plants also allow evapo- 2004a, 2004b) and Portland (Liptan and Murase, 2000;
transpiration to occur on the roof. By introducing these Huchinson et al., 2003) showed that phosphorus was pres-
features, storm runoff is prevented from entering the sewer ent in green roof runoff in greater amounts than control roof
system. This can help to reduce the number of sewer over- runoff. Similarly, Moran found that nitrogen concentrations
flow events and in turn improve the water quality of the were not significantly different in green and control roof
region’s three principal rivers. samples.
A number of previous studies have indicated that green The objectives of this research study were to investigate
roofs have improved storm water conditions. Two studies in the effects of an extensive green roof in Pittsburgh, Penn-
Ottawa, Ontario (Liu 2003; Bass and Baskaran, 2003), for in- sylvania, on storm water runoff quantities and selected wa-

FIG. 1. Extensive green roof on a commercial building in Pittsburgh approximately 3 months after installation.
GREEN ROOF 409

FIG. 2. Control roof adjacent to the green roof on a commercial building in Pittsburgh.

ter quality characteristics. A green roof and an adjacent con- drain entered the sewer system, a Tracom 60-degree Extra
trol roof on a commercial building were instrumented and Large Trapezoidal Flume was installed. A flume is a chan-
monitored for storm water runoff quantity, duration, and nel with a trapezoidal area where the depth of water was
water quality and data from 13 storms were collected. measured with a Greyline LIT25 ultrasonic sensor. The ul-
trasonic sensors were suspended above the designated mea-
Materials and Methods surement point in the flume. The sensors take a depth mea-
surement by bouncing a sound wave off of the surface of the
An 1150 square meter (12,300 square foot) extensive green
water. The flow rate was calculated from the sensor mea-
roof (Fig. 1) was constructed over the new portion of an ex-
surement with calibration data provided by the manufac-
panded supermarket in Pittsburgh. The roof was a hot-
turer. A cumulative volume of runoff for each roof was cal-
mopped bitumen built-up membrane with root barrier,
culated from the flow rate values. The equipment used on
drainage, and filter fabric layers beneath a 14-centimeter (5.5
both roofs was calibrated by their respective manufacturers
inch) thick synthetic growing medium, made primarily of
before they were installed on the roof.
expanded shale, perlite, and coconut husk. The plants are
A National Instruments Fieldpoint Datalogging system
largely sedums. Adjacent to the green roof was a 1950 square
was installed on the roof. The Fieldpoint consists of a series
meter (21,000 square foot) conventional ballasted membrane
of modules for different types of inputs from the monitor-
roof (Fig. 2), separated from the green roof by a parapet wall.
ing instruments and a network module connected via Eth-
This served as the control for all measurements.
To measure the effect of an extensive green roof on man-
agement of storm water runoff, a number of pieces of equip- TABLE 1. TESTING PROCEDURES USED FOR THE WATER
ment were installed on both the green and control roofs to QUALITY PARAMETERS IN THE STUDY
measure water quantities and a system was put in place to
collect water samples for laboratory testing. Water quality
A Hydrologic Services RG703 8 inch (2 cm) tipping bucket parameter Test procedure
rain gauge was installed on the control roof to measure the
Sulfate Hach Method 8051a
depth and calculate the intensity of rainfall. Two Campbell Phosphorus Hach Method 8048a
Scientific SM616 Soil Water Content Sensors were placed on COD Hach Method 8000a
opposite sides of the green roof. The sensors measured the Total Nitrogen Hach Method 10071a
volumetric water content of the soil at 5-min intervals. This Lead Standard Methods Section 3111b
allowed for monitoring of the amount of water in the soil Zinc Standard Methods Section 3111b
before, during, and after a rainstorm. Cadmium Standard Methods Section 3111b
One roof drain from both the green and control roof was Turbidity Standard Methods Section 2130 Bb
separated from the remainder of the roof drainage system so aHach Water Analysis Handbook, 4th ed., Revision 2. (2003).
that runoff could be monitored. The surface area draining bStandard Methods for the Examination of Water and Wastewater, 18th
into each drain was equal, and measured approximately 330 ed. (1992).
square meters (3530 square feet). Before the runoff from each COD, chemical oxygen demand.
410 BLISS ET AL.

