Designing Green Roofs for Low Impact Development: What ... · Past and current practices in green...

Designing Green Roofs for Low Impact Development:

What Matters, and Why?

by

Jenny Charlotte Hill

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Graduate Department of Civil Engineering University of Toronto

© Copyright by Jenny Hill 2017

ii

Designing Green Roofs for Low Impact Development:

What Matters, and Why?

Jenny Charlotte Hill

Doctor of Philosophy

Graduate Department of Civil Engineering

University of Toronto

2017

Abstract

This thesis assesses the performance of green roofs primarily as hydrologic systems and as components in

biogeochemical cycles; asking: What are they made from? Which design parameters are most influential?

What are the relative impacts? Can the findings be explained?

A multivariate experiment (n = 24) at the Green Roof Innovation Testing laboratory (GRITlab) between

May 2013 - April 2015 revealed that irrigation has the greatest effect on annual runoff coefficient,

compared to type of planting medium or planting depth (�̅� = 0.5). Switching between Sedum. or native

species planting made no significant difference. NRCS curve numbers (�̅� = 90) and peak runoff

coefficients (�̅� = 0.1) were considered robust, unchanging with any of the design factors.

Water extractable total phosphorus in 3.5-year-old media had been unaffected by irrigation, depth or

planting compared to the overall difference between the compost or mineral basis (90 ppm and 46 ppm

respectively). Electrical conductivity was higher in water discharged from the mineral media; in situ

measurements are highly variable, complicated by the heterogeneity of the materials. Higher

concentrations of humic acids were found at in the water discharged from the compost.

The water retention curve (WRC) of ten media components and mixtures were used to explore the bi-

modal distribution of inter-particle voids and intra-particle pore spaces and to explain why the non-linear

storage capacity of the materials. The water retention capacity was inversely related to saturated

iii

hydraulic conductivity; both macroscale properties were highly dependent on the size of the lowest decile

fraction of particle sizes. Media components with high organic matter content were assessed for

wettability using contact angle measurements.

Past and current practices in green roof construction were considered by sampling media from thirty-three

green roofs. Most planting media were compost based with high organic matter (OM), or mineral based

with very low OM. Bulk density, particle density and porosity were all dependent on OM, as were the

hydrological properties of water retention capacity and permeability. An average 10% loss of depth was

observed across all installations regardless of their age or organic matter content.

iv

Acknowledgments

Firstly, a big shout out to D.Dub, Primary Sponsor of my dreams (again) and to my parents who

successfully taught me that I could be anything (whilst failing to mention I couldn’t be everything).

Sincere thanks to both of my supervisors, and excellent support team: Dr. Jennifer Drake who tolerated

my most frustrated and frustrating moments and Dr. Brent Sleep who invited me to join the school and

refrained from expelling me in early 2016. I am also indebted to Dr. Bryan Karney, for the existential

crisis incited to make my research more meaningful, and to Professor Liat Margolis, who had the

extraordinary vision and energy to create the GRITlab.

It has been a great joy reconnecting with Terry McGlade, who successfully brought in Flynn Canada as

sponsors. Every week I looked forward to the industry gossip and coffee. At Flynn I also valued highly

the practical advice of Becky Murphy and colleagues, and appreciated the interest that Mark Agius placed

in the partnership. I must also acknowledge the Natural Sciences and Engineering Research Council of

Canada (NSERC) for providing the federal scholarship that supported three years of this work.

At the GRITlab and in Civil Engineering I have taken great pleasure in sharing the sunshine with (and

benefitted from research collaboration with): Matt, Catherine, Eli, Gabrielle, Allan, Michael, Humberto,

Raquel and Scott. The GRITlab is sponsored by: DH Water Management Services Inc., GroBark, IRC

Group, Toro, and Tremco Roofing and supported by grant funding from the City of Toronto Environment

Office, Ontario Centers of Excellence, RCI Foundation, the Connaught Fund and the Landscape

Architecture Canada Foundation.

I am delighted to have received many interesting opinions and useful guidance from the professional

friends I acquired in environmental engineering and allied industries during my studies. Finally, I am

grateful to all of the owners and custodians of green roofs for facilitating physical access and in sharing

their insights.

v

Table of Contents

ABSTRACT ............................................................................................................................................... II

ACKNOWLEDGMENTS ........................................................................................................................ IV

TABLE OF CONTENTS ........................................................................................................................... V

LIST OF TABLES .................................................................................................................................... IX

: INTRODUCTION ......................................................................................................1

RESEARCH OBJECTIVES ...............................................................................................................7

Context .....................................................................................................................................7

Objectives ................................................................................................................................8

BACKGROUND ............................................................................................................................10

The Green Roof as a Reservoir ..............................................................................................10

The Green Roof as an Orifice ................................................................................................18

The Green Roof as an Evaporation Pan ................................................................................19

THESIS ORGANIZATION ..............................................................................................................21

AUTHORSHIP ..............................................................................................................................22

: INFLUENCES OF FOUR EXTENSIVE GREEN ROOF DESIGN VARIABLES

ON STORMWATER HYDROLOGY .....................................................................................................23

INTRODUCTION ...........................................................................................................................24

METHODS ...................................................................................................................................26

Green Roof Innovation Testing laboratory ............................................................................26

Theory and Calculations........................................................................................................28

RESULTS AND DISCUSSION ........................................................................................................30

Local Climate ........................................................................................................................30

Volumetric Runoff Coefficients ..............................................................................................31

Event-based Analysis .............................................................................................................36

Peak Flow ..............................................................................................................................37

CONCLUSIONS ............................................................................................................................39

: INFLUENCES OF FOUR EXTENSIVE GREEN ROOF DESIGN VARIABLES

ON ANNUAL WATER BALANCE ........................................................................................................41

INTRODUCTION ...........................................................................................................................42

METHODS ...................................................................................................................................44

vi

Green Roof Innovation Testing laboratory ............................................................................44

Theory and Calculations........................................................................................................45


Irrigation and Water Retention .............................................................................................46

Winter Climate and Snow Accumulation ...............................................................................48

Winter Cvol ..............................................................................................................................54

Annual Cvol .............................................................................................................................56

CONCLUSIONS ............................................................................................................................57

: PHYSICOCHEMICAL PROPERTIES OF EXTENSIVE GREEN ROOF

PLANTING MEDIA .................................................................................................................................59

INTRODUCTION ...........................................................................................................................60

BACKGROUND ............................................................................................................................60

Phosphorus ............................................................................................................................60

Electrical Conductivity ..........................................................................................................62

METHODS ...................................................................................................................................64

Green Roof Experimental Set up ...........................................................................................64

Phosphorous ..........................................................................................................................64



Phosphorous ..........................................................................................................................67


CONCLUSIONS ............................................................................................................................74

: THE INFLUENCE OF DEPTH AND POROSITY ON THE HYDRAULIC

PROPERTIES OF GREEN ROOF PLANTING MEDIA ....................................................................76

INTRODUCTION ...........................................................................................................................77

METHODS ...................................................................................................................................80

Medium property measurements ............................................................................................80

Water Retention Parameters ..................................................................................................81


Density and porosity ..............................................................................................................83

WRC parameters ....................................................................................................................85

System water storage .............................................................................................................92

Hydrophobicity, wetting and shrink/swell characteristics.....................................................95

vii

CONCLUSIONS ............................................................................................................................96

: COMPARISONS OF EXTENSIVE GREEN ROOF MEDIA IN SOUTHERN

ONTARIO 98

INTRODUCTION ...........................................................................................................................98

METHODS .................................................................................................................................100

System Properties ................................................................................................................101

Physical Properties ..............................................................................................................102

Chemical Properties ............................................................................................................103

RESULTS AND DISCUSSION ......................................................................................................104

Age of installation ................................................................................................................104

Particle composition ............................................................................................................105

Planting medium/Water Interactions ...................................................................................107

Chemistry .............................................................................................................................116

CONCLUSIONS ..........................................................................................................................117

: CONCLUSIONS .....................................................................................................118

THE EXTENSIVE GREEN ROOF AS A RESERVOIR ......................................................................118

Irrigation .............................................................................................................................118

Planting Medium .................................................................................................................120

Depth ...................................................................................................................................120

Planting type ........................................................................................................................121

THE EXTENSIVE GREEN ROOF AS AN ORIFICE .........................................................................121

THE EXTENSIVE GREEN ROOF AS AN EVAPORATION PAN .......................................................122

FURTHER WORK .......................................................................................................................124

Irrigation .............................................................................................................................124

Cisterns ................................................................................................................................125

Nutrition versus pollution ....................................................................................................125

Development of Organic Matter ..........................................................................................125

FINAL COMMENTS: THE ‘BEST’ EXTENSIVE GREEN ROOF? ....................................................125

: REFERENCES .......................................................................................................128

APPENDIX A: GLOSSARY ..................................................................................................................150

APPENDIX B: DATA RELATING TO CHAPTER 2 .........................................................................154

APPENDIX C: DATA RELATING TO CHAPTER 3 ........................................................................166

viii

APPENDIX D: DATA RELATING TO CHAPTER 4 ........................................................................169

APPENDIX E: DATA RELATING TO CHAPTER 6 .........................................................................171

APPENDIX F: SUCCESS AND SUCCESSION...................................................................................174

Introduction ......................................................................................................................................175

APPENDIX G: GONE WITH THE WIND ..........................................................................................186

ix

List of Tables

Table 1-1 Summary of green roof papers resulting in SCS curve numbers. *Getter et al (2007) were

studying effect of slope, hence the range of CN reflecting 2% to 25% slope. ............................................13

Table 2-2 Physical data for grit planting media, according to manufacturer’s ASTM 2399 report (Bioroof

Systems, 2011) ............................................................................................................................................27

Table 4-1 Levels of four experimental variables being considered at the GRITlab ....................................64

Table 4-2 Subset of green roof modules tested for TP in discharged water ................................................65

Table 4-3 Group mean volumetric runoff coefficients for six extensive green roof design combinations.

