EPA/600/R-14/180 | September 2014 | www.epa.gov/research
United States
Environmental Protection
Agency
Evaluation of Green Roof
Water Quantity and Quality
Performance in an Urban Climate
Office of Research and Development
-------
EPA/600/R-14/180
September 2014
Evaluation of Green Roof Water
Quantity and Quality
Performance in an Urban Climate
By
Patricia Culligan
Tyler Carson
Stuart Gaffin
Rebecca Gibson
Raha Hakimdavar
Diana Hsueh
Nadine Hunter
Daniel Marasco
Wade McGillis
Columbia University
New York, NY 10027
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. AE-83484601-0
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
September 2014
-------
Notice
The U.S. Environmental Protection Agency, through its Office of Research and Development, funded and
managed, or partially funded and collaborated in, the research described herein. It has been subjected to
the Agency's peer and administrative review and has been approved for publication. Any opinions
expressed in this report are those of the authors and do not necessarily reflect the views of the Agency,
therefore, no official endorsement should be inferred. Any mention of trade names or commercial products
does not constitute endorsement or recommendation for use.
-------
Abstract
Green roofs are increasingly seen as an established 'green infrastructure' technology that confers many
environmental benefits. This is especially the case in urban areas where rooftops comprise a large fraction
of the landscape, are typically low albedo and add to widespread impervious surfaces. The benefits of green
roofs include urban heat island mitigation, reduced or eliminated roof facade heat transfer with associated
building energy benefits, stormwater retention and detention, ecosystem service benefits and aesthetic
amenity value, to name a few. Stormwater mitigation and subsequent receiving water quality improvement
are increasingly perceived as an important function of this technology.
In this report we present an analysis of water benefits from an array of observed green roof and control
(non-vegetated) roof project sites throughout New York City, where average annual precipitation in New
York's Central Park is over 1200 mm for the 40-year historic period 1971-2010. The projects are located
on a variety of building sites and represent a diverse set of available extensive green roof installation types,
including vegetated mat, built up, and modular tray systems, as well as plant types. Moreover the projects
have been monitored for a few years and are being observed in an urban climate.
For water retention performance, we monitored runoff from four full-scale green roofs, including one built
up system, one modular tray system and two vegetated mat systems. We gathered roof runoff data for over
100 storm events for each green roof over a period of 23-months. Our main findings for water quantity
performance include: (i) runoff from green roofs has a quadratic relationship to precipitation depth, where
the percent retention decreases as storm size increases; (ii) the relationship between precipitation depth and
green roof runoff depth (runoff volume divided by rooftop drainage area) can be described by a
Characteristic Runoff Equation (CRE) for each roof; (iii) the CRE can be used with historic rainfall data to
reduce bias in reported green roof retention performance, which might arise due to a bias in the distribution
of storm events during a monitoring period; (iv) the modular tray system captured the lowest percentage of
precipitation among all green roof systems for storms 0-20 mm in depth, and the highest for storms above
30 mm; (v) multi-year predictions show that on an annual basis, the built up system will retain more rainfall
than the modular tray system, which will retain more rainfall than the vegetated mat systems. Our findings
reveal the importance of green roof technical design, as well as substrate capacity, for stormwater retention
at different storm sizes. The Natural Resources Conservation Service curve number (CN) method, while
providing similar average results to the CRE, could not capture observed relative differences between the
retention performance of the built up, modular tray and mat systems in different storm categories.
For water quality performance with respect to stormwater runoff, we undertook a 16-month survey of
stormwater runoff quality from five full-scale green roofs, including two built up systems, one modular tray
system and two vegetated mat systems. For comparison, we also surveyed the chemical composition of
runoff from five non-vegetated (control) roofs as well as local precipitation. In total we collected and
analyzed over 100 water samples. Our results show that the pH of runoff from green roofs was consistently
higher than that from the control roofs and precipitation with observed average pH's equal to 7.28, 6.27
and 4.82 for the green roofs, control roofs and precipitation, respectively. Thus, the green roofs neutralized
the acid rain. In general, we observed lower NOs (nitrate) and NH4+ (ammonium) concentrations in green
runoff than control roof runoff, with the exception of runoff from the built up system, which had higher
NOs concentrations than the control roof runoff. Overall, total P (phosphorus) concentrations were higher
in green roof runoff than control roof runoff. Finally, with respect to micronutrients and heavy metals: we
either detected these constituents at very low concentrations or not at all (concentrations were below the
detection limit), with a few exceptions. One exception related to the detection of boron in runoff from one
of the vegetated mat systems, and another related to the detection of Ca (calcium) and Na (sodium) in runoff
from all five green roofs. Based on our results, we estimated that annual mass loading per unit rooftop area
iv
-------
of NOs", NH4+ and total P discharging from all five green roofs was considerably less than that from their
respective control roofs, due to the ability of green roofs to retain precipitation. Thus, green roof
implementation could improve urban stormwater and subsequently urban receiving water quality if
achieved at large areal scales.
In order to investigate monitoring schemes that could be used on a wider scale of study, a new method for
green roof runoff and evapotranspiration estimation was derived. Termed the Soil Water Apportioning
Method (SWAM), this is a water balance approach which analytically links precipitation to substrate
moisture, and enables inference of green roof runoff and evapotranspiration from information on substrate
moisture changes over time. Twelve months of in situ rainfall and soil moisture observations from two
green roofs, both vegetated mat systems, were used to test the reliability of the proposed approach using
two different low-cost soil moisture probes. SWAM estimates of runoff were compared with observed
runoff data for the entire duration of the study period. Preliminary results indicate that SWAM can be an
effective low-cost and low-maintenance alternative to the custom made weir and lysimeter systems
frequently used to quantify runoff during green roof studies. The method may also provide a simple way of
estimating green roof evapotranspiration.
-------
Table of Contents
Notice ii
Abstract iii
Contents v
List of Figures vi
List of Tables vii
Acronyms and Abbreviations viii
Acknowledgements x
Executive Summary 1
Chapter 1 Introduction 1-1
Chapter 2 Overall Conclusions and Recommendations 2-1
2.1 Conclusions 2-1
2.2 Recommendations for Further Study 2-2
Chapter 3 Monitoring Sites and Systems 3-1
3.1 Green Roofs and Monitoring Equipment 3-1
3.2 Green Roof Site Descriptions 3-3
Chapter 4 Water Quantity Study 4-1
4.1 Introduction 4-1
4.2 Methodology 4-2
4.3 Results 4-4
4.4 Results from BDCA and Regis 4-9
4.5 Discussion of Results 4-10
4.6 Conclusions 4-12
Chapter 5 Water Quality Study 5-1
5.1 Introduction 5-1
5.2 Methodology 5-1
5.3 Results 5-3
5.4 Discussion of Results 5-3
5.5 Conclusions 5-9
Chapter 6 Soil Moisture Water Balance Study 6-1
6.1 Introduction 6-1
6.2 Methodology 6-1
6.3 Results 6-4
6.4 Discussion of Results 6-8
6. 5 Conclusions 6-9
Chapter 7 References 7-1
Appendix A Site Equipment A 1
Appendix B Water Quality Results B 1
Appendix C Water Quality Analysis C 1
VI
-------
List of Figures
Figure 3-1: Locations of monitored green roof sites and priority combined sewer sheds in New York
City 3-2
Figure 3 -2: W115 Roof (A) Satellite photograph (B) Weir device (C) Roof photograph 3-3
Figure 3 -3: W118 Roof (A) Satellite photograph (B) Weir device (C) Roof photograph 3-3
Figure 3-4: Fdston Roof (A) Satellite photograph (B) Equipment (C) Roof photograph 3-6
Figure 3-5: BDCA Roof (A) Satellite photograph (B) Weir device (C) Roof photograph 3-7
Figure 3-6: Regis Roof (A) Satellite photograph (B) Weir device (C) Roof photograph 3-8
Figure 4-1: (A) Runoff monitoring weir device (B) Calibration chamber used to simulate rooftop
runoff. 4-2
Figure 4-2: Monitored total rainfall retention for the Wl 15, Wl 18, USPS, and ConEd green roofs by event
size 4-5
Figure 4-3: Rainfall by event size for monitored and historic (NYC) data as percent of (A) events and (B)
depth 4-6
Figure 4-4: Characteristic runoff equations and events for (A) Wl 15, (B) Wl 18, (C) USPS, and (D) ConEd.
4-7
Figure 4-5: Residuals between characteristic runoff equations and curve number and observed depth for
(A)W115, (B)W118, (C) USPS, and (D) ConEd 4-8
Figure 4-6: Predicted rainfall retention over historic period using (A) Characteristic runoff equations (B)
Curve number 4-9
Figure 4-7: Characteristic runoff equations and events for (A) BDCA (B) Regis 4-10
Figure 5-1: Distribution of pH, conductivity, turbidity, and apparent color for all roof types and
precipitation 5-4
Figure 5-2 Distribution of nutrient and heavy metal concentration (mg/L) from all roof types 5-5
Figure 6-1: NSE and PBIAS for the (a) W118, and (b) W115 roofs at different time aggregates 6-5
Figure 6-2 Precipitation, soil moisture, observed and simulated daily runoff and error for green roof Wl 18.
6-5
Figure 6-3: Precipitation, soil moisture, observed and simulated daily runoff and error for green roof Wl 15.
6-6
Figure 6-4: Comparison between green roof Wl 18 and W115 from March to May 2012 only 6-7
Figure 6-5: Comparison of measured versus modeled evapotranspiration (Dome ET to SWAM ET) on
W118 6-8
VII
-------
List of Tables
Table 3-1: Summary of Monitored Green Roof Sites 3-1
Table 4-1: Summary of Monitoring Duration and Observed Events 4-3
Table 5-1: Water Quality Measurement Parameters 5-2
Table 5-2 Summary of Mean Water Quality Results with Standard of Deviation 5-3
Table 5-3: Average Concentration (AC) and Annual Mass Loading (AML) of NH4+, NOs", and P 5-8
Table 5-4: Projected Reduction in Nutrient Loading per Year from Retrofitting Available NYC
Rooftops 5-9
Table 6-1: Time Aggregate Optimization for W118 and W115 Green Roofs 6-4
VIII
-------
Acronyms and Abbreviations
AC = Average Concentration
Al = Aluminum
AML = Annual Mass Loading
ANOVA = Analysis of Variance
As = arsenic
ASCE = American Society of Civil Engineers
ASTM = American Society for Testing and Materials
B = boron
Ba = barium
BDCA = Bronx Design and Construction Academy
BMP = Best Management Practices
C = control roof
C3 = C3 photosynthesis
C4 = C4 photosynthesis
Ca = calcium
CAM = crassulacean acid metabolism
Cd = cadmium
CGRC = Columbia Green Roof Consortium
CN = curve number
ConEd = Con Edison Building
Cr = chromium
CRE = characteristic runoff equation
CSO = Combined Sewage Overflow
CSS = Combined Sewer System
CU = Columbia University
Cu = copper
DEC = Department of Environmental Conservation (of NYS)
DEP = Department of Environmental Protection (of NYC)
DIA = digital image analysis
EB = energy balance
EPA = U.S. Environmental Protection Agency
ET = evapotranspiration
Fdston = Ethical Culture Fieldston School
Fe = iron
GI = green infrastructure
GR = green roof
ISA = impervious surface area
K = potassium
L =losses
LID = low impact development
Mg = magnesium
Mn = manganese
n = substrate effective porosity
N = runoff nitrogen content
NA = not applicable
Na = sodium
NH4+ = ammonium
IX
-------
Ni = nickel
NOs = nitrate
NOAA = National Oceanic and Atmospheric Administration
NRCS = Natural Resources Conservation Service
NSE = Nash-Sutcliffe Efficiency
NYC = New York City
NYS = New York State
P = total phosphorus or precipitation
Pb = lead
PBIAS = percent bias
Fobs = observed precipitation
Q = green roof runoff
QC = location of green roof runoff quality measurement
QG = location of control roof runoff quality measurement
Qobs = observed runoff
Qpred = estimated runoff from SWAM
RARE = Regional Applied Research Effort
SE = standard error
SM = substrate moisture
SMnorm = normalized soil moisture
SWAM = Soil Water Apportioning Method
SSO = Sanitary Sewer Overflow
Tukey HSD = Tukey Honestly Significant Difference
UHI = urban heat island
UN = United Nations
US = United States
USPS = US Postal Service Morgan Processing and Distribution Center
VG# = location of green roof runoff quantity measurement
VMC = volumetric moisture content
Wl 15 = West 115th Street, NY, NY
Wl 18 = West 118th Street, NY, NY
WWF = Wet Weather Flow
WWTP = Waste Water Treatment Plant
Zn = zinc
-------
Acknowledgements
This project would not have been possible without the U.S. Environmental Protection Agency's (EPA)
Regional Applied Research Effort (RARE) grant. In particular, we are grateful for the support provided by
Thomas O'Connor of the EPA's Urban Watershed Management Branch, Edison, NJ and Rabi Kieber,
Region 2 Green Building/Sustainability Coordinator at EPA, who enabled this research to be undertaken.
Support for the research was also provided, in part, by the National Science Foundation (NSF) grants
CMMI-0928604 and DGE-0903597. The authors also wish to thank the following for their contributions to
the work: Kevin O'Connor from Con Edison; Helen Bielak, Janice Erskine and Emilio Trejo from the
Columbia University Facilities Office and the Columbia University Office of Environmental Stewardship;
Jamie Cohen, Luis Correa, Bhupendra Patel and David Stoff from the U.S. Postal Service; Howard
Waldman and Beselot Birhanu from Ethical Culture Fieldston School; Frank Barona, Philip Judge and
Donald Allison from Regis High School; Nathaniel Wight from the Bronx Design and Construction
Academy; GreenGrid; Smartroofs; Liveroofs; Prides Corner Farms; Tecta Green; Greensulate; Xero Flor;
Clare Buck; and Ellen Lee. The report was reviewed by Dr. Kimberly DiGiovanni of Drexel University and
Dr. Olyssa Starry of the Portland State University. We gratefully acknowledge their helpful comments. Any
opinions, findings, and conclusions expressed in this report are those of the authors and do not necessarily
reflect the views of any other person or entity.
XI
-------
Executive Summary
This report is the result of an U.S. Environmental Protection Agency (EPA) Regional Applied Research Effort (RARE)
project that documents the quantity and quality of runoff from a suite of urban green roofs located in New York City
(NYC). An overall research goal was to assess green roof performance on actual urban rooftops, which have realistic
runoff dimensions to drains and are subject to more realistic urban environmental conditions, as opposed to test plots
at academic research campuses or laboratories.
Green roofs are known to provide multiple urban ecosystem services, foremost of which include urban heat island
mitigation, reduced rooftop winter and summer heat flows, and stormwater management, including runoff retention and
detention. A growing number of different green roof designs are commercially available, although most include the
following components: waterproof membrane, geotextile layer, drainage layer, substrate (also known as growth
medium) and plants. Such layers can be delivered and installed in various ways, but three common types have emerged:
(i) vegetated mat systems that utilize a geo-composite structure to hold substrate, which is then installed like a carpet
over other desired geosynthetics; (ii) built up systems that are constructed on the rooftop itself starting from the
waterproof layer at the bottom and ending with the plant installation at the top, and (iii) modular tray systems comprising
relatively small easy-to-move and manage trays that combine all layers in a single unit and are simply placed tile-like
within the desired roof area.
The material design and composition of each layer in a green roof can vary widely but two of the most important
variables are the growing substrate composition and depth and the plant selection. With regard to growing substrate
depths, a generally accepted definition is that shallow systems, called 'extensive,' are usually 100 mm or less, while
deeper systems, called 'intensive' are usually 150 mm or more. The deeper systems, which are commonly constructed
as built up, offer a much greater opportunity for variable plant choices, including native plant options. Shallow systems
are mostly planted with hardy Sedums, which can thrive in depths as shallow as 25 mm or even less. Generally, extensive
green roofs are cheaper, require less maintenance, and are lighter than intensive systems. Therefore, they are
implemented more frequently and most especially on existing building stock where rooftop weight limitations come
into play. Due to their wider applicability in dense urban environments like NYC, extensive green roofs were the focus
of this study.
With respect to stormwater water quantity performance, we have been monitoring six full-scale green roofs in NYC,
including two vegetated mat systems (named Wl 15 and Wl 18), two built up systems (named USPS and Regis), and
two modular tray systems (named ConEd and BDCA). This report focuses on the analysis of four of these systems
where monitoring equipment has been in place the longest: Wl 15, Wl 18, USPS and ConEd. Continuous rainfall and
runoff data were collected from each green roof between June 2011 and April 2013, resulting in 520 rainfall events,
ranging from 0.25 to 180 mm in rainfall depth, which were used for analyses. Rainfall retention over the entire study
period was found to be 62% for W115, 42% for W118, 56% for USPS, and 59% for ConEd. However, results also
demonstrated that the percent of rainfall retained by the green roofs decreased with increasing rainfall depth and, as a
result, the distribution of rainfall during a study period plays a significant role in reported water retention values. To
extend the analyses of observations made during the study period to longer (decadal) time periods, we explored the
utility of two empirical models, both of which predict runoff from a set of precipitation events. One model used the
characteristic runoff equation (CRE) proposed by Carson et al. (2013), and the other used the widely adopted curve
number (CN) approach developed by the Natural Resources Conservation Service. Both models were applied to 40-
years of historic rainfall events generated from hourly precipitation data recorded in Central Park (NOAA 1971-2010).
During the 40-year period, the CRE method estimated total rainfall retention to be 51% for Wl 15, 43% for Wl 18, 57%
for USPS, and 54% for ConEd; whereas the CN method estimated rainfall retentions of 53% for Wl 15, 48% for Wl 18,
58% for USPS, and 59% for ConEd. Correlation between predicted and observed runoff was high for both methods,
with r-squared values of 0.82 or greater. A major difference between the two models, however, is that the CRE method
accounted for observed changes in relative retention performance of the different green roof systems with storm size:
for example, the ConEd modular tray system captured the lowest percentage of precipitation among all green roof
systems for storms 0-20 mm in depth, and the highest for storms above 30 mm. In contrast, the CN method assumes
- 1 -
-------
that a green roof with a lower CN will always outperform a green roof with a higher CN in every storm category, which
belies our monitoring results. Overall, the CN method predicts higher annual rainfall retention than the CRE method
for each of the green roofs studied.
To determine the impact of urban green roof establishment on stormwater quality, we performed a 16-month water
quality survey of stormwater runoff from five full-scale green roofs, including the four systems where water quantity
performance was being monitored (the W115 and W118 vegetated mat systems, the USPS built up system and the
ConEd tray system) and a second built up system (named Fdston). We also concurrently surveyed water quality at five
non-vegetated (control) roofs, which were located near each green roof study site. Over the study period, we collected
and analyzed more than 100 water samples. We found the measured pH of green roof runoff to be consistently higher
that of the control roof runoff and precipitation, with observed average pH equal to 7.28, 6.27 and 4.82 for the green
roofs, control roofs and precipitation, respectively. All micronutrients (with the exception of sodium) and heavy metals
(with the exception of boron at Wl 18) in the green roof runoff were either detected at very low concentrations or were
below instrument detection limits. Despite variability in the average concentration of nitrate (NOs), ammonium (NH/f)
and total phosphorous (P) in green roof runoff across the different roof types, we estimated that the annual mass loading
of NOs, NH4" and total P per unit rooftop area discharging from the green roof types was less than that from the nearby
control roofs due to the ability of green roofs to retain precipitation. Based on estimated annual mass loadings (mg/m2)
and an assumption that about 20% of rooftops in NYC could be retrofit with an extensive green roof, we project that
widespread green roof installation in NYC could decrease annual stormwater nutrient discharge of total P by over 600
kg, NHzf by over 7,000 kg and NOs by over 150,000 kg. Although these amounts are not large in comparison to the
annual nutrient discharges from waste water treatment plants (WWTP), this still provides evidence that green roof
implementation may improve urban stormwater and subsequent receiving water quality.
Green roof stormwater attenuation performance has often been studied using either lysimeters or custom-made weirs,
as were used in this study. While these systems have yielded very good results, issues of cost and labor regarding their
application in larger monitoring studies are a concern. To address these issues, a new cost-effective method for green
roof hydrologic monitoring that could be implemented on a broad scale was explored. A water balance method - called
SWAM - was derived which relies solely on the monitoring of precipitation and green roof substrate moisture content.
Eleven months of measurements from the Wl 18 and Wl 15 vegetated mat systems were used to validate the results of
this approach. Various statistical tools were used to assess SWAM's performance, with particular focus on maximizing
Nash-Sutcliffe Efficiency (NSE) coefficients, which indicate how well the simulated data fit observed values on a 1:1
line. Volumetric substrate moisture was collected with two low-cost soil moisture probes: a CS615 water content
reflectometer and an ECH2O EC-5 soil moisture probe. The results from SWAM were compared with observed runoff
data from the two green roofs. The method was able to successfully predict daily runoff from the two study sites and
appears to have potential for estimating green roof evapotranspiration.
-2-
-------
Chapter 1 Introduction
According to the U.S. Environmental Protection Agency (EPA), non-point source pollution is the currently the Nation's
largest water quality problem, leading to the impairment of tens of thousands of rivers, lakes and estuaries across the
nation (Berghage et al. 2009). A recent report by the National Oceanic and Atmospheric Administration (NOAA)
(Bricker et al. 2007) also cites non-point source pollution as the single, largest threat to US coastal water quality, Non-
point source pollution is degrading the functionality of the wetlands, marshes and riparian areas that make up many
coastlines, causing eutrophication that is resulting in fish kills, algal blooms and limited growth of sea-grass, threatening
commercial shell-fish beds, and causing major human-health concerns related to the presence of water-borne pathogens.
In urban areas, the leading cause of non-point source pollution is urban storm water discharges caused by wet-weather
flow (WWF). Problem constituents in WWF include visible matter, pathogenic microorganisms, oxygen-demanding
materials, suspended solids, nutrients, and toxicants (EPA 2004). The three most common sources of urban WWF are
stormwater runoff, combined sewer overflow (CSO) and sanitary sewer overflow (SSO). A 2008 report by the New
York State Department of Environmental Conservation (NYS DEC), which highlights poor water quality in much of
the Lower Hudson Watershed, cites urban stormwater runoff and combined sewer overflows as major concerns within
this watershed, which incorporates the dense urban area of New York City (NYC) (NYS DEC 2008).
Because of the magnitude of water quality impairment arising from urban WWF, abatement of WWF pollution is a
major focus of many US government agencies and municipalities. Abatement options for WWF pollution include
control at the source by land management, both in-line and off-line storage, or end of pipe treatment (i.e., in the treatment
plant or with satellite treatment facilities at outfall points). According to Field and Sullivan (2001) traditional
engineering solutions for WWF abatement, such as off-line storage or end of pipe solutions, are difficult for
municipalities and other stakeholders to implement because of design and cost challenges that arise due to the need for
low footprint solutions at the ground or in the sub-terrain. This is particularly true in dense urban environments, like
NYC, where land costs are high, land availability is scarce and subsurface construction can be prohibitively expensive.
