BookPDF Available

Green Roofs for Stormwater Runoff Control

Authors:

Abstract and Figures

This project evaluated green roofs as a stormwater management tool. Specifically, runoff quantity and quality from green and flat asphalt roofs were compared. Evapotranspiration from planted green roofs and evaporation from unplanted media roofs were also compared. The influence of media type, media depth and drought during plant establishment on plant growth and long-term management of media pH were investigated. The goal of the project was to provide high-quality replicated data which could be used to develop and refine reliable anticipated runoff volumes and loadings from green roofs, respectively, as well as evaluate factors which impact plant growth and establishment. Results indicate that the green roofs are capable of removing 50% of the annual rainfall volume from a roof through retention and evapotranspiration. Rainfall not retained by green roofs is detained, effectively increasing the time to peak, and slowing peak flows for a watershed. There are seasonal considerations as more runoff is generated during winter and for many summer storms there was no runoff. Green roof runoff does contain concentrations of some nutrients and other parameters, but values are in line with other planted systems. Due to the volume reduction, actual nutrient loadings from green roofs are less than asphalt roofing runoff or otherwise manageable at the downspout.
Content may be subject to copyright.
EPA/600/R-09/026 | February 2009 | www.epa.gov/ord
Green Roofs for Stormwater
Runoff Control
Office of Research and Development
National Risk Management Research Laboratory - Water Supply and Water Resources Division
EPA/600/R-09/026
February, 2009
Green Roofs for Stormwater
Runoff Control
By
Robert D. Berghage
David Beattie
Albert. R. Jarrett
Christine Thuring
Farzaneh Razaei
The Pennsylvania State University
University Park, PA 16802
Thomas P. O’Connor
Urban Watershed Management Branch
Water Supply and Water Resources Division
National Risk Management Research Laboratory
Edison, NJ 08837
Cooperative Agreement No. 83124401
NATIONAL RISK MANAGEMENT RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OH 45268
Notice
The U.S. Environmental Protection Agency (EPA) through its Office of Research and Development performed and
managed the research described here. It has been subjected to the Agency‘s peer and administrative review and has
been approved for publication as an EPA document. Any opinions expressed in this report are those of the author and
do not, necessarily, reflect the official positions and policies of the EPA. Any mention of products or trade names
does not constitute recommendation for use by the EPA.
ii
Abstract
This project evaluated green roofs as a stormwater management tool. Specifically, runoff quantity and quality from
green and flat asphalt roofs were compared. Evapotranspiration from planted green roofs and evaporation from
unplanted media roofs were also compared. The influence of media type, media depth and drought during plant
establishment on plant growth and long-term management of media pH were investigated. The goal of the project was
to provide high-quality replicated data which could be used to develop and refine reliable anticipated runoff volumes
and loadings from green roofs, respectively, as well as evaluate factors which impact plant growth and establishment.
Results indicate that the green roofs are capable of removing 50% of the annual rainfall volume from a roof through
retention and evapotranspiration. Rainfall not retained by green roofs is detained, effectively increasing the time to
peak, and slowing peak flows for a watershed. There are seasonal considerations as more runoff is generated during
winter and for many summer storms there was no runoff. Green roof runoff does contain concentrations of some
nutrients and other parameters, but values are in line with other planted systems. Due to the volume reduction, actual
nutrient loadings from green roofs are less than asphalt roofing runoff or otherwise manageable at the downspout.
iii
Foreword
The U.S. Environmental Protection Agency (EPA) is charged by Congress with protecting the Nation’s land,
air, and water resources. Under a mandate of national environmental laws, the Agency strives to formulate and
implement actions leading to a compatible balance between human activities and the ability of natural systems to
support and nurture life. To meet this mandate, EPA’s research program is providing data and technical support for
solving environmental problems today and building a science knowledge base necessary to manage our ecological
resources wisely, understand how pollutants affect our health, and prevent or reduce environmental risks in the future.
The National Risk Management Research Laboratory (NRMRL) is the Agency’s center for investigation of
technological and management approaches for preventing and reducing risks from pollution that threaten human
health and the environment. The focus of the Laboratory’s research program is on methods and their cost-
effectiveness for prevention and control of pollution to air, land, water, and subsurface resources; protection of water
quality in public water systems; remediation of contaminated sites, sediments and ground water; prevention and
control of indoor air pollution; and restoration of ecosystems. NRMRL collaborates with both public and private
sector partners to foster technologies that reduce the cost of compliance and to anticipate emerging problems.
NRMRL’s research provides solutions to environmental problems by: developing and promoting technologies that
protect and improve the environment; advancing scientific and engineering information to support regulatory and
policy decisions; and providing the technical support and information transfer to ensure implementation of
environmental regulations and strategies at the national, state, and community levels.
This publication has been produced as part of the Laboratory’s strategic long-term research plan. It is
published and made available by EPA’s Office of Research and Development to assist the user community and to link
researchers with their clients.
Sally C. Gutierrez, Director
National Risk Management Research Laboratory
iv
Contents
Notice…………………………..………………………………….…………………………………………………ii
Abstract………….……………...……………...……………….……………………………………………………iii
Foreword………….……..……………………….…………..………………………………………………………iv
Contents………………………………………………………………………………………………………….......v
Acknowledgements………………………………………………………………..…………………………………xi
List of Figures….…..…………………………………………….…………………………………………………..vii
List of Tables…..…………………………………………….…………….…………………………………...…….ix
Acronyms and Abbreviations…………………………………………………………..…………………………….x
Executive Summary……………………………………………………………..…….………………………..………1
Chapter 1 Introduction.................................................................................................................................................. 1-1
Field Data .............................................................................................................................................................. 1-3
Greenhouse and Laboratory Experiments ............................................................................................................. 1-4
Chapter 2 Conclusions and Recommendations ............................................................................................................ 2-1
Conclusions............................................................................................................................................................... 2-1
Monitoring Stormwater Runoff from Green Roofs............................................................................................... 2-1
Monitoring Runoff Water Quality from Green Roofs........................................................................................... 2-1
Evaluating Evaporation and Evapotranspiration Rates of Green Roofs................................................................ 2-2
Factors Affecting Green Roof Establishment and Maintenance ........................................................................... 2-2
Implications........................................................................................................................................................... 2-2
Recommendations..................................................................................................................................................... 2-3
Chapter 3 Green Roof Effects on Runoff Quantity ...................................................................................................... 3-1
Introduction............................................................................................................................................................... 3-1
Methods..................................................................................................................................................................... 3-1
Precipitation Summary.............................................................................................................................................. 3-2
Analysis..................................................................................................................................................................... 3-5
Roof Runoff Volumes ........................................................................................................................................... 3-5
Peak Hydrograph Analysis.................................................................................................................................. 3-12
Summary................................................................................................................................................................. 3-14
Chapter 4 Green Roof Effects on Runoff Quality ........................................................................................................ 4-1
Introduction............................................................................................................................................................... 4-1
Runoff Summary....................................................................................................................................................... 4-1
Methods and Analyses .............................................................................................................................................. 4-3
Color and Turbidity............................................................................................................................................... 4-3
pH.......................................................................................................................................................................... 4-5
Electrical Conductivity.......................................................................................................................................... 4-7
Nitrate.................................................................................................................................................................... 4-9
Other Water Quality Constituents ....................................................................................................................... 4-13
Summary................................................................................................................................................................. 4-16
Chapter 5 Green Roof Plant and Media Management Considerations......................................................................... 5-1
v
Evapotranspiration Rates from Green Roofs ............................................................................................................ 5-1
System Description ............................................................................................................................................... 5-1
Evapotranspiration Rates....................................................................................................................................... 5-2
Effect of Media Depth, Type, and Drought during Establishment ........................................................................... 5-5
Data Analysis ........................................................................................................................................................ 5-5
Effects of Drought................................................................................................................................................. 5-6
Effects of Media Depth ......................................................................................................................................... 5-7
Effects of Media Type........................................................................................................................................... 5-9
Discussion ............................................................................................................................................................. 5-9
Evaluating Long-Term Media Buffering Capacity................................................................................................... 5-9
Media Buffering Capacity................................................................................................................................... 5-10
Cation Exchange Capacity .................................................................................................................................. 5-11
Chapter 6 References.................................................................................................................................................... 6-1
Appendix A Pictures of Experimental Set-ups………………..……………………....…..…………….…………...A-1
Appendix B Cost and Implementation Information………………………….…......……..…………………………B-1
Appendix C Preliminary Data for Experimental Design…………………………..…………………………………C-1
vi
List of Figures
Figure 1-1 Typical cross-section of a green roof system (Berghage et al., 2007) ....................................................... 1-2
Figure 3-2 Monthly average precipitation and runoff for green and asphalt roofs including unplanted media roof
Figure 3-6 Precipitation and green and asphalt roof runoff from snow and mixed precipitation events in March, 2005
Figure 5-2 Average water loss during the daylight or night from planted and unplanted green roof lysimeter test beds
Figure 5-3 Average hourly water loss from planted and unplanted boxes for the 2nd day after watering during the hot
Figure 5-4 Average hourly water loss from planted and unplanted boxes for the 6th day after watering during the hot
Figure 5-5 Mean shoot dry weight (g) (n=10) for Petrorhagia saxifraga in three depths of two media under early
Figure 5-6 Mean shoot dry weight (g) (n=10) for Petrorhagia saxifraga in three depths of two media under late
Figure 5-7 Mean shoot dry weight (g) (n=10) for Sedum album in three depths of two media under early drought
Figure 3-1 Retention and runoff from green roofs (percentage of average monthly precipitation) ............................ 3-6
section........................................................................................................................................................................... 3-6
Figure 3-3 Measurements for June 6, 2005 ................................................................................................................. 3-7
Figure 3-4 Plot of precipitation versus runoff from green roofs for paired events...................................................... 3-8
Figure 3-5 Probability plot of precipitation versus retained precipitation by green roofs........................................... 3-9
.................................................................................................................................................................................... 3-10
Figure 3-7 Precipitation and runoff from green and flat asphalt roofs on August 16, 2005...................................... 3-11
Figure 3-8 Precipitation and runoff from green and flat asphalt roofs on August 30, 2005...................................... 3-11
Figure 3-9 Precipitation and runoff from green and flat asphalt roofs on October 7-8, 2005 ................................... 3-12
Figure 3-10 Hydrograph for August 30-31, 2005...................................................................................................... 3-13
Figure 3-11 Hydrograph of most intense period of August 30-31, 2005 event......................................................... 3-13
Figure 4-1 Average runoff color from all roof types................................................................................................... 4-3
Figure 4-2 Runoff color comparison between green roofs and flat asphalt roofs ....................................................... 4-4
Figure 4-3 Average turbidity of runoff samples from all roof types ........................................................................... 4-4
Figure 4-4 pH of runoff from green and flat asphalt roofs .......................................................................................... 4-5
Figure 4-5 Average pH of runoff from all roof types.................................................................................................. 4-6
Figure 4-6 United States map of acid rain (National Atmospheric Deposition Program, 2006)................................. 4-7
Figure 4-7 Electrical conductivity of runoff from all roof types ................................................................................. 4-8
Figure 4-8 Electrical conductivity of runoff from green and asphalt roofs ................................................................. 4-8
Figure 4-9 Electrical conductivity from green and media roofs.................................................................................. 4-9
Figure 4-10 Average nitrate concentration in runoff from all roof types .................................................................. 4-10
Figure 4-11 Nitrate concentration in runoff from green and flat asphalt roofs ......................................................... 4-10
Figure 4-12 Nitrate (pounds per acre) in runoff from green and asphalt................................................................... 4-11
Figure 4-13 Calculated annual loading of nitrate from test roofs.............................................................................. 4-12
Figure 4-14 U. S. map of atmospheric nitrate wet deposition (National Atmospheric Deposition Program, 2006). 4-12
Figure 5-1 Average cumulative water loss from planted and non-planted green roof lysimeter test beds in the fall . 5-2
in the fall....................................................................................................................................................................... 5-3
summer condition (Rezaei 2005).................................................................................................................................. 5-4
summer condition (Rezaei 2005).................................................................................................................................. 5-4
drought conditions (Thuring, 2005).............................................................................................................................. 5-6
drought conditions (Thuring, 2005).............................................................................................................................. 5-7
conditions (Thuring, 2005)........................................................................................................................................... 5-8
vii
Figure 5-8 Mean shoot dry weight (g) (n=10) for Sedum album in three depths of two media under late drought
conditions (Thuring, 2005)........................................................................................................................................... 5-8
Figure 5-9 Accelerated acid aging trial for clay-based media................................................................................... 5-10
Figure 5-10 Accelerated acid aging trial for slate based media................................................................................. 5-11
viii
List of Tables
Table 3-1 Precipitation Event Summary...................................................................................................................... 3-2
Table 3-2 Synopsis of Rain Events Greater Than Half an Inch................................................................................... 3-1
Table 4-1 Precipitation Events Sampled for Water Quality Monitoring..................................................................... 4-1
Table 4-2 Volumes for Water Quality Monitoring...................................................................................................... 4-2
Table 4-3 Summary of Concentrations and Volumes of Runoff for Five Sampled Events....................................... 4-14
Table 4-4 Estimated Loadings from Flat Asphalt and Green Roofs.......................................................................... 4-15
Table 5-1 Mean Shoot Dry Weight for Sedum album under Early and Late Drought Conditions.............................. 5-9
ix
Acronyms and Abbreviations
BMP = Best Management Practice
CAM = Crassulacean Acid Metabolism
CSO = Combined Sewer Overflow
CV = Coefficient of Variation
EC = Electrical Conductivity
ET = Evapotranspiration
EMC = Event Mean Concentration
EPA = U.S. Environmental Protection Agency
LID = Low Impact Development
NURP = National Urban Runoff Program
SM = Standard Methods
SWMP = Stormwater Management Plan
WWF = Wet-Weather Flow
x
Acknowledgements
A major undertaking of this type requires the dedication and cooperation of many individuals.
The Pennsylvania State University project team which performed the sampling and analysis was led by Dr. Robert D.
Berghage, Dr. David Beattie and Dr. Albert. R. Jarrett of the Penn State Center for Green Roof Research. Funding
from this project led to two master theses of Christine Thuring and Farzaneh Razaei.
EPA Region 3 identified this project for funding through the EPA’s Regional Applied Research Effort (RARE). The
technical direction and coordination for this Cooperative Agreement No. 83124401 was provided by the Office of
Research Development (ORD) Project Officer, Thomas P. O’Connor.
Many others assisted in making this product available to the public. Special recognition is extended to Dr. Ronald
Landy, ORD Regional Scientist, Region 3, for his help in coordinating and initiating the project. Dr. William Hunt,
P.E., Dr. Bradley Rowe, and Dr. Scott Struck performed reviews of this report. Carolyn Esposito reviewed the
quality assurance project plan and this report. Dr. Swarna Muthukrishnan, former Post Graduate Researcher, ORISE,
provided technical assistance.
Dr. David Beattie passed away on February 11, 2008. Dr. Beattie pioneered green roof research while at Penn State
University, and distinguished himself as a great supporter of the green roof industry.
xi
Executive Summary
This report documents efforts to quantify and monitor the performance of green roofs. This project investigated green
roofs as an approach to control stormwater discharge and pollutant content. The work was carried out by the Penn
State Green Roof Center of The Pennsylvania State University at University Park, PA. Prior to this project, EPA had
limited information available on the beneficial use of green roofs as a stormwater control due to the limited
application of this technology in the United States.
The Urban Watershed Management Branch of Edison, New Jersey, as part of the EPA National Risk Management
Research Laboratory seeks to develop new urban technologies and assist municipalities in selection of appropriate
technologies to control urban wet-weather discharges. This report provides the data and background municipalities
and private entities need to move forward with decisions to implement green roofs as a stormwater control
technology, or best management practice (BMP). Green roofs fit into a subcategory of on-site BMP, termed low-
impact development (LID) technology, that also includes, biofiltration, swales and rain gardens. Green roofs can be
incorporated with other stormwater controls and included in a municipal stormwater management plan (SWMP).
Urban development has led to large areas of impervious surfaces such as parking lots and building roofs. Runoff
from these areas is causing problems for many urban and suburban communities. Not only is total volume of WWF
increased, but peak flow rates are also increased. Implementation of traditional stormwater BMPs in urban areas may
not be practical in all circumstances due to limited available surface area and other concerns. Green roofs have been
suggested as a means to reduce the stormwater impacts of development because they have been shown to both detain
and retain stormwater. The purpose of this study was to quantify runoff reductions resulting from the use of extensive
green roof systems.
This project investigated design specifications and materials of green roofs used as stormwater control devices. This
was done through replicate experimental setups using small-scale buildings exposed to the elements and in controlled
environment greenhouse test bed systems. The size of the buildings allowed the complete volume of runoff to be
captured analyzed. Performance data derived from side-by-side field and laboratory tests provide clear performance
criteria for stormwater volume control by green roofs. Key water quality parameters evaluated included: real-time,
flow, turbidity, electrical conductivity (EC), pH and nitrate; and, grab sample pollution assessment, for nutrients,
metals and hardness.
- 1 -
This report addresses the following:
1. Monitoring stormwater runoff from green roofs
Stormwater runoff was monitored and analyzed from January, 2005 through November, 2005. Replicated data were
collected for 72 precipitation events from three green roofs and two flat asphalt roofs. Unreplicated data were
collected from an unplanted, media-only roof section and a rooftop detention section. A range of events were
monitored including a high-intensity short-duration [1 in. (25 mm) in 30 min] event and a high-total precipitation
steady-rate [2.65 in. (67.3 mm) over 8 hr] event. Data were also collected from winter snow events.
These data confirm previous data from other studies’ conclusions that a 3.5 - 4 in. (8 - 10 cm) deep green roof can
retain 50% or more of the annual precipitation. The replicated data from this study provide estimates of expected
differences in performance from similarly constructed green roofs. Green roof performance was quite consistent
during the warm summer months (limited runoff) but was more variable during winter months. Flow rates were
reduced in runoff from green roofs until the systems were saturated at which point runoff flow roughly equaled the
rate of precipitation input. Even when rainfall led to saturated conditions, green roofs significantly increased the time
to peak prior to producing runoff as compared to the flat control roofs. These data support the use of green roofs as a
stormwater mitigation tool.