*n=2, encompassing both types of vegetation, apart from case E, where n = 1 (meadow planting only). .69

Table 5-1 Identity and shared sources of ten sample materials for analysis and comparison ....................80

Table 5-2 Density, porosity and organic matter content of ten porous test materials. ................................85

Table 5-3 the van Genuchten parameters from the fitted curves arising from the evaporative drying of ten

test materials ................................................................................................................................................87

Table 5-4 System static and dynamic air and water properties for ten samples ..........................................93

Table 5-5 Dynamic contact angle data from the analysis of the biologically derived materials E-G .........96

Table 6-1 Equations used to summarize physical characteristics of the porous media. ............................103

Table 6-2 Six independent and fourteen dependant variable measured on the surveyed roofs. ................104

Table 6-3 Chemistry of water extracts prepared from thirty three green roof media samples. .................116

Table 0-1 Details of study green roofs in Toronto ....................................................................................176

Table 0-2 Selected planting details study 2 Genus only identified where the species is unknown (in a

proprietary seed mixture) or where multiple species have been used. ......................................................177

Table 0-1 Combinations of erosion control measures and planting methods. Red not recommended,

yellow may present some difficulty, green represents recommended combinations. ...............................189

x

List of Figures

Figure 1-1 Urbanized population, data from (United Nations, Department of Economic and Social Affairs,

2014) ..............................................................................................................................................................1

Figure 1-2 Where we live: The two densest urban centres are the Greater Toronto Area in the lower centre

of image and Montreal in the top right (Simmon, 2012) ...............................................................................2

Figure 1-3 The area of combined sewers within the City of Toronto, from (Di Gironimo et al., 2013). ......2

Figure 1-4 The huge increase (55 % from 10 %) in runoff water resulting from storm events on a post-

development landscape with a high degree of pervious cover. Image abridged from (Federal Interagency

Stream Restoration Working Group (FISRWG), 1998) .................................................................................3

Figure 1-5 The intersection and overlap between Low Impact Development (left) and Green Infrastructure

(right) .............................................................................................................................................................5

Figure 1-6 The rise in 'green roof' articles from Figure 1 in a hydrological review paper by Li and

Babcock (2014) .............................................................................................................................................7

Figure 1-7 Summary of previous studies assessing volumetric retention: A (Teemusk and Mander, 2007);

B (Cronk, 2012); C (Hilten et al., 2008; Prowell, 2006); D (Schroll et al., 2011); E (Berghage et al.,

2010); F (Gregoire and Clausen, 2011); G (VanWoert et al., 2005); H (Ma et al., 2012); I (Moran et al.,

2004); J (Van Seters et al., 2009); K (Hathaway et al., 2008); L (Voyde et al., 2010); M (Uhl and Schiedt,

2008); N (Starry, 2013); O (Carter and Rasmussen, 2006); P (Burszta-Adamiak, 2012); Q (Getter et al.,

2007); R (Palla et al., 2012).........................................................................................................................10

Figure 1-8 Toronto rainstorm depth distribution, from 1937 – 1983 Bloor Strreet rain gauge data. ..........11

Figure 1-9 Conceptual relationship between maximum theoretical storage capacity (mm) and annual Cvol

.....................................................................................................................................................................12

Figure 1-10 Derived from (e.V., 2008), with the caption “All figures relate to locations with annual

precipitation values of 650 – 800 mm where monitoring has been performed over a period of several

years.” ..........................................................................................................................................................16

Figure 1-11 Conceptual illustration of effects on hydrograph ....................................................................18

Figure 1-12 Conceptual closed system model combining a green roof with a cistern ................................19

xi

Figure 1-13 One part of the irrigation system on the Rottman School of Management, University of

Toronto. This system uses recycled and/or harvested rainwater to irrigate extensive green roofs. ............20

Figure 2-1 Typical layering of a built-up extensive green roof system .......................................................24

Figure 2-2 Schematic of GRITlab, illustrating the randomized layout of the four experimental variables.

Key - colours in each rectangular module can be read from west to east. Vegetation: dark = Meadow,

light = Sedum. Planting medium: dark = biological, light = mineral. Irrigation: dark = daily, light =

sensor, mid = none. Construction depth: dark = 15 cm, light = 10 cm. ......................................................26

Figure 2-3 Local weather at GRITlab, Toronto between May-October 2013 and May-October 2014, the

duration of the green roof study. .................................................................................................................30

Figure 2-4 Annual exponential cumulative distribution of storm depths in Toronto, ON. According to

2013 and 2014 GRITlab data and historical records (1937-1983) from Bloor St. ......................................31

Figure 2-5 Regression tree for the runoff coefficients (Cvol) determined on twenty-three extensive green

roofs over 12 summer months encompassing May-October 2013, and May-October 2014 . .....................32

Figure 2-6 Monthly group mean volumetric runoff coefficients for 23 green roof modules during the

periods between May-October 2013 and May-October 2014. ....................................................................33

Figure 2-7 Monthly group mean volumetric runoff coefficients for four design factors during the during

the periods between May-October 2013 and May-October 2014. ..............................................................34

Figure 2-8 Box plot of volumetric runoff coefficients over rainstorm events in 2013 and 2014, grouped

according to medium type and antecedent volumetric water content over the range 0 – 0.55 v/v. Group

means indicated with ‘X’ and connected within the medium type. .............................................................36

Figure 2-9 Regression tree for the NRCS Curve Numbers determined on twenty three extensive green

roofs over 12 summer months encompassing May-October 2013, and May-October 2014. ......................37

Figure 2-10 Validation of peak based runoff coefficients using Rational method to calculate peak flow

(Qp) and compared to experimental data for twelve, 2015 rainstorm events. Error bars represent the

standard deviation of all twenty-four module’s peak flows per event. ........................................................38

Figure 3-1 GRITlab modules raised above the roof deck to accommodated monitoring equipment. .........44

xii

Figure 3-2 Extensive green roof annual total water retention for months October 2013 –September 2014,

grouped by irrigation program.....................................................................................................................47

Figure 3-3 Monthly water retained group means for three levels of irrigation between October 2013 to

September 2014. Reference ET from the GRITlab weather station. ...........................................................47

Figure 3-4 Input and output volumes associated with two irrigation programs ..........................................48

Figure 3-5 Winter months climate normal snow cover, daily minimum temperatures and daily

precipitation depth from 1981-2010 data in Toronto, Ontario (Environment Canada, 2013). ....................49

Figure 3-6 Mean daily air temperature (dashed line) from GRITab and precipitation record (bars) from

Toronto City weather station for the periods encompassing November 2013 to April 2014, and November

2014 to April 2015. .....................................................................................................................................50

Figure 3-7 Twenty-four modules accumulated snow depth throughout winters 2013-14 and 2014-15,

plotted over ground level data. ....................................................................................................................51

Figure 3-8 Moran’s I from winter 2013-14 centred about zero, indicate no significant geospatial clustering

or trends in the snow depth across the GRIT lab experiment. .....................................................................52

Figure 3-9 Mean snow depth, grouped by medium type (top), and irrigation (bottom) throughout winters

2013 and 2014. ............................................................................................................................................53

Figure 3-10 Native meadow vegetation mix grown on: a) biological medium with daily irrigation, b)

mineral medium with daily irrigation, and c) mineral medium without irrigation. Photographs taken 20

September 2013 (University of Toronto, 2013). .........................................................................................54

Figure 3-11 Mean volumetric runoff coefficients from 23 modules, over 12 months of summertime events

May-Oct in 2013 and 2014 and 12 months of wintertime balance, Nov-April in 2013-2014 and 2014-

2015. ............................................................................................................................................................54

Figure 3-12 Group mean runoff coefficients per month through May 2013 to April 2015. .......................55

Figure 3-13 Group mean runoff coefficients by irrigation program, for months through November – April

2013-14 and 2014-15. ..................................................................................................................................56

xiii

Figure 3-14 Annual volumetric runoff coefficients for extensive green roofs, calculated from 24 months

of data spanning May 2013- April 2015. Each cell contains: Design factor ‘level’, group mean value, and

(# modules). .................................................................................................................................................57

Figure 4-1 Summary of previous studies assessing the total phosphorous discharge from extensive green

roofs: A (Gregoire and Clausen, 2011); B (Toland, 2010); C (Berndtsson et al., 2006); D (Teemusk and

Mander, 2007); E (Van Seters et al., 2009); F (Harper et al., 2015); G (Beck et al., 2011). Many of the

mixtures contain lightweight expanded aggregate (LEA). ..........................................................................61

Figure 4-2 Summary of previous studies which state the electrical conductivity of discharge from

extensive green roofs: A (Beecham and Razzaghmanesh, 2015); B (Gnecco et al., 2013); C (Göbel et al.,

2007); D (Buffam et al., 2016); E (Van Seters et al., 2009); F (Buccola and Spolek, 2011) .......................63

Figure 4-3 The water extractable total phosphorous in twenty-four, 3.4-year-old green roof modules is

distinguished only by the type of planting medium. ...................................................................................68

Figure 4-4 TP in discharge water from six green roof modules. ................................................................68

Figure 4-5 Regression tree illustrating the relative influence of three design factors on the TP

concentrations in samples taken March/April 2016. ...................................................................................69

Figure 4-6 The influence of green roof medium type on the physicochemical parameters, pH and

electrical conductivity. ................................................................................................................................71

Figure 4-7 Calibration of 5TE sensor in biological planting medium (top), and mineral based medium

(bottom) .......................................................................................................................................................72

Figure 4-8 Range of ε0 in eleven green roof modules containing bioloigcally derived planting medium

(left), and mineral based green roof planting medium (right). ....................................................................73

Figure 4-9 Irrigation makes a more significant impact on pore water electrical conductivity in April 2016,

than any other design factor: planting medium type, depth or planting type. .............................................74

Figure 5-1 Green roof matric pressure as a function of medium depth under static equilibrium with

maximum water storage. Where θ = volumetric water content, and θs = saturated volumetric water content

.....................................................................................................................................................................78

Figure 5-2 Drying curve data from the analysis of ten samples. Grey circles are raw data, lines are the

fitted curves: Bulk materials A, C, E, and G are grouped as having significant (w1 > 0.9) weighting on the

xiv

inter-particle voids (top); bulk materials B, D, and F are grouped as having distinctly separate and more

evenly weighted van Genuchten parameters (middle); blended materials H, I, and J (bottom) ..................86

Figure 5-3 The largely unimodal pore size distributions (line) plotted over the particle size distributions

(bars) found in: A: Sand, C: Poorly-graded LEA, E: ¼” Screened composted wood, and G: Shredded

Pine. .............................................................................................................................................................88

Figure 5-4 The largely bimodal pore size distributions (line) plotted over the particle size distributions

(bars) found in: B: Well-graded LEA, D: Crushed brick, and F: Bark fines. ..............................................89