TABLE 2. SUMMARY OF RAINFALL DATA, MAXIMUM FLOW RATE, AND TOTAL STORM WATER
RUNOFF VOLUME FOR STORMS BETWEEN AUGUST 27, 2006 AND JANUARY 15, 2007

Rainfall
Maximum flow rate (m3/s) Total volume (m2)
Depth Length
Date (in) (HM) Green Control Reduction Green Control Reduction

8/27/06 0.59 1525 0.0011 0.0013 21% 10.83 13.68 21%


10/17/06 1.97 2341 0.0012 0.0017 29% 38.49 47.76 19%
10/19/06 1.73 1631 0.00085 0.0013 32% 41.72 43.74 5%
10/27/06 1.2 2347 0.00051 0.00061 16% 26.19 31.97 18%
10/31/06 0.21 958 0.000028 0.000099 71% 1.04 3.31 69%
11/1/06 0.07 208 0.000014 0.000025 38% 0.32 0.94 66%
11/11/06 0.57 1513 0.00046 0.00067 30% 8.75 12.48 30%
11/15/06 1.5 1940 0.00061 0.00098 38% 31.06 38.02 18%
11/19/06 0.17 1110 0.000020 0.000059 65% 0.50 1.55 68%
12/1/06 0.83 647 0.00082 0.0011 28% 6.85 8.68 21%
1/5/07 0.69 2024 0.00023 0.00024 5% 14.98 18.37 19%
1/8/07 0.48 1647 0.00018 0.00026 28% 8.85 11.24 21%
1/12–1/15/07 2.2 3 days, 0.00034 0.00033 5% 52.87 56.91 7%
349

ernet through a digital subscriber line (DSL). Software on a dures utilized. Approximately half of the parameters utilized
computer at the University of Pittsburgh campus remotely spectrophotometry as the method of detection. Metals were
controlled the equipment, and recorded and displayed data measured using atomic absorption spectrometry (AAS). A
through a series of programs written in the LabVIEW pro- portion of each sample was filtered through a 0.45-m mem-
gramming language. Data were also displayed on websites brane filter to determine both soluble and particulate phases
in real time. of each constituent. pH and turbidity levels were determined
A series of six solenoid valves connected to the stormwa- with pieces of equipment specifically designed for those pa-
ter drain after each flume were used to collect grab samples rameters: an Oakton pHTestr 30 and HACH Model 2100A
of runoff. The valves were controlled by the Fieldpoint data Turbidimeter, respectively.
logger and LabVIEW computer program. The valves were A 3- to 5-point calibration curve over the measurement
programmed to open in series, after predetermined cumu- range was calculated before each set of spectrophotometry
lative amounts of total runoff had been reached. Sampling or AAS tests. Calibration standards for pH and turbidity tests
at regular intervals throughout a storm allowed for first- were also measured with each set of samples.
flush effects to be studied. The grab samples were taken at
the same volumetric flow point on the green and control Results and Discussion
roofs and chemically analyzed at the University of Pitts-
Water quantity
burgh’s Environmental Engineering laboratories. Samples
were collected after 5, 10, 15, 30, 45, and 60 cubic feet of runoff Data were obtained from 13 storms in the Pittsburgh area
had flowed from each roof. from August 2006 through January 2007. Complete infor-
The water samples were analyzed for a number of chem- mation is available from Bliss (2007). Table 2 summarizes
ical parameters. Table 1 summarizes the specific test proce- data for each storm. Two sample storms will be discussed

0.07 2.5
0.06
2
0.05
Rainfall (in)

0.04 1.5
Green Roof
0.03 Control Roof
1
Rain
0.02
0.5
0.01
0 0
0:00 4:00 8:00 12:00 16:00
Duration (H:M)

FIG. 3. Runoff flow rates for the green and control roofs and cumulative rainfall for the October 17, 2006 storm.
GREEN ROOF 411

0.004 0.2
Green Roof 0.18
0.0035
Control Roof
0.16
0.003 Rain
0.14

Rainfall (in)
0.0025 0.12
0.002 0.1

0.0015 0.08
0.06
0.001
0.04
0.0005 0.02

0 0
0:00 2:00 4:00 6:00 8:00 10:00 12:00
Duration (H:M)

FIG. 4. Runoff flow rates for the green and control roofs and cumulative rainfall for the October 31, 2006 storm.