2. Monitoring runoff water quality from green roofs
Stormwater runoff samples were collected from green and flat asphalt roofs and analyzed for water quality
parameters. Twenty three samples were evaluated for pH, EC, color, turbidity, and nitrate. A limited set of five
samples was evaluated for additional nutrients, hardness, salts and metals. This small sampling of green roof runoff
indicated the runoff was similar to what might be expected as leaching from any other planted system in the
landscape.
Green roof runoff was colored yellow, had higher pH and EC, and generally had equal or higher concentrations of the
nutrients and ions measured in solution. Loadings (in lb/acre) of various nutrients, with the exception of nitrate, and
hardness from green roofs were greater than from flat asphalt roofs. However, other ion loadings in the green roof
runoff were not statistically different from flat asphalt roofs. In summer, when green roofs retained nearly all
precipitation, there was limited nitrate loading from the green roofs. Some of the water quality impacts of a green
roof are thus seasonal, and depend on concentrations from a planted system, precipitation concentrations, and runoff
rates. Established green roofs may discharge less nutrients than green roofs being established.
Green roofs can improve water quality with regard to pH and nitrate. The data suggest that green roof discharge from
downspouts should be routed to another LID practice such as vegetated filter strips, rain gardens or stormwater
collection systems with end-of-pipe BMPs for further treatment, when possible; however, green roofs are often used
because there is limited space for other technologies. Green roof runoff should not be discharged directly to the
receiving water without further treatment, if possible. Runoff discharged to stormwater collection systems that have
water quality BMPs, is preferred; however, the time delay and volume reduction provided by green roofs still offer
receiving water quality benefits for stormwater systems that discharge without treatment. For this reason, discharge
from green roofs to a combined sewer system is appropriate and desirable, due to the significant reduction of volume
discharge and extension of time to peak over flat asphalt roofs. Green roofs are an important stormwater technology
for urban areas with limited space for retrofitting BMPs into the existing conveyance system.
3. Evaluating evaporation and evapotranspiration rates of green roofs
Eight 0.5 m
2
weighing lysimeters were planted with a mixture of Delosperma nubigenum and Sedum album or
remained unplanted with green roof media and drainage geotextile. Weighing lysimeters have been used for many
years to measure evaporation and evapotranspiration and these eight were monitored in a greenhouse during dry-
down cycles during warm actively growing periods and cool dormant periods. Drying cycles lasted 21 days.
- 2 -
Green roof plants lose water rapidly following irrigation, after which they reduce water loss rates. This is a new
finding. Evapotranspiration (ET) rates from green roofs were similar to other measured systems and could be
described using normal ET prediction equations such as Pennman-Monteith. Unplanted media lost water at a similar
rate initially but water loss declined below that of the green roof after a couple of days. Plants are thus essential to the
system and move water from down in the media to the surface quickly, resulting in faster return of the retention
potential for the green roofs. This is an important new finding.
These data demonstrate the superiority of a planted roof over an equivalent ballast roof for retention of stormwater
during the summer months. Rapid initial loss of water from these plants followed by drought adaptation is a new
finding that provides an important component of any model or design tool to predict the effectiveness of a green roof
as a stormwater tool.
4. Factors affecting green roof establishment and maintenance
Media type, depth, and early drought were evaluated as factors affecting establishment and early management of a
green roof. Early drought is very detrimental to survival and establishment of green roof plants particularly with
shallow media depths. Sedum species may survive, but other green roof plants may not survive. The results suggest
that 3 - 4 in. (8 - 10 cm) of media, with the potential for supplemental irrigation during establishment, will result in
improved plant establishment for most species.
A test procedure for evaluating long-term pH buffering of the roof was developed and evaluated. Tests of pH
buffering capacity of green roofs suggest that media can buffer acid precipitation for approximately 10 years, after
which it may be necessary to amend the media with lime to maintain pH buffering capacity.
- 3 -
Chapter 1 Introduction
Significant water quality and quantity issues are caused by stormwater runoff from developed areas in North America.
For the five years from 1997 to 2001, the rate of urban development averaged 890,000 ha/year (2,400 ha/day) (NRCS,
2003). Development results in water quality impairment and quantity management issues throughout the affected
watersheds. For example, nutrient loading (a widespread result of agricultural runoff) may be replaced as the critical
impairment issue for a watershed by increased peak flows, flooding, and urban pollutant loads as runoff is collected
from impervious pavement and roof surfaces.
Wet weather flow (WWF) including combined sewer overflow (CSO), sanitary sewer overflow (SSO), and
stormwater discharges is one of the leading causes of water quality impairment in the United States and improvement
of controls is one of two priority water focus areas cited by the EPA's Office of Water in its National Agenda for the
Future (Perciasepe, 1994). Pollution problems stemming from these WWFs are extensive throughout the country.
Problem constituents in WWF include visible matter, pathogens, biochemical oxygen demand (BOD), suspended
solids (SS), nutrients, and toxicants (e.g., heavy metals, pesticides, and petroleum hydrocarbons). National estimates
have projected costs for WWF pollution abatement in the tens of billions of dollars (APWA, 1992). Therefore,
municipalities need alternatives to control the high costs of WWF treatment prior to release. This report presents data
showing that green roofs are effective best management practices (BMPs) for mitigation of the environmental impacts
to receiving waters associated with urban runoff.
Greening of rooftops, by incorporating plants into the design of roofing systems, has been suggested as a method to
reduce the impacts of stormwater runoff by reducing the impervious surface within a developed zone (Scholz-Barth,
2001). The benefits of green roofs (sometimes called Eco roofs) for stormwater control include direct retention of a
portion of the rainfall, and delaying and decreasing the peak rate of runoff from the site (PACD, 1998). Most
extensive green roofs currently being installed in North America consist of four distinct layers: an impermeable roof
cover or roofing membrane; a “drainage net;” lightweight growth media, i.e. 3 in. (8 cm); and adapted vegetation
(PACD, 1998).
Figure 1-1 is a cross-section of a typical extensive green roof system. The drainage layer is an open, highly
permeable material that quickly channels gravitational water off the roof. Growth media, in addition to providing a
suitable rooting zone for vegetation, is typically a low-density aggregate with high-water holding capability while also
providing good drainage. A lightweight media from 3 - 6 in. (8 - 15 cm) deep allows for retrofit installation on
existing buildings, and reduces the need for extra structural support in new buildings. Media depth and porosity play
an important role in stormwater retention and plant growth. Plants provide shade to the surface below foliage,
intercept rainfall, and slow direct runoff from sloped roofs (Miller, 1998). Plant size and selection depend on the
depth of the roof overburden (growing media) and local climate, but almost always consists of winter-hardy, drought-
tolerant, perennial plants, e.g., sedums which are a type of succulent, cactus-like plant.
1-1
Figure 1-1 Typical cross-section of a green roof system (Berghage et al., 2007)
The use of green roofs in Germany is widespread and has been promoted in many cities through financial incentives
(Pederson, 2001). Economies of scale, contractor experience, and specialized equipment have reduced the cost of
installing a green roof in Germany and throughout Europe. In contrast, installing a green roof in the United States can
be very expensive, adding from at least $6/ft
2
($65/m
2
), to more than $30-$40/ft
2
($320-$430/m
2
), to the cost of the
roof. Other barriers also limit widespread use of green roofs in the US. Engineers, architects, developers, and policy
makers are unsure of the actual quantifiable benefits of a green roof. In the U.S., Peck (2007) observed as recently as
2006, only 70 acres (28.5 ha or 285,000 m
2
) of green roofs had been planted, as compared to Germany where in 1996
alone, 2,500 acres (1 Kha or 10,000,000 m
2
) of green roofs were installed (Peck et al., 1999); however, the 2006 totals
represent a 24% increase over the previous year, and only account for reported projects.
Annual reductions of runoff of 38 - 54% and 38 - 45% have been reported for 3 in. (8 cm) deep media (Miller, 1998).
A media depth of 2.5 in. (6.5 cm) can retain 40% of the rain for an individual 2-in. (50-mm) storm (Scholz-Barth,
2001). The City of Portland, Oregon, has developed guidelines for green roofs that state that some jurisdictions may
reduce water and sewer charges or may provide financial incentives to developers who retain stormwater on site, and
that green roofs can help reduce the size of stormwater management ponds. Much of the existing published
information on green roof performance in North America has been collected from pilot-scale or sometimes
commercial-scale green roofs without replication. For example, recently Van Woert et al. (2005) performed studies of
three simulated roof platforms with dimensions of 2.44 x 2.44 m (8 x 8 ft), divided into three sections, to quantify the
effects of various treatments on stormwater retention. The mean precipitation retention ranged from 48.7% for gravel
test beds to 82.8% for vegetated test beds. Roof slope and green roof media depth also impacted stormwater retention
with the combination of reduced slope and deeper media reducing the runoff the most. They also observed
moderation of peak flows. Little scientifically based replicated data have been collected in North America at the
building scale.
EPA is emphasizing the use of BMPs to capture and treat runoff from small storms, especially the use of onsite
BMPs, often termed low-impact development (LID), such as bioretention, swales, or rain gardens. Green roofs offer
1-2
a practical alternative for new construction and for retrofitting existing structures. Implementation of green roofs in
European countries like Germany is a regulatory driven technology in the municipalities that have adopted mandates
for green roofs on new buildings. With more municipalities in the United States looking for flexible ways to control
stormwater, including the use of stormwater credits or watershed-based trading, developing new stormwater controls
such as green roofs is a vital initiative for the EPA. Green roofs appear to be a suitable technology for urban areas, as
there is limited space to implement traditional stormwater controls. Land values are too high to devote much surface
area to stormwater control devices. In addition, surface-based stormwater control devices can be vandalized and may
pose public access and safety issues. Green roofs can slow the runoff from roofs during larger storms and during
smaller storms are capable of absorbing a majority if not all of the rainfall. In essence, the impervious area is
decreased when planted roofs are installed on, or retrofitted to, buildings. The plants also act as a bio-filter in
reducing the pollutant content of the rainfall. This technology reduces the heat island effect of standard roofs by
replacing the low albedo surface, as well as by increasing evapotranspiration, which helps cool the air by several
degrees. Furthermore, because of reduced air temperature, less energy is needed for air conditioning (Osmundson,
1999).
Field Data
Six similar structures, three with, and three without green roofs, were used for this study so that statistically valid,
replicated demonstrations could be performed. The six 6 x 8 ft (1.8 x 2.4 m) buildings were constructed in 2001, and
have nearly flat (1:12 pitch facing south) plywood roofs. All structures were insulated, heated, and air conditioned.
Each is covered with standard rolled, asphalt roofing. Three of the structures have green roof systems constructed on
top of the standard roof layer. The three green roofs had a 1 x 4 wooden frame mounted to the top of the roof to
contain the green roof system. Each green roof consisted of a drainage mat covered with 3.5 - 4 in. (9 - 10 cm) of
green roof media. The media consisted primarily of expanded clay with some compost amendment. The green roofs
were originally planted in the fall of 2001 and had 95% plant coverage (90% Sedum spurium and 5% Sedum album) at
the beginning of this study.
The third roof without a green roof also had a 1 x 4 wooden frame mounted on the roof like the green roofs, but was
modified to provide two experimental setups. One half of this rooftop provided detention with slow release, while the
other was lined and filled with the same expanded clay-based green roof mixture that was used on other roofs. Media
without compost, i.e., used strictly as a ballasted roofing system, was not tested, as previous studies have indicated
green roofs work better than ballasted roofs. A “V” notch in the board of the detention rooftop allowed this section of
the roof to retain the same volume of runoff as is retained by the media filled portion. This building was used for
housing the computers for automated data collection, and did not provide a suitable replicate to the other flat asphalt
roofs due to heat generated by the computer and data loggers. Modifications to this roof were completed in 2004.
A covered gutter and downspout was attached to the lower end of each roof to collect all runoff into a 55 gal (208 L)
plastic barrel. Each barrel was connected to a pressure transducer which allowed continuous measurement of the
water level in the barrel. Pressure transducers were connected to a Campbell data logger which recorded pressure
every five minutes and downloaded to a computer hard drive every couple of days. Each barrel and associated
pressure transducer was calibrated by adding water to the barrel in 1.3 gal (5.0 L) increments. Individual calibration
curves (output from the transducers is linear) were used to process the raw pressure data. Barrels and associated
gutter systems were initially wrapped with insulation and heat tape to prevent freezing; however, this was insufficient
to collect winter data. In June, 2004, a lightening strike caused the failure of the weather station, several pressure
transducers, and data loggers, requiring replacement. While the automated data collection system was not functional,
several modifications to increase precision, accuracy, and reliability of stormwater collection were performed. New
60 gal (230 L) barrels with conical bottoms were installed and fitted so that complete drainage could be achieved, in
addition to easier removal of water samples and elimination of standing water which caused subsequent algae growth
that had been occurring in the old system. These barrels were also housed in wooden substructures with 1 in. (2.5 cm)
thick insulation and a small heater to improve measurements during winter.
1-3
Rooftops were configured with thermisters to measure temperature in and over the green roof systems. A weather
station including a tipping bucket rain gage, temperature, light, humidity, and wind speed and direction sensors was
located on the corner of the roof of one of the buildings. Weather station data were recorded every 5 min. to
correspond with data obtained from the pressure transducers. Data from the weather station was downloaded once
each week. The measurement limit for the weather station was 0.01 in. (0.25 mm) of rainfall (one bucket tip). The
pressure transducer systems also had a sensitivity which allowed determination of 0.02 in. (0.5 mm) of water runoff
from the roofs. For the purpose of runoff data, a storm event was considered to be any event which resulted in
measurable runoff (0.02 in.; 0.5 mm).
A sampling tube was installed in each gutter near the downspout to allow discrete grab sampling of runoff when
needed. Routine grab samples of runoff water were obtained by sampling the water in the barrel following rain
events. A simple field laboratory for water quality analysis was located at the site for field sample analysis. Field
equipment included a Hanna Instruments of Woonsocket, RI, model HI9803 pH and electrical conductivity (EC)
meter, Hach of Loveland, CO, model 52600-00 pocket turbidity meter, and a Hach model 890 portable colorimeter.
Reagents for nitrate and ammonia analysis, standards, and deionized DI water were kept at the field site. Nitrate was
measured using Standard Method (SM, 1998) 8039.
Additionally, discrete samples were collected using manual techniques and brought back to Penn State’s analytical
laboratory for further analysis. Five samples were evaluated for calcium (Ca), iron (Fe), potassium (K), magnesium
(Mg), manganese (Mn), sodium (Na), phosphorous (P), and zinc (Zn). Two samples were evaluated for sulfur (S).
Experimental design of this project was based on preliminary data collected in the fall of 2002 and summer of 2003
where three building replicates of each type of roof allowed statistical analysis of runoff from green compared to flat
asphalt roofs. Replicated buildings allowed estimation of variance of runoff volume from each roof type within each
event so statistically significant differences can be described for any given rain event. Preliminary water quality data
(Appendix C) also suggested that for measurements of pH, EC, color and nitrate, statistical significance for main
effects of green versus flat asphalt roofs and individual rain events could be achieved.
Greenhouse and Laboratory Experiments
Controlled test bed systems located in the Penn State Horticultural Science greenhouses were used to obtain
background information and to demonstrate differences in water storage, and retention and detention characteristics of
the media, with and without the presence of plants. Rehydration, evapotranspiration (ET), and plant growth rates
under controlled environmental conditions of temperature, light, and vapor pressure deficit were determined.
Seasonality effects were observed by evaluating the characteristics of the plants year-round.
Test beds consisted of small modules (6.1 ft
2
; 0.5 m
2
) with a drainage layer and media identical to field-site buildings.
Each of the eight planting beds was suspended from a load cell to provide continuous measurement of weight change
(primarily water gain or loss). Load cells were connected to a datalogger and recorded individual weights. This
method of testing is also termed weighing lysimeters. Lysimeters have been used for centuries, but measurement and
instrumentation have been improved greatly during the past 50 years (Howell et al., 1991). As described above,
lysimeters are simple containers or tanks filled with soil in which plants are growing. There are two different
methods of using lysimeters to measure ET: weighing lysimeters and non-weighing lysimeters. The weighing
lysimeters have been used for many years to measure the evaporation and ET from bare and cropped soils (Howell et
al., 1991).
Each planting bed drained to a gutter system into a plastic bucket with a pressure transducer also connected to the
datalogger. Weighing lysimeters were maintained with a 1:12 pitch to simulate the slope of the field rooftops. Four
weighing lysimeters were planted with Sedum album and Delosperma nubigenum. Vegetation covered 80-100% of
the surface of weighing lysimeters at the start of the study. The remaining four weighing lysimeters contained the
drainage layer and media, but were not planted.
1-4
Simulated rainfall was applied using a low volume overhead irrigation system. The amount of simulated rain applied
was controlled by timing the irrigation application. Irrigating at full volume for 1 hr resulted in approximately 6 in. of
simulated rain. A known quantity of simulated rain of either 0.75 or 1.5 in. (1.9 or 3.8 cm) was applied over a 2 hr
period based on observed rain events during the summer of 2003.
Weight change and runoff were measured. The weighing lysimeters were then allowed to dry for a series of fixed
periods (1, 4, and 16 days). These times were selected based on preliminary data from both field and greenhouse
projects. Weight change, light, temperature, humidity, wind speed, and flat pan evaporation were also recorded
during this period. Plant conditions were monitored and changes in size, percent coverage, color or other visible
changes were recorded. After drying periods, the weighing lysimeters were rehydrated with simulated rain and the
process was repeated. This process was repeated four times for each hydration and simulated rain volume. Day length
was maintained at a minimum of 16 hr with high-intensity discharge (HID) lights.
A drought and establishment study tested the responses of five plant species to three depths of two media under three
drought regimes, in two replications. Two media were compared for their influences on plant performance in
combination with treatments of depth and drought. Three watering regimes were used: no drought, early drought, and
late drought.
In addition to these greenhouse studies, an accelerated aging study to determine the effects of acid rain on the length
of roof life was conducted. In the course of this laboratory study, acid was delivered to different media, i.e., expanded
slate or clay, to determine the buffering capacity of the media.
1-5
Chapter 2 Conclusions and Recommendations
Conclusions
Green roofs retained over 50% of total precipitation during the study period. During the summer months, nearly 95%
of the precipitation was retained. During winter, retention was smaller (<20%) and not significant. Peak flows were
delayed by green roofs and in many cases peak flow rates were also reduced. Green roofs were most effective at
delaying and reducing peak flows when they were not fully saturated. Rapid peak flows, i.e. high-intensity, short-
duration rainfalls were attenuated more than lower intensity, high-total volume longer period flows.