Figure 5-5 Surface detail visible under 100x magnification: left) B: LEA, centre) D: Brick particle, right)

F: Bark fragment .........................................................................................................................................90

Figure 5-6 The pore size distributions (line) plotted over the particle size distributions (bars) found in

commercial green roof planting media blends: H: Compost based - Manufacturer A, I: Mineral based -

Manufacturer A, and, J: Mineral based - Manufacturer B. ..........................................................................91

Figure 5-7 Modelled water storage in three 5 cm increments of green roof profile depth, for seven bulk

materials (A-G) and three commercial blended materials (H-J). ................................................................92

Figure 5-8 Regresison trees for prediction of container capacity (θ(h15)) and wilting point (θ(h15296), from

predictors ρd, ϕ, and OM. ............................................................................................................................94

Figure 5-9 Binary image from x-ray of material I particle (left), results of surface fractal analysis to show

the network of connected pores (right) ........................................................................................................95

Figure 6-1: Age of thirty-three green roofs at the time of surveying and sampling. .................................100

Figure 6-2 Schematic (not to scale) and photograph of the infiltrometer used for in situ measurements .102

Figure 6-3 Organic matter content of plating media recovered from thirty-three extensive green roofs;

roofs are alphabetical from oldest to most recently constructed, the dashed line crosses at 8 % ..............106

Figure 6-4 Multifactor box plot of bulk and solid particle densities, divided between low (< 30%) and

high (≥ 30%) OM content..........................................................................................................................107

Figure 6-5 Relationship between maximum water content and organic matter content of green roof media

...................................................................................................................................................................108

xv

Figure 6-6 Multifactor box plot of particle size distribution coefficients, divided between low (< 30%)

and high (≥ 30%) organic matter content. .................................................................................................110

Figure 6-7 Particle size distirbution curves from green roof planting media recovered from green roofs

with CU ≈ 16. Dashed line media CC = 0.5; solid line media CC = 5.5. .....................................................111

Figure 6-8 Regression tree of MWC demonstrating the relative importance of OM and interaction with

particle size parameters CU and CC. ...........................................................................................................112

Figure 6-9 Relationship between free air space and organic matter content in green roof planting media

...................................................................................................................................................................113

Figure 6-10 Relationship between infiltration and permeability rates in eighteen green roof media samples

...................................................................................................................................................................115

Figure 7-1 GRITlab module 6E, 23 June 2015 .........................................................................................126

Figure 7-2 Toronto Botanical Garden Extensive Green Roof, 28 May 2014 ............................................127

Figure 0-1 Stakeholder rankings of the importance of green roof functions (n=7). ..................................179

Figure 0-2 Earth Rangers Southern roof: 2005, after 2 years establishment (left), and 2013 (right). .......180

Figure 0-3 George Vari Engineering Building roof, 2013 (left), and 2014 (right)....................................181

Figure 0-4 Toronto Botanical Garden sloped section, 2006 (left) and 2014 (right). .................................181

Figure 0-5 Arts and Administration green roof, University of Toronto, 2005 (left) and 2013 (right). .....182

Figure 0-6 Royal Ontario Museum scorched section detail, 2013 (left) and 2014 (right) ........................183

Figure 0-7 Depth of planting substrate on eight green roofs .....................................................................184

Figure 0-1. Preparation of pre-grown Sedum ‘mats’ (left), root penetration after two years growth on a

green roof (right). ......................................................................................................................................188

Figure 0-2. Anchors to retain pre grown mats in high wind velocity situations. ......................................188

Figure 0-3 Tenting of polymer mesh over native wildflower seed mixture on an extensive green roof. ..190

xvi

Figure 0-4. Evaporation of water from; a) psyllium husk amended compost, b) Polyacrylamide amended

compost. ....................................................................................................................................................192

Figure 0-5 Change in percolation rate after replicated measurements in a) psyllium husk, and b)

polyacrylamide amended compost. ...........................................................................................................193

Figure 0-6. Water retention as a proportion of the material dry weight in psyllium husk (PH) amended

compost and polyacrylamide (PAM) amended compost. ..........................................................................194

xvii

List of Appendices

Appendix A: Glossary 150

Appendix B: Data relating to Chapter 2 154

Appendix C: Data relating to Chapter 3 166

Appendix D: Data relating to Chapter 4 169

Appendix E: Data relating to Chapter 6 171

Appendix F: Success and Succession 174

Appendix G: Gone with the Wind 186

xviii

Symbols and Abbreviations A Area (m2)

Abs400 Absorbance at 400 nm

Agr Area of green roofs (m2)

b Molality (mol/kg)

CC Coefficient of curvature

CN Curve number

Cpeak Peak runoff coefficient

CU Coefficient of uniformity

Cvol Volumetric runoff coefficient

DOM Dissolved organic matter

dx Particle size (mm) such that x % of the mixture comprises particles finer than dx

ET Evapotranspiration

F F-distribution parameter

FAS Free air space

FEEM Fluorescence Excitation-Emission Matrix

FLL Forschungsgesellschaft Landschaftsentwicklung Landschaftsbau

g Acceleration due to gravity (m/s2)

h Capillary pressure head (cm)

H Hydraulic head (m or cm)

Ht Hydraulic head at time t (cm)

i Rainfall intensity (mm/min)

i.d. Internal diameter

Ia Initial abstraction (mm)

Irr. Irrigation

Kf Permeability (field conductivity) (mm/s)

Ksat Saturated hydraulic conductivity (m/hr)

L Depth of the medium sample in the column (cm)

LEA Lightweight expanded aggregate

LID Low Impact Development

MWC Maximum water capacity (%)

NTU Nephelometric Turbidity Unit

OM Organic matter (%)

n Sample size

xix

ni Unitless pore size distribution parameter

nm Nanometers

NRCS Natural Resources Conservation Service (U.S.)

p p value: the smallest level of significance for which the null hypothesis is rejected

pann. Annual precipitation (mm)

P Precipitation depth (mm)

PAW Plant available water

PCC Pearson product-moment correlation coefficient

pF Negative log10 of head in cm

pH Negative log10 of [H+]

PSD Particle size distribution

PWP Permanent wilting point (= 1.5 MPa)

Q Discharge (mm)

Qp Peak flow rate (mm/min)

r Pore radius (μm)

R Universal gas constant

RH Relative humidity (%)

S Theoretical storage (mm)

WRC Water retention curve

t Time (s)

T Temperature (°C)

TP Total phosphorous (mg/L)

V Volume (L)

WETP Water extractable total phosphorous (mg/kg)

wi Weighting factor

�̅� Arithmetic mean of the sample

Z Statistical Z-score

z Elevation head (cm)

αi Fitting parameter (cm-1)

γ Interfacial tension (N/m2)

δ Receding contact angle (°)

ɛ0 Theoretical dielectric permittivity of dry media

ɛb Bulk dielectric permittivity

ɛp Pore water dielectric permittivity

xx

ζ Parameter used for fitting exponential annual rainfall depth distribution

θ Volumetric water content (v/v)

θr Irreducible water content (v/v)

θant. Antecedent volumetric water content (v/v)

θs Saturated water content (v/v)

λ Ratio between S and Ia in NRCS curve number calculations

ρw Density of water (kg/m3)

μS/cm MicroSiemens per centimeter

ρm Maximum (wet, saturated) medium density (g/cm3)

ρd Dried bulk medium density (g/cm3)

ρs Mean solid particle density (g/cm3)

σ Standard deviation

σb Bulk electrical conductivity (μS/cm)

σp Pore water electrical conductivity (μS/cm)

σw Discharge water electrical conductivity (μS/cm)

Σ Sum

ϕ Porosity

ψm Pressure potential, syn. matric potential (kPa)

ψo Osmotic potential (kPa)

ψt Total water potential (kPa)

ψz Gravitational potential (kPa)

1

: Introduction

In 2014, the United Nations announced that, since 2007, over half of the world’s population were living

in urbanized environment (Figure 1-1). Owing to geographical and climatic factors the Canadian

population are well ahead of this trend, with over 80 % of us living in urban areas, as of 2014 when the

data was last collated (United Nations, Department of Economic and Social Affairs, 2014).

Figure 1-1 Urbanized population, data from (United Nations, Department of Economic and Social Affairs,

2014)

The City of Toronto (Figure 1-2) is the fourth most populous in North America (Contributors, 2016) and

the largest within the Great Lakes Basin, a watershed governed under a Joint Commission with the United

States of America (International Joint Commission, 2016). Whilst the topography of Toronto’s inner city

core determines that the landscape drains almost directly into Lake Ontario to the south (Toronto Region

Conservation Authority, 2016), urban development has driven stormwater to be rerouted into drains to

reduce flooding in the streets. Like many other urban areas established over a century ago, Toronto has

aging and somewhat undersized wastewater/stormwater infrastructure for the population now depending

upon it.

0

25

50

75

100

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

2010

2015

2020

2025

2030

2035

2040

2045

2050

Popula

tion U

rban

ized

(%

)

WORLD

Canada

2

Figure 1-2 Where we live: The two densest urban centres are the Greater Toronto Area in the lower centre of image and Montreal in the top right (Simmon, 2012)

The combined wastewater/stormwater sewer system outlined in Figure 1-3, serves the oldest and densest

parts of the city, and is prone to overflowing contaminated water directly into natural watercourses during

heavy rainstorm events (City of Toronto, 2016a). Approximately 25% of the city is served by the

combined sewer, which has 80 outfalls where direct overflows can occur (Podolsky, 2013). Annual

statistics regarding the number of overflow events is not available to the public, but 42 events were

reported between April – October 1991 (City of Toronto, 2010; Podolsky, 2013).

Figure 1-3 The area of combined sewers within the City of Toronto, from (Di Gironimo et al., 2013).

3

The demands upon the combined sewer system are increasing for two reasons. Firstly, through urban

densification, a greater population are producing wastewater within the area served. The population of the

Council Area ‘Toronto and East York’ (encompassing most of the combined sewer area) rose over 8%

between 2001 and 2011, to 7.32 thousand people/km2 (City of Toronto, 2016b). The base flow arising

from people’s activities fluctuates with their daily and seasonal activities. Overlaid onto this is the sudden

additional flow during and after rainstorm events. In Southern Ontario the intensity of rainstorm events is

expected to increase under current climate change predictions (Bates, 2008; SENES Consultants Ltd.,

2011). As most urban areas comprise a high proportion of impervious surfaces compared to undeveloped

landscapes, this stormwater runoff flows hot, fast and dirty (Figure 1-4).