below to illustrate the effect of green roofs on stormwater than the control roof. The reduction in peak flow dropped
runoff. The first storm occurred on October 17, 2006, where to 52% for the second. This performance was typical for the
5 cm (1.94 inches) of rain fell on the roof in approximately observed storms.
10 h (0.5 cm [0.19 inches] per hour average intensity). This The green roof also significantly attenuated the maximum
storm was one of the heaviest observed during the study pe- flow rate for each storm. For the October 17, 2006 storm, the
riod. The second storm, on October 31, was much lighter maximum flow rate of the green roof was 29% lower than
with 0.5 cm (0.19 inches) of rain in 5.25 h (0.1 cm [0.04 inches] the control roof. For the lower rainfall intensity October 31,
per hour average intensity). 2006 storm, the maximum flow rate of the green roof was
The results show that the rate of runoff from the green 70% lower than the control. For all storms during the mon-
roof was substantially lower than that of the control roof, es- itoring period, the maximum flow rate reduction ranged
pecially in the early in the storm. As the storm progressed from 5% to 70%.
and the green roof substrate reached saturation, the mea- Figures 4 and 5 show the delays and extensions of runoff
sured flow rates from both the green and control roofs con- during precipitation events. There was usually a delay be-
verge. For instance, after 0.6 cm (0.25 inches) of rain fell on tween the time rainfall started and runoff began to flow on
October 17 (Fig. 3), the green roof flow rate was 88% lower the green roof, and runoff usually began to flow from the
than the control roof. By the end of the storm, the flow rates control roof first. This was due to the time that it takes for
were nearly identical. A similar pattern occurred on October the soil to become saturated. For instance, the control roof
31. Figure 4 shows two peaks in runoff flow caused by produced runoff 16 min before the green roof during the Oc-
changes in rainfall intensity during the course of the storm. tober 17 storm. At the time runoff began to flow from the
During the first, the green roof peak flow rate was 85% lower green roof, it had been raining for just over 3 h. Similarly,

2.5

2.0

1.5

1.0
Green Roof
Control Roof
Rainfall
0.5

0.0
0:00 4:00 8:00 12:00 16:00
Duration (H:M)

FIG. 5. Cumulative rainfall and runoff volume for the green and control roofs shown as equivalent rainfall depth for the
October 17, 2006 storm.
412 BLISS ET AL.

0.20 Green Roof


0.18 Control Roof
0.16
Rainfall
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
0:00 2:00 4:00 6:00 8:00 10:00 12:00
Duration (H:M)

FIG. 6. Cumulative rainfall and runoff volume for the green and control roofs shown as equivalent rainfall depth for the
October 31, 2006 storm.

there was a 26-min delay between the control and green roofs the flow from both roofs stopped at approximately the same
during the October 31 storm and runoff began to flow from time.
the green roof 3 h after the storm began. While there were There were differences in the timing and magnitude of the
instances where the green roof produced runoff first, this is peak flow rates from the green and control roofs. The dif-
the exception to the rule and these instances were usually ferences in timing depended on rainfall intensity, while the
associated with irrigation. Shortly after the construction was differences in magnitude were independent of intensity. The
completed on the green roof, it was regularly irrigated. At peak flow rate of the green roof occurred later in the storm
times, the roof was even watered during rainstorms. The soil compared to the control roof during storms with low inten-
was often saturated during this period, which affected the sity rainfall. The green roof runoff was at its peak value ap-
performance of the roof. proximately two hours after the control roof on October 31.
Generally, runoff continued to flow from the green roof During the October 17 storm, which was of greater intensity,
for an extended period of time after the conclusion of the there was no delay. Both peaks occurred at virtually the same
storm and after runoff has stopped flowing from the control time. In both low and high intensity storms, however, the
roof. This delay was because water entering the green roof magnitude of the green roof peak flow rate was lower.
must filter through the soil, drainage and filter layers before Total volume of runoff is an important parameter when
it can exit the roof as runoff. During the October 17 storm, measuring green roof performance. The green roof consis-
runoff flowed from the green roof for 4 h and 45 min longer tently reduced the volume of runoff. Figures 5 and 6 show
than the control roof. However, during low rainfall intensity the cumulative runoff volumes from the green and control
and duration storms that do not exceed the water retention roofs as equivalent inches of rainfall, and the cumulative
capacity of the growing medium, e.g., the October 31 storm, rainfall over time for the October 17 and October 31, 2006