Monitoring Stormwater Runoff from Green Roofs
Stormwater runoff was monitored and analyzed from January, 2005 through November, 2005. Replicated data were
collected for 72 precipitation events from three green roofs and two flat asphalt roofs. Unreplicated data were
collected from an unplanted, media-only roof section and rooftop detention section. Events included high-intensity,
short duration (1 in. (25 mm) in 30 min) events and high total precipitation steady rate (2.65 in. (67.3 mm) over 8 hr)
events. Unique data were also collected from winter precipitation events, including snow and ice.
These data and data from other studies at this site confirm that under ambient conditions, a 3.5 - 4 in. (8 - 10 cm) deep
green roof can retain 50% or more of the annual precipitation. The replicated data from this study provide the only
available estimate of expected differences in performance from identical green roofs. Green roof runoff was quite
consistent during the warm summer months (almost no runoff), but was more variable during winter months when
runoff from buildings varied during some storm events from 80% for one building to 100% for others. Flow rates
were reduced in runoff from green roofs until systems were saturated, at which point runoff flow roughly equaled the
rate of precipitation input; however, peak flows were reduced and time-to-peak increased.
Monitoring Runoff Water Quality from Green Roofs
Stormwater runoff samples were collected from green and flat asphalt roofs and analyzed for water quality
parameters. Twenty-one precipitation events were evaluated for pH, EC, color, turbidity, and nitrate. Additionally,
discrete samples were collected using manual techniques and brought back to Penn State’s analytical laboratory for
further analysis. A limited data set of five sampling events was analyzed for nutrients, hardness, and other ions.
Analysis of the 21 precipitation events revealed that green roof runoff was colored yellow and had higher pH and EC.
The increased pH was a benefit in an area of such acid precipitation. The smaller data set of five samples indicated
that green roof runoff generally had equal or greater concentrations of nutrients (phosphorous and potassium) and
hardness (calcium and magnesium) measured in solution than flat asphalt roof runoff. The concentration of green
roof phosphorous release was comparable to that of known residential landscape values. Loadings of nutrients (to
sustain plants) and hardness (a property of the clay based media) were significantly greater for the green roofs,
approximately 300% for phosphorous and potassium, and as much as 1000% for magnesium. Analysis for other ions
2-1
did not statistically discern whether the loadings were greater or lesser from green roofs. Partly, this is due to the
small sampling size, but also indicates that beyond proper management of the planting media to reduce excess
nutrient release, loadings from green roofs are not significant. Results based on this smaller, limited water quality
monitoring data set (five samples) should be used cautiously.
Green roofs appeared to be beneficial for the removal of atmospheric nitrate. In the summer when green roofs
retained nearly 100% of the precipitation almost no nitrate ran off the green roofs. Water quality impacts of a green
roof are thus seasonal plant-related mechanisms and depend on both the input concentration and the precipitation and
runoff rates.
The data collected suggest that the best use of a green roof is probably in conjunction with other stormwater BMPs
such as bio-infiltration and rain gardens, where possible. Runoff discharged to stormwater collection systems that
have water quality BMPs, is preferred; however, the time delay and volume reduction provided by green roofs still
offer receiving water quality benefits for stormwater systems that discharge without treatment. For this reason,
discharge of green roof runoff to a combined sewer system is appropriate and desirable, due to the significant
reduction of volume discharge and extension of time to peak, regardless of discharge concentration. Green roofs are
an important stormwater technology for urban areas with limited space for retrofitting BMPs into the existing
conveyance system. Direct discharge of roof runoff to the receiving water is not recommended.
Evaluating Evaporation and Evapotranspiration Rates of Green Roofs
Eight 0.5 m
2
(6.1 ft
2
) weighing lysimeters planted with a mixture of Delosperma nubigenum and Sedum album were
compared to unplanted media. These lysimeters were monitored during 21-day dry-down cycles during warm actively
growing periods and cool dormant periods. Drying cycles lasted 21 days.
Green roof plants rapidly lose water following irrigation after which water loss rates decline. This is a new finding.
Initial ET from green roofs was similar to other measured systems and could be described using normal ET prediction
equations such as Pennman-Monteith. Unplanted media lost water at a similar rate initially, but after several days,
water loss rates declined below that of the green roofs. Thus plants are essential to the system, while the unplanted
media are limited to evaporating water from the surface, the plants continue to remove water from down in the media,
resulting in quick recharge of the stormwater runoff reduction potential.
These data demonstrate the superiority of a planted roof over an equivalent ballast roof for retention of stormwater
during the summer months. Rapid initial loss of water from these plants followed by drought adaptation is a new
finding that provides an important component of any model or design tool to predict the effectiveness of a green roof
as a stormwater tool.
Factors Affecting Green Roof Establishment and Maintenance
Media type, depth, and early drought were evaluated as factors affecting establishment and early management of a
green roof. A test procedure for evaluating long-term pH buffering of the roof was developed and evaluated.
Early drought is very detrimental to the survival and establishment of green roof plants particularly with shallow
media depths. Sedum species may survive but other green roof plants may not survive. The results suggest that 3 - 4
in. (80 - 100 mm) of irrigation with the potential for supplemental irrigation during establishment will result in better
plant survival rates.
Tests of the pH buffering capacity of the planting media suggest that the green roof media can buffer acid
precipitation for approximately 10 years, after which it may be necessary to amend the media with lime to maintain
the pH buffering capacity.
Implications
Green roofs can attain an annual 50% reduction in roof runoff. From a practical standpoint, this potentially translates
into a reduction in area and volume control needed for the typical suite of water quality BMPs. In terms of practice,
2-2
the stormwater volume and increased time to peak control offered by green roofs could result in more building space,
additional parking spaces or additional and usable open space. However, this concept would need to be field tested at
a larger scale and the actual percent reduction in stormwater BMPs would need to be evaluated, e.g., potential BMP
reduction may be between 5 and 20%, not a full 50% annual capture, particularly in areas with dormant (winter)
seasons.
Clay-based media may be better in areas affected by drought due to water retaining capacity of the media. Shale-
based media may be better for areas subject to more frequent precipitation, particularly acid precipitation.
Recommendations
In this project, several constituents of concern from the green roof were studied. Results demonstrated that green
roofs may reduce certain pollutants, e.g., acid precipitation and nitrate, but that it may increase loadings directly
related to these planted systems, e.g., phosphorous, potassium, calcium, and magnesium. Due to the variability in
results, continued sample collection and analysis to minimize the variability may be warranted. Further testing of
materials used for green roof construction and planting should be conducted to determine loadings coming from roofs.
Also, other constituents from atmospheric deposition and building materials for standard roofing should be tested
under controlled systems. The nitrate results should not be viewed as a surrogate for all nitrogen and future studies
should look at total nitrogen, and potentially ammonia and total Kjeldahl
Sampling for some of the water quality parameters was only represented by five storms; additional sampling is
warranted due to the small size of this data set. In addition, only analyzing five paired events for green and flat
asphalt roof runoff may have biased results toward lower loadings from the flat asphalt roofs as not as much rain was
required to produce runoff, i.e., diluted flat asphalt runoff was compared to higher concentration, lower volume green
roof runoff. If further comparison tests are performed to standard roofing materials and systems, paired analysis of
loadings should include storm results from other roofing systems and should include a full range of precipitation
events, from when green roofs are not producing runoff to events with large amounts of runoff. The greatest benefit
green roofs can provide is the reduction in runoff, which is also a water quality benefit not adequately represented by
the five paired data points.
The field site provided statistically valid results for runoff volumes and much of the concentration data; however, the
calculated loadings are based on relatively small experimental rooftops, when compared to an urban watershed. Due
to the variability observed in this study, modeling loadings for green roofs for watershed management may require
additional monitoring with full-scale roofs or multiple roofs in an urban setting.
The size and time interval limited analysis of peak flows from the green roofs. Additional hydrology monitoring may
be warranted on larger roofs to better determine potential peak flow values from green roofs.
The half-media and half-detention roofs, while providing insight to the experiment, were of limited value for
individual storms because of the absence of replicates. Splitting the roof into two media sections or two detention
sections would have provided additional replicates. Another rain gauge or triangulation of rain gauges around the
buildings would have provided more insight to rainfall totals.
The laboratory studies indicate that ET can be modeled using standard equations; however, further testing should be
conducted. Data suggest there may be a need to develop unique water loss model factors to account for water loss
patterns in Sedum carpet roofs to accurately predict rate of recharge for water detention capacity.
The drought studies indicated some potential limitations without the use of irrigation. Green roofs need to be tested in
other climates so that further design specifications on plant mixtures, media depth and amendments, and potential
irrigation requirements can be determined. Other climatic conditions should also include year-to-year or long-term
studies, as it seems very likely that in dry years the green roof runoff would be far less than in wet years.
2-3
The effects of green roof runoff discharge on receiving waters or the potential for additional treatment of green roof
discharge were not addressed. For suburban or agricultural areas, green roof runoff treatment may be as simple as
directing the downspouts to grassed areas (vegetated filter strips or swales) or collecting green roof runoff in rain
barrels to be used for irrigation, but this may not be practical for urban areas where there is limited room for
stormwater controls. For urban areas that have combined sewers, green roofs should be viewed as a benefit due to the
volume reduction to the combined system and the delay in time to peak. The same can be said for stormwater
conveyance systems that drain to stormwater BMPs. The effects of mixing and the delay in time to peak may be
sufficient to allow discharge to stormwater conveyance systems that discharge to a receiving water even without
treatment; however, further studies or modeling exercises may be warranted. Directly discharging green roof runoff
to a receiving water is not recommended.
Additional lysimeter studies should be conducted to identify more plant species suitable for green roofs, especially
varieties that are drought resistant and require minimal nutrient supplements.
2-4
Chapter 3 Green Roof Effects on Runoff Quantity
Introduction
Runoff samples were collected from six small buildings at the Center for Green Roof Research at Rock Springs, PA
during the period from January 2005 through November 2005. Runoff from three green roofs, two flat-asphalt
control roofs, and one roof divided between detention and a green-roof system without plants, was collected in rain
barrels. Comparisons of runoff volumes are presented among roof types. Runoff water volume varied with both
sample event and roof type. Analyses were performed on the study period, monthly and individual storm basis.
Methods
Six similar structures, three with green roofs, two with standard asphalt rolled roofing and one with half of the roof in
a custom-built detention section, and the other half with an unplanted green roof media were used for this study. All
green roofs were constructed to be identical replications of each other to allow statistical analysis and comparisons.
The rooftop storage section was designed to provide approximately 1.0 in. (25 mm) of short-term water detention
with slow release, and the unplanted media section containing lightweight expanded shale and compost amendment,
was included as a control to evaluate the effects of plants on roofs in the field. Structures were insulated, heated, and
air conditioned. Each structure was a 6 x 8 ft (1.8 x 2.4 m) building with a nearly flat roof (1:12 pitch). A covered
gutter and downspout was attached to the lower end of each roof to collect all runoff from the roof. Each downspout
drained into a 60 gal (230 L) plastic barrel. Total barrel capacity was approximately 2 in. (50 mm) of rain over the
roof surface. Rain exceeding barrel capacity could overflow the top of the barrel to the ground, so in anticipation of
overtopping during the largest storm monitored, the barrels were emptied to allow complete monitoring of the storm
(as noted in comments Table 3-1). This size roof and barrel size combination allowed for complete capture for all but
one event in the study. This was a unique approach to monitoring of green roofs.
Each barrel was connected to a pressure transducer to provide continuous measurement of the water level in the
barrel. Pressure transducers were connected to a Campbell data logger which recorded the pressure every five
minutes. The pressure transducer systems have a sensitivity which allows determination of approximately 0.02 in.
(0.5 mm) of water runoff from the roof area. This sensitivity was more than adequate for assessing the total volume
of the storms, but did introduce some uncertainty and variation in time series analysis or instantaneous results, as the
transducer might toggle between values or experience drift. The data logger was downloaded to a PC hard drive
every couple of days. Barrels were calibrated and standard curves developed to relate pressure to gallons of water in
the barrel by adding water in increments until the barrel was full.
Barrels were periodically emptied, usually after each rain event, but during the winter for snow events and during
periods where the end of a rain event was unclear, several defined events were combined. Barrels were enclosed in
insulation and heat tape inside the barrels and gutters was used to prevent freezing of the barrel and associated gutter
systems. Flat asphalt roofs were covered with standard asphalt roofing. Green roofs had a 1 x 4 wooden frame
mounted to the top of the roof to contain the green roof system. The green roofs consisted of 3.25 in. (8.5 cm) of an
3-1
expanded, clay-based mineral placed on top of a 0.75 in. (2.0 cm) drainage layer (Enkadrain 9615, Colbond, Enka,
NC) for a total roof profile of 4 in. (10.5 cm) deep. The green roofs were first planted with Sedum spurium in the fall
of 2001. Sedum album and Delosperma nubigenum were added at about 10% coverage per species in the spring of
2003. Delosperma nubigenum did not survive the winter. During the study the growth distribution on the three roofs
was 90% Sedum spurium and 5% Sedum album. Approximately 5% of the roof surface, mostly near the edges, was
not covered by plants and remained bare media throughout the study period. A weather station including a tipping
bucket rain gage, temperature, light, humidity, and wind speed and direction sensors was located on the corner of the
roof of the building with the split roof that housed the data storage. Weather station data were recorded every 5
minutes to correspond with data obtained from the pressure transducers. Data from the weather station were
downloaded once each week. The measurement limit for the weather station was 0.01 in. (0.25 mm) (one bucket tip).
For the purpose of runoff data, a storm event was considered to be any event which resulted in more than 0.02 in. (0.5
mm) of rain preceded and followed by a minimum of 6 hr without measurable precipitation. Data used in the analysis
included 1 hr before to 1 hr after the end of the precipitation event.
Measurements were terminated in November, 2005 when a cold spell froze and ruptured five of seven pressure
transducers. Barrel connections to the pressure transducers also froze in January, 2005 and February, 2005, but the
transducers were not damaged for these periods. A lightning strike in early July, 2005 damaged several transducers,
the datalogger PC interface, and destroyed the PC. All transducers were replaced and calibrated, the datalogger
interface was repaired and the PC was replaced. No runoff data were collected for the period from July 6, 2005
through July 23, 2005. There were several other events during the evaluation period where data were not complete
either because of data-logging errors (Februrary 14, 2005 and March 28, 2005) or because the drain valves for the
barrels were left open (November 9, 2005).
Precipitation Summary
A total of 111 events during the monitoring period resulted in 35.5 in. (902 mm) of precipitation measured by the
weather station on the site (Table 3-1). Of the 111 total events 83 events resulted in precipitation greater than 0.02 in.
(0.5 mm) for a total precipitation of 34.9 in. (902 mm). There were 54 events with greater than 0.10 in. (2.5 mm) of
precipitation and seven events with more than 1 in. (26 mm) of precipitation. The largest single event occurred
October 7, 2005 when 3.2 in. of precipitation was recorded. Roof runoff data were collected from 72 events with a
total precipitation of 26.9 in. (683 mm). Winter (January - March) included snow, freezing rain and rain events.
Table 3-1 Precipitation Event Summary
Event
number
Event
date
start
Start
time
Event
date
end End time Total Precipitation
(in.) (mm) Comments
1 01/02/05 10:05 PM 01/03/05 5:05 AM 0.03 0.8 Rain
2 01/03/05 11:15 AM 01/04/05 7:25 AM 0.81 20.6 Rain
3 01/05/05 1:35 AM 01/06/05 5:13 PM 2.02 51.3 Rain
4 01/08/05 4:48 AM 01/08/05 9:33 AM 0.43 10.9 Rain
5 01/11/05 4:28 PM 01/12/05 10:52 AM 0.64 16.3 Snow - precipitation total as measured by
rainguage, not actual snow depth.
6 01/13/05 10:47 PM 01/14/05 5:52 AM 0.48 12.2 Snow/rain
7 01/30/05 11:15 AM 01/30/05 12:10 PM 0.05 1.3 Snow
8 02/04/05 9:55 AM 02/04/05 11:55 AM 0.07 1.8 Snow
9 02/08/05 2:51 AM 02/08/05 8:01 AM 0.21 5.3 Snow
10 02/09/05 1:41 PM 02/09/05 8:36 PM 0.74 18.8 Snow/rain
11 02/14/05 12:56 PM 02/14/05 7:46 PM 0.73 18.5 Incomplete data set - logging error. Data
only available for first 0.3 in.
12 02/16/05 8:41 AM 02/16/05 10:16 AM 0.05 1.3 Snow/Rain
3-2
Event
number
Event
date
start
Start
time
Event
date
end End time Total Precipitation
(in.) (mm) Comments
13 02/21/05 11:56 AM 02/21/05 6:06 PM 0.25 6.4 Snow
14 02/25/05 12:55 PM 02/25/05 3:15 PM 0.08 2.0 Snow
15 03/01/05 1:00 PM 03/01/05 1:00 PM 0.01 0.3 Snow
16 03/03/05 10:45 AM 03/03/05 11:20 AM 0.02 0.5 Snow
17 03/04/05 9:55 AM 03/04/05 11:35 AM 0.07 1.8 Snow
18 03/07/05 11:02 PM 03/08/05 2:47 AM 0.11 2.8 Snow/rain
19 03/11/05 10:47 AM 03/11/05 5:42 PM 0.13 3.3 Snow/rain
20 03/12/05 1:07 PM 03/12/05 4:17 PM 0.02 0.5 Snow
21 03/20/05 7:38 AM 03/20/05 10:48 AM 0.07 1.8 Snow
22 03/23/05 3:57 AM 03/23/05 12:42 PM 0.42 10.7 Snow
23 03/23/05 7:52 PM 03/23/05 10:17 PM 0.02 0.5 Snow
24 03/24/05 8:22 AM 03/24/05 3:22 PM 0.5 12.7 Snow
25 03/25/05 11:22 AM 03/25/05 12:07 PM 0.03 0.8 Snow
26 03/26/05 9:17 AM 03/26/05 9:17 AM 0.01 0.3 Snow
27 03/27/05 5:17 PM 03/28/05 2:52 AM 0.58 14.7 Rain
28* 03/28/05 9:07 AM 03/29/05 3:27 AM 1.28 32.5 Incomplete data set - logging error. Data
only available for first 063 in.