As a result, alternatives to traditional solutions to WWF abatement have been aggressively pursued in recent years both
by the academic (Montalto et al. 2007) and professional communities (GeoSyntec Consultants Inc. 2012).
One potential low cost and effective strategy for WWF abatement that is rapidly gaining acceptance in the US is Low
Impact Development (LID). The intent of LID is to mimic predevelopment hydrology (Coffman 2000) which is based
on a combined strategy of conservation to reduce hydrologic impacts and the incorporation of distributed micro-scale
Best Management Practices (BMP's) throughout a subcatchment area. These BMPs are intended to ensure that
stormwater generated in the subcatchment has the chance to infiltrate or evapotranspire, rather than being transported
without reduction to a centralized system. Although LID techniques might not completely eliminate the need for
centralized stormwater management systems, they can greatly reduce the dependence of WWF abatement on costly,
and often aging, centralized systems (Williams 2003; Williams and Wise 2006).
A large number of US cities, including Portland, Seattle, New York and San Francisco (Erlichman and Peck 2013) as
well as cities around the world such as Melbourne, Australia, Shanghai, China and Copenhagen, Denmark
(GreenRoofs.com 2014) are investing in urban green infrastructure as part of larger BMP programs to mitigate the
detrimental impacts of WWF. Urban green infrastructure, such as green roofs, green streets, advanced street tree pits,
rain gardens and bio-swales, introduce vegetation, depression storage, and perviousness back into city landscapes,
thereby enabling local capture and management of stormwater and stormwater pollution. In cities with combined sewer
systems (CSS) the re-introduction of these landscape features reduces both the volume and peak-rate of flow of
stormwater into the CSS, thereby also reducing the occurrence and severity of CSO events.
It has to be recognized, however, that the physical and time scales of green infrastructure implementation needed to
help mitigate the impacts of urban WWF are substantial. For example, NYC's 2010 plan to manage WWF generated
by an inch of rain falling on 10% of the city's impervious area by incorporating green infrastructure into 52% of urban
land served by CSSs at a projected cost of $2.4 billion over the next 18 years (NYCDEP2010). Opportunities to better
1-1
-------
understand and advance the performance of current, conventional green infrastructure are thus important, and such
advances could have significant, positive impact on urban WWF abatement programs for years to come.
In urban areas where land availability for surface area stormwater controls, such as rain gardens and bioswales, is scarce,
green roof technology is an important component of many green infrastructure programs. Rooftops can make up to 40
to 50% of impervious urban land area (Mayor's Office of Long-Term Planning and Sustainability 2008), providing a
meaningful opportunity for implementation of local stormwater controls.
Green roofs, also known as vegetated roofs, eco-roofs, or living roofs, are typically constructed by placing a drainage
course, growing substrate, and vegetation on top of a roofs waterproof membrane. Green roofs have been used for
stormwater management for over thirty years in Germany (Kohler and Keeley 2004) and, according to February 2013
information available in the Green Roof and Wall Projects Database (GreenRoofs.com 2014), the number of green roof
projects in the US today now exceeds 1,000, with many projects concentrated in metropolitan areas. At present, the US
has only recently begun to develop national green roof standards and, as a result, the materials, configuration, and
installation methods for green roofs can vary widely from site to site. For example, in some installations, green roofs
may also have additional geosynthetic layers for preventing plant root penetration damage of the roof membrane,
limiting sediment intrusion into the drainage course, and/or improving water storage.
It is common for green roofs to be classified as either extensive or intensive based on the thickness of the growing
substrate layer. Extensive roof substrates are typically 100 mm thick or less and feature short rooting, drought resistant
plants, whereas intensive roof substrates are greater than 150 mm thick and may be sowed with deeper rooting plants
including shrubs and trees. Generally, extensive green roofs are cheaper, require less maintenance, and are lighter than
intensive systems. Thus, they are implemented more frequently, most especially on existing building stock where
rooftop weight limitations come into play. Due to their wider applicability in dense urban areas like NYC, extensive
green roofs were the focus of this study.
Within the extensive green roof classification, three major construction types have emerged: vegetated mat, built up,
and modular tray systems (Oberndorfer et al. 2007). Typically, both the vegetated mat and built up systems require a
specialized drainage course to prevent ponding and surface flow that would otherwise cause substrate erosion. The two
systems differ, however, in how the substrate is installed. In mat construction the growing substrate is bound within a
geo-composite used for off-site pre-planting, whereas the growing substrate for a built up system is placed within
bordered rooftop regions and landscaped on site. In contrast, the walls of the modular trays already restrict surface
runoff, in turn limiting erosion, and therefore may be placed directly on a roofs waterproof membrane. Each
construction type imposes a unique set of boundary conditions on the growing substrate layer that affects the roofs
drainage behavior and runoff characteristics. For example, the mat and built up systems promote lateral runoff
movement to varying degrees, whereas the unconnected modular trays generally facilitate vertical percolation. The type
of construction might also determine the non-vegetated area required for maintenance activities and the feasibility of
different vegetation types. As a result, the installation method might be a significant factor in overall green roof
performance.
Recently, a number of studies have helped to better understand the role green roofs might play in mitigating CSO
pollution and minimizing problems associated with urban runoff in general (Berndtsson 2010). These studies report a
wide range of hydrologic behavior due to differences in, among other parameters, green roof construction type, growing
substrate depth, vegetation type, and areal coverage. Even similar systems may have significant performance variation
since the water retention ability of green roofs is heavily influenced by local climate; where the distribution, size, and
intensity of rainfall events (Stovin 2010), as well as seasonal evapotranspiration (ET) rates (Bengtsson et al. 2005), are
thought to play a key role.
The goal of this project was to collect data on the hydrological performance of a suite of full-scale extensive green roofs
located in an urban area within EPA Region 2, specifically NYC, in order to provide better understanding of the potential
role of green roofs in urban WWF abatement. An important research objective was to assess green roof hydrological
performance on actual urban rooftops with realistic runoff dimensions to drains, as opposed to test plots at academic
1-2
-------
research campuses or laboratories. The seven green roofs that were studied are located on a variety of building sites and
represent all common extensive green roof installation types, including the vegetated mat, built up, and modular tray
systems. Monitored drainage areas ranged from 38 m2 to about 1275 m2. Water quantity performance was measured
and analyzed over a 22-month period for six green roofs, while water quality performance was measured and analyzed
over a 16-month period for five green roofs. In what follows, overall conclusions and recommendations from the study
are summarized in Chapter 2. Next, Chapter 3 describes the monitoring sites and systems used in the work, while
Chapters 4 and 5 provide results from the water quantity and water quality performance monitoring, respectively.
Chapter 6 then presents an investigation into a new cost-effective method for green roof hydrologic monitoring, based
on substrate moisture (SM) measurements, which has potential for wide scale implementation. SM measurements
collected from two green roofs over an 11-month period were used to test the method. Finally, a list of references cited
in the text is provided in Chapter 7. Additional tables and figures are contained in the appendices of this report.
Appendix A contains a list of equipment at each green roof monitoring site, Appendix B contains a summary of the
water quality measurements, while Appendix C contains water quality statistical analysis results.
1-3
-------
Chapter 2 Overall Conclusions and
Recommendations
2.1 Conclusions
The following conclusions are based on evaluations over two consecutive growing seasons that were obtained from
measurements made on seven extensive green roofs in NYC, which shares the cold dominant climate of the continent
but has hot summers (> 22 °C) and no discernible dry season (Koppen-Geiger climate classification) (Peel et al. 2007).
The roofs that were part of this study are: Wl 15 (a vegetated mat system with a monitored drainage area of 99 m2 and
a substrate depth of 32 mm), W118 (a vegetated mat system with a monitored drainage area of 310 m2and a substrate
depth of 32 mm), USPS (a built up system with a monitored drainage area of 390 m2 and a substrate depth of 100 mm),
Fdston (a built up system with an estimated monitored drainage area of 1275 m2 and a substrate depth of 100 mm),
BDCA (a tray system with a monitored drainage area of 112 m2 and a substrate depth of 115 mm) and Regis (a built up
systems with two monitored drainage areas of 38 m2 and a substrate depth of 100 mm).
Water Quantity Study
Our data indicate that total green roof runoff is strongly correlated to total event precipitation, where the coefficients of
determination (r-squared values) are 0.83 or higher depending on the roof and modeling approach. Our data also show
that that the percent of green roof rainfall captured decreases with increasing event size. The CRE (Characteristic Runoff
Equation) and CN (Curve Number) methods described in Chapter 4 may be used to empirically model long-term green
roof water retention performance based only on historic rainfall records, or even projected events from climate change
modeling.
During our study period, continuous rainfall and runoff data were collected from six of the seven green roofs between
June 2011 and March 2013 (all but Fdston). An in-depth analysis was conducted for four of these roofs: Wl 15, Wl 18,
USPS and ConEd. Reliable runoff data from these roofs were obtained from 520 rainfall events ranging from 0.25 to
180 mm in rainfall depth. From our analysis of these events, we determined 62% overall rainfall retention during the
23-month period for W115, 42% for W118, 56% for USPS, and 59% for ConEd. Using the CRE methodology in
conjunction with 40-years of historic rainfall records for New York City's Central Park, we estimated the range of
annual rainfall capture of the four green roofs to lie between 43-60% for Wl 15, 37-51% for Wl 18, 49-66% for USPS,
and 47-61% for ConEd during the modeling period. The CN method predicted higher retentions than the CRS method.
Differences between the observed retentions during the study period and those predicted using 40-years of historic
rainfall data are attributed to different frequencies in storm sizes between the monitoring period and the historic data.
For example, during the study period the average rainfall retention on W115 was higher than that modeled due to a
lower frequency of large storms than average on this roof during the study period. The curve number (CN) method,
which is a widely used method developed by the Natural Resources Conservation Service (NRCS), could not predict
observed variation in comparative green roof performance between storm sizes, whereas the CRE method did. Factors
driving differences in performance between the four green roofs are thought to be substrate depth, water holding
capacity, and the size and location of non-vegetated areas, i.e. effects of impervious area or other hydrological factors,
on each rooftop area.
Water Quality Study
We performed a 16-month water quality survey of precipitation, and stormwater runoff from five of the seven green
roofs (W115, W118, USPS, ConEd and Fdston) and nearby nonvegetated (control) roofs to determine the impact of
increased green roof establishment on stormwater runoff quality in an urban environment. The water quality indicators
measured included: pH, conductivity, turbidity, apparent and true color, ammonium, nitrate, calcium, potassium,
magnesium, phosphorus, aluminum, arsenic, boron, barium, cadmium, chromium, copper, iron, manganese, sodium,
nickel, lead and zinc.
The key findings of this portion of the study are i) green roofs neutralize acidic precipitation, ii) nitrate concentration
in runoff from the green roofs was lower than that from the control roofs, iii) concentrations of macronutrients, such as
2-1
-------
calcium, potassium and magnesium, were higher in green roof runoff than control roof runoff, iv) concentrations of
micronutrients observed in green roof and control roof runoff were very low, with the exception of sodium and iv) no
significant concentrations of heavy metals were detected in green roof or control roof runoff with the exception of boron
at the W118 green roof, which is attributed to pesticide use at the site. While there appears to be more chemical
constituents present in green roof runoff than control roof runoff, there is an overall reduction in the volume of runoff
from green roofs. Thus, the total mass of nutrient runoff from green roofs is less than that from non-vegetated roofs. As
a result, the water quality benefits of green roofs are favorable in urban environments.
Soil Moisture Water Balance Study
The results from the investigation into a new cost-effective method for green roof hydrologic monitoring indicate that
a soil water balance approach using monitored precipitation and SM content, such as the proposed method introduced
in Chapter 6, can provide a low-cost and low-maintenance alternative to typical systems used for quantifying green roof
runoff. This approach also has potential for estimating green roof ET. The preliminary case study conducted on the
Wl 15 and Wl 18 green roofs, which was evaluated using the Nash-Sutcliffe statistical method, yielded Nash-Sutcliffe
Efficiency coefficients for estimated runoff ranging between 0.72 and 0.88 at daily time aggregates. While any time
aggregate can be used with SWAM, time spans of 24 hours yielded the best results. It was observed that there might be
biases in the soil moisture probe readings - affecting the runoff and ET estimates - caused by instrument temperature
sensitivities as well as instrument location. While these biases were observed, they have not yet been systematically
investigated nor corrected.
2.2 Recommendations for Further Study
The work presented in this report confirms that deploying green roofs on existing buildings can reduce the negative
impacts of urban WWF, including water quality and water quality impacts. Nonetheless, meaningful urban WWF
abatement in many municipalities will require implementation of green infrastructure options beyond extensive green
roofs because aggregate street level impervious area significantly exceeds aggregate rooftop area. Study of the
hydrological performance of green streets, expanded tree pits, bioswales and other green infrastructure interventions,
which are rapidly being adopted in many US cities, is therefore critical to fully understanding the role of green
infrastructure in addressing urban stormwater issues. Expansion of the work presented here to wider green infrastructure
systems is therefore recommended, including exploration of the SWAM, or an alternative method, for wide spread
monitoring of urban green infrastructure performance. In addition, continued monitoring of the urban green roofs that
were part of this project would provide the data needed to understand the evolving performance of urban green roofs
with age, as well as the role of seasonality in green roof hydrology. Other recommendations include undertaking relative
cost-benefit analysis of green roofs versus other stormwater management technologies, more research experiments
considering driving factors for water control such as substrate depth and water holding capacity, and continued studies
that will optimize design with respect to maintenance and performance.
2-2
-------
Chapter 3 Monitoring Sites and
Systems
3.1 Green Roofs and Monitoring Equipment
The following table provides a summary of the green roof installations that were monitored as part of this study. The
installations span a variety of commercially available green roofs. The green roof components in each case were all
provided and installed by various and independent green roofing companies (Table 3-1). Figure 3-1 illustrates the
locations of the green roof sites, which are distributed throughout NYC with a maj ority of the sites falling within priority
combined sewer sheds.
Table 3-1: Summary of Monitored Green Roof Sites
Site
Construction
Type
Manufacturer
Year Built
Substrate
Depth (mm)
Vegetation
Type
Monitored
Watershed
Area (m2)
Watershed
Vegetated (%)
Monitoring
Conducted
W115
Vegetated
mat
Xero Flor
America
2007
32
Sedum mix
99
58
Runoff
Quantity and
Quality
W118
Vegetated
mat
Xero Flor
America
2007
32
Sedum mix
310
53
Runoff
Quantity and
Quality
USPS
Built up roof
Tecta Green
2009
100
(200 berms)
Sedum mix
and natives
390
67
Runoff
Quantity and
Quality
ConEd
Modular tray
GreenGrid
Roofs
2008
100
Sedum mix
940
52
Runoff
Quantity and
Quality
Fdston
Built up roof
American
Hydrotech
2007
100
Sedum mix
and natives
1275
(assumed)
50
Runoff
Quality
BDCA
Modular tray
Liveroofs
2010
115
Sedum mix
112
65
Runoff
Quantity
Regis
Built up roof
Greensulate
2010
100
Natives
38
65
Runoff
Quantity
The investigation into the hydrological performance of the urban green roofs involved three monitoring components:
monitoring of (i) environmental conditions, (ii) the quantity of roof runoff and (iii) the water quality of roof runoff.
3-1
-------
1. W115
2.W118
3.ConEd
4. USPS
5. Fdston
6. BDCA
7. Regis
G.R. Monitoring
CSOshed
Priority Levels
1.5
Figure 3-1: Locations of monitored green roof sites and priority combined sewer sheds in New York City
The monitoring of environmental conditions was conducted using weather stations installed on each roof.
Measurements of precipitation, runoff, temperature, radiation, humidity, and wind speed and direction were undertaken
at the studied green roofs. Monitoring equipment budgets varied for each roof and thus sensor selection was not identical
everywhere. Tables in Appendix A provide a description of monitoring equipment installed at each green roof site. Data
from the equipment was stored by on-site data loggers at 5-minute intervals. For equipment connected to Wi-Fi or GSM
Cellular HOBO U30 data loggers, sample readings were taken every second and five minute averages were recorded
and wirelessly uploaded the to the Onset Hobolink data service every hour. Data were then accessible on-line via the
Hobolink service. For the equipment connected to the Campbell Scientific data loggers, the data were stored on site and
needed to be downloaded from the data logger at periodic intervals. The quantity of roof runoff was measured through
use of custom drainage pipe weir devices created for each roof. The quality of roof runoff was measured through lab
analysis of manually collected water samples. The methodologies for runoff quantity and quality measurements are
further discussed in Chapters 4 and 5, respectively.
3-2
-------
3.2 Green Roof Site Descriptions
The following paragraphs provide images and descriptions of the study's monitored rooftops. Figure 3-2 through Figure
3-9 consist of: (A) satellite images of each roof (Courtesy Google Maps), with locations of green roof water quantity
measurements (VG#), as well as, green roof and control roof water quality measurements (QG and QC, respectively)
and drainage areas monitored for green roof water quantity are denoted by the dotted lines; (B) photographs of the weir
devices used to measure green roof water quantity (if present); (C) photographs of the roof taken on-site.
Figure 3-2: W115 Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
Figure 3-3: W118 Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
W115 and W118: The 635 West 115th Street building (Wl 15 - Figure 3-2) houses the Columbia University Office of
Environmental Stewardship, while the 423 West 118th Street building (W118 - Figure 3-3) is a Columbia University
graduate student residence. In 2007, a pre-vegetated mat, Xero Flor America's XF301+2FL green roof system, was
retrofitted on both buildings. This system consists of a 32 mm thick pre-vegetated mat, supported by two 6 mm thick
3-3
-------
water retention fleeces created from recycled synthetic fibers, a 19 mm non-woven polymer drainage mat, and an 0.5
mm polyethylene root barrier. A variety of Sedum species, such as Saxifraga granulata, Sedum acre, Sedum album,
Sedum ellacombianum, Sedum hybridum 'Czars Gold', Sedum oregonum, Sedum pulchellum, Sedum reflexum, Sedum
sexangulare, Sedum spurium var. coccineum, Sedum stenopetalum, are present on these roofs. The growing substrate
on these roofs has a water-saturated density of 1.37 g/cm3, water storage capacity of 37.1%, and a saturated hydraulic
conductivity of 0.021 cm/s, as reported by Hummel and Co., Inc in April 2007. The Wl 15 green roof has a single 99
m2, 58% vegetated watershed connected to an exterior parapet downspout. The 600 m2 Wl 18 total roof area consists of
two watersheds connected to exterior parapet downspouts, of which the 310 m2, 53% vegetated drainage area of the
Southeast watershed was monitored for rainfall and runoff. Gravel walkways, parapets, and the raised rooftop above
the elevator shaft comprise the non-vegetated areas of both rooftops.
Figure 3-4: USPS Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
USPS: Figure 3-4 shows the 10,000 m2US Postal Service Morgan Processing and Distribution Center (USPS) green
roof in mid-Manhattan, which was installed in 2009 by TectaGreen of Tecta America. The roof was built in-place. Roof
edges were established with 100 mm tall metal brackets, and an expanded shale based substrate of varying depth was
placed in the bounded area. A majority of the green roof is comprised of 100 mm of substrate depth and was planted
with Sedum species, including: Sedum acre, Sedum album 'Coral Carpet', Sedum album murale, Sedum reflexum,
Sedum sexangulare, Sedum reflexum 'Blue Spruce', Sedum grisebachii, Sedum kamtschaticum, Sedum 'Matrona', Sedum
pluricaule 'Rosenteppich', Sedum spurium 'Roseum', Sedum telephium 'Autumn Joy'. Additionally, the 200 mm deep
berms throughout the roof, usually about 2 m wide, have the following larger plant species: Achilea fllipendula
'Moonshine' Alium schoenoprasum, Coreopsis vert 'Moonbeam' Silenecaroliniana ssp. wherryi, Talinum calycinum,
Tradescantia ohiensis. The growing substrate has a water-saturated density between 1.15-1.35 g/cm3, water storage
capacity between 35-65%, and a saturated hydraulic conductivity between 0.001-0.120 cm/s, as reported by Skyland
USA LLC in March 2011. Monitoring equipment was installed in a 390 m2 watershed in the Northwest corner of the
roof. The watershed has one 6 m long berm and a single internal downspout. The watershed is 67% vegetated with the
remaining area consisting of gravel ballast.
3-4
-------
QG
%
QC
Figure 3-5: ConEd Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
ConEd: ConEdison Learning Center (ConEd) green roof in Queens (Figure 3-5), which was installed in 2008, consists
of GreenGrid-G2 modular trays with dimensions 61 cm x 122 cm x 10 cm. The trays were packed with a proprietary
expanded shale substrate and then placed in adjacent rows on the 2,700 m2 roof area. The growing substrate has a water-
saturated density of 1.18 g/cm3, water storage capacity of 31.8%, and a saturated hydraulic conductivity of 0.326 cm/s,
as reported by Penn State University's Agricultural Analytical Services Laboratory in July 2008. Plugs and cuttings
used to plant were comprised of the following 15 Sedum varieties: Sedum oreganum, Sedum kamtschaticum
Weihenstephaner Gold', Sedum kamtschaticum, Sedum ternatum, Sedum 'John Creech', Sedum spurium 'Album
Superbum', Sedum spurium 'Fulda Glow', Sedum spurium 'Dragons Blood', Sedum spurium 'Bronze Carpet', Sedum
angelina, Sedum sexangulare, Sedum 'Ruby Glow', Sedum 'pachclados', Sedum 'Bertram Anderson', Sedum 'Vera
Jameson'. Monitoring equipment for this study was installed in the 52% vegetated, 940 m2 Eastern watershed. Both of
the watershed's internal downspouts were monitored for runoff. The non-vegetated sections of this roof are comprised
of rubber mat walkways, gravel ballast transitions and raised glass skylights.
3-5
-------
Figure 3-4: Fdston Roof (A) Satellite photograph (B) Equipment (C) Roof photograph
Fdston: The Ethical Culture Fieldston School (Fdston), which teaches K-12, has been operating in the Bronx location
since 1929. In early 2007 the school began construction of a new middle school building and this provided the
opportunity to install two different built up green roofs (Figure 3-4(A)). The green roof installer was the Town and
Garden landscaping firm. The larger green roof, which is 5100 m2 in area with 100 mm of substrate depth, was
monitored for water quality only. This roof has four drains, and it is assumed (but not confirmed) that the monitored
watershed is 1275 m2, i.e. one quarter of the 5100 m2. Plants on the roof were installed as plugs in August 2007. The
six planted species are: Sedum album, Sedum sexangulaire, Sedum reflexum, Sedum floriferum, Sedum hybridum and
Sedum spurium. The growing substrate has a water-saturated density of 1.34 g/cm3, water storage capacity of 42%, and
a saturated hydraulic conductivity of 0.119 cm/s, as reported by Penn State University's Agricultural Analytical
Services Laboratory in May 2007. Growing substrate particle sizes ranged from 0.02 mm to 9.5 mm with the maximum
percentage (30%) ranging from 3.2 to 6.3 mm and had an organic mass content of 4.2%. The Fieldston roof is about
50% vegetated, with non-vegetated areas consisting of paved walkways and mechanical equipment.