100% 2.5

80% 2
Rainfall (in)

60% 1.5

Green/Control
40% Rainfall 1

20% 0.5

0% 0
0:00 4:00 8:00 12:00 16:00
Duration (H:M)

FIG. 7. Cumulative rainfall and runoff volume reduction over time for the October 17, 2006 storm.
GREEN ROOF 413

100% 0.2
0.18
80% 0.16
0.14

Rainfall (in)
60% 0.12
0.1
40% 0.08
Green/Control
0.06
Rainfall
20% 0.04
0.02
0% 0
0:00 2:00 4:00 6:00 8:00 10:00 12:00
Duration (H:M)

FIG. 8. Cumulative rainfall and runoff volume reduction over time for the October 31, 2006 storm.

storms. The runoff volume was normalized by dividing the reduction was calculated using the following equation,
total volume of runoff from each roof by the surface area of where VGreen and VControl were the total runoff volume from
the portion of the roof being monitored. Equation 1 shows the green and control roofs, respectively.
the calculation, where D is the equivalent inches of rainfall,
 
VGreen
SA is the surface area of the roof and V is the total runoff Runoff reduction  1   (2)
VControl
volume.
For both storms, the reduction was initially 100% during
 
V
D  (1) the period when the control roof was producing runoff, but
SA
the green roof was not. The October 31 storm declined at a
Of the 4.93 cm (1.94 inches) of rain that fell on October 17, more constant and gradual rate compared to the October 17
3.6 cm (1.41 inches) became runoff on the green roof and 4.4 storm because of the lower rainfall intensity, lower total rain-
cm (1.75 inches) became runoff on the control. This 0.9-cm fall, and the lack of extension of flow from the green roof dur-
(0.35-inch) difference (19% reduction) means that the green ing this storm. The October 17 storm was typical of one of the
roof reduced the runoff volume by approximately 9.85 m2 larger observed storms, with a final reduction of approxi-
(9850 L or 2600 gallons). If the control roof was covered with mately 20%. Likewise, the reduction of approximately 70% in
a green roof this would increase to approximately 26.9 m2 total runoff volume from the October 31 storm was similar to
(26,875 L or 7100 gallons). For the October 31 storm, 0.5 cm storms with lower total rainfall. These graphs also indicate
(0.19 inches) of rain fell on that date. Only 0.1 cm (0.04 inches) that the performance of the green roof was similar at the same
became runoff on the green roof, while 0.3 cm (0.12 inches) cumulative rainfall point in different storms. There was ap-
became runoff from the control roof. proximately a 90% reduction in runoff volume after 0.5 cm
The runoff reduction over the course of the storm is plot- (0.19 inches) of rain fell on October 17. At that same relative
ted in Figure 7 (October 17) and Figure 8 (October 31). Runoff point on October 31, there was a reduction of 85%.

FIG. 9. Runoff volume reduction versus storm duration for all storms observed during the monitoring period.
414 BLISS ET AL.

FIG. 10. Runoff volume reduction versus rainfall depth for all storms observed during the monitoring period.

Several other observations can be made by looking at the change in water content during the October 17 and 31 storms.
three summary graphs showing results from all storms dur- During the October 17 storm, the water content was at 16%
ing the monitoring period (Figures 9, 10, and 11). First, in by volume just prior to the start of the storm and peaked at
Figure 9, the duration of a storm had an effect on the over- approximately 20% during the storm. On October 31, the wa-
all performance of a green roof. Shorter storms reduced the ter content started at 19% and peaked at 28%. The water con-
runoff volume by greater amounts than longer storms. This tent declined after both storms, but declined more gradually
is because the majority of the storms during the monitoring after the October 31 storm. Since the total rainfall in this
period were of a similar intensity (Fig. 11). Figure 10 shows storm was much less, an extension of flow from the green
that storms with lower total rainfall had the greatest runoff roof did not occur.
reduction. Figure 10 also shows the range of performance The soil matrix has a finite moisture saturation value and
observed for the green roof. For a storm of 1.5 cm (0.60 drier soil retains significantly more water than wet soil. For
inches) or less, a reduction in runoff volume of at least 20% instance, when there is approximately 25% water by volume
could be expected. This accounted for approximately 75% of in the soil prior to the start of a storm, only 0.1 cm (0.04 inches)
the storms that occurred during the observation period. of rain were absorbed. This is near saturation (approximately
However, a strong relationship between intensity and runoff 30% by volume). When the water content was approximately
reduction (Fig. 11) was not found. A wide range of runoff 15%, however, the green roof was able to absorb about 0.5 cm
reductions occurred for storms of similar intensities. (0.20 inches) of rain before producing runoff.
The volumetric water content in the green roof synthetic As shown by the effect of the volumetric water content on
soil also had an affect on performance. Figure 12 shows the the runoff volumes, most of the water that did not become