29 04/01/05 9:45 PM 04/03/05 12:45 AM 0.7 17.8 Snow/Rain
30 04/03/05 4:05 PM 04/03/05 10:15 PM 0.07 1.8
31 04/20/05 9:09 PM 04/20/05 9:44 PM 0.08 2.0
32 04/22/05 5:58 PM 04/23/05 7:58 AM 0.28 7.1
33 04/23/05 3:23 PM 04/23/05 4:38 PM 0.22 5.6
34 04/24/05 3:18 PM 04/24/05 6:38 PM 0.05 1.3
35 04/27/05 12:00 AM 04/27/05 1:00 AM 0.02 0.5
36 04/29/05 4:15 AM 04/29/05 10:40 AM 0.04 1.0
37 04/30/05 2:25 AM 04/30/05 6:20 PM 0.18 4.6
38 05/14/05 4:39 AM 05/14/05 4:44 AM 0.04 1.0
39 05/14/05 6:54 PM 05/15/05 3:39 AM 0.26 6.6
40 05/19/05 9:44 PM 05/20/05 11:04 AM 0.46 11.7
41 05/21/05 4:44 AM 05/21/05 4:44 AM 0.01 0.3
42 05/21/05 7:44 PM 05/21/05 8:14 PM 0.03 0.8
43 05/22/05 2:44 AM 05/22/05 2:44 AM 0.01 0.3
44 05/23/05 7:54 AM 05/23/05 3:13 PM 0.32 8.1
45 05/24/05 12:03 PM 05/24/05 5:08 PM 0.12 3.0
46 05/25/05 5:43 AM 05/25/05 5:43 AM 0.01 0.3
47 05/28/05 10:51 AM 05/28/05 6:41 PM 0.37 9.4
48 05/30/05 5:56 PM 05/30/05 10:36 PM 0.03 0.8
49 06/03/05 2:32 AM 06/03/05 5:52 PM 0.28 7.1
50 06/04/05 10:57 AM 06/04/05 10:57 AM 0.01 0.3
51 06/06/05 11:45 AM 06/06/05 5:10 PM 0.67 17.0
52 06/07/05 7:25 AM 06/07/05 7:25 AM 0.01 0.3
53 06/09/05 3:50 PM 06/09/05 4:00 PM 0.15 3.8
54 06/10/05 3:25 PM 06/11/05 4:15 AM 0.21 5.3
55 06/13/05 2:35 PM 06/13/05 2:50 PM 0.05 1.3
56 06/16/05 3:48 AM 06/16/05 3:48 AM 0.01 0.3
57 06/17/05 1:18 PM 06/17/05 1:23 PM 0.06 1.5 No water in barrels
58 06/22/05 5:38 AM 06/22/05 5:38 AM 0.01 0.3
3-3
Event
number
Event
date
start
Start
time
Event
date
end End time Total Precipitation
(in.) (mm) Comments
59 06/28/05 9:42 PM 06/28/05 10:17 PM 0.02 0.5
60 06/29/05 6:32 AM 06/29/05 6:32 AM 0.01 0.3
61 06/30/05 3:47 AM 06/30/05 3:47 AM 0.01 0.3
62 07/05/05 6:18 PM 07/05/05 11:18 PM 0.75 19.1 Runoff estimated from barrel content
63 07/06/05 3:18 PM 07/06/05 3:18 PM 0.03 0.8 NDC - system down from lightning strike
64 07/07/05 8:53 PM 07/08/05 4:18 PM 1.13 28.7
65 07/09/05 4:38 AM 07/09/05 4:38 AM 0.01 0.3
66 07/14/05 5:14 PM 07/14/05 5:54 PM 0.03 0.8
67 07/16/05 2:29 PM 07/16/05 2:49 PM 0.64 16.3
68 07/16/05 9:34 PM 07/17/05 5:49 AM 1.31 33.3
69 07/17/05 2:09 PM 07/17/05 3:29 PM 0.21 5.3
70 07/21/05 2:49 PM 07/21/05 3:19 PM 0.08 2.0
71 07/21/05 6:09 PM 07/21/05 6:09 PM 0.01 0.3
72 07/22/05 8:44 PM 07/22/05 8:44 PM 0.02 0.5
73 07/23/05 2:49 AM 07/23/05 2:49 AM 0.01 0.3
74 07/25/05 1:19 AM 07/25/05 9:09 AM 0.68 17.3 Resumed data collection.
75 07/26/05 9:32 PM 07/26/05 9:52 PM 0.04 1.0
76 07/27/05 2:07 PM 07/27/05 3:37 PM 0.03 0.8
77 08/07/05 6:47 AM 08/07/05 7:57 AM 0.04 1.0
78 08/08/05 6:52 AM 08/08/05 12:47 PM 0.29 7.4
79 08/11/05 2:41 PM 08/11/05 3:06 PM 0.31 7.9
80 08/16/05 6:56 AM 08/16/05 4:36 PM 0.98 24.9
81 08/17/05 6:26 AM 08/17/05 6:26 AM 0.01 0.3
82 08/19/05 3:11 AM 08/19/05 10:51 AM 0.27 6.9
83 08/20/05 2:06 AM 08/20/05 2:06 AM 0.01 0.3
84 08/20/05 9:26 PM 08/20/05 10:36 PM 0.03 0.8
85 08/27/05 3:27 PM 08/28/05 4:02 AM 0.21 5.3
86 08/29/05 7:32 PM 08/29/05 9:07 PM 0.05 1.3
87 08/30/05 7:27 AM 08/30/05 10:37 AM 0.04 1.0
88 08/30/05 11:42 PM 08/31/05 4:52 PM 1.43 36.3
89 09/15/05 2:33 AM 09/15/05 2:38 AM 0.04 1.0
90 09/16/05 2:43 PM 09/16/05 4:38 PM 0.14 3.6
91 09/18/05 7:13 AM 09/18/05 7:33 AM 0.02 0.5
92 09/24/05 5:21 PM 09/24/05 5:41 PM 0.02 0.5
93 09/26/05 9:21 AM 09/26/05 6:33 PM 0.48 12.2
94 09/29/05 7:18 AM 09/29/05 9:08 AM 0.25 6.4
95 10/07/05 6:46 AM 10/08/05 11:21 AM 3.2 81.3 Barrels emptied to prevent overflowing.
96 10/08/05 5:36 PM 10/08/05 6:21 PM 0.02 0.5
97 10/12/05 2:44 AM 10/12/05 9:34 AM 0.07 1.8
98 10/13/05 8:19 PM 10/14/05 8:24 AM 0.04 1.0
99 10/14/05 5:39 PM 10/14/05 5:59 PM 0.02 0.5
100 10/20/05 10:21 AM 10/20/05 10:21 AM 0.01 0.3
101 10/20/05 8:41 PM 10/21/05 6:01 AM 0.16 4.1
102 10/21/05 10:31 PM 10/22/05 7:31 PM 0.6 15.2
103 10/24/05 4:46 AM 10/24/05 1:21 PM 0.22 5.6
104 10/24/05 7:56 PM 10/25/05 9:21 AM 0.48 12.2
105 10/25/05 5:31 PM 10/26/05 3:26 PM 1.35 34.3
3-4
Event
number
Event
date
start
Start
time
Event
date
end End time Total Precipitation
(in.) (mm) Comments
106 11/01/05 5:37 PM 11/01/05 8:02 PM 0.2 5.1
107 11/02/05 9:12 AM 11/02/05 9:12 AM 0.01 0.3
108 11/06/05 4:03 PM 11/06/05 4:18 PM 0.02 0.5
109 11/09/05 7:38 AM 11/09/05 9:03 PM 0.23 5.8 NDC - barrel valves open.
110 11/14/05 11:42 PM 11/15/05 9:47 AM 0.69 17.5
111 11/16/05 7:27 AM 11/16/05 4:57 PM 0.51 13.0
NDC - No data collected
Analysis
Roof Runoff Volumes
For 26.9 in. (683 mm) of recorded precipitation, there was a corresponding mean value of 12.7 in. (323 mm) with a
standard deviation of 2.8 in. (71 mm) of green roof runoff compared to a mean of 23.1 in. (587 mm) with a calculated
standard deviation of 1.7 in. (43 mm) for the flat asphalt roofs. (Note: Standard deviation values were also calculated
for two replicates of the flat asphalt roof, while recommendations are typically for a minimum of three). The green
roofs retained 52.6% while flat asphalt roofs retained 14.1% of the precipitation. It is not surprising that some
retention should occur with rolled asphalt roofing; all surfaces have “nooks and crannies” where initial abstractions
(losses) of precipitation fill before runoff can occur. Additionally, some precipitation likely evaporated from roof
surfaces, and some most likely was lost to splashing over the edges. The detention roof had 21.8 in. (554 mm) of
runoff based on 26.1 in. (554 mm) of precipitation (excluding events from May 19, 2005 through May 24, 2005 when
transducer not working properly), only retaining 16.3 %. The media-only roof had runoff of 15.8 in. (400 mm) based
on 22.4 in.(569 mm) of precipitation (excluding February and events after October 25, 2005 when transducer not
working properly) retaining 29.7%.
Rainfall retention by green roof buildings varied from month to month. Retention in cool weather months (January,
through March, October and November) was less than in the warm weather months (April through September)
(Figure 3-1). During the summer months almost no runoff from green roofs was observed; however, as might be
expected, during wetter periods some runoff from the green roofs was observed. For example, runoff in August was
18% of the 3.65 in. (93 mm) of precipitation, as compared to September, which was only 4% of the 0.91 in. (23 mm)
of precipitation. In the cool months, drier periods increased retention and produced less runoff, than during wetter
periods. For example, only 44% of the 1.7 in. (43 mm) of precipitation in February generated runoff, while runoff of
71% and 82%, respectively was generated in the month of March with 2.54 in. (65 mm) of precipitation, and January
with 4.46 in. (113 mm) of precipitation.
In the warm months, nearly all precipitation was retained by the green roofs. In contrast there was no real difference
in retention by flat asphalt roofs between cool and warm season periods (Figure 3-2). Error bars represent the
standard deviation for the three green roofs and two flat asphalt roofs for each month. These results are similar to
results from monitoring of these roofs before the start of this study (Denardo et al., 2005) and other stormwater
retention studies in the U.S. and Europe which also suggested that about 50% of the annual precipitation in certain
climates can be retained by a green roof.
In addition to the green and flat asphalt roofs, the unplanted media roof section had similar runoff to the planted green
roofs in cool months (within one standard of deviation green roof runoff), but much more runoff from April through
September (exceeded standard of deviation of green roof runoff in Figure 3-2). Because the difference between the
media section and green roofs is the plants, it is not surprising that the difference in runoff would be limited to periods
when the plants are actively growing and transpiring. The media appear to play a role in some runoff retention when
compared to flat roofs in six of nine directly comparable months, though this data is not as robust due to limited
sample size.
3-5
Figure 3-1 Retention and runoff from green roofs (percentage of average monthly precipitation)
Figure 3-2 Monthly average precipitation and runoff for green and asphalt roofs including unplanted media roof section
3-6
This suggests that although some of the stormwater retention of a green roof can be obtained by simply putting an
equivalent amount of media on the roof, the effectiveness will be less than if plants are used. Clearly, the plants serve
more than an aesthetic purpose on the roof.
The runoff from the green roofs is much more variable with a cumulative monthly standard deviation (error bars in
Figure 3-2) of 2.82 compared to the runoff from the flat asphalt roof with a standard deviation of 1.68. Even though
the green roofs in the study were constructed to be identical, there were sometimes fairly large differences in retention
for any given precipitation event. There were no obvious reasons for these differences. Plant cover and health was
similar for all the green roof buildings, and the retention differences were not related to a specific building. Green
roof design contributes to this variation, as it relies on non-uniform media and random survival rates of individual
plantings.
Error bars may not capture all the potential variation in measurements of the precipitation and runoff monitoring
system of the different roof types during the experiments. There was some drift in the pressure transducer values.
One prominent example of drift is that of the flat-roof pressure transducers which appears to have decreased total
accumulation on May 15, 2005 by at least -0.2 in. (5 mm) Another case of drift by the pressure transducer of the
media roof for June 6, 2005, resulted in a 0.6 in.(15 mm) gain, as pictured below in Figure 3-3. This explains, at least
in part, why the unplanted media roof total exceeded the flat roof totals for May and June in Figure 3-2. The results
show that while drift may be a concern particularly for individual events, in terms of annual values, drift is only one
source of error potentially contributing to the overall variability. Another factor is that some data were not duplicated
for each storm event. There are no duplicates for the media and retention rooftop half-roof sections, this limits the
degree of confidence for individual storm totals from these roofs. Precipitation measurements, typically a stand alone
measurement, were also not duplicated, though this does not completely explain why, for some events, runoff totals
unexpectedly exceeded precipitation totals. This was most likely due to antecedent or freezing precipitation
conditions; the latter is discussed in more detail below. As noted earlier, however, there were precipitation events that
were not monitored for runoff; therefore, discerning effects of antecedent conditions of several storms for all roofs
was not always possible. For seasonal and annual assessments, individual storms are treated as replicates, while for
comparison of individual events, only the asphalt and green roofs are compared with rainfall totals as a baseline.
Figure 3-3 Measurements for June 6, 2005
3-7
The retention and runoff from green roofs also varied from one precipitation event to another. Figure 3-4 shows
precipitation events for which runoff was generated. Trend line analysis fit using second order polynomial shows that
the lines are parabolic in nature (Microsoft
®
Office Excel 2003). Green roof runoff data were not collected for events
between July 5, and July 22, 200505, or after November 16, 2005 and error bars represent the standard deviation for
three green roofs for each event measured. The green roof runoff trend line reinforces the observed seasonal effect
demonstrating that there was little or no runoff from the green roof during the warm summer months. The regression
coefficients were quite low (r
2
~ 0.1), due to the randomness of the quantity and interval between storm events.
Figure 3-4 Plot of precipitation versus runoff from green roofs for paired events
An analysis of cumulative probability was performed, using 72 of the events and sorting based on the size of the
precipitation events. The graph in Figure 3-5 is seasonally independent and used a rise-to-maximum function. A
major conclusion from the graph is that green roofs have little or no runoff for 65% of the precipitation events up 0.4
in.(10 mm) of precipitation, regardless of season (with few exceptions, e.g., due to wet antecedent conditions on
October 24, 2005 at the 0.51 percentile). A regression analysis for the precipitation event totals resulted in an r
2
value
that is very high, 0.98, while the r
2
for the retained precipitation is 0.61. This regression value for retained
precipitation is lower for two reasons. First, larger precipitation events have more runoff and greater variability in the
green roof retention, and secondly, this sorting did not account for period between storms. This graph shows there
will be more green roof runoff for larger storms; however, this is a volume assessment only, and does not quantify
initial abstraction and time to peak that a green roof provides in delaying runoff. Two points of the retained
precipitation were not used, as these values were negative, i.e., there was more runoff than precipitation. As both
occurrences were in winter, January and March, they were most likely the result of the additional release of snow or
ice melt from the roofs.
3-8
Figure 3-5 Probability plot of precipitation versus retained precipitation by green roofs
During winter storms the media in a green roof may freeze and slowly release moisture over an extended period.
Snow may also accumulate and melt over an extended period. Figure 3-6 is a graph of several events during the
period from March 23 through March 29, 2005, which included three storms consisting of snow (March 23 through
March 25) and mixed precipitation (March 28). Precipitation and runoff (with out error bars, which would increase
obscuration of the precipitation line) were plotted against each other on the same scale to show the retention and
release as runoff of precipitation. Liquid precipitation (measured by the weather station tipping bucket) was about 1.6
in. during this period. Essentially, all the precipitation became runoff from the green roofs, but the runoff was clearly
delayed as the snow melted. Interestingly, the runoff from the flat asphalt roofs followed very closely the runoff from
the green roofs. All structures in this study were heated in winter and air-conditioned during the summer. The
controlling mechanism for runoff in the time period of this graph was probably ambient temperature and sunlight as
the predominant form of precipitation was snow and freezing rain. While these structures may not have had the most
representative roof in terms of more sophisticated buildings, this tends to confirms anecdotal observations that snow
on green roofs will generally melt with time.
3-9
Figure 3-6 Precipitation and green and asphalt roof runoff from snow and mixed precipitation events in March, 2005
In addition to seasonal effects, runoff from green roofs was also greater for large storms or from precipitation events
in close succession. The effect of the period between storms can be observed by looking at the results of select
individual storms. For example (Figure 3-7), on August 16, 2005 there was 0.98 in. (25 mm) of precipitation and
nearly all the precipitation was retained by the green roof (96%). In contrast (Figure 3-8), the August 30, 2005
precipitation was about 1.4 in. (36 mm) and the green roofs only retained 57% of the precipitation. Since the green
roof media holds about 30% water by volume and the roofs were 3.5 - 4 in. (8 - 10 cm) deep the total water holding
capacity of the roof should have been about 1.0 to 1.5 in. (25 to 38 mm), of which some is tightly bound and some is
readily exchangeable, i.e., to be taken up by plants or evaporated. As media moisture is recharged through rain or lost
to evaporation and taken up by plants, the potential to retain water for the next storm changes. The implications
become clear in a closer examination of these two August precipitation events. Almost all the 0.98 in. (25 mm) rain
was retained on August 16 due to no rain for five days. On August 30, with 1.4 in. (36 mm) of new precipitation and
0.3 in. (8 mm) from the preceding days, only about 1.0 in. (25 mm) was retained, while 0.6 in. (15 mm) of runoff was
released from the green roofs. This is also the case for the large storm on October 7-8, where 2.2 in. (56 mm) of the
3.2 in. (81 mm) of precipitation were not retained by the green roofs (Figure 3-9).
Note, that there is a 0.2 in. (5 mm) gap between the rainfall total and asphalt roof discharge at 3:00 AM on Figure 3-7.
Some of the rain falling on the flat roof at peak intensity must have either splashed off or potentially spilled over the
size due in part to the flatness of the surface. Because of the more complex three-dimensional nature of the green
roof, i.e. plants and media, it is assumed the green roof actually captured more rainfall than the asphalt roof. This is
supported by data (not shown) from the split roof where the media and detention sections approached 1.2 in. (30 mm),
of capture at 3:00 AM.