Figure 1-4 The huge increase (55 % from 10 %) in runoff water resulting from storm events on a post-development landscape with a high degree of pervious cover. Image abridged from (Federal Interagency

Stream Restoration Working Group (FISRWG), 1998)

As combined sewers contain a mixture of stormwater, blackwater (sewage from toilets) and greywater

(e.g. washing water), the entire wastewater stream must be treated as hazardous and usually receives

costly multi-stage treatment at municipal plants before discharge into natural systems. For this reason,

reducing the volume of all three sources is desirable from economic and environmental perspectives.

In Toronto’s contemporary stormwater management policies, there are three primary, overarching

principles which development plans must achieve (City of Toronto, 2006). These may be summarized as:

Water Balance:

1. That ≤ 50% of annual precipitation becomes runoff water. i.e. ≥ 50% annual precipitation

must be retained on site,

2. That a 24 hour, 5 mm rainfall event must be entirely (100%) retained on site.

4

Water Quality:

1. That ≥ 80% of total suspended solids are removed from runoff water leaving the site,

2. For lakefront discharges there are also variable, seasonal E. coli limits.

Water Quantity:

1. Variable guidelines exist for the maximum permitted flow limits from sites according to

their size and location. For sites < 2 ha, the Rational method is permitted to make the

necessary calculations.

2. Erosion control criteria focus on larger development sites, particularly adjacent to

sensitive areas, such as ravines.

3. Peak discharge flow to municipal infrastructure during a 2 year return period storm event

must not exceed the lower of:

a. the capacity of the downstream municipal system, or

b. the flow resulting from a runoff coefficient of the ‘pre-developed conditions’,

capped at 0.5.

Whilst these targets may be achieved by the construction of grey, concrete infrastructure including larger

pipes and ponds, or vaults in dense urban settings; there are infrastructural benefits from adopting a

decentralized approach to managing city stormwater. Low Impact Development (LID) is both a

conceptual approach promoting source control, and a selection of tools used to reduce stormwater flow

and protect developed watersheds in urbanized landscapes. By minimizing the imperviousness of a site

and treating rainwater as a commodity rather than a nuisance, the burden on trunk infrastructure can be

reduced and the hydrologic behaviour of the watershed could more closely mimic that of a natural

ecosystem. Some LID techniques are based on the use of transpiration by vegetation, resulting in some

aesthetic or amenity benefit from the planting, and so creates an overlap with the term ‘Green

Infrastructure’ (GI) (see Figure 1-5). Green Infrastructure Ontario provide the following broad definition:

“…natural vegetative systems and green technologies that collectively provide society with a multitude of

environmental, social and economic benefits.”(Cirillo and Podolsky, 2012)

The US EPA focus their definition of green infrastructure on the nexus with LID:

“…systems and practices that use or mimic natural processes to infiltrate, evapotranspirate (the return of

water to the atmosphere either through evaporation or by plants), or reuse stormwater or runoff on the

site where it is generated.”(United States Environmental Protection Agency, 2014)

5

Figure 1-5 The intersection and overlap between Low Impact Development (left) and Green Infrastructure (right)

Subsurface infiltration is often a primary objective of LID; wherever possible projects are installed on the

land surface and include some sort of subgrade enhanced infiltration component. Examples include

permeable paving and vegetated bio-retention cells. Where there is a constraint on the available land, or

infiltration is otherwise hindered, cisterns are an alternative to retain stormwater for reuse, or later

discharge. Another option for sites that have limited space to provide for subsurface infiltration is to

design for retention of stormwater on roof tops. Where this is simply conducted using a weir/overflow

system, this is termed a blue roof and is relatively uncommon in Ontario (Cheung, 2016; Crawford, 2013;

Duncan, 2015). A more popular option is a green roof, in which vegetation in a supporting planting media

are assembled to emulate a naturalized setting on a building rooftop. As green roofs create the potential

for habitat, reduce urban heat island effect, provide amenity value and insulate their supporting building

(Castleton et al., 2010), they are often included as a key element of urban green infrastructure; they share

some of these characteristics with parks and urban forest.

A number of countries in Western Europe have a long standing tradition of using cut sod and other basic

materials to construct green roofs (Almssad and Almusaed, 2015; van Hoof and van Dijken, 2008).

However, in North America, the interest in building vegetation is more recent, popularized by the

Urban forest

Green walls

Parks

Green roofs

Swales

Bioretention cells

Wetlands/

Ponds

Rainwater

harvesting

Perforated pipes/

Soakaways

Permeable paving

Low Impact Development Green Infrastructure

6

environmental movement of the late 20th Century (Mentens et al., 2006). The later onset of construction is

evident in the carefully chosen and specifically engineered products found in most local green roofs. The

development of green roof systems in the North American market is still a rapidly advancing field for

several reasons:

- In the 1980s, the German Forschungsgesellschaft Landschaftsentwicklung Landschaftsbau e.V.

(FLL) prepared the first comprehensive document detailing many aspects of green roof planning,

maintenance and monitoring. Since then, the easiest course was to follow their recommendations

(e.V., 2008; Philippi, 2005).

- The industry is still dominated by aesthetic concerns, with many stakeholders viewing the

infrastructural benefits as convenient bonuses. There is rarely a clear intention what the ‘primary’

function of most installations should be, so that designs are not optimized to meet specific

objectives. This factor is exacerbated by the current green roof bylaw in Toronto which mandates

the construction of green roofs without specifying performance objectives other than vegetation

survival (Toronto, 2009).

- Within the field of stormwater management, there is no agreed method on how (or why) to model

extensive green roofs as part of a site-wide stormwater management strategy.

Consultants in both municipal engineering and in landscape architecture, and policy makers with the City

of Toronto, have to date, relied quite heavily upon information given to them from product manufactures,

who obviously have vested interests. In considering the stormwater management functions of an extensive

green roof, they may be conceptualized as one or more of the following traditional hydraulic structures:

- in retaining excess stormwater, they perform the function of a reservoir,

- in reducing or restricting peak flow rates, they emulate an orifice,

- or, they may be viewed as a conduit through which to empty a cistern or vault through

evapotranspiration, emulating an evaporation pan.

7

Research Objectives

Context

This thesis does not make the case for or against the construction of green roofs, for that argument has

already been successfully made, albeit sometimes grudgingly (Lstiburek, 2011). On a global scale, the

recent (2004 to date) general public interest in green roofs was at its maximum around ten years ago, with

peak Google searches in April 2006, 2007 and 2008 (Google Trends, 2016). Interest in all other Countries

and Cities are scaled against Canada and Toronto respectively, as the global centre of searches for the

term ‘green roof’. Toronto not only has the bylaw mandating construction under some development

circumstances (Toronto, 2009), but is also home to one of the most active industry advocacy

organisations, Green Roofs for Healthy Cities (2016). In the academic literature, interest levels continue

to rise (Figure 1-6)(Li and Babcock, 2014). The search term ‘green roof’ returns over 2,000 academic

journal articles published within the University of Toronto holdings so far this year (8th August)

(Univeristy of Toronto Libraries, 2016).

Figure 1-6 The rise in 'green roof' articles from Figure 1 in a hydrological review paper by Li and Babcock (2014)

Within this context, this thesis is based on the presupposition that extensive green roofs will continue to

be constructed for the foreseeable future, and focuses on how their design might be optimized for

stormwater control. Extensive green roofs are those constructed with the non-biotic components up to 15

8

cm in total depth, as these are the lightest systems, most commonly employed for stormwater

management and most suitable for retrofit installations (Czemiel Berndtsson, 2010). Green roofs > 15 cm

are typically constructed with amenity benefits as a driving factor, and the opportunities for their

installation are limited owing to the load they present. There is no upper limit to the depth and weight of

this type deeper ‘intensive’ type of green roof, so that in a dense urban environment so that many city

parks may double as underground parking or conceal other subterranean infrastructure. For additional

definitions associated with rooftop vegetation, see the Glossary at the end of this document.

Objectives

There are three primary research objectives of this thesis:

1. To produce responsive research regarding construction practices of extensive green roofs

Connecting with industry and considering current construction practices and beliefs and maintaining an

awareness of the multidisciplinary teams involved in the design and implementation of green roofs are

essential to producing influential findings. This work aims to produce simple parameters and interpretive

figures as decision making tools to help connect different disciplines, and academia with industry and

policy. Supporting predictions of the performance of design and maintenance configurations will lead to

recommendations to optimize green roofs according to specific storm water management objectives,

context and overarching functional priorities. In so doing, to increase the uptake and development of

useful extensive green roofs as part of our urban infrastructure.

Connection with industry is facilitated by the NSERC IPS funding mechanism supporting three of

the four years, but is expanded by regular social and professional engagement with designers,

manufacturers, installers, maintenance crews, owners, custodians and policy makers.

Awareness of the current practices will be gained by visiting, inspecting and sampling from as

wide spectrum of extensive green installations as is possible within the duration.

2. To characterize extensive green roof stormwater management performance in the context of the

local climate, encompassing both rainstorms and snowfall events

By determining robust coefficients for the modelling of extensive green roofs as one of three conceptual

hydraulic structures, they may be easily incorporated into site stormwater models:

As a reservoir, the important parameters are:

the capacity of the system to retain stormwater and hold that as a plant available source to

maintain the health of the vegetation, in effect water balance. On a monthly, seasonal, and annual

basis this will be defined as a volumetric runoff coefficient (Cvol). This is calculated as the

9

fraction of water from summed discharge volume in relation to precipitation. These are often

translated into a % retention value by simply finding the ‘missing’ fraction by subtraction:

% 𝑟𝑒𝑡𝑒𝑛𝑡𝑖𝑜𝑛 = 100 × (1 − 𝐶𝑣𝑜𝑙)

Equation 1-1

the capacity of the system to retain stormwater on a per event basis, and in so doing contribute to

the control of stormwater volumes and helping to reduce urban flooding. For this NRCS (Natural

Resources Conservation Service) curve numbers (CN) will be determined for the extensive green

roofs based on individual event discharge volumes (Q, mm) as a fraction of the precipitation

volumes (P, mm). Details follow in Chapter 2.

As an orifice, the important parameter is:

the reduction in peak flow provided by the system. This will be determined as a peak runoff

coefficient (Cpeak), by rearrangement and regression of the Rational method equation, in which

peak flow (Qp) is determined from peak rainfall intensity (i) and catchment area (A).