FIG. 11. Storm intensity versus storm water runoff reduction for all storms observed during the monitoring period.
GREEN ROOF 415

27

Oct. 17, 2005 25


Oct. 31, 2005

23

21

19

17

15
-12:00 -8:00 -3:59 0:00 4:00 8:00 12:00 16:00 20:00
Duration (H:M)

FIG. 12. Soil moisture content versus storm duration for the October 17 and 31, 2006 storms.

runoff was absorbed by the soil. Evapotranspiration would (COD), and heavy metals. Tables 3 and 4 summarize test re-
have accounted for some of this volume of water as well, but sults from the October 17, 2006 and December 1, 2006 storms.
it could not be estimated with the equipment on the roof. Phosphorous and COD showed the most consistent pat-
Likewise, the drainage layer of the green roof is assumed to tern. Both were found to be elevated on the green roof rela-
have retained water, but both parameters are assumed to tive to the control roof. Phosphorus levels in green roof
have been insignificant compared to the retention by the soil. runoff samples were in the range of 2 to 3 mg/L, while rain-
water and control roof runoff samples had only negligible,
if any, phosphorus. COD levels varied by storm, but the
Water quality
green roof levels were consistently higher than the control
Water quality tests were performed on a select number of roof.
storms (including the previously discussed October 17, 2006 Turbidity, measured in nephelometric turbidity units
storm) for several parameters. These included turbidity, pH, (NTU), was higher in runoff from both roofs than from rain-
phosphorus, sulfate, nitrogen, Chemical Oxygen Demand fall (1.5 NTU). Control roof samples were also consistently

TABLE 3. WATER QUALITY RESULTS FOR THE UNFILTERED SAMPLES FROM THE DECEMBER 1, 2006 STORM

Concentration (mg/L)

10 20 30 45 60
Parameter 5 cf cf cf cf cf cf

Phosphorus Green 2.1 0.8 2.0 1.4 2.1 2.1


Control 0.1 0.1 0.0 0.0 0.1 0.0
Rain —
Sulfate Green 27.6 30.0 34.8 28.8 28.2 28.2
Control 30.0 27.0 24.0 23.4 18.0 16.8
Rain 4.2
Nitrogen Green 0.0 0.0 0.0 0.0 0.0 0.0
Control 1.7 1.7 0.0 0.0 3.4 3.4
Rain 0.1
COD Green 31.7 40.9 40.9 30.8 26.2 26.2
Control 15.1 11.5 9.6 16.1 15.1 4.8
Rain 12.4
pH Green 8.0 7.9 7.9 7.9 8.0 8.0
Control 8.3 8.0 7.7 7.9 7.8 7.8
Rain 6.4
Turbidity Green 5.5 5.0 4.5 7.3 8.2 9.2
Control 72.0 44.0 34.0 26.0 20.0 9.4
Rain 1.3

Note: Phosphorus concentrations are from 10/17/06 storm.


416 BLISS ET AL.

TABLE 4. WATER QUALITY RESULTS FOR THE FILTERED SAMPLES FOR THE DECEMBER 1, 2006 STORM

Concentration (mg/L)

10 20 30 45 60
Parameter 5 cf cf cf cf cf cf

Phosphorus Green 2.10 — — — 2.0 —


Control 0.00 0.0 0.0 0.0 0.0 0.00
Rain —
Sulfate Green 26.40 — — 25.2 27.0 35.40
Control 20.40 19.8 20.4 20.4 15.6 15.60
Rain 8.40
Nitrogen Green 0.00 — — 0.0 0.0 0.00
Control 0.00 0.0 0.7 0.7 3.4 0.70
Rain 1.20
COD Green 22.50 — — 26.2 22.5 27.10
Control 19.70 5.0 15.1 7.8 15.1 7.80
Rain 1.30
Lead Green 0.20 — — — — —
Control — — — — — 0.20
Rain 0.40
Cadmium Green 0.00 — — — — —
Control — — — — — 0.00
Rain 0.00
Zinc Green 0.02 — — — — —
Control — — — — — 0.08
Rain 0.02