3-10
Figure 3-7 Precipitation and runoff from green and flat asphalt roofs on August 16, 2005
Figure 3-8 Precipitation and runoff from green and flat asphalt roofs on August 30, 2005
3-11
Figure 3-9 Precipitation and runoff from green and flat asphalt roofs on October 7-8, 2005
Peak Hydrograph Analysis
In a larger storm, the green roof typically will retain nearly all precipitation at the start of the event until storage
capacity is exceeded. Once storage capacity, i.e., initial abstraction, is exceeded and runoff begins to occur, nearly all
the additional precipitation is released as runoff as previously demonstrated in Figures 3-8 and 3-9. Runoff from
green roof is thus delayed from the start of the storm by a period which corresponds to filling the media storage
capacity. Runoff is also delayed because precipitation must pass through the vegetation canopy (interception and
stem flow), root zone and the media before it reaches the drainage system. Delay for flow through the green roof
system will be influenced by the size of the roof, vegetation cover, the path, hydraulic conductivity of the media, and
the drainage layer. Once precipitation reaches the drainage layer, the water flows through the system to the gutter and
downspouts. The vertical component is responsible for the initial abstraction and runoff delay, while the lateral
component appears similar to other roofing systems.
During the second most intense rain of the study, the green roof runoff peak flow was not only delayed but generally
was attenuated as well, reaching only about half of the peak rainfall as demonstrated in Figure 3-10 and 3-11 (these
figures used runoff values that were corrected to increase monotonically). The peak intensity of this rainfall event
approached 3 in/hr (~ 75 mm/hr) as can be seen in Figure 3-11, with peak asphalt roof coming at the next five-minute
time interval and the green roof runoff peak coming at the next five-minute interval after that. The green roof also
released runoff more gradually than the asphalt roof, which dropped off as rapidly as the precipitation intensity in this
event.
3-12
Figure 3-10 Hydrograph for August 30-31, 2005
Figure 3-11 Hydrograph of most intense period of August 30-31, 2005 event
3-13
Table 3-2 lists events over 0.5 in. (13 mm) excluding storms where snowfall and freezing were an issue (storm
number can be cross-referenced to Table 3-1). The clear implication is that in many ways the green roof will act to
retain, detain and attenuate WWF; however, some events can exceed this capacity, reducing the performance of green
roofs. Calculated time of concentration values for the asphalt roof ranged from 1.0 to 2.6 minutes for these storms,
and were all less than the five-minute time interval used for the runoff measurements; therefore, only rudimentary
peak flow analysis could be performed, as demonstrated in Figures 3-10 and 3-11. As stated in the
“Recommendations”, larger roofs should be monitored for hydrology routing. Also, decreasing the time interval may
help, but it may also create more noise.
While green roof runoff performance is subject to the rainfall pattern, seasons and interval between storms, peak flow
attenuation was demonstrated in most storms. Short, intense rainfall events, especially the events of July 15 and June
6, had short times to peak green roof runoff but still reduced runoff volume, while some of the longer events in the
fall (October 25, 2005 and November 14 and 16, 2005) had slow times to peak but limited attenuation. In very large
sustained events, exemplified by the August 30 and October 7 events, which demonstrated a typical S curve rainfall
intensity pattern (highest rainfall intensity in the middle of the storm), a delay in peak rate was still evident, though
muted, and there was still runoff reduction, though less than 50%.
Summary
Green roofs in the study retained over 50% of the total precipitation during the study period. During summer months
nearly all the precipitation was retained. During the winter, retention was smaller (<20%). Seasonal effects appear to
be a result of snow or freezing conditions, otherwise green roofs effectively retained up 0.4 in. (10 mm) regardless of
season. For storms producing green roof runoff, peak flows rates were delayed by the green roofs and in many cases
peak flow volumes were also attenuated. Green roofs were most effective at delaying time to peak and reducing peak
flows when they when not fully saturated by rainfall events on previous day. The response of green roofs is subject to
season, seasonal rainfall pattern, and interval between rains, but still offers benefits of attenuation and
evapotranspiration even during non-growing seasons, and performed exceptionally well during the summer.
3-14
Table 3-2 Synopsis of Rain Events Greater Than Half an Inch
Peak Runoff
Event Event date Rainfall
Green
Roof
Runoff
Total
Peak
Rainfall
Intensity
(in/hr)
Intensity (in/hr)
Asphalt
Roof Green
Roof Rainfall
Time until Peak
Runoff After Rainfall (hr:min)
Comments
Number start Total (in.) (in.) (hr:min) Asphalt Roof Green Roof
95 10/7/2005 3.2 2.2 0.96 0.96 0.92 4:25 0:05 0:15 Barrels emptied to prevent overflows
88 8/30/2005 1.43 0.61 2.88 2.04 1.64 2:15 0:05 0:10
105 10/25/2005 1.35 1.0 0.36 - - 0.16 10 - - 19:10 Steady rain, asphalt data not
representative.
80 8/16/2005 0.98 >0.1 0.6 0.6 0.08 4:15 0:10 0:00
110 11/14/2005 0.69 0.43 0.24 0.36 0.24 0:10 0:10 2:30 Steady rain
74 7/25/2005 0.68 0.18 2.76 1.38 0.32 0:10 0:05 0:05 0.33 in. in 10 min at start of storm
Double peak, 2nd peak 2.04 in/hr,
51 6/6/2005 0.67 >0.1 1.44 1.62 0.2 0:10 0:05 0:05 1.74 in./hr asphalt roof 0.12 in/hr
green roof runoff.
102 10/21/2005 0.6 0.39 0.24 0.24 0.24 0:45 19:20 20:25 Rain on and off for 21 hrs.
111 11/16/2005 0.51 0.46 0.36 0.3 0.24 0:50 0:05 1:20 Rainfall previous day
3-15
Chapter 4 Green Roof Effects on Runoff Quality
Introduction
Runoff samples were collected from six small buildings at the Center for Green Roof Research at Rock Springs, PA
during the period from January 2005 through May 2006. Runoff from three green roofs, two flat-asphalt control
roofs, and one roof divided between detention and a green-roof system without plants, was collected in rain barrels
and sampled for various water constituents following various precipitation events. Comparisons of runoff
concentrations are presented among roof types. Runoff nutrient loadings in pounds per acre (lb/acre) were calculated
from measured concentrations and measured runoff quantities of the flat asphalt and green roofs for comparison.
Runoff water quality varied with both sample event and roof type.
Runoff Summary
Sampled precipitation events listed in Table 4-1 covered from February 14, 2005 to May 15, 2006, ranged from 0.07
in. (2 mm) to 1.31 (52.1 mm) for individual events (some multiple events are include din this analysis) and included
rain, snow and freezing rain. Total precipitation from the sampled storms was 16.15 in. (410.2 mm), though, as noted
in Table 1, lesser amounts are used for comparative volume analyses. Table 4-2 lists individual storm and total runoff
values. Runoff volumes were not obtained for all events due to errors in collection systems, (e.g., a valve left open
for detention runoff on May 23, problem with media runoff measurements in October and November) or data
collection system problems (i.e., February 15, March 29 and July 16 through 17). Total runoff collected from
sampled precipitation events was 202.9 gal (768.0 L) from green roofs, 400.3 gal (1515.1 L) from flat asphalt roofs,
175.9 gal (665.8 L) from the detention section (half roof), and 109.2 gal (413.3 L) from the media roof section (half
roof). These runoff volumes correspond to 48% of the total precipitation for green roofs and 94% for flat roofs based
on 425 gal (1610 L) of precipitation, 88% for the detention section based on 401 gal (1520 L) and 67% for the media
roof section based on 328 gal (1240 L), respectively. These values are similar to the ratios cited for the full study
runoff volume analysis presented in Chapter 3. Total runoff collected also varied seasonally as in the larger runoff
study with a reduction in runoff from green roofs in April through September (4% of the precipitation). In contrast
runoff from flat asphalt roofs was fairly consistent throughout the year, only dipping to 89% during this period.
Table 4-1 Precipitation Events Sampled for Water Quality Monitoring
Sample Number Precipitation Sample Total Field Laboratory Comments
(Event
Numbers)
1
Dates Date Precipitation
(in.) (mm) Analysis Analysis
1 (11) 2/14/05 2/15/05 0.73 18.5 Yes System failure (no volumes)
2 (17) 3/4/05 3/7/05 0.07 1.8 Yes Snow - may include previous
snow
3 (18 -20) 3/7 -12/05 3/15/05 0.26 6.6 Yes Snow - may include previous
snow, problem with volumes
4 (22-28) 3/23-28/05 3/28/05 1.75 44.5 Yes Yes Snow/rain -freezing rain
4-1
Sample Number Precipitation Sample Total Field Laboratory Comments
(Event
Numbers)
1
Dates Date Precipitation
(in.) (mm) Analysis Analysis
5 (28*) 3/28/05 3/29/05 0.85 21.6 Yes System failure (no volumes)
6 (32-34) 4/22-24/05 4/26/05 0.55 14.0 Yes
7 (40-43) 5/19-22/05 5/23/05 0.51 13.0 Yes No data for detention roof
8 (44) 5/23/05 5/24/05 0.32 8.1 Yes No data for detention roof
9 (51) 6/6/05 6/7/05 0.67 17.0 Yes
10 (53) 6/9/05 6/10/05 0.15 3.8 Yes
11 (62) 7/5/05 7/6/05 0.75 19.1 Yes Runoff estimated by barrel
content
12 (67) 7/16/05 7/16/05 0.64 16.3 Yes Lightning strike (no volumes)
13 (68) 7/17/05 7/17/05 1.31 33.3 Yes Lightning (system down)
14 (80) 8/16/05 8/17/05 0.98 24.9 Yes Yes
15 (82) 8/19/05 8/19/05 0.27 6.9 Yes
16 (93) 9/26/05 9/27/05 0.48 12.2 Yes
17 (94) 9/29/05 9/29/05 0.25 6.4 Yes
18 (95) 10/7/05 10/8/05 3.2 81.3 Yes
19 (103-105) 10/24-26/05 10/27/05 2.05 52.1 Yes
20 (110-111) 11/14-16/05 11/16/05 1.2 30.5 Yes
21 (NA) 4/24-25/06 4/25/06 0.14 3.6 Yes Yes
22 (NA) 5/11/06 5/13/06 0.61 15.5 Yes No field measurements
23 (NA) 5/14-15/06
2
Totals Precipitation 5/16/06 0.4
14.22 10.2
361.2 Yes No field measurements
1 - Event number from Table 3-1.
2 - Excludes flat and green roof samples for which no volume data were generated; 13.4 in. (340 mm) for detention roof and 11.0
in. (279 mm) for media roof.
NA - Events not listed in Table 3-1.
Table 4-2 Volumes for Water Quality Monitoring
Sample Sample Green roof Runoff Flat roof runoff) Detention runoff Media runoff
Number Date (Average of 3 ) (Average of 2) (Stdev)
(gallon) (liter) (gallon) (liter) (gallon (liter) (gallon) (liter)
(Stdev) Roof dimensions (ft
2
) (m
2
)
48 14.6 48 14.6 24 7.3 24 7.3
1 2/15/05 N/A N/A N/A N/A
2 3/7/05 0.13 (0.06) 0.49 1.7 (0.85) 6.4 3.3 12.49 0 0.00
3 3/15/05 N/A N/A N/A N/A 0.00
4 3/28/05 47.50 (8.13) 179.8 51.0 (0.99) 193.0 28.7 108.3 21.4 81.0
5 3/29/05 N/A N/A N/A N/A
6 4/26/05 0.37 (0.12) 1.4 10.45 (0.35) 39.6 8.8 33.31 0.6 2.27
7 5/23/05 0.43 (0.15) 1.6 11.4 (1.27) 43.2 N/A 1.2 4.54
8 5/24/05 0.43 (0.12) 1.6 6.50 (1.41) 24.6 N/A 1.9 7.19
9 6/7/05 1.43 (0.87) 5.4 23.0 (0.42) 88.2 9.5 35.96 19.5 73.81
10 6/10/05 0 0 4.2 (1.56) 15.9 1.9 7.19 1.5 5.68
11 7/6/05 0.5 (0) 1.9 13.5 (2.12) 51.1 2.0 7.57 3.0 11.36
12 7/16/05 N/A N/A N/A N/A
13 7/17/05 N/A N/A N/A N/A
14 8/17/05 1.17 (1.36) 4.43 28.9 (1.13) 109.4 12.8 48.45 0.4 1.51
15
16
3
8/19/05
9/27/05 0.37 (0.31)
0.67 (0.25) 1.40
2.5 7.8 (0.71)
15.7 (0.85) 29.5
59.4 2.5
7.2 9.46
27.25 0.4
1.9 1.51
7.19
17 9/29/05 0.7 (0.3) 2.7 7.0 (0.28) 26.5 2.7 10.22 1.1 4.16
18 10/7/05 65.1 (15.21) 246.4 89.3 (1.91) 343.1 39.4 149.13 39.9 151.0
19
20
3
10/27/05
11/16/05 49.47 (11.25)
29.07 (11.74) 187.2
110.0 44.9 (0.71)
36.25 (0.78) 167.0
137.2 25.7
15.8 97.27
59.80 NA
NA
21 4/25/06 0.77 (0.12) 2.9 3.65 (0.49) 13.82 1.0 3.79 1 3.79
22 5/13/06 3.4 (1.83) 13 31.4 (0.85) 118.9 11.7 44.28 12.7 48.07
23 5/16/06 1.4 (0.46) 5.3 12.0 (0.85) 45.4 2.9 10.98 2.7 10.22
Totals 202.9 (52.3) 768.0 400.3 (17.5) 1515.1 175.9 665.8 109.2 413.3
NA - not applicable.
4-2
Methods and Analyses
Samples were either analyzed in the field for pH, EC, turbidity, color, and nitrate or were sampled and tested for
nutrients and other ions in the Penn State Agricultural Department Analytical Laboratory. An Orion electrode and
meter measured pH, EC was measured using a Hach portable meter and probe, turbidity was measured with a Hach
portable meter, color was measured using a Hach portable spectophotometer, and nitrate was measured with a Hach
specific ion electrode and Orion meter.
Color and Turbidity
The most noticeable difference between runoff from green and asphalt roofs was the color of the water. The runoff
appeared to be very similar to the leachate obtained from potted plants or other containerized planting systems. Color
analysis was conducted because green roof runoff had a distinct yellow color. Green roof media contributed
inorganic and humic substances to the runoff resulting in a yellow-to-brown coloration. Color is a function of the
organic components (humic acids), primarily compost incorporated in the media to enhance plant growth and
establishment, and will likely decrease with time as the media ages. Field analysis was conducted using a Hach DR
890 colorimeter which report in units of 0 to 500 Platinum-Cobalt (Pt-Co) color units (based on the 500 Pt-Co
standard solution). Samples over the limit were diluted to within the limit, usually requiring a 50% to 75% dilution
and were then calculated for full concentration. Expected standard accuracy is plus or minus 15 Pt-Co color units
with a method detection limit of 25 Pt-Co color units. A yellow color was obvious in all runoff from the green roofs
and the media roof section (Figure 4-1). Although the green roof runoff was colored and was detected by the
colorimeter, it was quite clear. The intensity of the runoff color varied with the storm event. Figure 4-2 shows that
events in the cooler seasons generally had less color while small events with little runoff, i.e., May 24, 2005, had the
most color.
Figure 4-1 Average runoff color from all roof types
4-3
Figure 4-2 Runoff color comparison between green roofs and flat asphalt roofs
The turbidity of the runoff was more variable for green roofs and the media roof section compared with the flat
asphalt roofs (Figure 4-3). Turbidity from green roofs, media roofs, and flat roofs was very similar ranging between
0.8 and 5.6, with the exception of two events May 23 and 24, 2005 when green roof values peaked at 24.4 and 18.1,
respectively. This coincides with the May 24, 2005 peak in runoff color.
Figure 4-3 Average turbidity of runoff samples from all roof types
4-4
Events where green roof turbidity was elevated were events where very little runoff was collected, e.g., May 23 and
24, 2005 events. It seems likely that the elevated turbidity in these samples was related to the small amount of runoff.
The color of green roof runoff is likely to limit the direct use of this water for anything other than landscape irrigation
without some additional treatment or filtration. If the green roof runoff was planned to be the primary source of a
grey water system for flushing toilets for example, the color might be aesthetically unpleasant to users.
pH
The average pH for three rain events was 4.7 (March 3, April 26 and May 24, 2005). Green roofs consistently raised
the pH of runoff compared to flat asphalt roofs. The pH of runoff from the asphalt roofs ranged from 4 to 7, while
runoff from the green roofs was consistently above 6.4 (Figure 4-4). This is clearly a function of the media. The
media roof section had runoff sample pH values similar to those obtained from the planted green roof systems (Figure
4-5). Runoff from the rooftop detention was consistently lower than the other roofs. It is not clear why this runoff
had a lower pH than runoff from the flat asphalt roofs. It is possible that the roofing tar used to waterproof the
detention basin influenced runoff pH, and conversely, the asphalt roofing of the flat roofs may have raised the pH of
the runoff due to the small granules of rock used. Note that the March 28, 2005 data point is several storms some of
which had frozen precipitation, so when precipitation could not contact stones of asphalt roof due to sheet flow over
frozen surfaces, buffering capacity was by-passed.
Figure 4-4 pH of runoff from green and flat asphalt roofs
4-5
Figure 4-5 Average pH of runoff from all roof types
Buffering pH is one of the most significant and consistent effects of a green roof on water quality. The ability of a
green roof to neutralize acid precipitation is a major potential benefit for using these roofing systems in areas like the
Northeastern U.S. which are subject to acid rain, as can be seen in Figure 4-6. Acid rain has been demonstrated to
have negative effects on surface waters and may contribute to metal concentrations in runoff. Acid in the rain comes
in contact with metals in ventilation systems, roofs, drainage systems, and other building materials. Green roofs, if
widely implemented in urban areas with acid precipitation, could potentially reduce these impacts. It should be noted
that the limited soil depth of an extensive green roof would have a finite ability to maintain this buffering capacity.
As acid precipitation is neutralized, the media buffering capacity will be expended and eventually the media will no
longer be able to buffer the runoff. This will not only influence pH of the runoff, but will also affect plant growth on
the roof. Many of the plants used on extensive green roofs do not grow well in acid soils, so replacing the buffer in
the media through liming may be a necessary part of the long-term management or maintenance of a green roof.