3-6
-------
BDCA
Figure 3-5: BDCA Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
BDCA: The Bronx Design and Construction Academy (BDCA) is a public career and technical education high school
located in the Bronx, NY. The 215 m2 hybrid modular tray green roof in the building's courtyard was constructed in
the fall of 2010 by SmartRoofs, LLC, a division of Sustainable South Bronx (Figure 3-5). The roof structural capacity
and waterproof membrane condition were evaluated prior to installation. Liveroofs' standard system, consisting of an
engineered substrate placed within 25 x 50 x 8 cm trays, was used for the tray system. The trays were overfilled with
substrate to 100 mm, with the help of temporary side walls. The plant species, grown from plugs and cuttings by Prides
Corner Farms, consist of: Euphorbia myrsinites, Sedum acre 'Aureum', Sedum album ('Coral Carpet', 'Green Ice', and
'Chloroticum), Sedum sexangulare, Sedum spectabile ('Brilliant', 'Stardust', and 'Neon'), Sedum spurium ('John
Creech', 'Summer Glory' and 'Royal Pink'), Allium senescens ssp. montanum, Sedum vera Jameson, Sempervivum
'Ruby Heart', Sedum immergrunchen, Sedum 'Angelina. Once filled and planted, the trays were transported to the roof
location, and placed directly on top of the roofs gravel ballast. After all trays were placed, the temporary side walls
were removed to create a seamless roof and a barrier was installed at the green roof borders to keep the substrate intact.
The growing substrate has a water-saturated density of 1.44 g/cm3, water storage capacity of 48.3%, saturated hydraulic
conductivity of 0.018 cm/s, organic matter content of 4.5%, and a small (<0.05 mm) particulate concentration of 6.1%,
as reported by Penn State University's Agricultural Analytical Services Laboratory in March 2008. The runoff quantity
was monitored from the southern 112 m2 drainage area, which is 65% vegetated. The non-vegetated roof areas consist
of gravel ballast.
3-7
-------
Figure 3-6: Regis Roof (A) Satellite photograph (B) Weir device (C) Roof photograph
Regis: Regis High School (Regis) is a private school established in 1914 and located on the Upper West Side of
Manhattan, NY. The school's 2000 m2 hybrid green roof was installed in August 2010 (Figure 3-6). The roof was
constructed onsite by Greensulate, who established roof layers before adding an extensive green roof substrate designed
and manufactured by Long Island Compost Corp. The monitored water quantity runoff sites consist of two 38 m2
elevated roof sections with a 100 mm substrate depth. These areas are 65% vegetated, where nonvegetated areas consist
of gravel ballast. Dr. Matthew Palmer of Columbia University chose the native plant species, by considering two native
shallow substrate environments. The species consist of: Asclepias tuberosa, Baptisia tinctoria, Eupatorium
hyssopifolium, Panicum virgatum, Schizachyrium scoparium, Solidago nemoralis, Sorghastrum nutans,
Symphyotrichum leave, Danthonia spicata, Dechampsia flexuosa, Dichanthelium clandestinum, Eupatorium
sesslifolium, Lespedeza hirta, Pycnanthemum tenuifolium, Rudbeckia hirta, and Solidago odora.
3-8
-------
Chapter 4 Water Quantity Study
4.1 Introduction
To date, the most cited aspect of green roof behavior is the ability of green roofs to capture rainfall and limit the
generation of stormwater runoff. Typically this behavior is quantified using a mass balance approach, where rainfall
and runoff are continuously measured and the performance of a system is reported as the total percent of rainfall
retention over a given monitoring period. Using this strategy, researchers have reported rainfall retentions of extensive
green roofs between 12% and 86% (Hutchinson et al. 2003; De Cuyper et al. 2004; Liu and Minor 2005; Moran et al.
2005; VanWoert et al. 2005; Connelly et al. 2006; Toronto and Region Conservation Authority 2006; Getter et al. 2007;
Teemusk and Mander 2007; Hathaway et al. 2008; Kurtz 2008; Spolek 2008; Berghage et al. 2009; Bliss et al. 2009;
DiGiovanni et al. 2010; Berghage et al. 2010; Voyde et al. 2010; Gregoire and Clausen 2011; Nardini et al. 2011; Palla
et al. 2011; Schroll et al. 2011; Stovin et al. 2012; Carson et al. 2013; Fassman-Beck et al. 2013; Morgan et al. 2013).
The large range of reported performance highlights the wide variety of green roof configurations and the multitude of
parameters that impact green roof water retention capacity. Considering this, there is a need to quantify the relationships
between green roof parameters, storm characteristics and rainfall retention in order to (1) develop industry standards
that optimize green roof systems, (2) create more accurate methods for evaluating stormwater management benefits
prior to green roof installation and (3) understand the overall potential contribution of widespread adoption of green
roofs to urban WWF abatement.
Studies have shown that rainfall capture is influenced by a green roofs substrate depth (De Cuyper et al. 2004;
VanWoert et al. 2005), rooftop slope (VanWoert et al. 2005; Getter et al. 2007), areal plant coverage (Berghage et al.
2009; Morgan et al. 2013), plant type (Nardini et al. 2011; Starry 2013), drainage configuration (Berghage and Gu
2009), orientation (Mentens et al. 2003), and location of non-vegetated spaces (Carson et al. 2013). These parameters
are generally well reported in green roof literature. However, in addition to characteristics of the green roof itself,
climate based factors, such as rainfall event intensity (Villarreal 2007), event size (Berghage et al. 2009), and
temperature (Schroll et al. 2011), were found to effect retention. Rainfall event size in particular has been shown to
significantly impact reported performance since rainfall retention may vary widely on a per event basis (Stovin et al.
2012). For instance, a study undertaken during a period of mostly small events (<10mm), where retention can be
between 80-100%, will report a much better performance than a study with mostly large events (50mm+), where
retention might only range from 20-40% (Carson et al. 2013). As a result, it is difficult to make comparisons between
studies, even those in similar climates, without a methodology that eliminates the impacts of event variability between
studies.
At present, there are two main empirically-based methods for describing the variation of green roof rainfall retention
by event size; namely, the curve number (CN) method developed by the NRCS and the characteristic runoff equation
(CRE) method recently proposed in Carson et al. 2013. Both methods relate rainfall to green roof runoff and, therefore,
may be coupled with historic rainfall data to simulate green roof performance for any number of years. In addition,
these methods could also be coupled with climate change forecasts to predict future green roof performance under
changing weather patterns. Theoretically, if the CN or CRE were known for multiple green roofs, the impact of
widespread green roof adoption on urban WWF mitigation could be explored.
In what follows, hydrologic observations related to water quantity are presented for four of the full-scale, extensive
green roofs described in Chapter 3, namely W115 (vegetated mat), W118 (vegetated mat), USPS (built up roof), and
ConEd (modular tray). The dataset includes information collected from 520 storm events occurring between June 2011
and April 2013, which were used to determine a CRE and CN for each green roof. To estimate the multi-year
performance of the green roofs, 40 years of NYC historic precipitation data were used as input to the CRE and CN
methods, and predictions from the methods were compared to the measured results as well as to each other. Additionally,
preliminary observations from monitoring of BDCA (modular tray) and Regis (built up roof), which do not have
sufficient data to perform a historic analysis, are presented. The Chapter ends with a discussion of results and summary
conclusions.
4-1
-------
4.2 Methodology
The W115, W118, USPS, and ConEd green roofs were instrumented to collect a variety of environmental data as
described in Chapter 3 and Appendix A. Stormwater attenuation behavior was evaluated using a mass balance approach
where rainfall and runoff were measured using a tipping bucket rain gauge and a custom-built weir device, respectively.
Figure 4-1: (A) Runoff monitoring weir device (B) Calibration chamber used to simulate rooftop runoff.
The weir devices for runoff measurement consist of a runoff chamber with an outlet weir and a Senix TSPC-30S1
ultrasonic sensor, Figure 3-2(A). The ultrasonic depth sensor measures the depth of water behind the weir face as water
discharges with a resolution of 0.086 mm, which translates to about 3 ml/min. As flow increases, the water level behind
the weir's face rises. The ultrasonic sensor detects the rise in water height and adjusts its output voltage accordingly.
The weir devices were sized to fit into existing rooftop downspouts (or parapet drain in the case of W118, Figure 3-
3(B)) and accommodate roughly 50 mm/hr of rainfall in saturated substrate conditions based on the drainage area.
Above this flow rate, water overflows the weir into the roof drain to prevent backup and ponding of water on the roofs.
Each weir was constructed by cutting acrylic parts and joining them with Scotch-Weld DP-810NS acrylic epoxy. The
weir face was cut from a flat piece of acrylic and attached to a vertical cutout on the side of the weir cylinder. A baffle
was installed at the top of the device and rubber based sealant was applied on all edges to minimize turbulence and
eliminate leaks without restricting water flow. The weir devices function between 0° to 70°C. To calibrate each weir, a
box was built that effectively simulates water flow conditions into roof drains, Figure 3-2(B). Weirs were sealed into
the simulation box, as they would be under field conditions, and calibrated up to their designated maximum capacity.
Water was pumped into the simulation box, flowed under the baffle, then rose up to enter the weir from all directions.
Repeat measurements were taken at incrementally increasing flow rates using an Armfield Fl-10 hydraulic bench,
which was supplemented with a 6 L/s pump at high flow rates. The corresponding voltage output was recorded from
the Senix ultrasonic sensor. The resulting data points were used to derive a calibration curve that related sensor output
voltage and flow rate. This calibration method significantly reduces errors compared to other techniques that rely on,
for example, a combination of measurements at low flow rates and reported weir equations. Once calibrated, weirs were
sealed into the rooftop drains to prevent water loss prior to measurement. Finally, the voltage output of the Senix
ultrasonic sensor was connected to the rooftop data logger for recording.
For the W115, W118, USPS and ConEd roofs, the rainfall and runoff measurement devices were connected to an Onset
Hobo U30 data logger. Sample readings were taken every second and five minute averages were recorded and wirelessly
uploaded to the the Onset Hobolink data service every hour. The unique calibration equation for each weir device was
then applied to the voltage readings and normalized by the monitored drainage area to determine average runoff depth
at each five-minute interval. Continuous data were collected between June 2011 and April 2013, with the exception of
several intermittent offline periods due to hurricane safety measures, power loss, or equipment failure.
Individual storm events were determined from resultant data considering a 6-hour "no rainfall" period between storm
events. This definition was selected based on Strecker et al. (2002) and has been used by VanWoert (2005), Getter et
4-2
-------
al. (2007), Berghage et al. (2009), Voyde et al. (2010) and Stovin et al. (2012), amongst others for green roof studies.
For the purpose of this study, an event begins when rainfall is first recorded and ends when no precipitation or runoff
has been recorded for six hours. Once individual storms were separated in this manner, storm events considered
unsuitable for analyses were discarded based on the following four exclusion criteria: (1) peak runoff rate exceeded
90% of the weir device's voltage output range, since turbulence at the upper limit of the recordable flow rates distort
depth recordings (36 Events); (2) precipitation included snowfall, since the time scale of snowmelt runoff prevents
reliable application of the NOAA event definition (19 Events); (3) total event runoff exceeded rainfall, which occurred
when leaves and other debris clogged the lower portions of the weir causing elevated runoff readings (29 Events); and
(4) sensor recording error, due to the loss of power or equipment failure (24 Events). Following removal of data based
on the above criteria, 520 so-called "reliable" storm events were identified from the original 628 recorded events across
all four monitored roofs over the approximately 2 year study period. From this point forward, all discussion of observed
storm events from the monitoring period is limited to this subset of recorded events deemed reliable for analysis. Event
details are provided in Table 4-1.
Table 4-1: Summary of Monitoring Duration and Observed Events.
Abbreviated Name W115 W118 USPS ConEd
Data Start
Data End
# Total Events
# Reliable Events
# Events (0-10 mm)
# Events (10-20 mm)
# Events (20-30 mm)
# Events (30-40 mm)
# Events (40-50 mm)
# Events (50+ mm)
7/11
10/12
127
105
77
20
4
2
1
1
6/11
4/13
161
134
84
20
17
3
3
7
6/11
4/13
199
179
119
30
13
7
1
9
7/11
4/13
141
102
67
19
10
1
4
1
Characteristic Runoff Equations (CREs) for each of the four green roofs were derived according to the methodology
presented in Carson et al. 2013. Specifically, for each green roof the recorded rainfall and runoff depth were related to
each other using a quadratic regression analysis of all events with non-zero roof runoff. Removal of the zero-runoff
storms in creation of the CREs limits the lower-bound overestimation of runoff caused by fitting a large number of
small events that generate zero runoff. Each of the CREs is considered applicable for event sizes ranging between the
equation's lowest rainfall value associated with roof runoff and 100 mm of rainfall. The 100 mm cap was selected
because of the limited number of events exceeding 100 mm that were recorded during the monitoring period. Thus,
green roof retention behavior above this threshold is outside the bounds of this analysis. For the purpose of the
evaluation that follows, when rooftop rainfall exceeds 100 mm it is assumed that rooftop storage capacity is reached
and a fixed attenuation depth (mm) applies, which is equal to the value of the CRE at 100 mm of rainfall. When rainfall
is less than the lowest rainfall value associated with roof runoff, runoff is set to zero.
The CN method uses a series of three equations (eq. 4-1 through 4-3) to estimate runoff depth for any rainfall event,
assuming that the CN of a particular watershed is known. In practice, the CN is typically selected from a list of
recommended CN, provided in the NRCS technical manual, which range from 30-100 depending on hydrologic soil
group, cover type, treatment, hydrologic condition, and antecedent runoff conditions of the watershed (NRCS 1986).
4-3
-------
<-
1000
S=— -10 [4-2]
;„ = 0.2(5) [4 - 3]
where:
Q = runoff (in)
P = rainfall (in)
S = potential maximum retention after runoff begins (in) and
Ia = initial abstraction (in)
In this analysis, a best-fit CN was selected for each of the four green roofs. To do this, a program was written which
iterates for all possible CNs and calculates runoff using observed rainfall events as inputs in equation 4-1. When input
precipitation was less than the initial abstraction, runoff was set to zero. For each CN, an r-squared value relating the
predicted performance and observed runoff was calculated. The CN with the highest r-squared value (i.e. the best fit)
for each green roof was then selected for use. It is important to note that only events with non-zero runoff were used
for generating the CNs presented in this report in order to maintain consistent methodology between the generation of
CREs and CNs for each of the roofs.
As the focus of this study was quantification of behavior of full-scale green roofs, and the CREs and CNs reflect full-
scale green roof behavior. The calculated CRE and CN represent composites of the vegetated and non-vegetated areas
as full-scale green roof systems always include non-vegetated areas.
Hourly precipitation data, recorded by the Belvedere Castle weather station in Central Park, NYC, were downloaded
from the NOAA National Climatic Data Center website (ncdc.noaa.gov) for the years 1971-2010. These records were
used to identify storm events based on the NOAA standard of 6-hour dry weather period between individual events.
The historic data were continuous with the exception of November 1983 and December 1983, when hourly data were
not available. This analysis of hourly precipitation records resulted in the identification of 4,291 historic precipitation
events over the 40-year period.
The historic rainfall events were used for two primary purposes. First, the distribution of events over the 40-year period
was compared to the distribution of observed events at each of the Wl 15, Wl 18, USPS and ConEd green roof sites.
This comparison helped identify whether the frequency of certain event sizes during the monitoring period were atypical
compared to "average" conditions in NYC. Since event size influences the percent of rainfall retained by a green roof,
atypical events during green roof monitoring periods need to be called out in reported results. Second, the historic events
were used as rainfall inputs to the CRE and CN empirical methods to estimate the total rainfall retention that would be
anticipated for each of the green roofs over a 40-year period that reflected rainfall patterns similar to those in Central
Park from 1971-2010. Because an identical set of rainfall events was used as input to each of the CRE and CN methods,
comparisons between the two methods were possible.
4.3 Results
The 520 reliable storm events (Table 4-1) were used for analysis. Rainfall depth of the monitored events lay between
0.25 and 180 mm, while green roof runoff depth (runoff volume divided by monitored drainage area) ranged from 0 to
159 mm. The W115, W118, USPS, and ConEd green roofs retained 62%, 42%, 56%, and 59% of the total rainfall in
the monitoring period, respectively. Percent of rainfall retention for individual events ranged widely from 28-100% for
Wl 15, 3-100% for Wl 18, 10-100% for USPS, and 13-100% for ConEd. The total number of storm events that generated
zero runoff was 33, 49, 83, and 20 for the Wl 15, Wl 18, USPS, and ConEd roofs, respectively; while the largest event
4-4
-------
with 100% retention was 3.56 mm, 7.62 mm, 16.51 mm, and 2.03 mm, respectively.
100
0-10
10-20
20-30 30-40
Storm Size (mm)
40-50
Figure 4-2: Monitored total rainfall retention for the W115, W118, USPS, and ConEd green roofs by event size.
Figure 4-2, shows the total rainfall retention (calculated as the total retention for all events within each size or seasonal
category) of each roof for the different storm categories. The figure illustrates that, generally, as rainfall depth increases,
the green roofs percent retention is reduced. These results agree with the reported findings of others (Stovin 2010;
Berghage et al. 2009; Getter et al. 2007). Total percent retention of storms up to 10 mm was 85%, 86%, 95%, and 82%
for the W115, W118, USPS, and ConEd green roofs, respectively. The ConEd modular tray system attenuated less
rainfall than the other roofs in the 0-10 and 10-20 mm category, but had better comparative retention performance
during larger events (30 mm+). For instance, during storms with 50 mm or more rainfall, the average retention for
W115, W118, USPS, and ConEd was 36%, 21%, 28%, and 34% respectively. W115, although the smallest roof, had
better overall performance than Wl 18, even though both systems were vegetated mats. This could be due to the higher
percent vegetation of W115 (58%) compared to W118 (53%), the differences between the locations of non-vegetated
portions of each roof, or the fact that there were fewer larger events recorded on Wl 15; possibly because the taller
buildings surrounding Wl 15 create a rain shadow on the Wl 15 roof. The limited numbers of recorded events on Wl 15
above 20 mm (8 Events) undermine the reliability of the W115 retention data for the larger storms. For storms below
20 mm, the Wl 15 and Wl 18 retention performance is similar, although Wl 15 still outperforms Wl 18.
4-5
-------
As noted above, the distribution of observed events must be comparable to that of historic precipitation for reported
retention to be representative of a green roofs multi-year performance, not accounting for other factors such as substrate
consolidation and depletion, changes in plant population and health, etc. Figure 4-3 compares the distribution of rainfall
by event size between the 40-year historic data period (Central Park, NYC 1971-2010) and observations during the
monitoring period. Wl 15's inflated performance in the 50+ mm category is most likely due to limited recorded events
(1%) in that range during the monitoring period, while the green roofs overall performance is influenced by a high
number of storms in the 0-10 mm range (73%) (Figure 4-3 (A)). Additionally, there is an increased percentage of total
volume of events with 50+ mm of rainfall for W118 (48%) and USPS (56%) compared to the historic data (26%).
Consequently, the reported overall rainfall retention performance of Wl 18 and USPS during the monitoring period is
likely reduced compared to expected multi-year performance.
I USPS
ConEd
I NYC
80
g 60
in
&
o
I- 40
20
— n_n _- -^ _«n_
3
0-10 10-20 20-30 30-40
Event Size (mm)
40-50
50+
1 USPS
ConEd
1 NYC
80
£
Q.
9
O
TR 60
05
cc
a 40
o
c 20
In lit 11
i.n.n
..Jn
0-10 10-20 20-30 30-40
Event Size (mm)
40-50
50+
Figure 4-3: Rainfall by event size for monitored and historic (NYC) data as percent of (A) events and (B) depth.
The CREs derived from the monitored storm events (Figure 4-4), were created from regression analyses of events with
non-zero runoff. The zero attenuation line in Figure 4-4 represents a hypothetical roof where all precipitation becomes
runoff. The CREs have coefficients of determination (r-squared values) of 0.95, 0.97, 0.97, and 0.90 for Wl 15, Wl 18,
USPS, and ConEd, respectively. Residuals are randomly distributed where error is likely due to the fact that the CREs
do not account for environmental conditions such as temperature, relative humidity and antecedent SM conditions that
also might impact green roof water balance (Figure 4-5).
4-6
-------
Ro = 0.0052(Rf)2 + 0.3393(Rf) - 0.6260
[for 1.80
-------
B
£_
_U5
to
2
T3
15 -
10 -
5 -
0
-5 -
-10 -
-15
0
W115CN
W115CRE
li
10
10
Rainfall (mm):
102
15
10
5
0
-5
-10
-15
•
o
W118CN
W118CRE
10
10
Rainfall (mm):
10
10
Rainfall (mrn):
15
10
I 5
J2
0
-15
A ConEd CN
A ConEd CRE
10"
10
Rainfall (mm):
10'
Figure 4-5: Residuals between characteristic runoff equations and curve number and observed depth for (A) W115, (B) W118, (C)
USPS, and (D) ConEd.
The CREs were applied to the 40 years of historic data from Central Park, NYC. The predicted rainfall retention during
the entire period for W115, W118, USPS, and ConEd was 51.3%, 43.6%, 57.0%, and 54.4%, respectively. As
hypothesized, annual performance changed each year due to variation in yearly storm size distribution, with annual
rainfall retention of 43-60% for W115, 37-51% for W118, 49-66% for USPS, and 47-61% for ConEd. Figure 4-6 (A)
is analogous to Figure 4-2, but now shows the predicted rainfall retention performance for the 40 years. As seen, the
ConEd green roof is predicted to retain the least runoff for storms 0-10 mm and the most for storm 20 mm or more.
However, USPS had the highest overall retention due to its better comparative retention in the 0-10 and 10-20 mm
category (94% and 68%, respectively) and the increased frequency of these storms in the historic data (66%).
Nonetheless, the performance difference between USPS and ConEd only varied slightly between 0-5% each year, based
on storm size distribution. Wl 15, while no longer showing the best retention among all roofs, still outperforms Wl 18
4-8
-------
in storms greater than 10 mm. Application of the CN method showed 0.9-4.5% higher overall retention than the CRE,
with predicted retentions of 53.4%, 47.8%, 57.9%, and 58.9% for Wl 15, Wl 18, USPS, and ConEd, respectively (Figure
4-6(B)). For reference, based on the total rainfall depth over the 40-year simulation period, a one percent change in
rainfall retention is the equivalent of a 500 mm (~20 inch) change in cumulative retention depth. Because the CN
method assumes that a green roof with a lower CN always outperforms a green roof with a higher CN, the CN method
did not capture observed changes in comparative performance between the four green roofs with storm size (see Figure
4-2). Instead, the CN method predicted ConEd to have the highest retention of all roofs for all storm size categories.