Note: Phosphorus concentrations are from 10/17/06 storm.

more turbid than the green roof runoff samples. The green represents the dissolved sulfate, so the test results indicate
roof runoff showed no signs of a first flush and demonstrated that the first flush was a result of increased particulate-bound
turbidities of 10 NTU or less. Soil particles would be expected sulfate that built up on the roof during the dry period of
to be washed out in the runoff especially during the early nearly 2 weeks prior to the storm.
phases of the green roof’s life. This was observed and some pH levels from both roofs were consistently slightly basic,
of the soil was deposited in the green roof flume during the with reading ranging from 7.5 to 8.3. Rain samples were
first few of months after construction was completed. As a slightly acidic, with a pH of 6 to 7. It appears that both the
consequence, the sampling and water quality data collection control and green roofs had the same effect of mitigating
was not initiated until 4 months after the green roof was fin- mild acid rain. Total nitrogen was measured using the per-
ished. By this point, the roof appears to have stabilized and sulfate digestion technique of Standard Methods for the Exam-
soil was no longer present in the runoff in significant ination of Water and Wastewater (1992). There was no mea-
amounts. sureable “total nitrogen” in runoff from either the green or
Control roof samples did show a first flush during some control roofs.
storms. The December 1, 2006 storm, a particularly intense Metal testing was conducted for zinc, cadmium and lead
storm after an 11-day period of no rainfall, demonstrated the on a limited number of samples using atomic absorption
most significant first flush effect. In this case, the first one spectroscopy. No detectible concentrations of cadmium were
third of an inch of rain fell on the roof in approximately 15 found in samples from any source. Zinc was found on oc-
min. It is hypothesized that the dry period preceding the casion, but in fairly low levels. The maximum level of zinc
storm allowed airborne particulates to accumulate on the was about 0.45 mg/L in one green roof sample, however, the
roof which were then dislodged and transported by the in- majority of samples had concentrations on the order of 0.10
tense rainfall at the onset of the storm. Turbidity levels mg/L with concentrations detected at about the same level
steadily declined from the first sample at approximately 70 in the green roof runoff, control roof runoff, and rainfall sam-
NTU to 10 NTU on the final sample. Corresponding to the ples. Likewise, lead concentrations were approximately 0.10
turbidity first-flush effect on the control roof for the De- mg/L for samples from all three sources and no clear pat-
cember 1 storm, there was a first flush effect detected in sul- tern was discernable.
fate. The unfiltered control roof samples showed a steady
decline in sulfate concentrations from 30 mg/L to about 15
Conclusions
mg/L. Vacuum filtering was also carried out on the samples
using a 0.45 m (nominal pore size) cellulose membrane. Sul- Thirteen storms were observed and field measurements
fate levels in the filtered samples were reduced, to a range recorded on an extensive green and a conventional control
of 20 mg/L to approximately 15 mg/L. A first flush was not roof from August 2006 through January 2007. The green roof
evident here, except for the final two samples, which had the outperformed the control roof in several key areas: total
same levels as the unfiltered samples. The filtered portion storm water runoff volumes were reduced; the runoff flow
GREEN ROOF 417