𝑄𝑝 = 𝐶𝑝𝑒𝑎𝑘𝑖𝐴

Equation 1-2

As an evaporation pan, the important parameters are:

the potential for evapotranspiration from the green roof system as a means of discharging excess

water, and

changes in the chemistry of the system associated with the evaporation of water from media and

plants.

3. To determine the relative influence of design and irrigation practices on the stormwater

management performance of extensive green roofs

By undertaking a multi-factorial study on green roofs with mixed design combinations, the factors

influencing each of the coefficients and performance indicators in section 1.1.1 above can be identified,

ranked and quantified. The design factors shall include the choice of vegetation between Sedum or an

alternative planting palette; the choice of planting medium type, whether a granular mineral mixture, or a

biologically derived, compost based product; the depth of planting material up to 15 cm; and the

provision of irrigation under different programming.

10

Background

The Green Roof as a Reservoir

Prior to the beginning of the 21st century, the vast majority of green roof construction and research was

undertaken in central Europe, resulting in little English language literature (Mentens et al., 2006). The

first phase of widespread green roof research began with comparisons made between green roof

installations and similar areas of ‘traditional roof’. Most studies, such as those shown in Figure 1-7, report

an aggregated precipitation retention, as this is easily measured and quite apparently a key strength of the

green roof in stormwater management. Per event, the amount retained depends largely upon the capacity

within the medium, which in turn depends on its material properties, depth, and existing moisture content.

This existing moisture in the system is determined by the climatic properties of depth of recent rainfall,

dry period since and the rate of evapotranspiration throughout that time. The publications from this type

of study vary widely in their conclusions, based on the variety of materials used and climates, in which

the test sites were located, see Figure 1-7.

Figure 1-7 Summary of previous studies assessing volumetric retention: A (Teemusk and Mander, 2007); B (Cronk, 2012); C (Hilten et al., 2008; Prowell, 2006); D (Schroll et al., 2011); E (Berghage et al., 2010); F (Gregoire and Clausen, 2011); G (VanWoert et al., 2005); H (Ma et al., 2012); I (Moran et al., 2004); J (Van Seters et al., 2009); K (Hathaway et al., 2008); L (Voyde et al., 2010); M (Uhl and Schiedt, 2008); N (Starry, 2013); O (Carter and Rasmussen, 2006); P (Burszta-Adamiak, 2012); Q (Getter et al., 2007); R (Palla et al., 2012).

32 events

62 events

153 events

31 events

84 events

48 events

91 events

15 months

154 events

9 months

3 events

83 events

97 events

72 events

18 events

70 events

63 events

3 events

0 10 20 30 40 50 60 70 80 90 100

R

Q

P

O

N

M

L

K

J

I

H

G

F

E

D

C

B

A

Average % storm water retention reported through entire study

11

The studies summarized were of varying durations, which accounts for the very low (study A – 3 events

(Teemusk and Mander, 2007)), and very high numbers (study R – 32 events (Palla et al., 2012). As

extensive green roofs have a finite capacity to store excess stormwater, there will usually be a proportion

of rainstorm events for which the capacity is insufficient. Whilst the rainstorm depth distributions vary

according to climate (Stovin et al., 2015), this remains a constraint to the overall potential retention for

any extensive system. The distribution of rainfall depths in Toronto is presented in Figure 1-8. The

coefficient for the exponential distribution (-0.119) of rainstorm depths in Figure 1-8 come from tables in

Adams and Papa (2000). A very similar distribution is in use within the City of Toronto Wet Weather

Flow Masterplan Guidelines, although the coefficient is not published (City of Toronto, 2006).

Figure 1-8 Toronto rainstorm depth distribution, from 1937 – 1983 Bloor Strreet rain gauge data.

In a conceptual reservoir, the inverse of this distribution can be used to understand the relationship

between storage (S, mm), and the proportion of annual rainfall which emerges as discharge, the

volumetric coefficient, Cvol:

𝐶𝑣𝑜𝑙 = 𝑒−0.119𝑆

Equation 1-3

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 5 10 15 20 25 30

% T

ota

l A

ver

age

Annual

Occ

ure

nce

s

Storm depth (mm)

Bloor St (1937-1983) 12 hr

Bloor St (1937-1983) 12 hr

12

Figure 1-9 illustrates this relationship and how the distribution of rainstorm depths creates difficulty in

reducing annual volumetric runoff coefficient of a limited storage system such as an extensive green roof.

Adding 4 mm storage to a system with Cvol of 0.5 will reduce Cvol to 0.3, but adding another 4 mm storage

only reduces Cvol to 0.2.

Figure 1-9 Conceptual relationship between maximum theoretical storage capacity (mm) and annual Cvol

Translating the cumulative distribution of storm depths (Figure 1-8) to the ability of a system to capture

and retain such events (Figure 1-9) requires a major assumption; that the system storage capacity is

entirely available as each storm occurs. The storage capacity of an extensive green roof depends upon

how wet the planting medium is at the storm onset. This is determined by the depth of the preceding

event, the time elapsed since that event and the rate of evapotranspiration in the intervening period.

In Ontario, three studies have examined the stormwater properties of green roofs in the last eleven years.

Liu and Minor (2005) reported a net retention of 57% (Cvol = 0.43) on two green roofs on Eastview

Community Centre. Linden and Stone (2009), reported net retention of 44% (Cvol = 0.56) by a green roof

in Waterloo, and Van Seters et al. (2009) reported the annual retention of water on the roof of the

Computer Sciences building at York University as 63% (Cvol = 0.37) after 154 precipitation events (> 0

°C periods) from May 2003 – August 2005. The net retention in each of the three studies is low compared

to the overall average shown in Figure 1-7 (61 %); notably, none of these local studies include water

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30

Volu

met

ric

runoff

co

effi

cien

t, C

vol

Precipitation depth captured and stored (mm)

Bloor St (1937-1983) 12 hr

13

balance calculations during the winter season, preventing the establishment of a clear net annual retention

value (Dhalla and Zimmer, 2010).

A widely used parameter, for continuous simulation modelling, is the NRCS curve number (CN). As the

empirical calibration methods and calculations were developed in the USA, a number of researchers in the

United States have used this metric to describe the hydrology of their green roofs. Two recent studies in

Michigan and Chicago have been conducted under climatic conditions similar to Southern Ontario are

presented in Table 1-1.

Table 1-1 Summary of green roof papers resulting in NRCS curve numbers. *Getter et al (2007) were studying effect of slope, hence the range of CN reflecting 2% to 25% slope.

Curve Number(s) Location Reference 86 Athens, Georgia, USA (Carter and Rasmussen, 2006)

84-90 East Lansing, Michigan, USA (Getter et al., 2007) 80 Chicago, Illinois, USA (Berghage et al., 2010)

For context, CNs for other land uses range between 98 for paved surfaces that have no storage or

infiltration capacity, to below 30 in forests on sandy soils (Mishra, 2003). CNs have local application as

they are used in single event or continuous simulations (Greenland International Consulting Inc., 2002),

which are recommended during development proposals in many Ontario conservation authorities (Central

Lake Ontario Conservation, 2010; Lake Simcoe Region Conservation Authority, 2013; Toronto Region

Conservation Authority, 2014).

The use of curve numbers for green roofs has been critiqued as: not being precise enough for runoff

simulation (Roehr and Kong, 2010), resulting in surprisingly high values and not representing the

physical differences from a natural hydrologic system (Elizabeth Fassman-Beck et al., 2015), and not

representing the variety of designs commonly employed (Roehr, 2010). However, this empirical model

provides a suitable mechanism to compact a large number of data pairs into a single metric per green roof

which can then be employed for statistical comparisons and assessment of retention performance.

1.2.1.1 Design Factors

As it became more widely accepted that green roofs retain more stormwater compared to a traditional

non-vegetated roof, the next generation of research developed, comparing green roofs against one

another. In an early example, a hydrological/thermal analysis of extensive green roofs at the University of

Texas, Simmons et al. (2008) concluded:

“Green roofs are not created equal”

14

However, this particular study of six different green roof systems used complete proprietary assemblies

preventing elucidation of the critical design factors. Other studies have focused on assessing the impact of

just one or two design parameters at a time, usually through a combination of fieldwork and laboratory

experiments (Czemiel Berndtsson, 2010; Li and Babcock, 2014). Commonly considered design factors

include: the type and depth of the planting medium, the species included in the planting and the provision

of irrigation.

Planting medium

Green roof planting medium is typically comprised of an engineered mixture of graded materials

including some composted organic matter and some granular mineral components. Inspired by the

German FLL guidelines (e.V., 2008) for green roof construction, many manufacturers produce a freely

draining mixture of graded aggregates, with a low total organic content. The composition of commercial

media blends is often considered proprietary information, but even branded products vary according to

materials available locally and at the required time (Rugh, 2013).

Desirable characteristics in a green roof medium are often conflicting, such as the requirement to be

freely draining to prevent ponding or roof membrane damage, impede the rate of percolation (low

permeability) to detain the peak flow discharge, and to retain water. In an effort to balance these demands,

and to improve triple bottom line sustainability indicators, academic research groups have trialed many

materials as components of green roof media. These have included: Rockwool cubes, coconut coir and

Styrofoam pellets, crushed shells, shredded tires, tumbled porcelain, and glass (Steinfeld and Del Porto

2008), crushed brick, clay pellets, paper ash pellets, and carbon8 pellets (Molineux et al. 2009). Despite

interest in the reuse and recycling of some of these more unusual materials, most products on the market

today use a combination of crushed recycle aggregate, lightweight expanded aggregate and composted

organic matter. The diversity of these mixtures, the commercial secrecy about their composition and the

lack of long-term studies leave many unanswered questions for researchers and contractors alike:

“My main problem with most substrates is the missing mid-range particle sizes in quite a few substrates,

and the inconsistency of composts, different compaction rates, and the lack of testing.”(Warmerdam

2013)

Detailed analyses of green roof planting media have been conducted by a few teams seeking to calibrate

their hydrological or thermal models (Hilten et al., 2008; Palla et al., 2012). Ouldboukhitine et al. (2012)

undertook analysis of five branded green roof media, assessing their thermal conductivities at a range of

saturation values. Having determined a ‘preferred’ product, this underwent a more detailed analyses using

15

Dynamic Vapor Sorption (DVS) and mercury porisometry to characterize the pore structure and water

retention properties for a subsequent model.