100
80
CD
tr
CL
75
o
60
40
20
0-10 10-20 20-30 30-40
Storm Size (mm)
40-50
50+
100
80
£ 60
40
20
I
II
[Bl
0-10 10-20 20-30 30-40
Storm Size (mm)
40-50
50+
Figure 4-6: Predicted rainfall retention over historic period using (A) Characteristic runoff equations (B) Curve number.
4.4 Results from BDCA and Regis
An analysis was also performed forthe BDCA and one of the Regis rooftops (labeled in Figure 3-6(A) as VGi). Reliable
storms were determined in the same manner as described for the other roofs, resulting in 32 and 23 reliable events for
BDCA and Regis, respectively. Results show that BDCA retained 55.6% and Regis retained 54.1% of rainfall during
their monitoring periods. Preliminary CREs were generated, but were not applied to the historic data for these roofs due
to the absence of large events. With regards to small event performance, the CREs show the BDCA and Regis roofs do
not generate runoff until 2.3 mm and 6.6 mm of rain, respectively (Figure 4-7). The fixed nature of the CN method
constrains large storm performance and allows a preliminary application of the CN method for these two roofs. The
CNs were found to be 97.4 and 95.5 for the BDCA and Regis roofs, respectively which, when applied to the 40-year
historic rainfall data, results in 30.5% and 42.8% retention over the simulation period, respectively.
4-9
-------
100
80
60
o
BDCA Events
10
(Rf) Rainfall (mm): Log Scale
100
80
60
o
Regis Events
B
10
(Rf) Rainfall (mm): Log Scale
10'
Figure 4-7: Characteristic runoff equations and events for (A) BDCA (B) Regis.
4.5 Discussion of Results
In what follows, discussion of results related to the green roof water quantity investigation is presented under three
separate headings: observations, predictive methods and findings from the multi-year predictions using historic rainfall
data.
Observations of Green Roof Retention Performance
The rainfall retention percentages observed for the four green roofs that were analyzed in this study (W115, W118,
USPS and ConEd) fall within the range of performance documented by others. In addition, the observed data also agree
with the general expectation that the percent of rainfall retained by a green roof will decrease as cumulative event
precipitation increases. However, perhaps the most significant finding from the data is that while the ConEd green roof
retained less rainfall than USPS in the 0-20 mm categories, it retained more rainfall for all larger event categories. As
green roof retention performance is thought to be controlled by substrate depth, it was expected that the comparative
performance between ConEd and USPS would be similar no matter the rainfall depth. It is hypothesized that this
counter-intuitive observation is due to differences in event-based runoff behavior caused by two main factors: (1) the
configuration of non-vegetated regions on the different green roofs; and (2) flow paths through the roof substrate and
drainage layers, which differ by construction method.
During small events, runoff from the green roofs is dominated by precipitation on non-vegetated surfaces since the
green roof substrate typically remains unsaturated. Compared to W115, W118 and USPS, a significant portion of the
non-vegetated area on ConEd is located adjacent to the roofs downspout; as a result, runoff from these sections flows
directly to the instrumented roof drain without intercepting greened portions of the roof. In contrast, the distance of
non-vegetated areas on W115, W118 and USPS promotes depression storage of runoff before it reaches a roof drain,
resulting in higher retention rates for small storms compared to ConEd.
As rainfall increases, the green roof systems reach maximum saturation, such that substrate storage capacity and the
conductivity of preferential flow paths begin to play a greater role in determining runoff volume. In all systems,
preferential flow paths likely develop within discontinuities of the substrate, through areas where vegetation is absent,
and/or along geo-composite planes. For Wl 15, Wl 18 and USPS the hydraulic conductivity along these flow paths are
thought to be high, whereas on ConEd the hydraulic conductivity along flow paths is restricted by small outlets at the
base of the modular tray system. During large rainfall events, the restriction likely causes the trays to fill, resulting in
a higher degree of substrate saturation. This observation is supported by modeling efforts of green roofs by She and
Pang (2010) that indicated that rain infiltrates through the substrate prior to saturation. Considering ConEd also has
large comparative storage capacity to begin with, this may explain the observed improved water retention performance
of this roof during the larger events.
4-10
-------
Rooftop configuration (e.g. size, slope, location of vegetated and non-vegetated areas, shading) also appears to impact
overall retention performance between rooftops of the same construction type. Specifically, W115 consistently
outperformed Wl 18, even though both roofs were constructed at the same time using the same vegetated mat system.
W115, although a smaller roof, has higher percent vegetation (58%) than Wl 18 (53%). Additionally, the largest section
of non-vegetated area, the elevated housing for the stairwell, drains into the green roof itself, allowing this runoff to be
intercepted, thus improving overall rainfall retention ability. However, as the number of events in the 20+ mm category
was limited for Wl 15, further monitoring data might change relative runoff performance between these roofs.
The findings above support a recently proposed idea that green roofs may have shape and/or installation factors
associated with their runoff attenuation behavior (Miller 2012). It also highlights the importance of considering the non-
vegetated regions, common for many full-scale installations, during the interpretation of results from full-scale green
roof studies and the development of generalized models for green roof behavior.
Predictions of Green Roof Retention Performance
The runoff quantity data collected during this study show that green roof runoff depth (defined as runoff volume divided
by drainage area) can be reasonably predicted using a quadratic relationship with rainfall, which is described by a CRE
for each rooftop. The hydraulic behavior of a green roof during different storm events defines the curvature of the CRE.
Relationships between rainfall depth and green roof runoff depth can also be described using the well-known CN
method. However, the CN method is limited in its ability to capture relative changes in rainfall retention performance
between the green roofs that occur with storm size (see Figure 4-1). This is because the method assumes that a green
roof with a lower CN will always outperform a green roof with a higher CN in every single storm category. The residuals
of monitored and modeled performance for ConEd (Figure 4-5 (D)) show that the CN method generally under predicts
runoff in smaller events (in the 0-30 mm categories), while the residuals from the CRE have a more random distribution
among event sizes. As, expected, the reduction in R2 from the CRE to the CN method is the most dramatic for ConEd
(0.90 to 0.83). Nonetheless, while the CREs provide a better predictive method based on green roof runoff data, the CN
method might be useful in predicting performance of roofs where monitored runoff data are limited.
Unexplained variance in monitored performance compared to the CRE predictions for each roofs is believed to be the
result of environmental conditions on the rooftop, such as, temperature, antecedent moisture conditions, vegetation
health, rainfall intensity, etc. A better understanding of the influence of these environmental conditions on the quantity
of green roof runoff would require expanded hydrological performance data, gathered over multiple years. Although
this project has observations from over 100 storm events for each of the green roofs studied, there are still insufficient
data to define well the role of seasonality, for example. However, based on the strength of correlation between rainfall
depth and runoff depth for each of the green roofs (0.95, 0.97, 0.97, and 0.90 for W115, W118, USPS, and ConEd,
respectively), it is uncertain that analyses involving other variables would significantly improve the overall predictive
accuracy of the CRE method.
Multi-year Predictions Using Historic Rainfall Data
Application of the derived green roof CREs to 40 years of historic rainfall data obtained from the Central Park NOAA
station revealed performance biases caused by the distribution of storm sizes during the monitoring period. For example,
W115' s rainfall retention performance from reliable events during the monitoring period was better than that predicted
using the 40 years of historic precipitation because of differences in the fraction of small (0-10 mm) storms between
the monitored (74%) and predicted (66%) periods. In the same vain, only 1% of events from W115 and ConEd were
large storms (50+ mm), compared to 4% recorded from the historic period, and both roofs saw lower predicted retention
performance for the 40-year period versus the monitoring period. Wl 18 and USPS, however, received a higher percent
of storm volume in the 50+ mm, 48% and 56%, respectively, during the monitoring period in comparison to the historic
period (26%). As a result, the retention performance of these green roofs was higher for the multi-decadal predictions
generated with the roofs' CREs than was actually observed. Nonetheless, it does need to be recognized that surrounding
landscape features in an urban environment can influence rainfall patterns on rooftops. Thus, it remains unknown
whether the Central Park NOAA weather data reflect, on average, the rainfall on the individual roofs over a 40-year
period.
4-11
-------
4.6 Conclusions
Governing entities, owners, and other stakeholders will ultimately decide what, if any, green roof system is preferred
for managing urban WWF. The results presented here show that for NYC's climate a variety of factors, most
importantly, growing substrate depth, installation method, and configuration of non-vegetated areas, impact green roof
rainfall retention. Among the three construction types, the modular tray system had the highest total rainfall retention
during the monitoring period (61%) and therefore was most effective at reducing runoff volume. However, the NYC
Mayor's Office estimates that 62% of CSO events in NYC are caused by storms under 12 mm (Mayor's Office of Long-
Term Planning and Sustainability 2008). Therefore, if limiting the number of small storm CSO events, rather than
reducing stormwater volume, were the goal, the built up roof system on USPS might be preferred since it had the highest
attenuation of 0-1 Omm storm events and fully captured 46% of all storms. Finally, it is important to note that while
USPS and ConEd had better predicted rainfall attenuation performance than W118 and W115 in the historic period,
due to thicker growing substrate depths, the vegetated mat system is often the least costly per area of the three extensive
roof types to install, and might also be the most constructible on a wider range of existing NYC building stock due to
its significantly lower weight.
Continued collection of data and study of urban green roof conditions will help improve understanding of the factors
influencing green roof runoff behavior, including seasonal factors such as ET, plant performance and freeze thaw
(substrate temp), and allow a more robust analysis of green roof hydrological performance. Nonetheless, the CREs or
CNs presented in this Chapter here can provide a good first-order method of estimating the water retention performance
of the three common extensive green roof construction types in NYC's urban environment.
4-12
-------
Chapter 5 Water Quality Study
5.1 Introduction
Green roofs can have important benefits for diminishing the environmental consequences of stormwater runoff. These
benefits are highly pronounced in urban environments, particularly where CSOs are present. While many research
studies have focused on the role of green roofs in reducing stormwater runoff quantity (refer to Chapter 4), fewer studies
have focused on the impact of green roofs on stormwater runoff quality. The quality of green roof runoff is an equally
important factor when determining the role of green roofs in mitigating the harmful pollution impacts of WWF.
In current literature, a debate exists over the role of green roof systems as a source or a sink for stormwater pollutants.
While many studies have shown that green roofs have the capacity to sequester pollutants, thereby reducing pollutant
concentrations in stormwater (Berghage et al. 2009; Berndtsson et al. 2009; Carpenter and Kaluvakolanu 2011; Gregoire
and Clausen 2011), several other studies have shown that the green roof growing media serves as a pollutant source
(Moran et al. 2005; Berndtsson et al. 2006; Hathaway et al. 2008; Aitkenhead-Peterson et al. 2010; Dvorak and Voider
2010). However, within all of these studies, there is variability based on pollutant type.
The project work reported in this chapter aimed to provide additional information on the contribution of green roofs to
stormwater runoff quality, with a focus on urban green roof behavior. Previous studies have employed a wide variety
of methodologies when testing the impact of established green roofs on stormwater quality. Several studies have used
test plots/modules (Monterusso et al. 2005; VanWoert et al. 2005; Berghage et al. 2009; Aitkenhead-Peterson et al.
2010; Vijayaraghavan et al. 2012; Morgan et al. 2013) and small pots (Alsup et al. 2010) as proxies for established
green roof systems. Others have simulated storm events with artificial precipitation (e.g. distilled water) (Alsup et al.
2010; Vijayaraghavan et al. 2012). Of the research that actually involved monitoring established extensive green roof
systems, many studies only incorporated one or two green roofs to represent the variety of extensive green roof types
in existence (Moran et al. 2005; Teemusk and Mander 2007; Hathaway et al. 2008; Berndtsson et al. 2009; Bliss et al.
2009; Gregoire and Clausen 2011; Carpenter and Kaluvakolanu 2011). Therefore, a comprehensive examination of
green roof runoff quality from a full range of established green roofs, covering the extensive roof typologies commonly
implemented in an urban environment, fills an important gap.
During the study, the quality of runoff was quantified for five of the full-scale, extensive green roofs described in
Chapter 3, namely W115 (vegetated mat), W118 (vegetated mat), USPS (built up), Fldstn (built up) and ConEd
(modular tray), as well as five neighboring control (i.e., non-vegetated) roof areas. Chemical analyses of rain and rooftop
runoff were conducted for multiple parameters including conductivity, turbidity, color, and heavy metals. However, a
particular emphasis was placed on understanding the impacts of pH and nutrient concentrations (specifically nitrate
NOs", ammonium NH4+, total phosphorus P) in runoff. The information gathered on nutrient concentrations in roof
runoff was used to estimate the potential water quality impacts associated with installing extensive green roofs on 20%
of roof areas in NYC.
5.2 Methodology
Stormwater runoff samples were collected over a 16-month period (March 10, 2011 -August 2, 2012) at the five green
roof sites: W115, W118, USPS, Fldstn and ConEd. Three different types of samples were collected: i) green roof (GR)
runoff, ii) control roof (C) runoff, and iii) precipitation (Rain). Runoff was hand sampled usually during mid-rain events
and collected from a previously specified gutter using a disposable 50 mL centrifuge tube. Specified sampling locations
were kept constant throughout the 16-month sampling period and are illustrated in Figures 3-2 to 3-6. During sampling,
the 50 mL centrifuge tube was placed below the gutter lip and filled six times. The first 50 mL was used to rinse out
the previously-used collection bottle to ensure the integrity of the current sample. The subsequent five fills topped off
a reusable 250 mL polyethylene collection bottle. The 250 mL collection bottle was then brought back to a Columbia
University laboratory for water quality analysis. Control roof (i.e., non-vegetated roof) samples were collected from
either a neighboring building, or drains coming down from elevated, non-vegetated parts of the green roof. Control roof
sampling points are also shown on Figures 3-2 to 3-6. Precipitation samples were taken from a reusable collection
bucket kept on the roof of a Columbia University building that is not part of the green roof monitoring system. The
bucket was rinsed and cleaned with de-ionized (DI) water just prior to a rain event to ensure that the sample was not
5-1
-------
contaminated by debris in the bucket. It is assumed that rainfall collected at this location is representative of rainwater
quality throughout NYC. The constituents in precipitation are mainly governed by regional patterns of air pollutants,
which are largely consistent between our study sites (National Atmospheric Deposition Program 2012). Over 100
samples were collected and analyzed over the study period. Appendix B provides a summary of the sampling results.
All samples were analyzed for pH, conductivity, turbidity, apparent color and true color in Columbia University's
Heffner Laboratory, located in the S.W. Mudd Building, NY 10027. If the samples could not be analyzed immediately
after they were brought to the laboratory following collection, samples were refrigerated; although a sample was never
refrigerated for more than 24 hours prior to analysis. A Fisher Scientific accumet Excel XL50 was used to measure
sample pH and conductivity, a La Motte 2020we Turbidimeter was used to measure turbidity, and a Hach DR890
Colorimeter was used to measure both the sample apparent and true color. In addition, samples were sent to the Auburn
University Soil Testing Laboratory in Auburn, Alabama for nutrient and heavy metal analyses. To prepare these
samples, 40 mL of each parent sample was extracted from the 250 ml collection bottles and then filtered using a syringe
and a 0.22 urn filter. The filtered aliquot was then frozen and shipped to the Auburn Laboratory. Table 5-1 summarizes
the measurement parameters used for assessing precipitation, green roof and control roof runoff quality during the study.
Table 5-1: Water Quality Measurement Parameters
Standard Parameters
Nutrients (mg/L)
Heavy Metals (mg/L)
Macronutrients
Micronutrients
pH
Conductivity ((iS/cm)
Turbidity (NTU)
Apparent Color (PtCo)
True Color (PtCo)
Nitrate (NO3-)
Ammonium (NH4+)
Calcium (Ca)
Potassium (K)
Magnesium (Mg)
Phosphorus(P)
Copper (Cu)
Iron (Fe)
Manganese (Mn)
Zinc (Zn)
Sodium (Na)
Barium (Ba)
Cadmium (Cd)
Chromium (Cr)
Nickel (Ni)
Lead (Pb)
Aluminum (Al)
Arsenic (As)
Boron (B)
5-2
-------
5.3 Results
The full assay and statistical results of the study are provided in Appendix B and C, respectively. Analyses with
significant findings are displayed as box-plots in Figures 5-1 and 5-2. Letters on Figures 5-1 and 5-2 indicate
significantly different groups. A one-way ANOVA and Tukey HSD (honestly significant difference) were used to test
for significance (p value < 0.05) between the Rain, GR and C sites for all of the water quality parameters. Statistical
tests were run using the R statistical package (http://www.r-project.org). Results of these tests are provided in Appendix
C. A comparative summary of the mean measurements for GR, C and Rain is provided in Table 5-2.
Table 5-2 Summary of Mean Water Quality Results with Standard of Deviation
Water Quality Green Roof Control Roof Precipitation
Measurement Mean Standard of Mean Standard of Mean Standard of
deviation deviation deviation
pH 7.28 ±0.51 6.27 ±0.69 4.82 ±0.39
Conductivity (uS/cm) 127.67 ±48.89 57.11 ±57.63 32 ±20.71
Turbid ity(NTU) 2.47 ±2.74 1.47 ±1.48 0.62 ±0.39
Color (PtCo) 162.53 ±90.24 28.45 ±32.42 5.32 ±9.79
Nitrate (mg/L) 0.27 ±0.59 0.87 ±1.31 0.6 ±0.53
Ammonium (mg/L) 0.86 ±1.86 1.47 ±2.55 1.19 ±1.85
Total phosphorous (mg/L) 0.47 ± 0.47 0.25 ± 0.38 0.21 ± 0.41
Calcium (mg/L) 13.59 ±6.8 3.93 ±5.23 0.74 ±0.50
Potassium (mg/L) 2.22 ±2.86 0.78 ±1.98 0.1 ±0.2
Sodium (mg/L) 3.58 ±3.47 1.8 ±3.01 0.98 ±0.88
Magnesium (mg/L) 2.92 ±1.03 1.31 ±2.30 0.2 ±0.24
Boron (mg/L) 0.58 ± 1.19 0.03 ± 0.1 0.0 ± 0.0
5.4 Discussion of Results
In what follows, the discussion of results from the water quality study is presented under separate headings for the three
categories of testing: standard parameters, nutrients, and heavy metals (see Table 5-1). In addition, an estimation of the
change in the annual mass nutrient loading in stormwater runoff that might be anticipated if green roofs were widely
adopted in NYC is provided.
Standard Parameters
An increase in pH from Rain to C runoff to GR runoff is an often reported benefit of green roof establishment (Teemusk
and Mander 2007; Clark et al. 2008; Berghage et al. 2009; Berndtsson et al. 2009; Bliss et al. 2009). This study provided
additional support for this reported benefit. At each of the five individual green roof sites monitored for runoff quality,
the pH of GR runoff was consistently higher than both the pH of C runoff and Rain, with measured average pH values
of 7.28, 6.27 and 4.82, for the GR, C and Rain during the study period, respectively (Appendix B). EPA designates 6.5-
8.5 as the standard pH range for safe drinking water (EPA, 2009). The spread of the pH of runoff samples taken from
all five green roofs consistently remained within the EPA standard range, while the spread for pH of control roof runoff
and precipitation samples fell out of the standard range (Figure 5-1). In the northeast region of the U.S where acid rain
is a problem, green roofs therefore might serve as a tool for protecting aquatic ecosystems in nearby water bodies from
the consequences of increased acidity. Note, average pH values were calculated by first transforming the pH
measurements into hydrogen ion concentrations: [H+] = 10"pH. Once averaged, the [H+] value was converted back to a
pH value: pH =-log [H+].
5-3
-------
Green Roof
Control Roof
Q.
B A A A A A
00
1
=t= 1 1 1 1 I—
Rain 115 118USPS Fid ConEd
B A A A A A
h-
Rain 115 118USPS Fid ConEd
o CM
co
•3- o
O 8-
A B.C B C B,C B
OT
•o
C
o
o
o .
tn •
CM
O
CM
O
O
A A A B A A
JBa PC
Rain 115 118 USPS Fid ConEd
Rain 115 118 USPS Fid ConEd
o
03
«-
T
CN-
0
A C A
! *
T
:
B.C
A.C A
i -T-
] — \-n-l <=l
— : — i
-e T
CM
o
A A ABA A
,BQ -
Rain 115 118 USPS Fid ConEd
Rain 115 118 USPS Fid ConEd
o o
1°
11
O
c 8
P CM
m
i§
0-
A A.B B B 8 B
Color (PtCo
O
O
tn
o
s
I
0
A A.B A.B B A.B A
Rain 115 118 USPS Fid ConEd Rain 115 118 USPS Fid ConEd
Figure 5-1: Distribution of pH, conductivity, turbidity, and apparent color for all roof types and precipitation.
5-4
-------
Green Roof
Control Roof
<* -
0) CO -
s
= CM-
o .
2"
"3» {O -
E
o
a. 10
3 <="
o.
o
o
co-
in
CM"
"a °-
3" CM
E £-
" 0
03 °-
m-
0
A A A A A A
0
o
7
0 4
Hi 1
^^^^
Rain 115 118USPS Fid ConEd
A A A A A A
0
0
* *
*
e
ES ^b s
Rain 115 118USPS Fid ConEd
-r
A A A "A A A
— -j-
Rain 115 118USPS Fid ConEd
A B,C B D C,D A.B
r
—*— ',
^B ^
(D -
in -
_i
~s> •* •
Q> CO ~
I
I ^-
^ "
o •
o .
5"
~$l co
E <0-
^
T5
i ^j
< CM-
sr> .
^i
0> o
in
3
O
Q. 10
S o-
Q.
O
O
CO "
ir>
CM "
_1
"fti O -
P (N
t
C **> -
0
ro ?-
tr> -
0 -
A A A A A A
!
j
0
rn ° ^^ °
— =
Rain 115 118USPS Fid ConEd
A A A A A A
4
0
o
4
oEi a^ •
Rain 115 118USPS Fid ConEd
A A A A A_ A
—
"~
:
Rain 115 118USPS Fid ConEd
A A A B A A
_,_
I
I
i^^a c^3^-
Rain 115 118USPS Fid ConEd Rain 115 118USPS Fid ConEd
Figure 5-2 Distribution of nutrient and heavy metal concentration (mg/L) from all roof types
5-5
-------
Green Roof
Control Roof
CM ,
O -
~O> CO -
I ••
m
1
CN '
O -
«O -
? ? -
£
3 «-
O -
o-
1 *
g •»-
-
c
s ™ -
o
CD
-•
*"*
A A A A B A
•
^ei^i
Rain 115 118USPS Fid ConEd
0
A A.B B,C A.B A.B A.B
4
i • 1
: 1 1 1 — ^-1 | 1 1 1
t^S _L J_ -
Rain 115 118USPS Fid ConEd
A B B B B B
•* ' 1 '
e
Rain 115 118USPS Fid ConEd
A A B A A A
—
CM .