rate was consistently lower; and peak flow rates were de- Bass, B., and Baskaran, B. (2003). Evaluating Rooftop and Vertical
layed. Results also show that the green roof affected the wa- Gardens as an Adaptation Strategy for Urban Areas. National Re-
ter quality of runoff. search Council Canada publication NRCC-46737.
The green roof reduced total volume of runoff produced Bliss, D. J. (2007). Stormwater runoff mitigation and water quality
during a storm by as much as 70%. Light storms and events improvements through the use of a green roof in Pittsburgh, PA.
with a short duration generally had the greatest reduction M.S. Thesis, University of Pittsburgh.
in storm water runoff volume. A storm of 1.5 cm (0.60 inches) DeNardo, J.C. (2003). Green roof mitigation of stormwater and en-
or less reduced runoff volume of the green roof by at least ergy usage. M.S. Thesis, Pennsylvania State University.
Gilbert, J.B., Hill, B.J., Lauria, J.M., Logsdon, G.S., Mccarty, P.L.,
20% compared to the control roof. The data from these 13
Miller, P., Moreau, D.H., Moyer, N.P., Platt, R.H., Schwartz,
storms did not show a clear relationship between storm in-
S.S., Shortle, J.S., Tarr, J.A., Vanbriesen, J.M., and
tensity and volume reduction.
Ziemkiewicz, P.F. (2005). Regional Cooperation for Water Qual-
Water that did not exit the green roof as runoff was either
ity Improvement in Southwestern Pennsylvania. Washington, DC:
absorbed by the growing medium and roof materials, tran- National Academies Press.
spired by the plants or evaporated into the atmosphere. The Hach Company. (2003). Hach Water Analysis Handbook. 4th ed.,
moisture content of the growing medium was measured. Revision 2. Loveland, CO: Hach Company.
During the observed storms, it was consistently found that Hutchinson, D., Abrams, P., Retzlaff, R., and Liptan, T. (2003).
the time that it takes for the growing medium to become sat- Stormwater monitoring two ecoroofs in Portland, Oregon,
urated delays the time that runoff begins to flow from the USA. Proceedings of Greening Rooftops for Sustainable Com-
green roof. Once the soil is the saturated the time required munities. Chicago, 2003.
for water to filter through the growing medium and roof ma- Green Roof Test Plot: 2003 End of Year Project Summary Report.
terials also reduced the flow rate throughout the storm. This (2004). Prepared by MWH for the City of Chicago Department
resulted in a reduced and delayed peak flow rate for the of Environment.
green roof compared to the control roof. Kohler, M., Schmidt, M., Grimme, F.W., Laar, M., and Gusmao,
Based on the water quality tests completed on runoff sam- F. (2001). Urban Water Retention by Greened Roofs in Tem-
ples collected from the site, a first flush was not evident in perate and Tropical Climate. Proc. 38th World Congress IFLA,
the green roof runoff. Some water samples from the control Singapore, U124-133.
roof indicated a first flush in quantities of turbidity and sul- Kohler, M., and Schmidt, M. (2004). Study of extensive ‘green
fate. Comparing the two roofs, the green roof had elevated roofs’ in Berlin. www.roofscapes.com/water_quality.
levels of phosphorous, which may have been caused by fer- Liptan, T., and Murase, R.K. (2000). Watergardens as Stormwa-
ter Infrasturcture in Portland, Oregon. Water Sensitive Eco-
tilizer applied during the first growing season. The COD of
logical Planning and Design Symposium, February 25–26,
green roof water samples was also greater than the control
2000.
roof. Turbidity levels, however, were found to be consis-
Lui, K. (2003). Engineering Performance of Rooftop Gardens
tently higher in control roof runoff. Both roofs neutralized
Through Field Evaluation. Proceedings of RCI 18th Interna-
the slightly acidic rain of the region. The green roof appeared tional Convention and Trade Show. Tampa, FL: March 13–18,
to have had no effect on sulfate, nitrogen, or heavy metal 2003, p. 1–15.
levels. Moran, A., Hunt, B., and Jennings, G. (2004a). A North Carolina
field study to evaluate greenroof runoff quantity, runoff qual-
Author Disclosure Statement ity, and plant growth. 2004 Green Roofs for Sustainable Com-
The authors state that no competing financial interests ex- munities Conference.
Moran, A., Hunt, B., and Jennings, G. (2004b). Greenroof re-
ist.
search of stormwater runoff quantity and quality in North
Carolina. NWQEP Notes, Number 114, August 2004.
References
New South Wales Department of Environment and Climate
3 Rivers Wet Weather Demonstration Program. (2006). Under- Change (NSW EPA). (2007). Stormwater first flush pollution.
standing the Collection System. www.3riverswetweather. www.environment.nsw.gov.au/mao/stormwater.htm.
com. Schueler, T.R. (2000). Is rooftop runoff really clean? In: The
American Public Health Association (1992) Standard Methods for Practice of Watershed Protection. T.R. Schueler and H.K. Hol-
the Examination of Water and Wastewater, 18th ed. Washington, land, Eds. Ellicott City, MD: Center for Watershed Protec-
DC: American Public Health Association. tion. p. 41.

Vous aimerez peut-être aussi