The percentage organic matter is another significant parameter which has been demonstrated to influence

both vegetation health and water holding capacity (Rowe et al., 2006; Yio et al., 2013). Nagase and

Dunnett (2011) combined their study of plant growth with hydrological characteristics, measuring the

effect of adding up to 50 % additional compost to a commercial crushed brick based product and

concluded that the addition of the compost had a significant, positive impact on the available moisture in

the planting media. They cautioned that application rates ≥ 25% may cause excessively ‘lush growth’ of

the vegetation, which would be unable to withstand later dry periods.

Concerns have been expressed over the shrink-swell characteristic of compost and the potential for

hydrophobicity to develop in green roof media, preventing rewetting in subsequent rainstorms

(Krzeminski, 2013). Doerr et al. (2000) undertook a review of water repellency in natural soils, and

determined that a higher percentage of organic matter, coarser particle sizes and high temperatures were

associated with the development of hydrophobicity; all three of these factors being notable in green roof

media. Multi-modal water retention functions of two related materials, peat and composted pine bark have

been have been described using a “modified van Genuchten-Durner approach” by Naasz et al. (2008).

They reported that the peat samples demonstrated significant hysteresis in the water retention curve (21

%) and concluded that this may be related to the shrink-swell characteristics of the material. The pine

bark, described as “quasi-rigid”, showed much less hysteresis (10 %) and did not show such a great

improvement in the multi-modal (4 pore domains) versus the uni-modal curve fit.

Depth

Mentens et al. (2006) derived a regression equation (R2 = 0.78) relating green roof runoff (RO, mm) to

precipitation (P, mm) and the depth of the planting medium (S, mm), following a literature review

including 125 observed events:

𝑅𝑂 = 693 − 1.15𝑃 + 0.001𝑃2 − 0.8𝑆

Equation 1-4

However, other laboratory controlled and single site studies have failed to concur on a relationship

between depth and retention. Spanning a range often considered to be ultra-light-weight, a study by

VanWoert et al. (2005) found no significant differences in retention over 14 months, when comparing

roofs with 2.5/4/6 cm planting medium depth. Looking at deeper systems, Nardini et al. (2012) reported

no significant difference in the retention of otherwise comparable roofs with 12/20 cm depth of medium,

16

when planted. However, they did find a significant difference between the two control plots at the same

depths, but without vegetation. Kelly (2008) studied paired roofs with 10/20 cm of medium and reported

no significant difference in net water retention. However, a contemporary study in Germany concluded

that medium depth dominated the overall retention of stormwater (Uhl and Schiedt, 2008).

The widely cited FLL guide indicates that the depth of extensive green roofs is not linearly related to net

retention of annual precipitation with increased depth of media failing to retain a proportional amount of

water (see Figure 1-10) (e.V., 2008).

Figure 1-10 Derived from (e.V., 2008), with the caption “All figures relate to locations with annual precipitation values of 650 – 800 mm where monitoring has been performed over a period of several years.”

The non-linear relationship has been attributed to the high proportion of macropores and subsequent

formation of preferential flow paths in the planting medium (She and Liu, 2013; Vergroesen et al., 2010).

However, observations on preferential flow are not universal, Palla et al. (2009) specifically reported that

the medium in their study demonstrated no significant preferential flows. In a detailed laboratory study of

unplanted green roof media (Yio et al., 2013) concluded that the depth has a significant effect, despite the

inherent heterogeneity of green roof planting medium, as:

“…preferential flow paths are more likely to be interrupted by zones of slower flow in deeper substrates

than shallower ones.”

Another significant factor in the development of the numbers illustrated in Figure 1-10, is the empirical

derivation. All of the monitoring periods will have encompassed storms of varying intensities and total

0

10

20

30

40

50

60

70

0 5 10 15 20

Annual

ret

enti

on (

%)

Medium depth (cm)

17

depths, both of which influence the retention properties of extensive green roofs on an event-basis and

thusly, an annual basis.

As depth range spanned by extensive green roofs is relatively narrow (typically ≤ 15 cm) and the

composition of the media varies between studies it has not been possible, to date, to conclude what the

effect of depth alone is on the hydrological performance of green roofs. However, deeper roofs are

associated with improved resiliency of the vegetation, which in turn protects the medium from wind scour

and erosion (Rowe et al., 2011).

Vegetation

The presence or absence of Sedum on the green roof was reported as playing a “minimal” role compared

to that of the planting medium by VanWoert et al. (2005). However, this assertion has since been

challenged; a report by Berghage et al. (2007) concluded that the vegetation can contribute up to 40% of

the hydrological performance of a green roof, depending upon the climatic factors. More recently, a

microcosm study by Bousselot et al. (2011) reported significant differences in the water retention of

medium planted with different species over 18 days. However, the difference between succulents (water-

storing) and herbaceous (broad-leaved) species was obscured to some degree, by variation within these

groups. In Italy, a study of green roofs with varying depth and plant types compared a 200 mm roof

planted with shrubs (multi-stemmed, woody), and a 120 mm roof planted with herbaceous species

(Nardini et al., 2012). They found no significant difference in the retention properties between the two

planting types despite the inclusion of depth of planting medium as a co-variable.

In addition to the contribution that the species have on evapotranspiration rate, by removing water from

the planting medium, the choice of vegetation may also have a significant impact on rainwater

interception. Standing water captured on leaf surfaces is available for direct evaporation, maximizing the

efficiency of removal from the system. The study of intercepted water is largely concerned with forested

ecosystems, but a few interception values exist for herbaceous ground covers. Gerrits (2010), reported

that a grass/moss layer intercepted between 4.1 ± 1.0 mm in the summer, and 2.0 ± 0.9 mm in the winter.

It was noted that the intensity of the individual precipitation events and the inter-event time each had

large effects on these numbers. Gerrits further concluded that whilst many papers detail observations on

vegetation interception, being able to model this process was important to reduce dependency on local

and specific field trials. This sentiment has been echoed by Taylor (2011), who reviewed literature

relating laboratory experiments and ‘early’ models to wetting of leaf surfaces in agricultural applications.

Given the number of complicating factors in the physiology of plants, in terms of their physical form and

metabolic processes, it is unsurprising that a conclusive answer to the role of vegetation has not yet been

18

reached. Even considering plant assemblies as complete ecosystems, conclusive distinctions between the

hydrologic performances of different classes of plants on green roofs remains to be demonstrated.

The Green Roof as an Orifice

Some studies report detailed event data including reduction in peak flow and peak flow detention

(increased lag time) (Carter and Rasmussen, 2006; Kelly, 2008; Liu and Minor, 2005; Moran et al., 2004;

Uhl and Schiedt, 2008; Voyde et al., 2010), Figure 1-11 illustrates these terms qualitatively.

Figure 1-11 Conceptual illustration of effects on hydrograph

Some reduction in peak flow is anticipated during any event, as some portion of the rainfall is retained by

the green roof. Often, shallow rainfall events result in no outflow from the media. Minimum storm depth

which results in outflow could be a way of comparing different installations, were it not for the large

effect of the antecedent conditions on the amount of water retained. This effect, whereby a large number

of events result in 100% peak flow reduction, strongly influences overall results. For example, an average

of 85% reduction in the peak flow was reported by (Moran et al., 2004) in North Carolina, USA. In the

more tropical climate of Auckland, NZ, (Voyde et al., 2010) reported a median peak flow reduction of

93%. Liu and Minor (2005) noted that peak flow reduction was affected by season, ranging between 25 -

60% in the summer and between 10-30% in the fall. Kelly (2008) found that 20 cm of planting medium

reduced the peak flow, compared to 10 cm of the same material (within a 30-minute period). However,

Uhl and Schiedt (2008) did not find the same, reporting instead that depth had no significance in reducing

peak flow. Peak lag increase has been reported by various parties, (Berghage et al., 2009; Carter and

Rasmussen, 2006; Moran et al., 2004), in all studies the delay time varies widely as a function of the

Time

Flo

w f

rom

ro

of

(Q)

Peak flow reduced and detained

Flow without green roof

19

intensity of the rain event. Green roofs are freely draining and constructed upon a base with a high

transmissivity, as such the peak detention function is determined only by the time taken for percolation

through the media.

An exception to these observations may be where a material of lower transmissivity (such as loose

aggregate) is used in place of the typical drainage board, slowing lateral, subsurface flow (Fassman-Beck

and Miller, 2016; Roofmeadow, 2013). As with all rooftops, the slope, the horizontal flow distance and

downspout design also contribute to the discharge hydrology.

The Green Roof as an Evaporation Pan

In dense urban settings where many forms of LID cannot be practiced, there is only the rooftop with

which to manage the ‘excess’ stormwater. In these cases, engineers, developers and policy makers locally

are often agreed that a subterranean vault or cistern beneath a proposed building is a preferred measure to

achieve annual stormwater retention targets (Cheung, 2016). As a building integrated system, the role of

the green roof is now changed into a means with which to empty this reservoir in between rain events

through irrigation with the harvested rainwater (Figure 1-12).

Figure 1-12 Conceptual closed system model combining a green roof with a cistern

This type of system has been considered by Hardin et al. (2012), who used a mass balance approach to

model a cistern of similar volume to their extensive green roof in Florida, USA. In so doing, the

researchers found that they could double the annual stormwater retention of their system. More recently,

in Shenzhen, China, Qin et al. (2016) used a 1-D HYDRUS model to simulate a green roof with

integrated storage beneath the porous media; the results of their study were framed on the irrigation

requirements of the vegetation rather than the volume of discharge from the system. Similarly, Chao-

Hsien et al., (2015) made conclusions regarding the amount of potable water required to top up their

Irrigation

Green roof

Cistern

Evapotranspiration Precipitation

Overflow discharge

20

cistern, after also producing a model which integrated a green roof with separate storage in Keelung,

Taiwan.

Retaining and recirculating discharged water provides a number of co-benefits including improved plant

health, increased evaporative cooling and prevents excess nutrient discharge from fertilizer or biological

components within the planting medium. Although this type of system is in use locally (Figure 1-13),

research is required to model the stormwater hydrology of such systems, and the long term effects of

recirculating water could result in increased salinity within the medium (Jones and Jr., 2012; Moritani et

al., 2013).