O
CB CO -
E
-------
Few studies report on the overall electrical conductivity levels of green roof runoff, however, the general consensus is
that there are higher ion concentrations in green roof runoff than in precipitation or control roof runoff (Teemusk and
Mander 2007; Berghage et al. 2009). This trend is consistent with this study, which revealed a statistically significant
difference in electrical conductivity between GR runoff samples and both C runoff and Rain samples, with the average
GR conductivity being higher than the average C conductivity, which in turn was higher than the average Rain
conductivity (Table 5-2).
Only a small number of studies have assessed turbidity in green roof runoff, and results from these studies have been
inconclusive. For this study it was found that, on average, green roof runoff had higher turbidity than control roof runoff
or precipitation. However, the only statistically significant difference in results lay between Rain and GR. (Appendix
C). The GR turbidity values measured during this study are similar to those reported by Berghage et al. (2009) (range:
0.8-5.6 NTU) from their evaluation of runoff from various vegetated test plots. In contrast, Bliss et al. (2009) report
their control roof runoff to be more turbid than both green roof runoff and precipitation. These authors hypothesize that
this can be explained by accruement of particulates on the control roof during the dry period prior to a rain event.
Color in green roof runoff is a function of the organic materials incorporated in the green roof growing media to enhance
plant growth and establishment (Berghage et al. 2009). Prior to making measurements, color was the only visible
difference observed between green roof samples (yellow tint), and control roof/precipitation samples (no visible tint).
The results of the color measurements showed a significant difference between GR-Rain and GR-C for apparent color,
with GR color being higher (Appendix C). This finding is consistent with that of Berghage et al. (2009).
Nutrients
Conflicting results regarding NOs in GR runoff have pervaded the literature. Some studies report no significant
difference between NOs concentrations in GR and C runoff (Clark et al. 2008; Carpenter and Kaluvakolanu 2011)
while others report GR to be a NOs sink (Berndtsson et al. 2009; Berghage et al. 2009) or a source (Teemusk and
Mander 2007; Carpenter and Kaluvakolanu 2011). Intriguingly, the concentrations of NOs" in GR runoff from
Monterusso et al. (2005) and Morgan et al. (2013) are much greater than those from other studies, including this one.
Variation in the literature with regards to runoff of nitrogen and phosphorus green roofs may be due to variations in the
amount of organic matter in green roof substrates. Studies have shown that nutrient runoff increases with increasing
compost content of substrates (Hathaway et al. 2008) with stronger evidence for the link between compost enrichment
of substrates and an increase in phosphorus runoff (Monterusso et al. 2005; Moran et al. 2005; Berndtsson et al. 2006;
Teemusk and Mander 2007; Bliss et al. 2009). Of the nutrients examined during this study that are typically associated
with eutrophication (NOs MV, P), there was only a significant difference between GR and C NOs levels, with GR
NOs concentrations being lower (Appendix C).
Consensus regarding P and MV concentrations in GR runoff is also lacking in the literature. In some cases, GR runoff
is reported as a statistically significant source of P (Clark et al. 2008; Berghage et al. 2009; Bliss et al. 2009; Fassman-
Beck et al. 2013), while in other cases, GR are reported to be sinks of P (Teemusk and Mander 2007; Gregoire and
Clausen 201 1). In one study, an extensive green roof in Sweden was cited as a source of P while an intensive green roof
in Japan was not (Berndtsson et al. 2009). Both green roofs discussed in the Berndtsson et al. (2009) study, however,
were reported to be sinks for NH4+. The data gathered during this study demonstrated that, on average, there was less
NH^+ in GR runoff than C runoff. The decrease in NH^+ did not correlate with an increase in NOs", which would have
indicated nitrification, the two-step process of converting NH4+to NOs (Berndtsson et al. 2009). Overall, this study
found no significant difference between the Rain, GR and C for both NH4+ and P (Appendix C).
For other macronutrients such as Ca, K and Mg, there was a significant difference between GR-Rain and between GR-
C, with GR concentration values being higher (Appendix C). The observed Ca, K and Mg concentration ranges are
similar to those observed in the studies conducted by Berghage et al. (2009) and Berndtsson et al. (2009). Overall, the
Ca, K and Mg measurements made during this study demonstrate that GR's role in leaching macronutrients into rooftop
runoff.
5-7
-------
The micronutrients Cu, Fe and Zn and Mn were either detected in very low amounts or below the detection limit (< 0.1
mg/L) for all runoff and precipitation samples, so these data are not provided in this report. For the micronutrients Na
there was a significant difference between GR-Rain, with Na GR concentrations being higher, possibly indicating Na
leaching from the green roof substrate. Measured Na GR concentration levels were similar to those reported by
Berghage et al. (2009); while the measured Na concentration of C samples were, on average, slightly higher than those
reported by these authors. For the other micronutrients, even though the data for this study cannot be directly compared
to other studies due to this study's higher detection limit, the measurements of Cu (Berndtsson et al. 2009; Gregoire
and Clausen 2011; Fassman-Beck et al. 2013), Fe (Berghage et al. 2009; Berndtsson et al. 2009), Mn (Berghage et al.
2009; Berndtsson et al. 2009) and Zn (Berghage et al. 2009; Berndtsson et al. 2009; Gregoire and Clausen 2011;
Fassman-Beck et al. 2013) reported by other researchers also indicate that low quantities of micronutrients would be
leached out of a green roof.
Heavy Metals
All heavy metal concentrations were either very low or below detection limits (< 0.1 mg/L) for all runoff and
precipitation samples, so these data are not provided in this report. The only exception to this was Boron (B) which was
detected in the runoff of the W118 GR (Figure 5-2). The W118 GR is a housing building for Columbia University
graduate students and various types of insecticide is applied to the building on a routine basis (Trejo 2013). One of the
products, Niban-FG, contains Boric Acid. It is thus hypothesized that the Boron measured in the water quality here is a
result of this insecticide.
Although the reported detection limits for other studies where heavy metals were measured in green roof runoff (Clark
et al. 2008; Berndtsson et al. 2009; Bliss et al. 2009) are much lower than this study, the other studies also conclude
that heavy metals are rarely present in green roof runoff. The only exception is Gregoire and Clausen (2011) who
observed slightly elevated chromium and lead concentrations in green roof runoff, which the authors attributed to nearby
construction and excavation.
Mass Nutrient Loading
Interpretation of the nutrient data obtained from this study specifically focused on understanding the potential impacts
of widespread green roof implementation on the eutrophication of NYC waterways. As a result, analyses were focused
on estimating how green roofs might change the annual mass loading of NOs", NH4+, and total P in NYC's stormwater
runoff.
The mass loading per unit area of NOs", NH4+, and P leaving the GR and C per rain event was calculated for three of the
monitored green roofs (and associated control roofs), namely: Wl 18 (vegetated mat), Con Ed (modular tray), and USPS
(built up). These green roofs represent the three commonly installed extensive green roof types and the CREs for these
roofs were based on the most runoff quantity data (see Table 4-1). The average annual runoff volume from each green
roof (L) was estimated based on historic rainfall data, the CRE for each roof (see Figure 4-4), and the area (m2) of the
roof. In the case of the control roofs, a blanket 10% retention of all storms was assumed. The annual runoff volume (L)
for each green roof or control roof was then multiplied by the average nutrient concentration (mg/L) in the roof runoff
and divided by the roof area (m2) in order to estimate the annual mass loading (mg/m2) of each nutrient from each roof
per unit rooftop area. Results are reported in Table 5-3, where bolding and shading indicate which values are higher.
Although Table 5-3 reports that concentrations of nutrients in green roof runoff were sometimes higher or the same as
nutrient concentrations in control roof runoff, the total mass (mg) of nutrients leaving each GR per unit area during a
storm event was less than that leaving each C per unit area because of GR capacity to hold and store precipitation (refer
to Chapter 4). As a result, the annual mass loading (AML) of nutrients from GR is lower than from that of the nearby
C.
Table 5-3: Average Concentration (AC) and Annual Mass Loading (AML) of NhU"1", NOs", and P.
Nutrient
W118
C GR
USPS
C GR
ConEd
C GR
5-8
-------
1
o
E
^i
E,
NH4+
N03-
Total P
NH4+
N03-
Total P
0.12 0.03
2.74 1.08
<0.26 <0.49
127.80 19.09
2992.77 612.33
<285.73 <276.44
0.27 0.34
1.76 1.16
<0.25 <0.47
290.02 158.49
1921.72 541.26
<276.15 <217.62
0.38 0.08
0.50 0.59
<0.47 <0.42
416.23 31.66
548.97 245.57
<513.43 <175.89
About 8% of NYC land area is estimated to qualify for an extensive green roof retrofit (Culligan 2011), which amounts
to a total area of approximately 6.4xl07 m2. To evaluate how the difference in AML between C and GR might impact
annual nutrient loading in NYC stormwater if all suitable rooftops were retrofit with an extensive vegetated mat green
roof, the lightest of the three systems, the AML (mg/m2) of Wl 18 GR and Wl 18 C was multiplied by this area. The
difference between the two mass loadings (kg) was then assumed to be the decrease in annual nutrient loading (kg) that
would be realized by a widespread green roof retro-fitting program in NYC (Table 5-4). According to this analysis,
approximately 600, 7,000 and 150,000 kg of P, NH4+ and NOs per year, respectively, would be removed from NYC
stormwater runoff if green roofs were widely adopted throughout the city.
Table 5-4: Projected Reduction in Nutrient Loading per Year from Retrofitting Available NYC Rooftops
Nutrient
NH4+
NO3-
Total P
Project Loading Discharge (kg)
Existing rooftop
8,182
191,610
18,294
Green roof retrofit
1,222
39,204
17,699
Reduction
6,960
152,406
595
Overall, the reductions in annual mass loadings reported in Table 5-4 appear minimal when compared to estimates of
annual nitrogen and phosphorus loads to the saline Hudson estuary. For example, Howarth et al. (2011) estimate that
24 x 103 metric tons N per year and 3.7 x 103 metric tons P per year were input annually into the Hudson River estuary.
Based on this fact, retrofitting all available rooftop in NYC with an extensive vegetated matt system would provide less
than a 1% decrease in annual nutrient loadings to local waterways. Nonetheless, although minimal, this study still
provides evidence that green roof establishment does have an overall benefit on stormwater quality by not exacerbating
nutrient loading in stormwater, and thereby provides support for the expansion of green LID in urban areas negatively
impacted by WWF.
5.5 Conclusions
The results of the water quality study on the five NYC green roofs show that the average pH of GR runoff was 7.28,
while the average pH of C runoff and Rain was 6.27 and 4.82, respectively. Statistical analyses show that the difference
between the GR runoff and Rain are significant. Thus, the study results confirm, as reported by others, that green roofs
neutralize acid rain. On average, lower NOs and NH4+ concentrations were observed in GR runoff than both C runoff
and Rain, although the only statistically significant difference lay between the NOs concentrations in GR and C runoff.
In general, total P concentrations were higher in GR runoff than either C runoff or Rain, but none of the differences in
concentrations between these three categories of water samples were significant. Finally, all micronutrients and heavy
5-9
-------
metals were either detected at very low concentrations or not at all (concentrations were below the detection limit). The
two exceptions relate to Boron concentrations on the Wl 18 green roof as well as Na and Ca across all five green roofs.
Although the concentration of nutrients in green roof runoff was sometimes observed to be higher than that in control
roof runoff, it was estimated that the annual mass (mg) of nutrients leaving a green roof is less than that leaving a control
roof because green roofs discharge less water volume per annum. Based on the work presented in Chapter 4 and the
work presented in this chapter, it is projected that universal installation of green roofs in NYC on all suitable rooftops
could save about 600 kg of Total P, 7,000 kg of NH4+ and 150,000 kg of NOs" from discharging into the sewer system
and/or local water bodies. Although these amounts are not large in comparison to the annual nutrient discharges from
waste water treatment plants, they still provide evidence that green roof establishment does have an overall benefit on
urban stormwater quality. However, improved management through lower fertilization of green roofs may also
contribute to reduced nutrient discharges (Oberndorfer et al 2007).
5-10
-------
Chapter 6 Soil Moisture Water
Balance Study
6.1 Introduction
Rainfall and antecedent SM are two key factors influencing the generation of green roof runoff (Crow et al. 2011). SM
affects the partitioning of precipitation into infiltration and runoff, thereby regulating surface and subsurface water flow
(Dorigo et al. 2011; Joshi etal. 2011). SM in the root zone is also crucial for the growth and development of green roof
vegetation, and plays a role in the apportioning of the surface energy budget (Joshi et al. 2011) which influences green
roof ET. Similar to other natural systems, SM thus plays a central role in the hydrologic behavior of green roofs,
indicating that direct studies of SM for assessing green roof water balance can be very valuable. Given the high
instrumentation costs, labor and technical expertise required for current green roof hydrologic monitoring schemes,
including those that were part of this study, the potential for SM to act as a proxy for green roof runoff and ET
measurements presents an important avenue for research.
Using a "backwards hydrology" method, Kirchner (2009) showed that if a water catchment could be represented by a
single storage element, where storage is assumed to be a function of water discharge alone, precipitation, ET, and
discharge could be linked without the need to explicitly account for changes in water storage. After constructing a first-
order nonlinear differential equation, the storage term was inferred from the resulting changes in water discharge,
allowing for streamflow hydrographs to be predicted from precipitation and ET time series alone. A similar method
was also adopted by Brocca et al. (2013) in order to estimate rainfall from SM observations using an "inverted soil-
water balance equation." Brocca et al. (2013) showed that the method was successfully able to reproduce daily rainfall
data based on in situ SM observations at three different sites in Italy, Spain and France.
A water balance approach was used by Jarrett et al. (2006) to estimate green roof stormwater retention based on daily
rainfall, ET, and the soil/substrate maximum water holding capacity in their Annual Green Roof Response (AGRR)
model. This model requires estimating ET rates, which are dependent on location, climate, vegetation, water storage
capacity, as well as SM levels. While conceptually the approach used by Jarrett et al. (2006) appears to effectively
predict green roof runoff when enough information is available, no formal validation was carried out via a comparison
of the predicted green roof runoff to observed runoff values. Sherrard and Jacobs (2012) present a vegetated roof water
balance model (VR-WBM) that uses daily time steps, requires precipitation and dew as inputs and outputs green roof
storage, runoff and ET. The model requires both vegetation parameters and soil/substrate characteristic parameters that
need to be obtained from field observations. Good agreement between the model and data from a lysimeter experiment
where shown following model calibration with several weeks of field data.
This Chapter reports a method for green roof runoff and ET estimation that was derived during the study and is termed
the Soil Water Apportioning Method (SWAM). Similar to the Kirchner (2009) and Brocca et al. (2013) methods, a
water balance model was constructed, analytically linking precipitation to SM, whereby runoff and ET were inferred
from the resulting changes in SM over time. Thus, SWAM relies solely on the monitoring of local precipitation and
green roof SM in order to predict green roof hydrological behavior and unlike the VR-WBM model does not require
field calibration. In situ rainfall, runoff and SM observations from the W118 and W115 vegetated mat systems (see
Chapter 3) were used to test the reliability of the proposed approach. Two different low-cost SM probes were compared
for accuracy in estimating green roof runoff using SWAM.
6.2 Methodology
Rainfall data needed for this portion of the study was collected with a tipping bucket rain gauge while runoff was
measured using a custom made weir device. For a detailed description of precipitation and runoff monitoring, see
Chapters 3 and 4. Volumetric SM was collected with a CS615 water content reflectometer (Campbell Scientific, Logan,
Utah, USA) on the Wl 18 roof, and with a ECH2O EC-5 soil moisture probe (Decagon Devices, Pullman, Washington,
6-1
-------
USA) on the W115 roof. The EC-5 probe was purchased from Onset Hobo® Dataloggers in order to be compatible
with other instruments used in this research. Substrate temperature was recorded using a 12-Bit Temp Smart Sensor (S-
TMB-M002). The study period for this work was from September 1, 201 1 to August 1, 2012.
Recent advances in monitoring techniques, have allowed for less destructive and more accurate means of quantifying
volumetric water content (Czarnomski et al. 2005). The most widely utilized sensors use either a time domain
reflectometry (TDR) method or the capacitance technique. The TDR method measures soil or SM content through a
relationship with the velocity of an electromagnetic wave that is passed along the waveguides, determined by measuring
the time of travel (Walker et al. 2004). Capacitance sensors determine soil water content by measuring the frequency
change induced by the changing permittivity of the soil permeated by the fringing fields of the capacitor sensor
(Baumhardt et al. 2000). The accuracy and precision of these instruments vary and their calibration is dependent on soil
type, electrical conductivity and temperature (Czarnomski et al. 2005).
The CS615 water content reflectometer consists of two 30 cm long stainless steel waveguides connected to a printed
circuit board: it measures soil moisture content using the TDR method (Campbell Scientific Inc. 2012). The CS615 is
specified to have an accuracy of ± 3.0% v/v when applied to typical mineral soils using the manufacturer's standard
calibration and an operating environment from -10 to 70 °C (Campbell Scientific Inc. 2012). Soils with different
dialectic properties show an error that appears as a constant offset (Campbell Scientific Inc. 2012). A single CS615
sensor was installed horizontally into the upper portion of substrate layer of the Wl 18 green roof within the monitored
drainage area.
The EC-5 soil moisture sensor calculates the apparent soil dielectric constant of a soil by measuring the charge time of
a capacitor in the soil (Czarnomski et al. 2005). The time required to charge the capacitor is related to the output voltage
of the instrument. An empirical equation is used to describe the relationship between the output voltage and SM. The
EC-5 is specified to have an accuracy of ± 2.0% v/v when applied to typical mineral soils using the manufacturer's
standard calibration and an operating environment from -40 to 60 °C (Decagon Devices Inc. 2012). However, according
to Onset Hobo® Dataloggers, the sensor is able to accurately measure SM only from 0 to 50 °C. A single EC-5 probe
was inserted at an angle, penetrating through the entire depth of the Wl 15 green roof substrate.
Soil or substrate electrical conductivity and temperature can influence the accuracy of these SM measurement
instruments (Baumhardt et al. 2000; Czarnomski et al. 2005; Campbell Scientific Inc. 2012). The factory calibration for
CS615 will accurately predict SM if the electrical conductivity is <3 dS-nv1 (Campbell Scientific Inc. 2012), while the
factory calibration is valid for electrical conductivities < 8 dS-m"1 for the EC-5 probe (Decagon Devices Inc. 2012).
However, the application of the sensors to conductive media, such as saline soils, certain clay and organic soils is
hindered due to significant attenuation effects of the desired signal (Hook et al. 2004). Moreover, SM readings may be
impacted by temperature variations due to effects on the dielectric permittivity of water, through soil-water interactions,
as well as through direct effects on the sensor circuitry (Bogena et al. 2007). Given these instrument sensitivities, it is
often necessary for individual users to calibrate the instruments for their specific measurement conditions if precise
estimates of SM are required (Czarnomski et al. 2005).
In this study, SM readings were normalized to reflect a saturation value ranging from 0 to 1, corresponding to the
minimum and maximum field SM conditions, respectively. Similar two point (dry-wet) TDR sensor field calibration
methods have been suggested and applied in past studies with success (Robinson et al. 2005; Sakaki and Rajaram 2006).
Due to the availability of several years of SM readings from each roof, the normalization calibration, i.e., obtaining the
dry and wet volumetric soil moisture values for this study, was carried out using field data.
SWAM incorporates a normalized substrate water balance approach. The water balance for a layer of substrate with
depth Z [L] can be described by the following expression:
[6-1]
at
6-2
-------
where S(t) [-] is the relative saturation of the substrate, t [T] is the time, n is the substrate effective porosity [unitless],
P(t), Q(t), ET(t), and L(t) [L/T] are the precipitation, runoff, ET, and losses, respectively.
Given the relative thinness of green roof substrates, particularly extensive green roofs, as well as their well-draining
nature (Getter and Rowe 2006), the dominant process governing water flow was considered to be vertical infiltration
through the growing media. Based on this, and an assumption that there was no water penetration through the roofs
waterproof membrane, the losses term in equation [6-1] was taken to be negligible (e.g. losses due to canopy interception
which is often neglected in many other green roof water balance approaches). Because green roofs promote rapid
drainage through the growing media (Passman and Simcock 2012), it was further assumed that the runoff from a green
roof was simply the excess of precipitation depth over the substrate's storage capacity; meaning that the possibility of
water ponding on the green roof was ignored. With these two assumptions, measurements of SM and precipitation can
effectively capture all of the water going into and out of a green roof system over a specified period of time (aggregate
period) with the following modified water balance equation:
[6-2]
where Si denotes the relative saturation of the substrate at the start of the aggregate period, 82 denotes the relative
saturation of the substrate at the end of the aggregate period, At is the time-step of the aggregate period, and the subscript
At denotes the parameter value associated with the aggregate period.
If precipitation occurs during the aggregate period, green roof runoff is estimated from:
Sl) [6-3]
where nZ(l-Si) is the water storage capacity of the green roof at the start of the aggregate period. If the precipitation
during the aggregate time period is less than nZ(l-Si), runoff is set to zero for the aggregate period.
If no precipitation occurs during the aggregate period, then no runoff is assumed and ET during the aggregate period is
estimated from:
Sl} [6-4]
6-3
-------
If precipitation occurs during the aggregate period, but no runoff is generated, ET during the aggregate period is
estimated from:
-$) [6-5]
If precipitation and runoff occur during the aggregate period, then ET during the aggregate period is estimated from:
ET&t=nZ(l-S2) [6-6]
where it is assumed that the reduction in substrate field saturation below 100% at the end of the aggregate period is the
result of ET during the aggregate period. Sometimes this assumption led to negative values of ET, primarily because
green roof runoff can be generated before the green roof medium is fully saturated. When this occurred, ET was set to
zero for the aggregate period.
Three different methods were used in assessing the performance of SWAM based on recommendations by Moriasi and
Arnold (2007): Nash-Sutcliffe efficiency (NSE) index, percent bias (PBIAS), and the root mean square error (RMSE).
The NSE is a normalized statistic that measures the relative magnitude of the residual variance in predicted values of
runoff versus observed data variance (Nash and Sutcliffe 1970) using equation [6-7]:
NSE = 1 -
n (nobs nPre
-
i=1(Qibs:-Qmean)
[6-7]
Where Qiobs is the rth value of the observed runoff data, Qipre is the rth runoff value arrived at using SWAM, Qmean is the
mean of the observed runoff data, and n is the total number of observations. NSE ranges from -co to 1.0, with 1.0 being
the optimal value. Values between 0 and 1 are generally viewed as acceptable model performance, while values below
0 indicate unacceptable performance (Moriasi and Arnold 2007). PBIAS measures the average tendency of simulated
data to be larger or smaller than observed data and RMSE is a measure of the average squared deviation of simulated
values from observed values. The maximization of the NSE and PBIAS statistics were selected as objective functions
for arriving at the optimal time aggregate. All three statistics were used for general model performance assessments.