Roofs monitored in this study were established in 2000 though 2001. To date, there is no evidence that buffering
capacity is running out; however, there is evidence from older, European green roofs that suggests that media
acidification will eventually occur. Determining the limiting buffering capacity and other maintenance of the green
roof is discussed later in Chapter 5.
4-6
Figure 4-6 United States map of acid rain (National Atmospheric Deposition Program, 2006)
Electrical Conductivity
The EC of runoff from green roofs was significantly higher than conductivity of runoff from flat asphalt roofs (Figure
4-7). Electrical conductivity, reported in units of milliSiemens per centimeter (mS/cm), is a measure of the total salts
or ions in solution. Since the green roof runoff was filtered through a media (soilless substrate) containing plant
nutrients, roots, organic matter, and clay-based aggregate, it is not surprising that some salts are leached into the
runoff.
Conductivity of the runoff also varied significantly with some precipitation events resulting in much higher EC than
others (Figure 4-8). The highest EC in green roof runoff occurred in events where only a small amount of runoff was
collected, i.e., May 23 and May 24. It is also interesting to note that there was a slight, general trend toward reduced
EC in runoff from the media section of roof from the beginning of the evaluation period to the end (Figure 4-9). The
average for first 10 media runoff events is 0.127 mS/cm versus 0.98 for remaining 10 events. Since this roof section
was relatively new at the beginning of the evaluation period, less than a year old, compared with the green roofs
which were five years old, it potentially suggests that new media of a new roof will leach more salts than the same
roof once established. This has implications for future evaluations of new roofs, where water quality of the runoff in
the first year may not be representative of the runoff of a mature established roof.
4-7
Figure 4-7 Electrical conductivity of runoff from all roof types
Figure 4-8 Electrical conductivity of runoff from green and asphalt roofs
4-8
Figure 4-9 Electrical conductivity from green and media roofs
Nitrate
On average, over the entire period sampled there was no significant difference between green and asphalt roofs for
runoff nitrate concentration (Figure 4-10). The media roof section runoff, in contrast was significantly higher than
any other sampled roof. It is likely that the nitrate in runoff from the media roof section was derived from breakdown
of the compost amendment to the media. In established green roofs, the organic content of the media is likely stable
with breakdown matched by growth of the plants. In the planted roofs, nitrate is being used by both the plants and
associated microbial communities, and is thus not as readily leached. The green roofs in this study had received only
small amounts of supplemental fertilizer on an annual basis from 2001 to 2004, and thus, the systems were likely
nitrogen starved and any available nitrogen would tend to be used by the biotic community. This suggests that new
roofs without full plant cover may contribute nitrate to runoff, while established roofs, if managed correctly, may not.
Some sampled events resulted in higher nitrate concentrations in runoff from green roofs while runoff from flat
asphalt roofs had higher nitrate concentrations for other events (Figure 4-11). Nitrate in the runoff from asphalt roofs
was generally higher in winter months, while summer runoff had lower nitrate concentrations. This trend was not
evident in runoff from green roofs. As a result, nitrate concentrations from green roofs were generally higher than
runoff from flat asphalt roofs in the summer months, while in general the reverse was true in the winter.
Loading calculations were performed two different ways for comparison of runoff from the green and asphalt roofs;
nitrate was the only constituent for which the detention and media roof was also compared, due to the limited data
sets for other constituents. Average loading was calculated by multiplying the volume of runoff by the concentration
per event and averaging events. For annual loading, the average concentration was multiplied by the annual runoff.
Annual runoff was calculated from the average rainfall, 44 in. (1100 mm), multiplied by the fraction of runoff from
the roofs derived from analysis presented in Chapter 3 (0.474 for green, 0.859 for asphalt, 0.837 for detention, and
0.703 for media). The roof area was scaled to acres (results in pounds per acre). In addition, ranges to these estimates
are provided, based on standard deviations of the concentrations. Individual event loading values are presented in
Figure 4-12 for green and asphalt roofs.
4-9
Figure 4-10 Average nitrate concentration in runoff from all roof types
Figure 4-11 Nitrate concentration in runoff from green and flat asphalt roofs
4-10
Figure 4-12 Nitrate (pounds per acre) in runoff from green and asphalt
Although runoff concentration was higher in the runoff from green roofs during summer when runoff was released,
the total mass of nitrate in runoff was less because there was far less runoff from the green roofs than from the flat
asphalt roofs during this period (Figure 4-12). Due to the decreased runoff, limited nitrate was released from green
roofs during the summer, and even in winter when runoff resulted in release of nitrate to the environment, the total
amount was less from green than asphalt roofs.
Over the study period for the sampled storms, the summation of nitrate in monitored events was 1.0 lb/acre of roof for
green roofs, 3.9 lb/acre (4.4 kg/ha) for flat asphalt roofs, 2.8 lb/acre (3.1 kg/ha) for the rooftop detention and 9.8
lb/acre (11 kg/ha) for the media roof section. The media roof section not only had the highest concentration of nitrate
in runoff but also resulted in the largest quantity of nitrate released, and this is with fewer monitored events. This is
further demonstrated in Figure 4-13 which presents the calculated annual loadings. In addition, the average loadings
per event were 0.06 lb/acre (0.07 kg/ha) of roof for green roofs, 0.24 lb/acre (0.27 kg/ha) for flat asphalt roofs, 0.16
lb/acre (0.18 kg/ha) for the rooftop detention and 0.51 lb/acre (0.57 kg/ha) for the media roof section.
The average background concentration of nitrate for the same three precipitation events as for pH was 0.87 mg/L.
Figure 4-14 is a national map of atmospheric nitrate deposition, and indicates, as expected, that the test roofs were in
an area of high nitrate deposition. Since nitrate is a major concern in surface runoff, green roofs offer the potential to
significantly reduce total nitrate being discharged from urban environments in certain parts of the country, due to
atmospherically deposited nitrate. In areas with lower nitrate concentrations in the precipitation, it is likely that the
use of green roofs may be less effective in nitrate reduction, and green roofs may even require fertilization, which
may contribute nitrogen to the runoff.
This study demonstrated that the green roof will remove and not leach nitrate with time, i.e., the media roof showed
that initial fertilization may leach nutrients, less than one year old at time of study, while the more mature green roofs,
about five years old, removed nitrate. However, further studies should be conducted on other nitrogen species.
4-11
Figure 4-13 Calculated annual loading of nitrate from test roofs
Figure 4-14 U. S. map of atmospheric nitrate wet deposition (National Atmospheric Deposition Program, 2006)
4-12
Other Water Quality Constituents
Water samples from storm-generated runoff were collected from the rain barrels and sent to the Penn State
Agricultural Analytical Laboratory for qualitative and quantitative analysis for elemental composition. The five
events ranged in individual rainfall totals from 0.14 in. (3.6 mm) to 0.98 in. (24.9 mm) with one sampling event being
comprised of several precipitation events totaling 1.75 in. (44.5 mm). Ions analyzed for five events and their
associated detection limits in mg/L were phosphorous (<0.06), potassium (<0.2), calcium, iron (<0.005), magnesium,
manganese (<0.001), sodium, and zinc (<0.005). In addition samples from two events were analyzed for sulfur.
Table 4-3 provides a summary of the analyses for five events. Not all samples provided concentration results as some
were below associated detection limits; these values have been reported at half the detection limit. It was difficult to
obtain green roof runoff samples in the growing season due to the limited runoff. Runoff from one of the green roofs
for the August 17, 2005 event was less than a tenth of a gallon as reported in Table 4-3. In general concentrations
from the green and media roofs were greater than the flat and detention roofs. The loadings were calculated in the
same way as for nitrate and are presented in Table 4-4 with comparisons between green and the flat asphalt roofs. For
values below detection limit values, one-half the value of the detection limit was used in the calculation of loadings
for comparison purposes. A discussion of the individual nutrients and ions follows Table 4-4.
Nutrients
Phosphorous (P) concentration and loading in runoff from all five sampled precipitation events was higher from the
green than the flat asphalt roofs. Phosphorous in runoff from flat asphalt roofs was low, with two events below
detection limits. Concentration in runoff from green roofs was consistent. Runoff phosphorous concentration from
the media roof section was higher than either green or flat asphalt roof runoff, mainly due to a very high concentration
in the first event sampled (March 28, 2005; 2.53 mg/L). It is likely that this was due to the media being less than a
year old. For the other measured events, phosphorous concentrations were similar for green and media roofs. For
measured events where there was little runoff from the green roofs, phosphorous in the green roof runoff was very
similar to the loading (April, 26, May, 13 and May 16, 2006) of phosphorous in flat asphalt roof runoff; however, in
storms where there was more runoff from the green roofs (March 28, 2005), phosphorous in the green roof runoff was
much higher, resulting in a higher overall average and higher sum for all the sampled storms and phosphorous from
the flat roof was below detection limits for both roofs. Calculated average and annual loadings from the green roof
clearly exceeded that of the flat asphalt roofs, by over 300%.
The average concentration of phosphorous 0.41 mg/L (CV = 0.45) in green roof runoff (Table 4-3) compares well to
the total phosphorous event mean concentration (EMC) reported for landscaped residential areas by the National
Urban Runoff Program (NURP) studies (EPA, 1983) 0.383 mg/L (CV = 0.69) and the National Stormwater Quality
Database (Pitt et al., 2004), 0.30 mg/L (CV = 1.1).
Potassium (K) in runoff from green and flat asphalt roofs was similar to results for phosphorous. Potassium
concentration was higher in runoff from green and media roof sections than flat asphalt roofs. Concentrations in
earlier events (2005) were higher than later events (2006). It is possible that because the green roofs were not
fertilized in 2005 and potassium is relatively mobile in the soil, that the available concentration decreased with time,
resulting in less potassium in the runoff in 2006. This suggests that if a green roof is to be used as a stormwater
management tool where water quality is an issue, care must be taken in the fertilizing rates of the roof. Enough
fertility must be present to support plant growth but excess will result in higher concentrations in the runoff than
might otherwise be necessary. As with phosphorous, the quantity of potassium in runoff was greatly influenced by
the amount of runoff from the green roofs, and the calculated loading from green roofs greatly exceeded the
calculated quantities from flat asphalt roofs.
4-13
Table 4-3 Summary of Concentrations and Volumes of Runoff for Five Sampled Events
Date Volume
runoff Concentrations (mg/L) (Detection limits)
(gal) (L) P
(<0.006) K
(<0.2) Ca Mg Na
(0.01) Mn
(<0.001) Fe
(<0.005) Zn
(<0.005) S
Green roofs
3/28/05 54.3 205.5 0.45 2.4 10.2 7.9 1.13 0.001 0.0025 0.0025 3.5
49.7 188.1 0.44 2.6 9.7 7.3 1.26 0.001 0.0025 0.0025 3.6
38.5 145.7 0.50 2.6 10.2 7.5 1.07 0.001 0.0025 0.0025 3.7
8/17/05 2.7 10.2 0.66 3.1 17.9 12.0 3.16 0.033 0.024 0.009 11.0
0.7 2.7 0.62 1.9 13.7 8.2 2.76 0.027 0.018 0.035 8.3
0.1 0.4 0.49 2.4 11.9 5.2 2.51 0.045 0.0025 0.026 5.4
4/26/06 0.7 2.7 0.18 0.8 13.0 9.2 0.91 0.008 0.094 0.010
0.9 3.4 0.14 0.7 12.1 8.5 1.16 0.010 0.079 0.009
0.7 2.7 0.75 2.6 25.6 12.1 1.89 0.004 0.079 0.006
5/13/06 5.0 18.9 0.52 1.5 8.2 5.1 0.70 0.005 0.045 0.014
3.8 14.4 0.35 1.8 6.3 3.9 0.89 0.012 0.021 0.018
1.4 5.3 0.25 1.3 6.1 2.8 0.51 0.029 0.008 0.015
5/16/06 1.3 4.9 0.16 1.0 12.4 7.8 0.91 0.005 0.074 0.022
1.9 7.2 0.26 1.0 12.3 8.6 1.03 0.006 0.069 0.014
1.0 3.8 0.43 1.3 18.8 11.7 1.67 0.001 0.087 0.014
Average 0.41 1.8 12.6 7.8 1.44 0.012 0.041 0.013 5.9
Standard of Deviation 0.19 0.8 5.1 2.8 0.80 0.014 0.036 0.009 3.1
Coefficient of Variation 0.45 0.4 0.4 0.4 0.55 1.136 0.882 0.697 0.5
Flat asphalt roofs
3/28/05 50.3 190.4 0.03 0.1 1.1 0.1 0.04 0.001 0.0025 0.007 0.9
51.7 195.7 0.03 0.1 1.0 0.1 0.04 0.001 0.0025 0.0025 0.8
8/17/05 29.7 112.4 0.03 0.1 2.0 0.2 0.01 0.001 0.0025 0.0025 1.4
28.1 106.4 0.03 0.1 7.3 0.7 0.06 0.004 0.0025 0.011 4.2
4/26/06 4.0 15.1 0.15 0.6 2.1 0.3 0.34 0.005 0.023 0.0025
3.3 12.5 0.03 0.1 2.4 0.3 0.09 0.001 0.013 0.0025
5/13/06 32.0 121.1 0.09 1.2 3.5 0.3 0.15 0.017 0.0025 0.0025
30.8 116.6 0.03 0.5 3.2 0.3 0.05 0.013 0.0025 0.0025
5/16/06 12.6 47.7 0.08 0.3 2.4 0.3 0.51 0.005 0.0025 0.0025
11.4 43.1 0.03 0.6 2.1 0.2 0.05 0.003 0.0025 0.0025
Average 0.05 0.4 2.7 0.3 0.13 0.005 0.006 0.004 1.8
Standard of Deviation 0.04 0.3 1.8 0.2 0.16 0.006 0.007 0.003 1.6
Coefficient of Variation 0.77 1.0 0.7 0.5 1.21 1.120 1.239 0.782 0.9
Detention section
3/28/05 28.7 108.6 0.03 0.1 0.5 0.2 0.06 0.001 0.0025 0.006 1.2
8/17/05 12.8 48.4 0.03 0.1 1.1 0.7 0.05 0.006 0.018 0.007 2.6
4/26/06 1.0 3.8 0.54 0.9 13.5 5.0 1.04 0.019 0.076 0.013
5/13/06 11.7 44.3 0.42 0.8 1.4 0.4 0.10 0.025 0.021 0.005
5/16/06 2.9 11.0 0.07 0.4 2.2 0.6 0.13 0.051 0.0025 0.010
Average 0.22 0.46 3.73 1.36 0.28 0.02 0.02 0.01 1.89
Media section
3/28/05 21.4 81.0 2.53 6.3 20.7 14.5 2.56 0.001 0.050 0.008 4.5
8/17/05 0.4 1.5 0.11 0.9 10.2 3.6 4.53 0.001 0.0025 0.022 4.8
4/26/06 1.0 3.8 0.54 1.0 13.3 4.9 1.03 0.018 0.073 0.014
5/13/06 12.7 48.1 0.60 1.5 12.8 4.2 0.98 0.065 0.131 0.013
5/16/06 2.7 10.2 0.15 0.5 4.8 1.6 0.39 0.027 0.0025 0.011
Average 0.79 2.0 12.35 5.77 1.90 0.022 0.051 0.014 4.65
Mann-Whitney Rank Sum Test between green and flat/asphalt roofs
P value* <0.001 <0.001 <0.001 <0.001 <0.001 0.183 0.010 0.004 0.042**
Note - values in bold and italic represent one half the detection limit value.
*A P-value of 0.05 or less typically indicates the difference in median values between the two concentrations is greater than would
be expected by chance and that there is a statistically significant difference between the results (Sigma Stat).
** The power of the t-test (0.484) was below the desired power of 0.800 (Sigma Stat).
4-14
Table 4-4 Estimated Loadings from Flat Asphalt and Green Roofs
Date Volume
runoff
(gal)
Loading (lb/acre) *
P K Ca Mg Na Mn Fe Zn S
Green roofs
3/28/05 54.3 0.185 1.001 4.2 3.3 0.47 0.0002 0.0010 0.0010 1.4
49.7 0.167 0.974 3.6 2.7 0.48 0.0002 0.0009 0.0009 1.4
38.5 0.147 0.754 3.0 2.2 0.31 0.0001 0.0007 0.0007 1.1
8/17/05 2.7 0.013 0.063 0.37 0.245 0.06 0.0007 0.0005 0.0002 0.23
0.7 0.003 0.010 0.07 0.043 0.01 0.0001 0.0001 0.0002 0.044
0.1 0.000 0.002 0.01 0.004 0.002 0.0000 0.0000 0.0000 0.004
4/26/06 0.7 0.001 0.004 0.07 0.049 0.005 0.0000 0.0005 0.0001
0.9 0.001 0.005 0.08 0.058 0.01 0.0001 0.0005 0.0001
0.7 0.004 0.014 0.14 0.064 0.01 0.0000 0.0004 0.0000
5/13/06 5.0 0.020 0.058 0.31 0.192 0.03 0.0002 0.0017 0.0005
3.8 0.010 0.052 0.18 0.113 0.03 0.0003 0.0006 0.0005
1.4 0.003 0.014 0.06 0.030 0.01 0.0003 0.0001 0.0002
5/16/06 1.3 0.002 0.010 0.12 0.077 0.01 0.0000 0.0007 0.0002
1.9 0.004 0.015 0.18 0.124 0.01 0.0001 0.0010 0.0002
1.0 0.003 0.010 0.14 0.088 0.01 0.0000 0.0007 0.0001
Average loading (lb/acre) 0.038 0.20 0.84 0.62 0.097 0.0002 0.0006 0.0003 0.69
Standard of Deviation 0.067 0.37 1.5 1.1 0.170 0.0002 0.0004 0.0003 0.67
Coefficient of Variation 1.8 1.9 1.7 1.8 1.8 1.0 0.7 1.0 0.97
Annual loading (lb/acre) 2.0 8.5 59.4 37.1 6.8 0.058 0.19 0.062 28
Estimate of variation for
annual loading 0.9 3.6 23.9 13.2 3.8 0.066 0.17 0.044 15
Average loading (kg/ha) * 0.042 0.22 0.94 0.69 0.11 0.0002 0.0007 0.0004 0.78
Annual loading (kg/ha) * 2.2 9.6 66.5 41.6 7.6 0.070 0.21 0.070 31
Flat asphalt roofs
3/28/05 50.30 0.011 0.04 0.43 0.05 0.016 0.0002 0.0010 0.003 0.011
51.70 0.012 0.04 0.41 0.06 0.015 0.0002 0.0010 0.0010 0.012
8/17/05 29.70 0.007 0.02 0.45 0.04 0.001 0.0001 0.0006 0.0006 0.007
28.10 0.006 0.02 1.55 0.14 0.013 0.0008 0.0005 0.002 0.006
4/26/06 4.00 0.005 0.02 0.06 0.01 0.010 0.0001 0.001 0.0001 0.005
3.30 0.001 0.00 0.06 0.01 0.002 0.00004 0.00034 0.0001 0.001
5/13/06 32.00 0.022 0.28 0.84 0.08 0.038 0.0041 0.0006 0.0006 0.022
30.80 0.007 0.11 0.75 0.07 0.012 0.0031 0.0006 0.0006 0.007
5/16/06 12.60 0.008 0.03 0.23 0.03 0.049 0.0005 0.0002 0.0002 0.008
11.40 0.003 0.05 0.18 0.02 0.004 0.0003 0.0002 0.0002 0.003
Average loading(lb/acre) 0.0081 0.061 0.50 0.050 0.016 0.0009 0.0006 0.0008 0.46
Standard of Deviation 0.0059 0.082 0.45 0.041 0.015 0.0014 0.0003 0.0009 0.28
Coefficient of Variation 0.73 1.4 0.92 0.81 0.96 1.5 0.46 1.1 0.60
Annual loading (lb/acre) 0.46 3.1 23.3 2.4 1.2 0.043 0.048 0.033 16
Estimate of variation for
annual loading 0.35 3.0 15.2 1.3 1.4 0.048 0.060 0.026 14
Average loading (kg/ha) * 0.0091 0.068 0.56 0.056 0.018 0.0011 0.0006 0.0009 0.52
Annual loading (kg/ha) * 0.51 3.4 26.1 2.73 1.3 0.048 0.054 0.037 17
Percent difference between green and flat roofs
Average loading 360 230 69 1100 500 -82 11 -61 49
Annual loading 330 180 160 1400 490 35 300 91 79
* Average and annual loading also presented in metric units of kilogram per hectare.