Figure 1-13 One part of the irrigation system on the Rottman School of Management, University of Toronto. This system uses recycled and/or harvested rainwater to irrigate extensive green roofs.

21

Thesis Organization

The body of the thesis is prepared as four proposed journal articles and a chapter of smaller, related

studies (Chapters 2 through 6). Chapters 2 though 4 describe measurements and calculations made on the

experimental Green Roof Innovation Testing laboratory (GRITlab) and are complimentary to one another.

Chapter 5 and 6 stand alone as a series of laboratory experiments and a field survey respectively:

Chapter 2 presents event-based analysis of rainstorms on experimental extensive green roofs at

GRITlab, to compare the influence of four design factors: type of medium, depth of medium,

planting and irrigation. The parameters calculated include monthly and seasonal volumetric

runoff coefficients (Cvol), NRCS curve numbers and peak runoff coefficients (Cpeak).

o A manuscript of this work has been submitted to the ASCE Journal of Hydrologic

Engineering, and is in review, as of August 2016.

Chapter 3 presents winter water balance data from GRITlab, and aggregates this with the event-

based analysis from the summer in Chapter 2 to determine annual Cvol values under the influence

of the same four design factors. Addressing the issue of irrigation discharge and the potential for

recirculation of excess water is undertaken by considering water balance from three different

irrigation programs.

o A manuscript of this work is currently being prepared for submission as an additional

paper to Chapter 2.

Chapter 4 considers the physicochemical properties of the GRITlab systems. Stormwater

retention is improved through the use of a compost based planting medium. But this causes

increased phosphorous to be discharged, a problem which could be countered with a recirculation

mechanism. Recirculating water until 100% evapotranspired could result in increased salinity

within the system, so methods of monitoring medium moisture content and dissolved solids in the

medium is presented.

o This chapter includes two separate studies which relate to the preceding chapters, but not

necessarily well to one another. This work was considered for a conference presentation,

but is not currently being prepared for publication.

Chapter 5 examines the interplay of green roof medium properties and the construction depth to

help explain what limitations simply increasing depth of planting medium has as a design

decision.

o A manuscript of this work has been submitted to the Elsevier Journal of Hydrology, and

is in review, as of August 2016.

Chapter 6 describes the properties of planting media recovered from thirty-three extensive green

roofs in Southern Ontario.

22

o This study has been accepted for publication in the Elsevier journal, Ecological

Engineering (Hill et al., 2016).

o A supporting photographic journal has been self published and is currently hosted online

(Lottie, 2016).

There are two complimentary studies included in the appendices, which have already been presented as

conference papers: Appendix D reviews a few of the attitudes towards evolving green roof landscapes

from their owners. Appendix E compares methods of erosion control applied on green roofs.

Authorship

The author (Jenny C. Hill) is the primary author of the work and writing presented in this thesis. Dr.

Brent Sleep and Dr. Jennifer Drake have served as co-authors for each of the substantive chapters, as has

Liat Margolis in Chapters 2 and 3.

23

: Influences of Four Extensive Green Roof Design Variables on Stormwater Hydrology

Abstract

This study assesses the relative influence of four independent variables on green roof hydrological

performance under rainstorm conditions. Twenty-four extensive green roofs representing all combinations

of four design factors were used: native meadow species versus Sedum, mineral-based versus

biologically-derived planting medium, 10 cm versus 15 cm depth, and irrigation provided daily, sensor

controlled, or not at all. From events covering the summer period May – October in 2013 and 2014, mean

values were determined for the seasonal volumetric runoff coefficient (Cvol = 0.4), peak runoff coefficient

(Cpeak = 0.12), and NRCS curve number, (CN = 90). Irrigation had the largest overall impact; daily

irrigation increased Cvol to 0.5 compared to 0.3 for systems with sensor controlled or no irrigation. The

biologically-derived planting medium, comprised of a high proportion of aged wood compost, made a

significant improvement, maintaining Cvol of 0.3 compared to 0.4 for the mineral-based product in the

modules without irrigation. A similar pattern was found in the NRCS curve numbers.

24

Introduction

Research on the hydrology of green roofs has established that a combination of lightweight planting

media and environmentally resilient vegetation on building roofs can improve the rainwater runoff

characteristics from buildings compared to ‘traditional’ non-permeable alternatives (Czemiel Berndtsson,

2010; Lundholm et al., 2010; Nagase and Dunnett, 2011; Van Seters et al., 2009). Widespread

deployment across an urbanized watershed of green roofs provides a valuable contribution to stormwater

control by making optimal use of the limited available catchment surfaces (Carter and Jackson, 2007). In

many cities, this ambition will require many retrofit installations and of the systems available, “extensive”

green roofs are thinnest (up to 15 cm), usually the most lightweight, cheapest and most likely to be

deployed (Oberndorfer et al., 2007). Therefore, it is extensive green roofs that are most likely to have the

highest impact potential on urban watersheds.

They typically are constructed from a number of layers, as shown in Figure 2-1. Each of these layer

components require design decisions, which are often driven by market forces owing to a plethora of

available proprietary products and solutions. From the rooftop up, the first hydraulically significant

component is a drainage/ retention layer, which reduces or eliminates pooling of water on the waterproof

roof membrane and ensures that the root zone is not saturated for extended periods.

Figure 2-1 Typical layering of a built-up extensive green roof system

A very common format for the drainage/retention layer is a pre-formed rigid polymer sheet or board, with

regularly placed drainage holes, such that depressions in the sheet form small reservoirs of water held

away from the underlying roof. A geotextile is usually employed on top of the drainage board to keep the

drainage board depressions free from excess particulate matter, because the next substantial component is

a lightweight, engineered, (usually soilless) porous planting medium. A large number of materials have

been trialed and blended for planting medium, with varying degrees of success in terms of stormwater

control and/or vegetation survival (Farrell et al., 2012; Molineux et al., 2009; Ouldboukhitine et al., 2012;

25

Steinfeld and Del Porto, 2008). Influences on the development of green roof planting media have come

from the horticulture and nursery industries, where potting mixtures have been studied for many years,

but even more so from the German FLL agency, who advocate for a much lower proportion of organic

material (< 6.5 %) than would typically be used in nursery potting mixtures (e.V., 2008). Two distinct

schools of thought about the role of organic matter in green roof media exist; those in favor of following

the FLL guidelines, state that biologically derived materials will biodegrade and lose porosity, have

reduced drainage and remain waterlogged, to the detriment of the planting. Those who prefer to specify

high organic matter planting media, containing a higher proportion of compost or other biologically

derived materials believe that these claims are unsupported and unjustified (Buist and Friedrich, 2008).

The choice of vegetation has also received much attention, as it is the most immediately visible,

contributing co-benefits such as aesthetic appeal and habitat for biodiversity and urban ecosystem

support. A significant body of green roof work has focused on the survival of plants, with fewer studies

assessing the hydrological impacts. Some research finds that plants play an important role in hydrology

(Berghage et al., 2007; Bousselot et al., 2011; Lundholm et al., 2010), whilst others have not discerned

any significant impact (Nardini et al., 2012; VanWoert et al., 2005). To support growth and survival of

vegetation, the use of supplementary irrigation is a widespread practice. Secondary benefits can include

aiding in fire prevention by keeping plants green rather desiccated in the height of summer, and reducing

loss of the granular planting medium from wind erosion or scour.

To support the growing application of extensive green roofs as effective tools in stormwater management

strategies, it is important to have accurate values for commonly used hydrologic parameters. In our study,

aggregated volumetric runoff coefficients (Cvol), have been determined, in keeping with other similar

studies (Fassman-Beck et al., 2015; Gregoire and Clausen, 2011). For individual event calculations US

Natural Resources Conservation Service (NRCS) Curve Numbers have been calculated. To provide the

most accurate parameter for peak flow calculations using the Rational method, peak flow runoff

coefficients (Cpeak) were derived using paired peak flow and peak storm intensity data (Young et al.,

2009). Many previous studies have reported observations and derivations of these hydrological

characteristics for green roofs (Czemiel Berndtsson, 2010); these parameters are dependent on the

climatic conditions under which they are measured (Fassman-Beck et al., 2015).

This study is designed to provide useful engineering information for extensive green roofs pertinent to a

humid continental climate (Dfa/Dfb) region (Kottek et al., 2006). The event-based analyses are

constrained to the summer period, encompassing May through to October; these are the months in which

all precipitation was received as rain. Within this context, the objectives are: to determine appropriate

values for the coefficients and parameters given above, to assess the robustness of such parameters with

26

respect to changes in green roof design with respect to: vegetation selection, planting medium type and

depth, and irrigation, and, to identify the preferred option for each of the most influential design factors.

Methods

Green Roof Innovation Testing laboratory

The experimental site, the Green Roof Innovation Testing laboratory (GRITlab) is located on the fifth

storey roof of the historic John H. Daniels Building, situated in the centre of the downtown St. George

Campus of the University of Toronto, Ontario. The lab has twenty-four individual green roof modules,

each with a 2.86 m2 drainage area (2.36 m x 1.21 m), constructed with 2% slope. The modules are

suspended 0.8 m above the roof deck to accommodate instruments and maintenance requirements

(Margolis, 2013).This study assesses four design variables, using a spatially randomized full factorial

(233) design; vegetation type, planting media type and planting media depth were considered at two

levels, whilst irrigation provision was tested at three levels. The grid layout of the modules and the

randomized distribution of the variables is presented in Figure 2-2.

Figure 2-2 Schematic of GRITlab, illustrating the randomized layout of the four experimental variables. Key - colours in each rectangular module can be read from west to east. Vegetation: dark = Meadow, light = Sedum. Planting medium: dark = biological, light = mineral. Irrigation: dark = daily, light = sensor, mid = none. Construction depth: dark = 15 cm, light = 10 cm.

Two types of vegetation were considered, a Sedum. blend initially containing 23 cultivars pre-

established onto mats, and a meadow mix of 19 species including grasses and forbs. Both the

meadow seeding and the Sedum. mats were installed in 2011. Further details regarding the plant

communities and their growth performance in previous years has been published (MacIvor et al., 2013).

27

The two types of planting media were selected as representative of the extremes in commercial use

locally: the mineral based medium comprises a large proportion of lightweight expanded aggregates and

crushed brick, and has low organic matter content in concordance with FLL recommendations (e.V.,

2008). The second type is a biologically based medium containing a matured, screened, pine bark

compost with < 5% additional components. The manufacturer’s specification for each product is

presented in Table 2-1. Each of these two materials were tested at 10 and 15 cm depth.