6.3 Results
A recent study by Passman and Simcock (2012) indicated that the soil-water relationship for extensive green roofs
varies greatly between laboratory and field measurements. For this reason, the maximum water holding capacity for
each roof was arrived at statistically using SWAM with a 24-hour time aggregate (At = 24 hours). The optimized
effective porosities were found to be 0.3 and 0.5 for the Wl 18 and Wl 15 roofs, respectively, from which the maximum
water holding capacities were calculated. Several time aggregates were considered before arriving at the optimal time
of 24-hours. Results of the time aggregate optimization are shown in Table 6-1 and Figure 6-1.
Table 6-1: Time Aggregate Optimization for W118 and W115 Green Roofs.
Time
Aggregate
(hr)
1
6
12
24
36
W1 18 Green Roof
RMSE
(mm)
1.77
3.91
5.34
6.22
7.96
NSE
0.27
0.79
0.85
0.88
0.88
Negative
PBIAS (%)
-16
-11
-13
-16
-18
Positive
PBIAS (%)
65
45
35
23
19
W1 15 Green Roof
RMSE
(mm)
3.25
5.21
8.4
7.76
13.33
NSE
-1.58
0.49
0.46
0.72
0.43
Negative
PBIAS (%)
-17
-23
-31
-35
-37
Positive
PBIAS (%)
64
36
32
19
30
6-4
-------
(a)
(b)
090 -
0 40 -
0 10 -
0.20
n nn
*•.
^ "
y^-
/ ~'--^r
/ ^ ' —
/
/
— * ••"•——..._.. — _.
60%
0%
• -10%
-1 10 -
-1 40
_? nn
-"***""' "^ —
^ r
>.
/ •-- .-•-
/
/
80%
60%
40%
0%
-20%
-40%
-Kn%
12 24
TimeAggreggate(hr)
36
12 24
TimeAggreggate(hr)
-MSE + PBIAS PBIAS
Figure 6-1: NSE and PBIAS for the (a)W118, and (b)W115 roofs at different time aggregates.
Figures 6-2 and 6-3 show the resulting hydrographs for the W118 and W115 roofs over the course of a year using
SWAM at daily time aggregates, respectively. Figure 6-4 shows a restricted time period, i.e. from March to May 2012
only, to provide a closer view for the purpose of comparison. Figure 6-5 compares SWAM ET estimates versus ET
measurements collected in 2009 using a chamber technique (Dome ET) (Marasco, 2014) on Wl 18.
E 100
aep
Oct
Nov
Dec
Jan
Feb
Mar
100
50
i
Jit Li*
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
5
III Q
0
(J
TS inn
I
i
i ;
i i
i !
:*,_ __i
- v yy . -v ' '^i -» -t f t
\ i
; i
^_ _. ^^^^^^^^
i i
>- Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul
Figure 6-2 Precipitation, soil moisture, observed and simulated daily runoff and error for green roof W118.
Aug
6-5
-------
en
E
:t 25
o
c
rz
* \ i i i i
j
: : _• :
4 L j i
n LuLLuLJ ii 1 A-
i i i
•i \
J :
Ls. i * 11 «i . JU j. A
Qpred
; _
y. J „ ^
aep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
E 50
LU o
-50
| 1 1 ! I |
^ i:»«^: «A: . i.
^^. . -V . "^
i 1 i i 1 i
1 | ! I
'.f, f.'- • ft r"
: > • v '^
i i i I
D-
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Figure 6-3: Precipitation, soil moisture, observed and simulated daily runoff and error for green roof W115.
Aug
6-6
-------
Green RoofW118
Mar
May
1 £
em
0,5 3i
1
«
o e
i
Runoff (mm)
S 8
— UQD1
Qpred
I
^ ,/V ^ ^ ^ AA
Apr
May
£ SO
,§
5
UJ 0
?-50
Mar
Green Roof W115
\
Apr
A^
May
May
£ SO
E
I
UJ 0
1-50
Mat
Apr
May
Figure 6-4: Comparison between green roof W118 and W115 from March to May 2012 only.
6-7
-------
?70
^60
§50
40
Df30
c
£20
4-1
0.10
,>, o
30
25
U
20
15
"
5 <
0
Jul Aug Sep Oct Nov
AvgAirTemp •DomeET BSWAMET
Figure 6-5: Comparison of measured versus modeled evapotranspiration (Dome ET to SWAM ET) on W118.
6.4 Discussion of Results
Increasing the time aggregate in the model improves the NSE and reduces the positive PBIAS for both roof and SM
probe types, yet has a minimal impact on the model's negative PBIAS (see Figure 6-1); positive PBIAS values indicate
model underestimation bias, and negative values indicate model overestimation bias. Thus, the runoff overestimation
bias appears to be an issue regardless of changes in time aggregate. Time aggregates under 6 hours produced
unacceptable results.
The analytical method proposed by SWAM is found to be capable of reproducing the observed daily runoff data with
NSE values of 0.88 and 0.72 for the W118 and W115 green roofs, respectively. There are a number of factors that can
be contributing to the performance differences, including the relative temperature sensitivities of each probe, seasonal
differences between the performance of each roof (possibly affecting the substrate maximum water holding capacity),
and instrument biases when collecting the observed data. However, it is believed that one of the greatest contributing
factors is the relative location of the sensors within the respective drainage areas. Specifically, the Wl 15 roof is a sloped
roof that might support ponding at lower elevations (where the probe is currently situated) during larger rainfall events.
As noted above, ponding is not considered in the water balance equations of the SWAM model. Another factor is a
single measurement point for SM, which relies on spatially uniform green roof behavior whereas spatial heterogeneity
could play a role in green roof runoff characteristics.
Figure 6-4 shows the performance of SWAM in predicting ET on monthly time aggregates. SWAM performs well in
the months of July, August and September, however, declines in performance in the months of October and November.
Average monthly air temperatures indicate that this drop in performance could be due to the sensitivity of the soil
moisture probe to sharp changes in air temperature, as well as the sensor sensitivity to non-ambient temperatures
(assumed to be temperatures above or below 25 °C). These preliminary results indicate that there is potential for this
methodology to be used in estimating green roof ET, however, more extensive analysis is required in order to validate
the approach. While there is some indirect indication of probe sensitivities to temperature changes, the effects are
inconclusive and require further study.
6-8
-------
6. 5 Conclusions
The results presented in this chapter show that a soil water balance approach using monitored precipitation and SM,
such as the proposed SWAM methodology, can provide a low-cost alternative to the custom made weir device or
lysimeter systems frequently used to quantify runoff during green roof studies. Thus, this work has provided an
important first step in developing a green roof runoff monitoring system that might be reliably deployed on a wider
scale. The discrepancies in runoff estimates have been attributed to SM probe sensitivities to ambient factors such as
temperature, which require further investigation, as well as possible effects of green roof spatial heterogeneity.
SWAM's ability to provide an indirect estimation of green roof ET is also considered important given the challenges
involved with obtaining direct ET measurements on urban roofs. Nonetheless, more work and more data are required
to validate the applicability of SWAM in estimating ET.
If deployed on new monitoring sites, this method would require knowledge of the water holding capacity of the soil as
well as the range of soil moisture probe voltage readings, which could be determined through simple calibration tests.
6-9
-------
Chapter 7 References
Aitkenhead-Peterson, J. A., Dvorak, B. D., Voider, A., and Stanley, N. C. (2010). Chemistry of growth medium and
leachate from green roof systems in south-central Texas. Urban Ecosystems, 14(1), 17-33. doi: 10.1007/sl 1252-010-
0137-4
Alsup, S., Ebbs, S., and Retzlaff, W. (2010). The exchangeability and leachability of metals from select green roof
growth substrates. Urban Ecosystems, 13(1), 91-111. doi:10.1007/s!1252-009-0106-y
Baumhardt, R. L., Lascano, R. J., and Evert, S. R. (2000). Soil material, temperature, and salinity effects on calibration
of multisensor capacitance probes, Soil Sci. Soc. Am. J. 64:1940-1946.
Bengtsson, L., Grahn, L., and Olsson, J. (2005). Hydrological function of a thin extensive green roof in southern
Sweden. Nordic Hydrology, 36(3), 259-268. Retrieved from http://www.iwaponline.com/nh/036/0259/0360259.pdf
Berghage, R. D., Beattie, D., Jarrett, A. R., Thuring, C., Razaei, F., and O'Connor, T. (2009). Green Roofs for
Stormwater Runoff Control. EP A/600/R-09/026. Cincinnati, OH. Retrieved from
http://nepis.epa.gov/Exe/ZyNET.exe/P1003704.TXT
Berghage, R. D., and Gu, J. (2009). Effect of Drain Layers on Green Roof Stormwater Performance. In Seventh Annual
Greening Rooftops for Sustainable Communities Conference, Awards and Trade Show (pp. 1-17). Atlanta, GA.
Berghage, R. D., Miller, C., Bass, B., Moseley, D., and Weeks, K. (2010). Stormwater Runoff from a Large Commercial
Roof in Chicago. In CitiesAlive! Eighth Annual Green Roof and Wall Conference (pp. 1-13). Vancouver, BC.
Berndtsson, J. C. (2010). Green roof performance towards management of runoff water quantity and quality: A review.
Ecological Engineering, 36(4), 351-360. doi:10.1016/j.ecoleng.2009.12.014
Berndtsson, J. C., Bengtsson, L., and Jinno, K. (2009). Runoff water quality from intensive and extensive vegetated
roofs. Ecological Engineering, 35(3), 369-380. doi: 10.1016/j.ecoleng.2008.09.020
Berndtsson, J. C., Emilsson, T., and Bengtsson, L. (2006). The influence of extensive vegetated roofs on runoff water
quality. The Science of the Total Environment, 355(1-3), 48-63. doi:10.1016/j.scitotenv.2005.02.035
Bliss, D. J., Neufeld, R. D., and Ries, R. J. (2009). Storm Water Runoff Mitigation Using a Green Roof. Environmental
Engineering Science, 26(2), 407-418. doi:10.1089/ees.2007.0186
Bogena, H. R., Huisman, J. A., Oberdorster, C., and Vereecken, H. (2007). Evaluation of a low-cost soil water content
sensor for wireless network applications. Journal of Hydrology, 344(1-2), 32-42. doi:10.1016/j.jhydrol.2007.06.032
Bricker, S., Longstaff, B., Dennison, W., Jones, A., Boicourt, K., Wicks, C., and Woerner, J. (2007). Effects of Nutrient
Enrichment in the Nation's Estuaries: A Decade of Change. NOAA Coastal Ocean Program Decision Analysis Series
No. 26 (p. 328). Silver Spring, MD.
Brocca, L., and Moramarco, T. (2013). A new method for rainfall estimation through soil moisture observations.
Geophysical Research, 40(October 2012), 1-6. doi:10.1002/GRL.50173
Campbell Scientific Inc. (2012). CS650 and CS655 Water Content Reflectometers Instruction Manual (p. 56).
Carpenter, D. D., and Kaluvakolanu, P. (2011). Effect of roof surface type on storm-water runoff from full-scale roofs
in a temperate climate. Journal of Irrigation and Drainage Engineering, 137(3), 161-169.
doi:10.1061/(ASCE)IR.1943-4774.0000185
Carson, T. B., Marasco, D. E., Culligan, P. J., and McGillis, W. R. (2013). Hydrological performance of extensive green
roofs in New York City: Observations and multi-year modeling of three full-scale systems. Environmental Research
Letters, 8(2), 024036. doi: 10.1088/1748-9326/8/2/024036
Clark, S. E., Long, B. V., Siu, C. Y. S., Spicher, J., and Steele, K. A. (2008). Early-Life Roof Runoff Quality: Green
vs. Traditional Roofs. In Low Impact Development for Urban Ecosystem and Habitat Protection. ASCE. (doi:
http://dx.doi.org/10.1061/41009(333)9)
Coffman, L. (2000). Low-Impact Development Design Strategies: An Integrated Design Approach (p. 150). Prince
George's County, MD.
Connelly, M., Liu, K., and Schaub, J. (2006). BCIT Green Roof Research Program, Phase 1 Summary of Data Analysis
(pp. 1-59). Vancouver, BC. Retrieved from http://publications.gc.ca/site/eng/389231/publication.html
Crow, W. T., van den Berg, M. J., Huffman, G. J., and Pellarin, T. (2011). Correcting rainfall using satellite-based
surface soil moisture retrievals: The Soil Moisture Analysis Rainfall Tool (SMART). Water Resources Research,
47(8), n/a-n/a. doi: 10.1029/2011WRO10576
7-1
-------
Culligan, P. J. (2011). Green roofs: A solution for urban stormwater management? Geo-Strata - Geo Institute of ASCE,
12(2), 30-32, 34.
Czarnomski, N. M., Moore, G. W., Pypker, T. G., Licata, J., and Bond, B. J. (2005). Precision and accuracy of three
alternative instruments for measuring soil water content in two forest soils of the Pacific Northwest, Can. J. For. Res.
35:1867-1876. doi:10.1139/X05-121
De Cuyper, K., Dinne, K., and Van De Vel, L. (2004). Rainwater Discharge from Green Roofs. In CIB W062 30th
International Symposium on Water Supply and Drainage for Buildings (p. 12). Rotterdam, Netherlands. Retrieved
from http://www.irb.fraunhofer.de/CIBlibrary/search-quick-result-list.jsp?AandidSuche=CIB+DC 10549
Decagon Devices Inc. (2012). EC-5 Soil Moisture Sensor User's Manual (p. 24).
DiGiovanni, K., Gaffin, S. R., and Montalto, F. (2010). Green Roof Hydrology: Results from a Small-Scale Lysimeter
Setup (Bronx, NY). In Low Impact Development 2010: Redefining Water in the City (pp. 1328-1341). San Francisco,
CA: ASCE. Retrieved from http://link.aip.Org/link/7ASCECP/367/114/l
Dorigo, W. A., Wagner, W., Hohensinn, R., Hahn, S., Paulik, C., Xaver, A., Gruber, A., Drusch, M., S. Mecklenburg,
S., van Oevelen, P., Robock, A., and Jackson, T. (2011). The International Soil Moisture Network: A data hosting
facility for global in situ soil moisture measurements. Hydrology and Earth System Sciences, 15(5), 1675-1698.
doi:10.5194/hess-15-1675-2011
Dvorak, B., and Voider, A. (2010). Green roof vegetation for North American ecoregions: A literature review.
Landscape and Urban Planning, 96(4), 197-213. Retrieved from http://www.scopus.com/inward/record.url?eid=2-
s2.0-77953324216andpartnerID=40andmd5=e4fa44b9f91f819d2b2f32cl02745332
EPA. (2004). Report to Congress: Impacts and Control of CSOs and SSOs. Washington, DC. Retrieved from
http://cfpub.epa.gov/npdes/cso/cpolicv report2004.cfm.
EPA (2009). National Primary Drinking Water Regulations. Water: Drinking Water Contaminants U.S. Environmental
Protection Agency, Report No. EPA 816-F-09-0004, pp. 6, http://water.epa.gov/drink/contaminants/upload/mcl-
2.pdf.
Erlichman, P., and Peck, S. W. (2013). Annual green roof industry survey for 2012 (p. 11).
Passman, E. and Simcock, R. (2012) Moisture measurements as performance criteria for extensive living roof substrates.
Journal of Environmental Engineering, 138(8), 841-851. doi:10.1061/(ASCE)EE.1943-7870.0000532.
Fassman-Beck, E., Voyde, E., Simcock, R., and Hong, Y. S. (2013). 4 Living roofs in 3 locations: Does configuration
affect runoff mitigation? Journal of Hydrology, 490, 11-20. doi:10.1016/j.jhydrol.2013.03.004
Field, R., and Sullivan, D. (2001). Overview of EPA's wet-weather flow research program. Urban Water, 3, 165-169.
GeoSyntec Consultants Inc. (2012). High performance green infrastructure for CSO Reduction and Advanced
Rainwater Harvesting. Retrieved from https://www.optirtc.com/CSO_ARWH.aspx
Getter, K. L., and Rowe, D. B. (2006). The role of extensive green roofs in sustainable development. HortScience,
41(5), 1276-1285.
Getter, K. L., Rowe, D. B., and Andresen, J. A. (2007). Quantifying the effect of slope on extensive green roof
stormwater retention. Ecological Engineering, 31, 225-31. doi:10.1016/j.ecoleng.2007.06.004
GreenRoofs.com. (2014). The International Greenroof and Greenwall Projects Database!
http://www.greenroofs.com/projects/plist.php
Gregoire, B. G., and Clausen, J. C. (2011). Effect of a modular extensive green roof on stormwater runoff and water
quality. Ecological Engineering, 37, 963-9. doi:10.1016/j.ecoleng.2011.02.004
Hathaway, A. M., Hunt, W. F., and Jennings, G. D. (2008). A field study of green roof hydrologic and water quality
performance. American Society of Agricultural and Biological Engineers, 51(1), 37-44.
Hook, W. R., Ferre, T. P. A., and Livingston, N. J. (2004). The effects of salinity on the accuracy and uncertainty of
water content measurement. Soil Science Society of America, 68 (1), 47-56, doi:10.2136/sssaj2004.4700.
Howarth, R. W., Marino, R., Swaney, D. P., and Boyer, E. W. (2011). Chapter 10: Wastewater and Watershed
Influences on Primary Productivity and Oxygen Dynamics in the Lower Hudson River Estuary. The Hudson River
Estuary. Edited by J. S. Levinton and J. R. Waldman, Cambridge University Press, ISBN: 9780521207980.
Hutchinson, D., Abrams, P., Retzlaff, R., and Liptan, T. (2003). Stormwater monitoring two ecoroofs in Portland,
Oregon, USA. In Greening Rooftops for Sustainable Communities (pp. 1-18). Chicago, IL. Conference sponsored
by Green Roofs for Healthy Cities, Toronto, and the City of Chicago, May 29-30, 2003.
Jarrett, A. R., Hunt, W. F., and Berghage, R. D. (2006). Annual and Individual-Storm Green Roof Stormwater Response
7-2
-------
Models. In 2006 ASAE Annual Meeting (p. 062310). St. Joseph, MI: American Society of Agricultural and Biological
Engineers, doi: 10.13031/2013.20793
Joshi, C., Mohanty, B. P., Jacobs, J. M., and Ines, A. V. M. (2011). Spatiotemporal analyses of soil moisture from point
to footprint scale in two different hydroclimatic regions. Water Resources Research, 47, W01508,
doi:10.1029/2009WR009002.
Kirchner, J. W. (2009). Catchments as simple dynamical systems: Catchment characterization, rainfall-runoff modeling,
and doing hydrology backward. Water Resources Research, 45, W02429, doi: 10.1029/2008WR006912.
Kohler, M., and Keeley, M. (2004). Chapter: The Green Roof Tradition in Germany: The Example of Berlin. Green
Roofs: Ecological Design and Construction. By Eath Pledge Foundation, Schiffer Publishing ISBN-10: 0764321897,
ISBN-13: 978-0764321894.
Kurtz, T. (2008). Flow Monitoring of Three Ecoroofs in Portland, Oregon. In 2008 Low Impact Development
Conference. Seattle, WA: ASCE.
Liu, K., and Minor, J. (2005). Performance evaluation of an extensive green roof. In Greening Rooftops for Sustainable
Communities (pp. 1-11). Washington, DC.
Marasco, D. E., Alternative Metrics of Green Roof Hydrological Performance: Evapotranspiration and Peak Flow
Reduction, PhD Thesis, Columbia University, 2014.
Mayor's Office of Long-Term Planning and Sustainability. (2008). Sustainable Stormwater Management Plan.
Management. New York. Retrieved from www.nyc.gove/PlaNYC2030
Mentens, J., Raes, D., and Hermy, M. (2003). Effect of Orientation on the Water Balance of Greenroofs. In Greening
Rooftops for Sustainable Communities (pp. 1-9). Chicago, IL. Conference sponsored by Green Roofs for Healthy
Cities, Toronto, and the City of Chicago, May 29-30, 2003.
Miller, C. (2012). The Need for Establishing Performance Metrics. In Mid-Atlantic Green Roof Science and
Technology Symposium. College Park, MD.
Montalto, F., Behr, C., Alfredo, K., Wolf, M., Arye, M., and Walsh, M. (2007). Rapid assessment of the cost-
effectiveness of low impact development for CSO control. Landscape and Urban Planning, 82(3), 117-131.
doi:10.1016/j.landurbplan.2007.02.004
Monterusso, M. A., Rowe, D. B., and Rugh, C. L. (2005). Establishment and persistence of Sedum spp. and Native
Taxa for green roof applications. HortScience, 40(2), 391-396.
Moran, A. C., Hunt, W. F., and Smith, J. T. (2005). Green Roof Hydrologic and Water Quality Performance from Two
Field Sites in North Carolina. In Managing Watersheds for Human and Natural Impacts (pp. 1-12). Reston, VA:
American Society of Civil Engineers, doi: 10.1061/40763(178)99
Morgan, S., Celik, S., and Retzlaff, W. (2013). Green roof storm-water runoff quantity and quality. Journal of
Environmental Engineering, 139(4), 471-478. doi:10.1061/(ASCE)EE.1943-7870.0000589
Moriasi, D., and Arnold, J. (2007). Model evaluation guidelines for systematic quantification of accuracy in watershed
simulations. Transactions of the ASABE, 50(3), 885-900.
Nardini, A., Andri, S., and Crasso, M. (2011). Influence of substrate depth and vegetation type on temperature and
water runoff mitigation by extensive green roofs: shrubs versus herbaceous plants. Urban Ecosystems, 15(3), 697-
708. doi: 10.1007/sl 1252-011-0220-5
Nash, J. E., and Sutcliffe, J. V. (1970). River flow forecasting through conceptual models part I—A discussion of
principles. Journal of Hydrology, 10(3), 282-290.
National Atmospheric Deposition Program, 2012. (2012). National Atmospheric Deposition Program 2011 Annual
Summary. Illinois at Urbana-Champaign, IL.
NRCS (1986). Urban Hydrology for Small Watersheds, TR-55 (p. 164). Washington, D.C.
NYC DEP. (2010). NYC Green Insfrastructure Plan: A sustainable strategy for clean waterways (pp. 1-154). New
York, NY.