Note - values in bold and italic represent one half the detection limit value.
4-15
Hardness
Calcium (Ca) concentration in runoff from green roofs was greater than that from flat asphalt roofs. Calcium
concentration in green roof runoff was similar to runoff from the unplanted media roof section. Although calcium
concentrations were similar between planted green and unplanted media, total calcium in the runoff was much greater
in unplanted media due to greater runoff than in the planted systems. It should not be surprising that there would be
significant calcium in the runoff from green roofs since the primary parent material in the media used was expanded
clay. Also, the ability of the media to buffer pH suggests that there must be significant quantities of calcium
carbonates in the green roof media.
As might be expected, magnesium (Mg) concentration in runoff from the green roofs was similar, but slightly lower
than calcium concentrations. In contrast, runoff from the flat asphalt roofs had much lower concentrations of
magnesium. Total magnesium loading in the runoff was 10 times greater in green compared to flat asphalt roofs,
although as with calcium, the total was largely a function of the first measured event.
Ions
With the exception of sodium, differences in concentrations and projected loadings of the remaining ions did not
provide as clear cut statistical difference. Sodium (Na) concentrations were much greater in runoff from green roofs
than flat asphalt roofs. Concentrations in runoff from green roofs and the unplanted media roof section were similar
but total sodium in the runoff was higher in the unplanted media due to higher total runoff. While the calculated
average and annual loading from the green roofs were significantly larger than that from the flat asphalt roofs.
Results of manganese (Mn), iron (Fe) and zinc (Z) are even less decisive, particularly when both roofs were often
below detection limits. The calculated average loading and the annual loadings provided results that indicate green
roofs can have either decreased or increased loadings. Most likely, more samples are necessary to provide a
conclusive result, but indications are that green roofs will not be a significant source for these ions over other standard
roofing systems.
Sulfur (S) concentration was analyzed for only two events. As with other ions tested, the concentration was higher in
runoff from green than flat asphalt roofs and indicated a greater loading. However, total sulfur in the runoff was not
statistically different in green versus flat asphalt roofs.
Summary
Runoff from green roofs had higher concentrations of most of the nutrients and ions evaluated. From the five
precipitation events monitored, green roof runoff appears similar to what might be expected as leaching from any
other planted system in the landscape. Concentration of phosphorous is comparable to that of the NURP and NSQD
database residential values for total phosphorous. An important exception for nutrients was nitrate, where the average
concentration in runoff from green and flat asphalt roofs was nearly the same.
Although the runoff concentrations were typically higher, the loading was not always higher. Since the total runoff
from green roofs was about 50% less than the runoff from flat asphalt roofs over the study period, total loading of
runoff was sometimes lower even when concentration was higher. Nitrate provides a good example of this loading
reduction from green roofs over asphalt roofs and especially in this case, the media roof, demonstrating the added
benefit of the plants. There were several ions which exhibited seasonal differences in either concentration or total
runoff or both. Many of these differences were related to reduced runoff from green roofs during warmer months.
Concentration of some ions, i.e., calcium and magnesium, in green roof runoff were very consistent, the concentration
most likely controlled by buffers and/or exchange in the media. Clearly, total loading for hardness or ions evaluated
depended greatly on the amount of precipitation retained and the amount of runoff from the different roof systems.
Relative effects of green roofs on total loadings and water quality of runoff will be a function of season and
precipitation totals. In wet years, green roofs might contribute loadings of nutrients and hardness to total runoff
during summer or winter, and in dry years, most of the green roof loading is likely to occur during winter months.
Although runoff from the green roof for other ions was higher in concentration than runoff from flat asphalt roofs, it
4-16
is not clear that the relatively low concentrations of ions found in green roof runoff should be a concern in most
applications.
Sampling only compared paired events that produced runoff for both green and flat asphalt roofs. This may have
biased results slightly, increasing the potential loadings from green roofs and decreasing the loadings from the asphalt
roofs. It is expected that the washoff concentration of pollutants for flat asphalt roofs could be higher for smaller
storms that may not produce any runoff from the green roofs. On the other hand, green roofs are biological systems,
and increased rainfall volumes and intensity may lead to higher loadings, while, after the initial flush from a flat
asphalt roof, loadings should diminish.
Runoff from the unplanted media roof section was considerably higher in both concentration and in total output than
the planted green roofs. Since this roof was relatively new (less than 1 year old), and since there were no plants to
utilize nutrients, it is not surprising that concentration in the runoff would be higher. This suggests that newly planted
roofs are likely to have much higher runoff loading rates than established roofs. Any water quality data collected
during the establishment phase of a green roof may not reflect steady state or long-term water quality. Also, during
periods of plant establishment, green roofs in these circumstances should not be irrigated to the point of saturation to
prevent leaching of nutrients.
Results indicate that a green roof will contribute more nutrients (except nitrate), hardness, salts, and other ions to the
roof runoff, due mostly to the use of media and initial composting to provide nutrients for plants. Increased hardness
appears to be related to pH of precipitation and neutralizing of runoff. Increased loadings of some water quality
constituents does not necessarily indicate that green roofs would increase loadings to receiving waters; however, this
would suggest that the green roof as a stormwater BMP should probably be integrated with other treatment
techniques. Appropriate additional treatment, such as routing through a centralized BMP or more appropriately
discharging from the downspout to LID type BMPs, (e.g., swale, bioretention system, rain gardens) could be
recommended for green roof runoff as a part of an overall stormwater system. Green roof runoff could also be
collected and used for ornamental landscape purposes. Due to the reduced volume, time lag from the green roof and
the lower pollutant content of the green roof discharge in comparison to that of combined sewage, discharge to a
combined sewer system in heavily urbanized areas would be a benefit. Directly connecting the green roof drainage to
a separately sewered storm system without additional treatment or without benefit of the mixing and routing (time
lag) may impact the receiving water. The extreme case of directly discharging to the receiving water without further
treatment should be avoided.
4-17
Chapter 5 Green Roof Plant and Media Management Considerations
In addition to field studies, additional greenhouse and laboratory studies were performed to provide the user
community with additional information to help implement and manage green roofs. Specifically, these studies
evaluated evapotranspiration rates and related to existing models, determined both establishment criteria for plants
and media depths required to resist drought and predicted the long-term buffering capacity of the green roofs though
accelerated testing.
Evapotranspiration Rates from Green Roofs
One of the most likely and cost-effective reasons for a developer or zoning board to promote the use of green roofs is
their ability to retain and detain stormwater, which could reduce or eliminate the need for increased stormwater
management infrastructure by a municipality or watershed manager, or the need to set aside part of a development for
onsite stormwater management basins. In either case, the costs and benefits are direct, easy to understand, and easy to
assign to an individual entity. One of the major limitations to promoting this use of green roofs is the lack of accepted
design tools or models to predict the stormwater reduction effects of the green roofs. A major limitation in current
models is the lack of a good understanding of ET by the green roof systems. To address this need, controlled
greenhouse studies attempted to develop accurate and dynamic estimates of ET for plantings with Delosperma
nubigenum and Sedum album.
System Description
A weighing lysimeter system was used to conduct ET studies in a computer-controlled greenhouse section on the
University Park Campus of the Pennsylvania State University. Eight test beds (3.5 ft long, 1.75 ft. wide and 4 in.
deep; 105 cm long, 54 cm wide and 10 cm deep) were each suspended by cables from an Omega Mini-beam LCEB-
150 lb (68 kg) load cell (Omega Mfg, Stanford, CT). Two Multiplexer 21X data loggers (Campbell Scientific, Inc.,
Logan, Utah), each with 16 input channels, were programmed to collect data from the aforementioned sensors,
collected and recorded output from the load cells and other measurement systems. Each test bed box contained a
multilayer green roof system consisting of a drainage mat (Enka-drain 9615, Cold Bond, Enka, NC) covered with
approximately 3.5 in. (9.0 cm) of a clay-based, commercial green roof media (Big River Industries, Alpharetta, GA).
Each weighing lysimeter was suspended at a pitch of 1:12 (approximately 8% slope) with a drainage slit in the box at
the lower end. Drainage water was collected in a gutter and directed to a collection bucket. Four lysimeters were
vegetated while the other four contained the green roof system without plants. The vegetated beds contained a
mixture of Sedum album and Delosperma nubigenum which covered 80-95% of the surface of the bed throughout the
project.
Air temperature and media surface temperature were measured with Omega Copper-Constantan HYPO-Series
thermocouples. Two thermocouples measured surface temperature of vegetated lysimeters and two measured surface
temperature of non-vegetated lysimeters. A wet and dry bulb thermocouple pair located in an aspirated sensor box
was used to measure air temperature and calculate relative humidity. In addition, two infra-red thermocouple sensors
5-1
(IRTC Exergen, Watertown, MA) were installed, with one directed to measure surface temperature of a vegetated and
the other directed to measure surface temperature of a non-vegetated lysimeter.
Light quantity was determined using a LiCor Pyranometer (Lincoln, NE) mounted two meters over the lysimeter
systems. Wind speed was measured with an Omega hot wire anemometer. A pan system was used to estimate
evaporation; it was composed of an open body of water in a 4.5 in. (11.5 cm) diameter and 0.2 in. (0.6 cm) depth
cylindrical container with an Omega PX26-PSI pressure transducer placed on the bottom and connected to a data
logger. The evaporation pan was calibrated by incrementally filling the cylinder and recording the data logger’s
response at the same time. Water was then incrementally removed from the cylinder and data logger readings were
measured for each new level as before. This process was reported three times and averaged to increase the accuracy
of the calibration. The calibration coefficient was 8.78 when pressure in mm of water was on y-axis and millivolts on
x-axis.
Evapotranspiration Rates
Data collected suggest, as expected, that water use was greater in planted than in unplanted systems. The average
water loss or measured ET rates before the plants responded to moisture stress was 0.41 in./d (10.5 mm/d), which was
about 40% of the field capacity moisture. After the plants had used 40% of the soil’s field capacity moisture, their
water loss rates reduced to an average of 0.014 in./d (0.036 cm/d) for the remainder of the 21-day experiment.
Example results of one dry-down period experiment are shown in Figure 5-1. In this 14-day dry-down period, 6 mm
(0.25 in.) of water was lost through evaporation from unplanted systems, while 10 mm (0.39 in.) was lost from
planted systems. In the beginning, water loss was similar for both systems, but after a few days water stored in the
plants and the plants ability to extract water from deeper in the media via their roots resulted in greatly increased ET.
Figure 5-1 Average cumulative water loss from planted and non-planted green roof lysimeter test beds in the fall
Not only was the water loss of the planted systems greater, but some interesting patterns in water loss emerged when
the data were examined more closely as demonstrated in Figure 5-2. In the beginning, water loss during the day
greatly exceeded water loss at night. This was true for both planted and unplanted roof systems. The rate of water
loss in the unplanted systems rapidly dropped over the first 4 to 5 days, presumably as the media surface dried. In
5-2
contrast, planted systems continued rapid and fairly consistent water loss. Water loss during the night was lower and
was roughly equal in green and flat asphalt systems during this period immediately following watering. After 3 to 4
days a change in the water loss patterns occurred in the planted systems. Water loss during the day dropped rapidly
while water loss at night remained fairly consistent. The result was that the planted systems lost much more than
unplanted systems at night during this phase, which lasted for about 4 to 5 days. This change may correspond to
plants utilizing crassulacean acid metabolism (CAM), opening stomata at night. After 7 to 8 days the water loss rate
difference between green and flat asphalt systems became increasingly small as plants became increasingly water
stressed. Data from summer conditions suggested that water loss from planted systems at night was, in fact, less than
from unplanted systems and that plants may in fact be extracting moisture from the air when the humidity was high.
Sedum album was chosen because it was anticipated to survive the stressful treatments of the experiment. This
popular green roof plant is known to switch to CAM when water stressed. CAM plants open their stomata at night,
taking up carbon dioxide and storing it as malic acid. During the day, they maintain closed stomata and
photosynthesize the stored carbon dioxide (Ting, 1985). This results in CAM plants having very high water use
efficiency because they open their stomata at night and close them during the day to minimize water loss through
evapotranspiration.
Figure 5-2 Average water loss during the daylight or night from planted and unplanted green roof lysimeter test beds in the fall
Figures 5-3 and 5-4 compare the average hourly water loss for the planted and unplanted test beds for a day when the
water loss was not limited (second day, Figure 5-3) and a day when the water loss was limited (sixth day, Figure 5-4)
for hot summer conditions. As can be seen from these figures, the planted test beds have less transpiration during
nighttime, since the maximum ET rate occurred during the day in both figures.
5-3
Figure 5-3 Average hourly water loss from planted and unplanted boxes for the 2nd day after watering during the hot summer
condition (Rezaei 2005)
Figure 5-4 Average hourly water loss from planted and unplanted boxes for the 6th day after watering during the hot summer
condition (Rezaei 2005)
Standard ET models, i.e., Penman Monteith, FAO Penman, Original Penman, and Blaney-Criddle, were applied to the
data and regression analyses indicated ET was fairly well predicted with R
2
values of 0.815, 0.644, 0.796, and 0.697
respectively, for winter conditions, when the systems were not too water stressed. During warmer periods, the models
adequately predicted water loss for the first seven days, with R
2
values of 0.855, 0.810, 0.794, and 0.646, respectively,
but did a poor job of predicting water loss after that as the systems became dry.
5-4
The data suggest there may be a need to develop unique water loss model factors to account for the water loss patterns
in these Sedum carpet roofs if there is a need to accurately predict the rate of recharge for water detention capacity of
green roof. The runoff reduction potential by green roofs for storm events may need to be modeled with
consideration to the ET rates. This may also require modeling of results in different areas of the country.
Effect of Media Depth, Type, and Drought during Establishment
Ideally, extensive green roofs do not require irrigation, and the growing media should be able to absorb and retain
water, but also drain readily (Miller, 2003). In addition to limitations of water, green roof plants must tolerate direct
solar radiation, high wind velocities, extreme temperatures and unpredictable precipitation (Durham et al., 2004).
Preferred green roof plants are shallow-rooted perennials, with limited growth to prevent toppling and desiccation
(White and Snodgrass, 2003). Well-designed ‘soilless’ media can provide comparable moisture holding properties to
organic-rich mixes, depending on particle size and the substrate (Miller, 2003).
The responses of five plant species to three media depths under three drought regimes were tested. In addition, two
different media, expanded clay and expanded shale, were compared for their influences on plant performance in
combination with treatments of depth and drought. All tests were performed in two replications. The experiment ran
for a period of 81 days, beginning June 16, 2004 and ending September 4, 2004. This period of time, after planting
but prior to the first year dormancy, was considered the plant establishment period. In late April of 2004, cuttings
from the three succulent taxa (Sedum album, Sedum sexangulare, Delosperma nubigenum) were taken from
greenhouse stock and propagated in modular plug trays (2 cm
2
, 0.5 in
2
, 273 per tray) in a peat-based potting soil. The
two herbaceous species (Dianthus deltoides, and Petrorhagia saxifraga) were provided, as plugs (8 x 4.5 cm, 3 x 1.75
in., 72 per tray), by Green Roof Plants (Emory Knoll Farms, Street, MD). On June 16, 2004, the plugs were
transferred to the experimental study flats, which are described below, along with the media. Before planting, the
flats were irrigated twice to saturation, or the point at which water drained freely from below. After planting, all flats
received the same amount of water again.
The expanded clay was expected to support plants better than the expanded shale, as clay has a good capacity for
holding water and exchanging cations (Nelson, 1998). The mineral aggregates were mixed with pelletized spent
mushroom compost (Laurel Valley Soils, Avondale, PA) to obtain a ratio of 85% mineral to 15% organic matter by
volume. The study flats, 1044 cm
2
(160 in
2
) propagation flats (Anderson Dye and Manufacturing, Portland, OR),
were filled and leveled to the desired depths of 3.0, 6.0, and 12.0 cm (1.2, 2.4 and 4.8 in.) to represent a range depth,
from too shallow to exceeding design depth for extensive green roofs.. The bottom of each flat was fitted with a piece
of Enkadrain (Colbond Geosynthetics, Enka, NC). Because the herbaceous plugs were longer than the depth of the
shallower flats of 3.0 and 6.0 cm (1.2 and 2.4 in.), the root masses of these plugs were cut into four sections and the
divided parts spread flat along the bottom of the flats, like octopus legs to completely bury the roots. The five plant
taxa were planted at a density of five individuals per box for each treatment, for samples of n = 10.