NYS DEC. (2008). The Lower Hudson River Basin Waterbody Inventory and Priority Waterbodies List. Bureau of
Watershed Assessment and Management, Division of Water, http: //www. dec .ny. gov/docs/water_pdf/pwllhudintr .pdf
Oberndorfer, E., Lundholm, J., Bass, B., Coffman, R R, Doshi, H., Dunnett, N., Gaffm, S., Kohler, M, Liu, K. and
Rowe, B. (2007). Green roofs as urban ecosystems: Ecological structures, runctions, and services. BioScience, 57(10),
823-833. doi:10.1641/B571005
Palla, A., Sansalone, J. J., Gnecco, I., and Lanza, L. G. (2011). Storm water infiltration in a monitored green roof for
7-3
-------
hydrologic restoration. Water Science and Technology, 64(3), 766. doi: 10.2166/wst.2011.171
Peel, M. C., Finlayson, B. L., and McMahon, T. A. (2007). Updated world map of the Koppen-Geiger climate
classification. Hydrology and Earth System Sciences, 11(5), 1633-1644. doi: 10.5194/hess-l 1-1633-2007
Robinson, D. A., Jones, S. B., Blonquist, J. M., and Friedman, S. P. (2005). A physically derived water
content/permittivity calibration model for coarse-textured, layered soils. Soil Science Society of America Journal,
69(5), 1372. doi:10.2136/sssaj2004.0366
Sakaki, T., and Rajaram, H. (2006). Performance of different types of time domain reflectometry probes for water
content measurement in partially saturated rocks. Water Resources Research, 42(7), n/a-n/a.
doi: 10.1029/2005 WR004643
Schroll, E., Lambrinos, J., Righetti, T., and Sandrock, D. (2011). The role of vegetation in regulating stormwater runoff
from green roofs in a winter rainfall climate. Ecological Engineering, 37(4), 595-600.
doi:10.1016/j.ecoleng.2010.12.020
She, N. and Pang, J. (2010). "Physically Based Green Roof Model." J. Hydrol. Eng. 15, Special Issue: Low Impact
Development, Sustainability Science, and Hydrological Cycle, 458-464.
Sherrard, J. A., and Jacobs, J. M. (2012). Vegetated roof water-balance model: Experimental and model results. Journal
of Hydrologic Engineering, 17(8), 858-868. doi:10.1061/(ASCE)HE.1943-5584.0000531
Spolek, G. (2008). Performance monitoring of three ecoroofs in Portland, Oregon. Urban Ecosystems, 11, 349-359.
doi:10.1007/s!1252-008-0061-z
Starry, O. (2013). The Comparative Effects of Three Sedum Species on Green Roof Stormwater Retention. University
of Maryland, Dissertation, URI: http://hdl.handle.net/1903/14622.
Stovin, V. (2010). The potential of green roofs to manage Urban Stormwater. Water and Environment Journal, 24, 192-
199. doi: 10.1111/j. 1747-6593.2009.00174.x
Stovin, V., Vesuviano, G., and Kasmin, H. (2012). The hydrological performance of a green roof test bed under UK
climatic conditions. Journal of Hydrology, 414-415, 148-161. doi:10.1016/j.jhydrol.2011.10.022
Strecker, E., B. Urbonas, M. Quigley, J. Howell, and T. Hesse (2002). "Urban Stormwater BMP Performance
Monitoring: A Guidance Manual for Meeting the National Stormwater BMP Database Requirements" Office of
Water, US EPA. Report No. EPA-821-B-02-001, April, 2002.
Teemusk, A., and Mander, U. (2007). Rainwater runoff quantity and quality performance from a greenroof: The effects
of short-term events. Ecological Engineering, 30, 271-277'. doi:10.1016/j.ecoleng.2007.01.009
Toronto and Region Conservation Authority. (2006). Evaluation of an Extensive Greenroof (p. 167). Toronto, ON.
Trejo, E. (2013). Insecticide application at W118. Personal Communication.
VanWoert, N. D. (2005). Watering regime and green roof substrate design affect sedum plant growth. HortScience,
40(3), 659-664.
VanWoert, N. D., Rowe, D. B., Andresen, J. a, Rugh, C. L., Fernandez, R. T., and Xiao, L. (2005). Green roof
stormwater retention: Effects of roof surface, slope, and media depth. Journal of Environmental Quality, 34(3), 1036-
44. doi:10.2134/jeq2004.0364
Vijayaraghavan, K., Joshi, U. M., and Balasubramanian, R. (2012). A field study to evaluate runoff quality from green
roofs. Water Research, 46(4), 1337-1345. doi:10.1016/j.watres.2011.12.050
Villarreal, E. L. (2007). Runoff detention effect of a sedum green-roof Nordic Hydrology, 38(1), 99.
doi:10.2166/nh.2007.031
Voyde, E., Passman, E., and Simcock, R. (2010). Hydrology of an extensive living roof under sub-tropical climate
conditions in Auckland, New Zealand. Journal of Hydrology, 394, 384-395. doi: 10.1016/j.jhydrol.2010.09.013
Walker, J. P., Willgoose, G. R., and Kalma, J. D. (2004). In situ measurement of soil moisture: A comparison of
techniques. Journal of Hydrology, 293(1-4), 85-99. doi: 10.1016/j.jhydrol.2004.01.008
Williams, E. S. (2003). Hydrologic and Economic Impacts of Alternative Residential Land Development Methods.
University of Florida.
Williams, E. S., and Wise, W. R. (2006). Hydrologic impacts of alternative approaches to storm water management and
land development. Journal of the American Water Resources Association, 42(2), 443-456.
7-4
-------
Appendix A Site Equipment Tables
Table 7-1: Monitoring Equipment at the W115 Office Building at Columbia University.
SITE NAME - W115
Description
Scientific Equipment
HOBO U30 Wi-Fi Data Logger W/ Analog Input - U30-WIF
2-bit Temperature/Relative Humidity (RH) Smart Sensor (2 m
cable) - S-THB-M002
Wind Speed and Direction Smart Sensor (3 m cable) - S-WCA-
M003
Range of Tolerance
Data Logger - Wi-Fi -20°C to 40°C
Air Temp/RH Sensor -40°C to 75°C
Wind and Gust
Speed/Wind Direction -40°C to 75°C
Sensor
Solar Radiation Sensor (Silicon Pyranometer) Smart Sensor - S-LIB-„ , „ ,. . „ Ar.0^ -,-0/^
*fnr,^ Solar Radiation Sensor-40 to 75 C
Barometric Pressure
Barometric Pressure Smart Sensor - S-BPA-CM10
EC-5 Soil Moisture Smart Sensor - S-SMC-M005
12-Bit Temp Smart Sensor (2 m cable) - S-TMB-M002
.2 mm Rainfall Smart Sensor (2 m cable) - S-RGB-M002
Custom Drainage Pipe Weir with Senix ToughSonic Ultrasonic
Distance Sensor - TSPC-30S1
„ -40° to 70°C
Sensor
Soil Moisture Sensor 0 to 50°C
Soil Temp Sensor -40° to 100°C
Tipping Bucket Rain no cnop
Gauge
Runoff sensor
0° to 70°C
-------
Table 7-2: Monitoring Equipment at the W118 Residence Hall at Columbia University.
SITE NAME - W118
Onset Scientific Equipment Description
HOBO U30 Wi-Fi Data Logger W/ Analog Input - U30-WIF
2-bit Temperature/Relative Humidity (RH) Smart Sensor (2m cable) - S-
THB-M002
Soil Moisture Smart Sensors - CS-615 L50
EC-5 Soil Moisture Smart Sensor - S-SMC-M005
12-Bit Temp Smart Sensor (2 m cable) - S-TMB-M002
.2 mm Rainfall Smart Sensor (2 m cable) - S-RGB-M002
EC-5 Soil Moisture Smart Sensor - S-SMC-M005
Custom Drainage Pipe Weir with Senix ToughSonic Ultrasonic Distance
Sensor-TSPC-30S1
Data Logger - Wi-Fi
Air Temp/RH Sensor
Soil Moisture Sensor
Soil Moisture Sensor
Soil Temp Sensor
Tipping Bucket Rain
Gauge
Soil Moisture Sensor
Runoff sensor
Range of
Tolerance
-20°C to 40°C
-40°C to 75°C
0 to 70°C
0 to 50°C
-40° to 100°C
0° to 50°C
0 to 50°C
0° to 70°C
Table 7-3: Monitoring Equipment at the US Postal Service Morgan Distribution Facility.
SITE NAME - USPS
Scientific Equipment
HOBO U30 GSM Data Logger W/ Analog Input - U30-GSM
2-bit Temperature/Relative Humidity (RH) Smart Sensor (2 m
cable) - S-THB-M002
Wind Speed and Direction Smart Sensor (3 m cable) - S-WCA-
M003
Solar Radiation Sensor (Silicon Pyranometer) Smart Sensor - S-
LIB-M003
EC-5 Soil Moisture Smart Sensor - S-SMC-M005
12-Bit Temp Smart Sensor (2 m cable) - S-TMB-M002
.2 mm Rainfall Smart Sensor (2 m cable) - S-RGB-M002
Custom Drainage Pipe Weir with Senix ToughSonic Ultrasonic
Distance Sensor - TSPC-30S1
Description
Range of Tolerance
Data Logger-GSM _20oCto40oC
Cellular
Air Temp/RH Sensor -40°C to 75 °C
Wind and Gust
Speed/Wind Direction -40°C to +75 °C
Sensor
Solar Radiation Sensor -40° to 75 °C
Soil Moisture Sensor 0 to 50°C
Soil Temp Sensor -40° to 100°C
Tipping Bucket Rain 0oto50oC
Gauge
Runoff sensor
0° to 70°C
A 2
-------
Table 7-4: Monitoring Equipment at the Con Edison Facility in Long Island City (HOBO Vendor).
SITE NAME - ConEd
Scientific Equipment Description Range of Tolerance
Data Logger - GSM
HOBO U30 GSM Data Logger W/ Analog Input -U30-GSM 55 -20°Cto40°C
Cellular
2-bit Temperature/Relative Humidity (RH) Smart Sensor (2 m cable) - . . „ /T,TT 0 ^o/-^ -7^0/^
c T-rm A,rnm Air Temp/RH Sensor -40 C to 75 C
--
Photosynthetic Light (PAR) Smart Sensor - S-LIB-M003 PAR Sensor -40° to 75°C
Barometric Pressure Smart Sensor - S-BPA-CM 1 0 Barometric Pressure _4QO fo 7QOC
Sensor
EC-5 Soil Moisture Smart Sensor - S-SMC-M005 Soil Moisture Sensor 0 to 50°C
12-Bit Temp Smart Sensor (2 m cable) - S-TMB-M002 Soil Temp Sensor -40° to 100°C
.2 mm Rainfall Smart Sensor (2 m cable) - S-RGB-M002 1PPmg U° G am 0° to 50°C
Gauge
2 Custom Drainage Pipe Weirs with Senix ToughSonic Ultrasonic
Runoff sensor 0°to70°C
Distance Sensors - TSPC-30S1
Table 7-5: Weather Monitoring Equipment at the Con Edison Facility (Campbell Scientific Vendor).
SITE NAME - ConEd
Scientific Equipment Description Range of Tolerance
Campbell Scientific (CS) CR3000-ST-SW-RC-NC Data Logger -25°C to 50°C
Temperature/Relative Humidity (RH) Sensor - CS215-L50 Air Temp/RH Sensor -40°C to 70°C
Wind and Gust
RM Young Wind Monitor - 05103-L65 Speed/Wind Direction -50°C to +50°C
Sensor
M j-r, j- ^ /-VNTTI ,IT Allwave Radiation .no. ono/^
Net Radiometer - CNR 4L „ -40 to 80 C
Sensors
Soil Moisture Smart Sensors-CS-615 L50 Soil Moisture Sensors 0 to 70°C
Temperature Probes - 107-L50 Soil Temp Sensor -35°to50°C
M T D • r> nm • u+- Tipping Bucket Rain ono „„<,„
Nova Lynx Ram Gage 0.01 inch tip „ ^ & -20 to 70 C
J & F Gauge
Apogee IR Temperature Radiometers „ c -30°to65°C
^ & ^ Temperature Sensors
A 3
-------
Table 7-6: Weather Monitoring Equipment at the Fieldston School in Riverdale, Bronx.
SITE NAME - Fldstn
Scientific Equipment Description
Campbell Scientific (CS) CR3000-ST-SW-RC-NC Data Logger
Range of Tolerance
-25°C to 50°C
Temperature/Relative Humidity (RH) Sensor - CS215-L50
RM Young Wind Monitor - 05103-L65
Kipp and Zonen Solar Radiation Sensors - CMP3
Soil Moisture Smart Sensors - CS-615 L50
Temperature Probes - 107-L50
Met One AC Rain Gage 0.01 inch tip
Apogee IR Temperature Radiometers
Air Temp/RH Sensor -40°C to 70°C
Wind and Gust
Speed/Wind Direction -50°C to +50°C
Sensor
Solar Radiation
Sensors
-40° to 80°C
Soil Moisture Sensors 0 to 70°C
Soil Temp Sensor -35° to 50°C
Tipping Bucket Rain
Gauge
IR Surface
Temperature Sensors
-20° to 50°C
-30° to 65 °C
Table 7-7: Monitoring Equipment at the Bronx Design and Construction Academy
SITE NAME - BDCA
Scientific Equipment Description Range of Tolerance
Data Logger - GSM
HOBO U30 GSM Data Logger W/ Analog Input - U30-Wi-Fi
Cellular
-20°C to 40°C
2-bit Temperature/Relative Humidity (RH) Smart Sensor (2 m cable) - . . „ /T,TT 0 ^o/-^ -7^0/^
c Ttro iv/rnm Air Temp/RH Sensor -40 C to 75 C
S-THB-M002
Soil Moisture Smart Sensor - EC-5 S-SMC-M005
.2 mm Rainfall Smart Sensor (2 m cable) - S-RGB-M002
Custom Drainage Pipe Weir with Senix ToughSonic Ultrasonic
Distance Sensors - TSPC-30S1
Soil Moisture Sensor 0 to 50°C
Tipping Bucket Rain
Gauge
Runoff sensor
0° to 50°C
0° to 70°C
A 4
-------
Table 7-8: Weather Monitoring Equipment at Regis High School
SITE NAME - Regis
Scientific Equipment Description Range of Tolerance
Campbell Scientific (CS) CR3000-ST-SW-RC-NC Data Logger -25°C to 50°C
Temperature/Relative Humidity (RH) Sensor - CS215-L50 Air Temp/RH Sensor -40°C to 70°C
Wind and Gust
RM Young Wind Monitor - 05103-L65 Speed/Wind Direction -50°C to +50°C
Sensor
XT j-r, j- ^ /-VNTTI ,IT Allwave Radiation .no. ono/^
Net Radiometer - CNR 4L „ -40 to 80 C
Sensors
Soil Moisture Smart Sensors-CS-615 L50 Soil Moisture Sensors 0 to 70°C
Temperature Probes - 107-L50 Soil Temp Sensor -35°to50°C
XT T or- nm • u+- Tipping Bucket Rain ono „„<,„
Nova Lynx Ram Gage 0.01 inch tip „ ^ & -20 to 70 C
J & F Gauge
Apogee IR Temperature Radiometers „ „ -30°to65°C
lemperature Sensors
3 Custom Drainage Pipe Weirs with Senix ToughSonic Ultrasonic „ .,., no ^ ™0/^
^.. . c Tcn/-onci Runoff sensor 0° to 70°C
Distance Sensors -TSPC-30S1
A 5
-------
Appendix B Water Quality Results
B 1
-------
Table 7-9: Summary of Results from Water Quality Monitoring (March 10, 2011 -August 2, 2012). The average of the pH was calculated by first transforming the
measurements using this equation: [H+] = 1 O-PH. Once averaged, the [H+] value was converted back via this equation: pH = -log[H+].
Sample ID
34-W115.C
35-W118.C
36-Fdston.C
37-W115.GR
38-W118.GR
39-Fdston.GR
40- Rain
41-W118.C
42-W118.GR
43-W115.C
44-W118.C
45-W115.GR
46-W118.GR
47- Rain
48-W115.C
49-W118.C
50-W115.GR
51-W118.GR
52-W118.C
53-Rain
54-W118.GR
55-USPS.GR
56-W115.C
57-W118.C
58-W115.GR
59-W118.GR
60-USPS.C
61-W118.GR
Sample
Date
3/10/2011
3/10/2011
3/10/2011
3/10/2011
3/10/2011
3/10/2011
3/10/2011
3/21/2011
3/21/2011
3/23/2011
3/23/2011
3/23/2011
3/23/2011
3/23/2011
4/5/2011
4/5/2011
4/5/2011
4/5/2011
4/12/2011
4/12/2011
4/13/2011
4/13/2011
4/28/2011
4/28/2011
4/28/2011
4/28/2011
6/17/2011
6/17/2011
PH
5.95
6.23
7.03
7.82
8.06
7.99
5.50
6.77
7.63
6.23
7.42
7.42
7.97
4.59
6.41
6.13
7.55
7.40
7.25
4.99
7.18
7.64
7.21
7.04
7.92
7.19
8.17
7.37
Conductivity
(uS/cm)
148.40
218.90
21.00
138.90
96.67
144.60
66.82
40.37
72.93
13.67
25.75
77.00
95.77
53.69
21.00
188.80
143.10
89.05
na
26.32
86.23
212.00
33.38
36.28
116.02
73.40
138.76
123.05
Turbidity
(NTU)
0.00
1.16
0.02
2.89
0.58
0.89
0.04
0.67
0.51
1.76
1.84
4.57
0.60
1.03
1.21
2.79
0.35
0.72
1.40
1.07
2.96
10.47
1.13
1.60
0.59
0.73
1.37
0.98
Apparent Color
(PtCo)
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
39.45
23.44
127.04
482.66
29.66
81.83
115.55
150.21
55.27
268.69
NO3
1.72
3.34
0.78
na
0.77
1.64
1.41
1.14
0.00
1.19
1.69
0.58
0.00
2.01
2.70
6.12
0.59
0.91
3.82
1.11
0.55
0.94
0.42
0.08
0.00
0.00
na
0.00
NH4+
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.60
0.32
0.11
0.10
na
0.12
P
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.00
0.00
0.73
0.00
na
0.00
Ca
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
2.51
2.69
12.31
7.07
na
10.16
K
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.28
0.40
0.33
0.13
na
0.92
Na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.54
1.37
3.38
2.45
na
16.60
Mg
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.28
0.51
2.31
2.08
na
5.69
B
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
na
0.00
0.47
0.00
2.62
na
3.26
B 2
-------
62-USPS.GR
63-Rain
64-W115.C
65-USPS.C
66-ConEd.C
67-USPS.GR
68-ConEd.GR
69- Rain
70-W115.C
71-W118.C
72-W115.GR
73-W118.GR
74- Rain
75-USPS.GR
76-Fdston.GR
77- Rain
78-USPS.C
79-USPS.GR
80- Rain
81-W115.C
82-W118.C
83-W115.GR
84-W118.GR
85-W115.C
86-W118.C
87-W118.GR
88-Rain
89-USPS.C
90-ConEd.C
91-USPS.GR
92-ConEd.GR
93-Rain
94- Rain
6/17/2011
6/17/2011
6/23/2011
6/23/2011
6/23/2011
6/23/2011
6/23/2011
6/23/2011
8/9/2011
8/9/2011
8/9/2011
8/9/2011
8/9/2011
8/15/2011
8/15/2011
8/15/2011
8/25/2011
8/25/2011
8/25/2011
9/6/2011
9/6/2011
9/6/2011
9/6/2011
9/23/2011
9/23/2011
9/23/2011
10/13/2011
10/19/2011
10/19/2011
10/19/2011
10/19/2011
10/19/2011
10/27/2011
7.67
4.72
5.71
7.78
6.22
7.49
6.80
4.37
4.94
6.47
9.28
7.38
4.73
7.29
7.45
lost
7.73
7.08
4.80
6.31
6.53
7.79
7.44
6.56
6.37
7.35
5.48
7.63
6.44
7.33
6.98
5.00
4.71
154.10
18.45
55.23
199.01
52.38
208.90
164.25
57.65
12.38
5.90
163.00
42.74
13.39
215.90
101.80
5.00
66.89
112.80
17.60
41.40
5.31
231.40
106.10
11.65
2.77
23.86
17.72
69.96
17.80
150.00
68.34
15.93
24.15
9.24
1.33
1.93
4.43
0.95
3.01
2.21
1.42
na
na
na
na
na
0.96
0.67
0.60
2.23
3.36
0.43
0.45
0.36
0.80
1.51
0.59
0.71
2.66
0.24
1.78
0.24
0.96
0.70
0.13
0.33
325.76
13.27
86.73
69.21
39.83
148.51
186.37
31.54
4.98
8.75
187.69
126.47
11.57
235.35
152.09
0.46
13.65
146.06
0.00
0.00
12.52
106.70
166.97
na
na
na
0.00
10.26
0.00
85.79
112.54
0.00
0.00
0.11
0.38
1.05
0.90
0.68
0.00
0.05
1.32
na
na
na
na
na
0.04
0.00
0.21
0.16
0.00
0.38
0.00
0.09
0.00
0.00
0.24
0.00
0.00
0.22
0.12
0.16
0.00
0.00
0.13
0.47
0.28
0.45
1.94
5.45
1.48
2.44
1.19
1.52
na
na
na
na
na
0.07
0.18
0.45
0.74
0.46
0.48
0.10
0.00
0.06
0.54
0.20
0.03
0.11
0.21
0.60
0.22
0.22
0.03
0.07
0.34
0.00
0.00
0.00
0.34
0.00
0.73
0.80
0.33
na
na
na
na
na
0.36
0.00
0.00
0.00
0.36
0.00
0.00
0.45
0.00
0.62
0.00
0.00
0.25
0.00
0.39
0.00
0.00
0.00
0.35
0.00
14.67
0.19
2.17
18.90
2.13
20.26
7.48
2.00
na
na
na
na
na
29.50
13.02
0.21
7.06
13.03
0.56
0.42
0.12
8.68
6.43
0.59
0.11
3.65
0.39
5.67
0.70
19.80
5.62
0.15
0.96
1.51
0.00
0.15
9.93
0.20
1.55
3.77
0.17
na
na
na
na
na
2.80
2.88
0.00
0.27
0.57
0.00
0.13
0.05
0.97
2.04
0.13
0.00
0.84
0.00
0.37
0.00
0.31
0.78
0.00
0.00
2.26
0.30
1.17
14.36
3.27
2.52
4.58
2.25
na
na
na
na
na
3.31
3.00
0.26
0.51
1.14
1.15
0.00
0.39
2.28
14.44
1.80
0.09
2.59
1.69
0.87
1.26
1.30
1.43
1.22
0.05
3.55
0.00
0.43
4.78
0.86
4.47
3.95
0.38
na
na
na
na
na
4.63
1.53
0.00
1.71
2.43
0.18
0.15
0.00
2.34
3.85
0.33
0.00
1.74
0.13
1.89
0.21
3.52
2.42
0.28
0.12
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
na
na
na
na
na
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.64
0.00
0.00
0.30
0.00
0.00
0.00
0.00
0.00
0.00
0.00
B 3
-------
95-ConEd.C
96-ConEd.GR
97-Rain
98-USPS.C
99-USPS.GR
100-USPS.GR
101-USPS.C
102-USPS.GR
103-Rain
104-W118.C
105-USPS.C
106-Fdston.C
107-W118.GR
108-USPS.GR
109-
Fdston.GR
110- Rain
Ill-Rain
112-Fdston.C
113-
Fdston.GR
114- Rain
115-USPS.C
116-USPS.GR
117-W118.C
118-W118.GR
119-ConEd.C
120-
ConEd.GR
121-Fdston.C
122-
Fdston.GR
123-Rain
124-USPS.GR
125-W118.GR
126- Rain
11/16/2011
11/16/2011
11/16/2011
11/17/2011
11/17/2011
11/17/2011
11/23/2011
11/23/2011
11/23/2011
12/7/2011
12/7/2011
12/7/2011
12/7/2011
12/7/2011
12/7/2011
12/7/2011
12/21/2011
1/12/2012
1/12/2012
2/29/2012
2/29/2012
2/29/2012
2/29/2012
2/29/2012
2/29/2012
2/29/2012
3/17/2012
3/17/2012
4/23/2012
5/1/2012
5/1/2012
5/9/2012
6.38
6.99
4.99
7.91
7.90
8.04
7.67
8.09
5.17
6.83
7.17
7.26
7.59
8.11
7.48
4.65
5.05
6.51
7.65
4.76
7.43
7.18
7.30
7.40
6.49
6.77
8.28
8.86
5.17
7.74
6.91
4.50
17.57
87.82
27.78
147.20
192.00
183.08
65.63
192.31
17.92
32.29
86.56
10.30
121.30
198.09
170.90
46.50
21.21
65.03
148.00
20.96
106.40
143.50
18.98
93.51
12.52
60.92
24.30
111.30
8.89
111.46
109.60
56.93
0.59
1.21
0.47
2.89
1.72
3.38
na
na
na
na
na
na
na
na
na
na
na
na
na
0.86
7.78
9.91
0.80
0.67
1.63
1.03
1.09
2.80
0.19
1.06
1.06
0.64
9.69
111.59
0.00
81.08
237.23
217.54
54.14
105.06
0.00
11.95
50.94
0.00
80.70
69.17
179.03
0.00
3.29
2.16
154.54
22.88
108.77
207.85
0.00
20.43
2.04
82.59
24.57
108.39
0.00
166.84
185.62
0.00
0.39
0.03
0.32
0.00
0.02
na
na
na
0.17
0.07
0.13
0.12
0.00
0.00
0.02
0.69
0.59
0.84
0.05
0.42
0.53
0.11
0.47
0.01
0.29
0.22
0.82
0.03
0.01
0.17
0.03
0.88
0.30
0.68
0.11
2.59
0.02
na
na
0.18
0.08
0.66
0.84
0.00
0.07
0.32
0.01
0.60
0.30
0.93
0.00
6.00
0.80
4.00
7.00
9.00
0.14
0.45
0.38
0.03
5.00
0.35
0.33
3.00
0.00
0.68
0.00
0.00
0.14
na
na
0.51
0.00
0.00
0.00
1.13
0.98
0.00
1.41
0.46
0.00
0.00
0.66
0.00
0.64
0.00
0.72
0.99
0.95
0.00
0.60
1.15
0.00
1.19
1.27
0.83
0.59
6.30
0.62
14.08
21.47
na
na
28.53
0.91
1.01
8.96
0.41
13.31
23.03
20.51
0.38
1.03
3.50
16.24
0.75
11.00
17.63
1.44
10.23
0.51
3.38
1.12
13.53
0.43
19.65
14.24
1.49
0.00
0.85
0.11
0.20
0.82
na
na
0.62
0.06
0.08
0.16
0.11
1.34
0.70
5.51
0.16
0.13
0.23
8.53
0.23
0.22
0.74
0.14
0.23
0.00
1.27
1.06
9.26
0.00
0.78
1.43
0.00
0.23
2.69
0.27
1.95
6.32
na
na
0.21
0.00
0.87
1.25
0.00
0.21
2.48
0.61
3.09
0.94
2.06
3.15
0.35
2.28
2.69
0.59
4.22
0.29
2.76
7.32
4.87
0.59
3.03
3.29
1.53
0.07
2.61
0.10
4.21
4.15
na
na
na
0.06
0.28
2.53
0.11
3.97
3.75
2.22
0.74
0.23
1.36
1.84
0.00
2.60
2.96
0.20
2.32
0.00
1.54
0.77
2.50
0.00
2.57
3.47
0.19
0.00
0.00
0.00
0.00
0.00
na
na
0.00
0.00
0.10
0.00
0.00
0.52
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.72
0.00
0.00
0.00
0.00
0.00
0.00
2.76
0.00
B 4
-------
127-W118.GR
128-W118.C
129-W118.GR
130-USPS.GR
131-W118.C
132-USPS.C
133-ConEd.C
134-W118.GR
135-Rain
136- Rain
137-USPS.GR
138-Rain
139-ConEd.C
140-
ConEd.GR
141-W118.C
142-W118.GR
143-W118.C
144-W118.GR
145- Rain
5/9/2012
5/15/2012
5/15/2012
6/2/2012
6/4/2012
6/4/2012
6/4/2012
6/4/2012
6/4/2012
6/13/2012
6/22/2012
7/20/2012
7/20/2012
7/20/2012
7/20/2012
7/20/2012
8/1/2012
8/1/2012
8/2/2012
6.67
6.77
6.79
7.60
7.35
7.51
7.39
7.22
na
na
7.51
na
7.08
6.65
6.78
7.11
6.77
7.26
5.91
106.90
8.84
74.51
120.35
13.10
129.30
19.60
61.03
9.23
59.62
176.88
64.93
76.71
142.00
20.92
159.90
59.44
150.90
53.30
0.67
0.33
1.03
2.23
0.75
3.63
0.57
3.93
0.56
0.67
9.69
0.58
0.28
1.85
0.81
5.38
1.70
0.76
0.54
90.87
8.18
95.96
78.97
0.00
0.00
0.00
98.79
0.00
0.00
73.73
0.00
10.07
243.26
12.59
368.71
82.02
237.42
0.00
0.03
0.10
0.11
0.39
0.02
0.01
0.38
0.16
0.12
0.58
3.23
na
na
na
na
na
na
na
na
0.06
11.00
0.24
0.60
0.14
1.29
0.37
0.18
0.00
0.36
4.93
na
na
na
na
na
na
na
na
0.00
0.00
0.34
0.53
0.26
0.00
1.10
0.00
1.45
0.00
1.24
na
na
na
na
na
na
na
na
13.83
0.41
6.21
13.59
0.70
14.26
1.05
6.22
0.74
1.05
15.44
na
na
na
na
na
na
na
na
0.23
0.64
0.72
1.08
0.39
2.38
2.64
11.63
0.79
0.00
5.82
na
na
na
na
na
na
na
na
6.68
1.25
3.17
1.99
0.90
na
0.35
1.51
1.59
0.35
3.39
na
na
na
na
na
na
na
na
3.60
0.00
2.05
2.13
0.00
10.62
0.20
1.88
0.77
0.00
2.38
na
na
na
na
na
na
na
na
4.94
0.00
1.32
0.04
0.16
0.00
0.00
1.35
0.00
0.00
0.00
na
na
na
na
na
na
na
na
Conductivity Turbidity Apparent Color
Avg±SD pH (uS/cm) (NTU) (PtCo) NO3 NH4+ P Ca K Na Mg B
GR
C
Rain
7.28 ±
0.51
6.27 ±
0.69
4.82 ±
0.39
127.67 ± 48.89
57.11 ±57.63
32.00 ±20.71
2.47 ±2.74
1.47 ± 1.48
0.62 ±0.39
162.53 ±90.24
28.45 ±32.42
5.32 ±9.79
0.27 ±
0.59
0.87 ±
1.31
0.60 ±
0.53
0.86 ±
1.86
1.47 ±
2.55
1.19 ±
1.85
0.47 ±
0.47
0.25 ±
0.38
0.21 ±
0.41
13.59 ±
6.8
3.93 ±
5.23
0.74 ±
0.50
2.22 ±
2.86
0.78 ±
1.98
0.10 ±
0.2
3.58 ±
3.47
1.80 ±
3.01
0.98 ±
0.88
2.92 ±
1.03
1.31 ±
2.30
0.20 ±
0.24
0.58 ±
1.19
0.03 ±
0.1
0.0 ±
0.0
B 5
-------
Appendix C Water Quality Analysis
The tables on the left on the following pages compare water quality within sample types (sample types = Rain, GR, and
C). The statistics for pH were calculated by transforming the measurements using this equation: [H+] = 10~pH Once
averaged, the [H+] value was converted back via this equation: pH = -log[H+]. Dark grey highlighted boxes indicate
comparisons among the different GRs, medium grey diagonal boxes compare GRand C within one site and white boxes
indicate comparisons among C roofs. Bolded and underlined p-values indicate significance. The smaller tables on the
right compare water quality between sample types. Other parameters that were measured but not listed below had no
significant difference between Rain, GR and C.
C 1
-------
Table 7-10: Summary of Statistical Analysis Results from Water Quality Monitoring (March 10, 2011 -August 2, 2012). One-way ANOVA tests were run. The tables
show the post-hoc Tukey HSD test to the seventh decimal place at the 0.05 level of significance.
PH
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
0.001
<0.001
<0.001
0.006
<0.001
W115
<0.001
0.227
0.988
0.988
0.997
0.996
W118
<0.001
1.000
0.001
1.000
1.000
1.000
USPS
<0.001
1.000
1.000
0.584
1.000
1.000
Fdston
<0.001
1.000
1.000
1.000
0.153
1.000
ConEd
<0.001
1.000
1.000
1.000
1.0000
0.151
Rain-C's: ANOVA F 5,59 = 11.12, p-value = <0.001
W115GR's-C's: ANOVA F 1,12 = 1.621, p-value = 0.227
W118GR's-C's: ANOVA F 1.36= 12.41, p-value= 0.001
USPSGR's-C's: ANOVA F 1,21 = 0.309, p-value= 0.584
FdstonGR's-C's: ANOVA F 1,7 = 2.577, p-value= 0.153
ConEdGR's-C's: ANOVA F 1,9 = 2.46, p-value= 0.151
Rain-GR's- ANOVA F 566 = 14 47 p-value = <0 001
PH
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
<0.001
<0.001
0.878
Rain-All GR-AII C:
ANOVA F 2,113 = 65.87
p-value <0.001
C 2
-------
Conductivity
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain-C's: AN(
W115GR's-C
W118GR's-C
USPSGR's-C
FdstonGR's-C
ConEdGR's-C
Rain-GR's: A
Rain
.
0.984
0.869
<0.001
1.000
1.000
W115
<0.001
0.002
1.000
0.013
0.997
0.998
W118
<0.001
0.076
0.003
0.004
0.985
0.987
USPS
<0.001
0.711
<0.001
0.003
0.021
0.008
Fdston
<0.001
0.998
0.340
0.436
<0.001
1.000
ConEd
<0.001
0.418
1.000
0.009
0.740
0.010
DVA Fs.ei = 5.404, p-value = <0.001
s: ANOVA F 1,12 = 15.57, p-value = 0.002
s: ANOVA F 1,35 = 9.905, p-value = 0.003
's: ANOVA F 1,21 = 10.96, p-value = 0.003
ys: ANOVA F 1,7 = 34.81 , p-value = <0.001
ys: ANOVA F 1,9 = 10.74, p-value = 0.010
NOVA F 5,69 = 32.47, p-value = <0.001
Conductivity
Turbidity
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
0.920
0.609
1.000
1.000
1.000
W115
0.811
0.294
1.000
<0.001
0.989
0.991
W118
0.770
0.999
0.404
<0.001
0.963
0.938
USPS
<0.001
0.092
<0.001
0.462
0.003
<0.001
Fdston
0.983
1.000
1.000
0.133
0.419
1.000
ConEd
0.967
0.999
1.000
0.032
1.000
0.074
Rain-C's: ANOVA F 5,50= 10.59, p-value = <0.001
W1 15GR's-C's: ANOVA F 1,10 = 1.227, p-value = 0.294
W1 1 8GR's-C's: ANOVA F 1,30 = 0.71 8, p-value = 0.404
USPSGR's-C's: ANOVA F 1,17 = 0.567, p-value = 0.462
FdstonGR's-C's: ANOVA F 1,3 = 0.874, p-value = 0.419
ConEdGR's-C's: ANOVA F 1,9 = 4.092, p-value = 0.074
Rain-GR's: ANOVA F 5,57 = 6.699, p-value = <0.001
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
<0.001
0.104
<0.001
Rain-All GR-AII C:
ANOVA F 2,115 = 43.78
p-value = <0.001
Turbidity
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.004
0.293
0.099
Rain-All GR-AII C:
ANOVA F 2,96 = 5.82
p-value = 0.004
C 3
-------
Apparent Color
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
0.399
0.267
<0.001
1.000
0.998
W115
0.058
0.021
0.998
0.768
0.842
0.777
W118
<0.001
1.000
<0.001
0.207
0.911
0.840
USPS
<0.001
0.911
0.719
0.003
0.129
0.034
Fdston
0.009
1.000
1.000
0.955
<0.001
1.000
ConEd
0.004
1.000
1.000
0.927
1.000
<0.001
Rain-C's: ANOVA F 5,49= 4.934, p-value = <0.001
W115GR's-C's: ANOVA F 1,5 = 11.14, p-value = 0.021
W1 1 8GR's-C's: ANOVA F 1,24 = 1 8.35, p-value = <0.001
USPSGR's-C's: ANOVA F 1,21 = 11.54, p-value = 0.003
FdstonGR's-C's: ANOVA F 1,5 = 54.43, p-value = <0.001
ConEdGR's-C's: ANOVA F 1,9 = 24.82, p-value = <0.001
Rain-GR's: ANOVA F 5,57 = 12.87, p-value = <0.001
Apparent Color
Nitrate
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
0.934
0.405
0.980
1.000
0.998
W115
0.927
0.165
0.997
0.748
0.990
0.894
W118
0.240
0.999
0.023
0.284
0.861
0.537
USPS
0.954
0.999
0.857
0.684
0.993
1.000
Fdston
0.954
1.000
0.988
1.000
0.486
0.999
ConEd
0.574
0.995
1.000
0.910
0.981
0.024
Rain-C's: ANOVA F 5,49= 1.378, p-value = 0.249
W115GR's-C's: ANOVA F 1,9 = 2.283, p-value = 0.165
W118GR's-C's: ANOVA F 1,27= 5.858, p-value = 0.023
USPSGR's-C's: ANOVA F 1,17 = 0.172, p-value = 0.684
FdstonGR's-C's: ANOVA F 1,7 = 0.541 , p-value = 0.486
ConEdGR's-C's: ANOVA F 1,7 = 8.229, p-value = 0.024
Rain-GR's: ANOVA F 5,54 = 1.253, p-value = 0.298
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
<0.001
0.390
<0.001
Rain-All GR-AII C:
ANOVA F 2,93 = 60.52
p-value = <0.001
Nitrate
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.377
0.571
0.013
Rain-All GR-AII C:
ANOVA F 2,93 = 4.189
p-value = 0.018
C 4
-------
Ammonium
Ammonium
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain-C's: AN(
W115GR's-C
W118GR's-C
USPSGR's-C
FdstonGR's-C
ConEdGR's-C
Rain-GR's: A
Rain
-
0.999
0.678
0.994
0.995
0.992
W115
0.971
0.382
0.725
0.978
1.000
1.000
W118
1.000
0.984
0.356
0.967
0.700
0.570
USPS
1.000
0.976
1.000
0.471
0.960
0.936
Fdston
0.894
1.000
0.943
0.915
0.160
1.000
ConEd
0.993
1.000
0.998
0.995
0.999
0.818
DVA Fs,36= 0.856, p-value = 0.520
s: ANOVA F 1,4 = 0.964, p-value = 0.382
s: ANOVA F 1,15 = 0.906, p-value = 0.356
's: ANOVA F 1,17 = 0.543, p-value = 0.471
ys: ANOVA F 1,5 = 2.71 9, p-value = 0.1 60
ys: ANOVA F 1,7 = 0.057, p-value = 0.818
NOVA F 5,42 = 0.373, p-value = 0.864
Calcium
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
0.990
1.000
<0.001
0.977
1.000
W115
0 006
0.003
0.999
<0.001
1.000
1.000
W118
<0.001
0.996
<0.001
<0.001
0.994
1.000
USPS
<0.001
0.013
<0.001
0.003
<0.001
<0.001
Fdston
<0.001
0.485
0.024
0.384
0.001
0.997
ConEd
0.127
0.598
0.547
<0.001
0.002
<0.001
Rain-C's: ANOVA F 5,35= 31, p-value = <0.001
W115GR's-C's: ANOVA F 1,4= 43.8, p-value = 0.003
W1 1 8GR's-C's: ANOVA F 1,15 = 31 .69, p-value = <0.001
USPSGR's-C's: ANOVA F 1,17 = 1 1 .57, p-value = 0.003
FdstonGR's-C's: ANOVA F 1,5 = 42.41 , p-value = 0.001
ConEdGR's-C's: ANOVA F 1,7 = 32.06, p-value = <0.001
Rain-GR's: ANOVA F 5,42 = 45.46, p-value = <0.001
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.868
0.911
0.526
Rain-All GR-AII C:
ANOVA F 2,71 = 0.594
p-value = 0.555
Calcium
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
<0.001
0.166
<0.001
Rain-All GR-AII C:
ANOVA F 2,71 = 37.7
p-value = <0.001
C 5
-------
Potassium
Potassium
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
1.000
1.000
0.117
0.999
0.991
W115
0.999
0.080
1.000
0.463
1.000
0.999
Rain-C's: ANOVA F 5,35 = 1.503, p-value
W115GR's-C's: ANOVA F 1.4 = 5.463, p-
W1 1 8GR's-C's: ANOVA F 1,15 = 1 .674, p
USPSGR's-C's: ANOVA F 1,17 = 0.174, p
FdstonGR's-C's: ANOVA F 1,5 = 1 1 .99, p
ConEdGR's-C's: ANOVA^^^^TjD
Rain-GR's: ANOVA F 5,42 = 6.924, p-valu
W118
0.214
0.957
0.215
0.333
1.000
0.999
USPS
0.498
0.995
0.991
0.682
0.739
0.658
Fdston
<0.001
0.016
0.004
<0.001
0.018
1.000
ConEd
0.721
0.991
1.000
1.000
0.014
0.240
= 0.213
value = 0.080
(-value = 0.215
-value = 0.682
-value = 0.018
-value = 0.240
e = <0.001
Magnesium
Rain-C's: ANOVA F 5,35= 10.06, p-value = <0.001
W115GR's-C's: ANOVA F 1,4 = 555.1, p-value = <0.001
W118GR's-C's: ANOVA F 1,15 = 35.88, p-value = <0.001
USPSGR's-C's: ANOVA F 1,16 = 0.548, p-value = 0.470
FdstonGR's-C's: ANOVA F 1,5 = 10.44, p-value = 0.023
ConEdGR's-C's: ANOVA F 1,7 = 25.19, p-value = 0.002
Rain-GR's: ANOVA F 5.41 = 25.91, p-value = <0.001
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
1.000
1.000
<0.001
0.984
1.000
W115
0.013
<0.001
1.000
<0.001
0.997
1.000
W118
<0.001
0.841
<0.001
<0.001
0.984
1.000
USPS
<0.001
0.595
0.976
0.470
0.009
<0.001
Fdston
0.003
0.998
0.266
0.084
0.023
0.996
ConEd
<0.001
0.998
0.943
0.687
0.890
0.002
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.008
0.611
0.044
Rain-All GR-AII C:
ANOVA F 2,71 = 5.725
p-value = 0.005
Magnesium
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
<0.001
0.065
<0.001
Rain-All GR-AII C:
ANOVA F 2,70 = 18.15
p-value = <0.001
C 6
-------
Phosphorus
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
-
0.916
1.000
1.000
0.664
0.917
W115
0.998
0.178
0.956
0.964
0.380
0.611
Rain-C's: ANOVA F 5,35 = 1.015, p-value
W1 1 5GR's-C's: ANOVA F 1.4 = 2.667, p-i
W1 1 8GR's-C's: ANOVA F 1,15 = 1 .362, p
USPSGR's-C's: ANOVA F 1,17 = 1.5, p-v
FdstonGR's-C's: ANOVA F 1,5 = 0.247, p
ConEdGR's-C's: ANOVA^^=(Hmj:)
Rain-GR's: ANOVA F 5,42 = 1.173, p-valu
W118
0.807
1.000
0.261
1.000
0.722
0.941
USPS
0.837
1.000
1.000
0.237
0.703
0.930
Fdston
0.213
0.874
0.769
0.699
0.640
0.991
ConEd
0.990
1.000
1.000
1.000
0.760
0.911
= 0.423
/alue = 0.178
(-value = 0.261
alue = 0.237
-value = 0.640
-value = 0.911
e = 0.339
Phosphorus
Boron
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
1.000
0.027
1.000
1.000
1.000
W115
1.000
NaN
0.198
1.000
1.000
1.000
W118
<0.001
0.017
0.009
0.087
0.289
0.144
USPS
1.000
1.000
<0.001
0.461
1.000
1.000
Fdston
1.000
1.000
<0.001
1.000
NaN
1.000
ConEd
1.000
1.000
<0.001
1.000
1.000
NaN
Rain-C's: ANOVA F5,36= 2.551, p-value = 0.045
W1 1 8GR's-C's: ANOVA F 1,15 = 9.091 , p-value = 0.009
USPSGR's-C's: ANOVA F 1,17 = 0.569, p-value = 0.461
Rain-GR's: ANOVA F 5,42 = 11.1, p-value = <0.001
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.134
0.955
0.145
Rain-All GR-AII C:
ANOVA F 2,71 = 2.671
p-value = 0.076
Boron
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.050
0.993
0.027
Rain-All GR-AII C:
ANOVA F 2,71 = 4.607
p-value = 0.013
C 7
-------
Sodium
Sodium
Site
Rain
W115
W118
USPS
Fdston
ConEd
Rain
.
1.000
1.000
0.230
0.693
1.000
W115
0.949
0.045
1.000
0.506
0.806
1.000
W118
0.003
0.815
0.042
0.305
0.697
1.000
USPS
0.682
1.000
0.157
0.553
1.000
0.523
Fdston
0.819
1.000
0.619
1.000
0.921
0.836
ConEd
0.831
1.000
0.604
1.000
1.000
0.079
Rain-C's: ANOVAF5,35= 1.571, p-value = 0.194
W115GR's-C's: ANOVA F 1,4 = 8.349, p-value = 0.045
W118GR's-C's: ANOVA F 1,15 = 4.96, p-value = 0.042
USPSGR's-C's: ANOVA F 1 ,16 = 0.367, p-value = 0.553
FdstonGR's-C's: ANOVA F 1 ,5 = 0.01 1 , p-value = 0.921
ConEdGR's-C's: ANOVA F 1 ,7 = 4.228, p-value = 0.079
Rain-GR's: ANOVA F 5,42 = 3.246, p-value = 0.01 4
Sample
Rain-All GR
Rain-All C
All GR-AII C
p-value
0.014
0.658
0.066
Rain-All GR-AII C:
ANOVA F 2,70 = 4.998
p-value = 0.009
C 8
------- |