The experiment was housed in a 12 x 100 ft (4 x 30 m) polyethylene greenhouse tunnel (located in Potter’s Mills, PA)
in order to control for drought. Three watering regimes were used: no drought, early drought and late drought.
Control plants (no drought) received 1.4 in. (3.6 cm) of water twice weekly, or 2.8 in. (7.2 cm) per week, for the 11-
week study period. Plants subjected to early drought received no water for the first two weeks after planting, while
those subjected to late drought received no water in the last two weeks of the study period. When drought was not in
effect, drought plants received the control amount of water. The experiment was designed in a split-block
arrangement by randomized complete blocks, with drought conditions separated as blocks of irrigation treatments.
The treatments of media type and depth were placed randomly within the blocks at factorial levels, and blocks were
also placed randomly within the replications.
Data Analysis
To assess species performance under the various treatments at the close of the 11-week study period, the shoot
biomass, defined as the above-ground plant material, including stems, foliage, flowers, was harvested and dried for 36
hr in an oven at 50ºC. Shoot dry weight for the means of two replications was analyzed with the SAS Mixed
5-5
Procedure and the General Linear Model (GLM) (Littell et al., 1996) at the 0.05 level of significance. The Mixed
Procedure suited this study because the model contained both fixed and random effects, i.e., mixed effects parameters,
and because the data set was unbalanced due to plant mortality. Dead plants were not used in the calculation of mean
shoot dry weight.
Effects of Drought
Since the deepest media did not cause apparent stress, the responses of plants grown in the 12.0 cm (4.8 in.) deep
media were attributed to drought and/ or media type, and were addressed first. As expected, the two herbaceous
species were significantly affected by drought, while the succulents were not.
When grown in 12.0 cm (4.8 in.) and subjected to early drought, Sedum sexangulare had the fewest survivors and
least shoot growth than control or late drought plants. Although mean shoot dry weight differed significantly between
control and early drought plants (GLM), early drought plants did not match the shoot growth of plants subjected to
late drought by the end of the study period. Delosperma nubigenum, Dianthus deltoides and Petrorhagia saxifraga
also produced less shoot dry weight after early, rather than late drought, but only in the expanded shale, behaving
oppositely in the expanded clay. In 12.0 cm (4.8 in.) expanded shale, these taxa produced less shoot biomass after
early drought, and also suffered the most mortality.
The hardiest of the bunch after early drought, Petrorhagia saxifraga in 12.0 cm (4.8 in.) shale had 83% less mean
shoot dry weight compared to control plants (or 17% of the biomass established by plants in control conditions),
while plants grown in clay had only had 24% less mean shoot dry (76% of biomass of control plants) as demonstrated
in Figures 5-5 and 5-6, respectively. Missing data points in these and following graphs signify that no plants survived
that treatment combination, e.g. in Figure 5-5, no plants survived the 3 cm (1.2 in.) depth and only plants in clay
survived the 6 cm (2.4 in.) depth. Under control conditions, this species established similar shoot growth in both
media and, compared to Sedum album, performed the most consistently. Rapid establishment can be promoted by
preventing transplant stress, i.e., providing water before the symptoms of water stress appear (Nelson, 1998). For
non-CAM plants, irrigation in the first few weeks after planting is essential. Other studies suggest that irrigation in
the first weeks of establishment promotes the longevity and robustness of a green roof plant community (Monterusso,
2003 and Dunnett and Nolan, 2004), and some recommend irrigation for the entire first growing season, unless
weekly rainfall exceeds 25 mm (1 in.) (White and Snodgrass, 2003).
Figure 5-5 Mean shoot dry weight (g) (n=10) for Petrorhagia saxifraga in three depths of two media under early drought conditions
(Thuring, 2005)
5-6
Figure 5-6 Mean shoot dry weight (g) (n=10) for Petrorhagia saxifraga in three depths of two media under late drought conditions
(Thuring, 2005)
Sedum album was the only species that actually had more shoot biomass after early, rather than late, drought in 12.0
cm (4.8 in.) of both media. Compared with control plants in 12.0 cm (4.8 in.) shale, plants subjected to early drought
had 47% less shoot dry weight, and in the expanded clay actually gained 5% shoot biomass. When subjected to late
drought, this species had 49% (shale) and 70% (clay) less shoot dry weight than control plants (Figures 5-7 and 5-8).
A similar study, testing media type, depth, and drought observed that Sedum album remained green after 100 days
without water (38 of which had temperatures higher than 18°C), and did not exhibit the red coloration that indicates
water stress (Lassalle, 1998).
Effects of Media Depth
Early drought conditions reduced survival or impacted biomass for most species, but especially so in combination
with shallow media depths. Depth influenced shoot growth for the overall experiment significantly (p<0.0001), and
all taxa studied, except for Dianthus deltoides, showed significant responses to depth. The results of Dianthus
deltoides, are consistant with a study by Dunnett and Nolan (2004) that demonstrated Dianthus deltoides had good
growth and survival when provided with regular and sufficient irrigation independent of media depth. Even when
plants survived, plants in the shallowest depth of 3.0 cm (1.2 in.) always produced the least shoot biomass. Under
early drought, Delosperma nubigenum, Dianthus deltoides and Petrorhagia saxifraga did not survive in 3.0 cm (1.2
in.) media, regardless whether it was clay- or shale-based. Delosperma nubigenum survived in 3.0 cm (1.2 in.) clay,
but did not survive in the shale. The two herbaceous species did not survive in 6.0 cm (2.4 in.) shale media under
early drought. Since the lifespan of perennials can be limited by media depth and other environmental factors, White
and Snodgrass (2004) concluded these plants require at least 10 cm (4 in.) media.
When subjected to early and late drought, Figure 5-7 and Figure 5-8, respectively, Sedum album established similar
mean shoot dry weight (g) across the three depths of expanded shale. In the deeper experimental depths of expanded
clay, 6 cm (2.4 in.) and 12 cm (4.7 in.), respectively, this species produced remarkably more shoot biomass after early
drought as compared to both late drought and control plants
5-7
Figure 5-7 Mean shoot dry weight (g) (n=10) for Sedum album in three depths of two media under early drought conditions
(Thuring, 2005)
Figure 5-8 Mean shoot dry weight (g) (n=10) for Sedum album in three depths of two media under late drought conditions
(Thuring, 2005)
The three hardiest taxa, Petrorhagia saxifraga and the two sedums, always produced more shoot biomass with
increasing media depths, regardless of water availability. While the sedums performed well in all depths, plants
grown in 6.0 and 12.0 cm (2.4 and 4.8 in.) media produced significantly more shoot biomass than those in 3.0 cm (1.2
in.) (GLM) as demonstrated in Table 1. If good performance by these hardy species indicates green roof success, then
unirrigated extensive green roofs should never be less than 6 cm (2.5 in.) deep; this minimal depth corresponds to
retaining 40% of the 2-yr storm (Scholz-Barth, 2001), as noted in Chapter 1. When it survived, Petrorhagia saxifraga
flowered continuously. The benefit of Petrorhagia saxifraga for green roofs is its prolific seed production and ability
to fill gaps in plant cover (Kolb et al. 1982). However, this species benefited from the deeper media depths.
5-8
Table 5-1 Mean Shoot Dry Weight for Sedum album under Early and Late Drought Conditions
Drought
Conditions Depth
(cm)
Expanded Shale Media Expanded Clay Media
mean dry
weight
(g)
% reduction
from
control
1
% survival
(n=10)
mean dry
weight
(g)
% reduction
from
control
% survival
(n=10)
Early
3.0 0.65 (a)
z
55 80 1.27 (a)
z
29 90
6.0 1.15 (a) 50 80 2.68 (a) -56 90
12.0 1.66 (a) 47 100 2.66 (b) -5 100
Late
3.0 0.57 (a)
z
39 100 1.49 (a)
z
83 90
6.0 1.42 (a) 62 100 1.69 (a) 98 100
12.0 1.55 (b) 49 100 1.78 (b) 70 100
1
The column “% reduction from control” shows how shoot growth was influenced by the drought conditions by contrast to control
conditions.
Negative numbers indicate that more dry weight was produced.
z - Different letters indicate significant differences in shoot dry weight as influenced by medium depth z. Mean separation by
Duncan’s multiple range test at P < 0.05
Effects of Media Type
Species responses to media depth and drought were sometimes complicated by their responses to the two media. The
capacity for the two media to support plants through drought was disparate. Since the herbaceous taxa performed
similarly to Sedum album under early drought conditions when grown in the expanded clay, and plants grown in the
shale either died or exhibited reduced biomass, it would appear that the clay is a better selection for green roofs that
will not be irrigated. Similar studies which used media dominated by expanded slate also found that herbaceous
perennials in green roof conditions required irrigation to thrive (Dunnett and Nolan, 2004; Durham et al. 2004; and
Monterusso et al., 2005). Geologically, slate is produced by the compression of shale, so if these aggregates are
considered similar, then those studies substantiate the poor plant growth observed in the expanded shale.
Discussion
Results from this study illustrate how factors such as type of media and depth and anticipated conditions such as
drought can direct species selection and planning for the design of extensive green roofs. Fundamentally, these
experiments revealed the variability in survival among drought tolerant species when subjected to other treatments.
Shoot growth often varied per species depending on the timing of drought or the type of media. However, early
drought apparently stunted the capacity for shoot growth, with the exception of Sedum album, since most species did
not match their late drought counterparts in shoot dry weight. Initial irrigation in the beginning of an establishment
period would benefit plant performance, especially for the long term.
Water stress has an enormous effect on plant cell development, and insufficient water during establishment can result
in stunted roots and a permanent decrease in leaf size. This inhibits water and nutrient uptake and reduces surface
area for photosynthesis, even after water is reinstated, further inhibiting a plant’s capacity to grow (Boodley, 1998).
In planning for extensive green roofs, therefore, it is valuable to know the original habitats and morphologies of
selected taxa. Sedum carpets will perform well in a minimum of 6.0 cm (2.4 in.) media without irrigation, while the
inclusion of herbaceous perennials should be provided with no less than 10 cm (4 in.) media. Sedum album is an ideal
green roof plant because it can avoid water stress by exhibiting CAM. For non-CAM plants, irrigation in the first few
weeks after planting is essential. Additionally, if the roof is not ot be irrigated clay-based media may improve plant
survivability.
Evaluating Long-Term Media Buffering Capacity
As presented in Chapter 4, the data suggested green roofs could effectively remediate acid precipitation. The green
roof neutralized the acid rain and increased the pH of the runoff (see Figure 4-5). In particular, the commercial green
5-9
roof media evaluated have tremendous potential to neutralize acid precipitation. Although the effects are clear and
consistent, this potential benefit needs to be fully explored or quantified. Acid precipitation is known to cause a
number of problems in urban runoff, including acid leaching of metal ions from rooftop flashings, downspouts, and
other exposed metals on the roof. A green roof has the potential to all but eliminate the effects of acid precipitation.
To manage a green roof for this purpose over the long term, both the inherent acid buffering capacity of the media,
and the frequency of routine maintenance procedures (liming) needed to maintain this capacity over the life span of
the green roof need to be better understood.
Results of this trial led to a more thorough investigation of the ability of green roofs to neutralize acid rain. Further
information can be obtained in Berghage et al. (2007).
Media Buffering Capacity
To examine the long-term acid rain buffering characteristics of a green roof media an accelerated aging trial was
conducted. Two media, one based on expanded clay and the other based on expanded slate were subjected to acid
additions (5 ml of 0.02 M sulfuric acid) each day for 47 days. Each media was 9 cm (3.5 in.) deep. The pH was
measured before and after the acid additions. Results are presented in Figures 5-9 and 5-10.
The buffering potential (pH) is presented in Figures 5-9 and 5-10 versus volume of simulated acid precipitation. Acid
titrations were converted to potential cumulative acid rain (in cm) based on the acidity determined for rainwater
collected from Rock Springs, PA, a pH of 4.5 (see Figure 4-6). When acid was added, the pH was reduced (lower
line in each of the figures), but typically within 24 hr the buffering capacity had greatly recovered (upper line). Over
time the ability of the buffering capacity to recover declined and there were significant differences in the recovery
between slate- and clay-based media. The rate of pH recovery (acid rain buffering) appeared to be lower in clay-based
media than the slate-based media. Further, the recovery seemed to be slower. Based on these accelerated aging
(buffering potential) titrations it appears that while these media have high levels of pH buffering capacity, the
capacity to neutralize acid precipitation becomes increasingly time-dependent as the materials age. It seems likely
that this trend and the differences in media responses will dictate different long-term management approaches if
neutralizing acid precipitation is a goal for the green roof. In either case, the buffering capacity appears to be
adequate to neutralize acid rain for approximately 10 yr before liming would be necessary to keep the plantings
functioning. This is based an annual rainfall of 110 cm/yr (44 in/yr.) and maintaining pH above 6 (about 1000 cm or
400 in. of simulated rain for each graph) resulting in a calculation of 8.9 yr.
Figure 5-9 Accelerated acid aging trial for clay-based media
5-10
Figure 5-10 Accelerated acid aging trial for slate based media
Cation Exchange Capacity
In order to develop properly, plants require mineral nutrients. Nutrient uptake by plants is controlled by the cation
exchange capacity (CEC) of the media in which they are grown (Boodley, 1998). CEC is normally associated with the
clay colloids of a mineral soil, although organic matter (e.g., humus, peat, plants roots) also has good CEC, but less
than that of clay (Bunt, 1988).
The negative surface charge of a clay colloid attracts positively charged ions (e.g., cations, like K
+
, Ca
++
). The
combination of CO
2
(from root respiration) with hydrogen from soil water, causes the formation of carbonic acid
(H
2
CO
3
), which releases hydrogen ions (H
+
) from the clay surface. When these hydrogen ions exchange positions
with other cations (between the soil solution and the surface of the clay colloid), those mineral nutrients become
available for root absorption (Boodley, 1998). Another important key to plant nutrient uptake is the pH of the
growing media. At the highest and lowest pH values, essential mineral nutrients may combine with other elements,
limiting their availability for plants (Boodley, 1998). Also, CEC is dependent on pH, since an acidic media (less than
pH 7) will contain more hydrogen ions than a neutral (pH 7) media, thereby inhibiting cation exchange and nutrient
absorption (Bunt, 1988). A green roof media should maintain a stable pH over time, and should moderate the pH of
rain to neutral, especially in areas like northeastern America where the pH of rain can be as low as pH 4.9 (Beattie
and Berghage, 2004). During the course of this study, four sampled precipitation events yielded an average pH of 5.1.
Roofs that become acidic lose the ability to neutralize acid precipitation, typically lose plantings and tend to favor
moss. Roofs dominated by mosses will not have the same hydrology of green roofs as detailed here, and will not
have the same aesthetic value.
5-11
Chapter 6 References
The American Public Works Association (APWA). 1992. Nationwide Costs to Implement BMPs.
Beattie, D.J. and R.D. Berghage. 2004. Design criteria for a green roof media. Proceedings from 2nd International
Green Roof Infrastructure Conference, Awards and Trade Show, Greening Rooftops for Sustainable Communities,
Portland, OR. June, 2004. The Cardinal Group, Toronto.
Berghage, R., A. Jarrett, D. Beattie, K. Kelly, S. Husain, F. Razaei, B. Long, A. Negassi, R. Cameron and W. Hunt.
2007. “Quantifying Evaporation and Transpirational Water Losses from Green Roofs and Green Roof Media
Capacity for Neutralizing Acid Rain,” National Decentralized Water Resources Capacity Development Project
(NDWRCP) Research Project, 2007, available at http://www.ndwrcdp.org/userfiles/04-DEC-10SG.pdf
Boodley, J.W. 1998. The commercial greenhouse. 2nd ed. Delmar Publishers. Albany, New York.
Bunt, A.C. 1988. Media and mixes for container-grown plants: A manual on the preparation and use of growing
media for pot plants. Second edition of Modern potting composts. Unwin Hyman Ltd. London, England.
DeNardo, J. C., A. R. Jarrett, H. B. Manbeck, D. J. Beattie, and R. D. Berghage. 2005. Stormwater mitigation and
surface temperature reduction by green roofs. Transactions of ASAE 48(4): 1491-96.
Durham, A.K., N.D. VanWoert, D.B. Rowe, C.L. Rugh, and D. Ebert. 2004. Evaluation of Crassulaceae species on
extensive green roofs. Proceedings from 2
nd
International Green Roof Infrastructure Conference, Awards and Trade
Show, Greening Rooftops for Sustainable Communities, Portland, OR. June 2004. The Cardinal Group, Toronto.
Dunnett, N. and A. Nolan. 2004. The effect of substrate depth and supplementary watering on the growth of nine
herbaceous perennials in a semi-extensive green roof. Acta Hort. 643:305-309.
Dunnett, N. and N. Kingsbury. 2004. Planting options for extensive and semi-extensivegreen roofs. Proceedings from
2nd International Green Roof Infrastructure Conference, Awards and Trade Show, Greening Rooftops for
Sustainable Communities, Portland, OR. June 2004. The Cardinal Group, Toronto.
Herris, A. 2005. Personal communication, 9th March, 2005,. Weston Solutions. Adapted from C.E. Thuring Thesis
(2005).
Howell, T. A., A. D. Schneider, and M. E. Jensen. 1991. History of lysimeter design and use for evapotranspiration
measurements. In Proc. International Symposium of Lysimeter. 1-9. New York, NY.: ASCE.
6-1
Kolb, W., T. Schwarz und P. Mansourie. 1982. Extensivbegrünung von Dachflächen: Vegetationstechnische
Eigenschaften und Kosten von 10 verschiedenen Substraten. Zeitschrift für Vegetationstechnik 5:106-112.
Lassalle, F. 1998. Wirkung von Trockenstreß auf xerophile Pflanzen. Stadt und Grün 6:437-443.
Liesecke, H.J. 1985. Ausstreuen von Sedum-sprossen mit Zusaaten zur Extensiven Begrünung von Flachdächern.
Zeitschrift für Vegetationstechnik 8:159-165.
Littell, R.C., G.A. Milliken, W.W. Stroup, R.D. Wolfing