Table 2-1 Physical data for grit planting media, according to manufacturer’s ASTM 2399 report (Bioroof Systems, 2011)

Mineral Biological Dry density g/cm3 >0.8 0.58

Saturated density g/cm3 1.28 1.1 Maximum water holding capacity 45% >60%

Saturated hydraulic conductivity cm/s >0.02 >0.01 Organic Matter (%) < 9% >70%

Irrigation was provided to the modules via drip lines, with 300 mm spacing of the emitters. The flow rate

was fixed, and the irrigation controlled by altering the timer program. The daily modules received

irrigation every morning, which maintained a high level of saturation in the media throughout the months

of application. The sensor controlled modules each had a custom adapted, fluid-filled tensiometer

installed (Irrometer) which was set to open the irrigation valve for media moisture tension < -25 kPa;

irrigation was only received by sensor modules, if the valve was open due to dryness of the media. Both

forms of irrigation produced measurable runoff. In 2013 the irrigation system was deployed between the

first week of May and the last week of October; and in 2014 this was reduced to include just the months

of June and September. Further details about the irrigation programming are given in Chapter 3.

Precipitation was measured on site using a tipping bucket rain gauge (TE525M Texas Electronics), whilst

parameters used for the automated calculation of reference evapotranspiration were measured using an

adjacent weather station (Allen et al., 2005): Wind monitor (RM Young), CMP 11 pyranometer (Kipp

and Zonen), HMP45C relative humidity and temperature probe (Campbell Scientific). On twelve

occasions between August 2014 and August 2015 manual rain gauges were placed adjacent to all green

roof modules and single event stormwater collected. These were used for spatial assessment of the rainfall

distribution across the laboratory roof. Planting media moisture content was recorded (5TE, Decagon

Devices) after recalibration for the dielectric properties of each of the two planting media types.

Discharged water from each module was measured using a rain gauge (TB6, Hydrological Services).

These rain gauges were adapted to handle the higher flows experienced, using customized 3D printed

funnels (J Hill et al., 2015). The data logger controlling all of the sensors recorded five-minute resolution

and the data presented here were collected during the months of May to October in 2013 and 2014.

28

Theory and Calculations

As extensive green roofs are relatively small catchments and have very short flow paths, they are highly

responsive to rainfall characteristics with discharge beginning and ceasing rapidly. In most cases at

GRITlab measurable drainage had ceased in less than an hour after a storm had passed, and peak lag

times were not discernable within the 5-minute resolution of the data logger. For this reason, an inter-

event time of one hour was used to determine separate rainfall events in the summer months. So, a storm

was considered to be any rainfall event of ≥ 0.2 mm of rainfall preceded and followed by a minimum of 1

hr without measurable precipitation. Spatial autocorrelation of rainfall patterns across the GRITlab was

assessed using Local Moran’s I values, generated using GeoDa (Anselin et al., 2006). Rainfall

distributions for each summer period were fitted to a single parameter exponential function using Easyfit

(Drokin, 2010):

𝑓(𝑝) = 휁𝑒(−𝜁𝑝)

Equation 2-1

Irrigation supply and discharge were not included in water balance calculations. Instead irrigation

provision was included as a categorical variable; none, sensor controlled, or daily. So, the aggregated

(monthly and seasonal) volumetric runoff coefficients (Cvol) have been calculated as the sum the total

discharge depth of the individual event discharge (Q, mm), as a proportion of the sum of the event total

precipitation depth (P, mm):

𝑪𝒗𝒐𝒍 = ∑ 𝑸

∑ 𝑷

Equation 2-2

Comparisons between group means of the independent variables were made using regression trees

(Demšar et al., 2013). At each level on the trees, the group mean value and number of contributing

modules are presented. The technique then identifies the single factor which provides the greatest

difference in group means and classifies data accordingly. This continues through successive branches

until no significant difference in the group means can be elucidated. The trees were pruned when no

practical difference was discerned in the parameter being fitted.

To assess the event-based stormwater retention and theoretical storage capacity of the modules, NRCS

curve numbers were generated for all twenty-four modules, for the summer months of 2013 and 2014.

The benefit of using curve numbers in this research is that this allows aggregation of the

precipitation/discharge volume of many storm events and reduces one dimension of the data to permit

statistical comparisons between the multivariate designs. Calculations were performed on natural data. i.e.

29

where the precipitation (P, mm) and discharge depths (Q, mm) were retained in event pairs (ASCE/EWRI

Curve Number Hydrology Task Committee, 2009). Observations of Q and P were fit to Equation 2-3,

using a least squares method to solve for storage (S, mm) and the ratio (λ) between storage and initial

abstraction (Ia). The abstraction ratio and storage parameters were not fixed, but bounded within limits (0

≤ λ ≤ 1, and 1 ≤ S ≤ 100).

𝑄 =(𝑃 − 𝜆𝑆)2

(𝑃 + (1 − 𝜆)𝑆), 𝑤ℎ𝑒𝑟𝑒 𝜆𝑆 = 𝐼𝑎

Equation 2-3

Storage values were transformed to curve numbers using the metric version of the NRCS curve number:

𝐶𝑁 =25400

(𝑆 + 254)

Equation 2-4

Further analysis of individual events included assessment of the relative impact of antecedent moisture

conditions on the overall discharge volumes. The individual event volumetric coefficient was calculated

for all modules and all rainstorm events. These values were then classified according to the antecedent

volumetric water content (θ) of the medium at event onset. The upper limit of the range considered was

0.35 v/v due to the porosity and free draining characteristic the mineral based medium; θ > 0.55 v/v was

recorded in some modules with biological medium. To enable statistical comparisons, break points were

set at 0.15 and 0.25 which created three classes of similar size within each medium type. The mean

individual event volumetric coefficient per class, per module was then calculated, and these numbers

further aggregated by grouping according to the independent variables.

Peak runoff coefficients (Cpeak) for each event were determined by rearrangement of the Rational method

equation, as determination of peak flow remains the most common application of a runoff coefficient;

peak flow rates (Qp, L/min) were divided by the product of peak rainfall intensity (i, mm/min) and

catchment area (A, m2).

𝐶𝑝𝑒𝑎𝑘 =𝑄𝑝

𝑖𝐴

Equation 2-5

30

Results and Discussion

Local Climate

Toronto is in Southern Ontario, Canada and enjoys a relatively temperate climate owing to the moderating

effect of the Lake Ontario to the south. During the study period weather conditions were typical for the

region (Environment Canada, 2013), although 2013 was a wetter year than 2014. Monthly values for

precipitation depth and reference evapotranspiration are presented in Figure 2-3.There was one exception

within the duration of the experiment; an event exceeding the 100 year return period storm occurred on

the 8th July 2013 (Di Gironimo et al., 2013). This event delivered more than 80 mm in a few hours,

overtopping the tipping bucket gauges so that neither the hydrograph characteristics, nor the total volume

discharged from the modules was recorded.

Figure 2-3 Local weather at GRITlab, Toronto between May-October 2013 and May-October 2014, the duration of the green roof study.

From the beginning of May to the end of October 2013, data were collected from 96 individual storm

events, with a total precipitation depth of 561 mm. From May 2013 to April 2014, the total precipitation

was 891 mm. In the following summer, 2014, observations were made on 80 storm events, with a total

depth of 314 mm, and with the following winter the total annual precipitation to end April 2015 was 599

mm. The ζ parameter describing the exponential distribution of rainfall depths, was 0.19 in summer 2013

and 0.26 in summer 2014. This indicates that not only did 2014 have less rainfall but that a greater

proportion of the rainfall arrived in lower volume events. These ζ values are calculated from 1 hour inter-

event definition of a storm; and are presented in Figure 2-4, alongside the local, historical distribution of

storms according to 1 and 12 hr inter-event times (Adams and Papa, 2000).

0

20

40

60

80

100

120

140

160

May

-13

Jun-1

3

Jul-

13

Aug-1

3

Sep

-13

Oct

-13

May

-14

Jun-1

4

Jul-

14

Aug-1

4

Sep

-14

Oct

-14

Dep

th /

mm

Precipitation (mm) Reference Evapotranspiration (mm)

31

Figure 2-4 Annual exponential cumulative distribution of storm depths in Toronto, ON. According to 2013 and 2014 GRITlab data and historical records (1937-1983) from Bloor St.

Spatial autocorrelation was performed on the manual rain depth data gathered after eleven rainfall events

between August 2014 and August 2015. Moran’s I, test results can range between 1 and -1, where higher

values indicate more clustered data and low/negative values indicate an even distribution of the parameter

across space. The Moran’s I data had a mean of 0.07 and a standard deviation of 0.11. This overall value,

close to zero, indicates that the rainfall depth was randomly distributed across the experimental area and

not influenced by the roof shape or aspect.

Volumetric Runoff Coefficients

The relative effects of the four design factors; irrigation, medium type, medium depth and vegetation, on

volumetric runoff were analysed using a stepwise regression, producing a regression tree. Technical

problems caused each data set to be reduced by one unique combination; as this was a full factorial

experimental design statistical elucidation of all factors remained valid (NCSS, 2015; Walpole, 2007). In

2013 a leaking irrigation line resulted in the exclusion of the module containing 15 cm mineral medium,

with Sedum planting and daily irrigation. In 2014, the module containing 15 cm biological medium, with

Sedum planting and daily irrigation was excluded, owing to the malfunction of the discharge tipping

bucket.

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 5 10 15 20 25 30 35

% T

ota

l A

ver

age

An

nu

al O

ccu

ren

ces

Storm depth (mm)

Bloor St (1937-1983) 1 hr

Bloor St (1937-1983) 12 hr

2013 GRITlab 1 hr

2014 GRITlab 1 hr

32

In the regression tree shown in Figure 2-5 (R2 = 0.947), the dominance of the irrigation in the water

balance is apparent. Green roofs that received no or sensor moderated irrigation each resulted in a group

mean Cvol of 0.3 and so retained approximately 70% of the rainfall. This retention value is comparable to

those from previous studies in similar climatic zones, both locally (Liu and Minor, 2005; Van Seters et

al., 2009) and further afield, including the USA, the UK and New Zealand (Hathaway et al., 2008; Moran

et al., 2004; Starry, 2013; Uhl and Schiedt, 2008; VanWoert et al., 2005; Voyde et al., 2010). However,

green roof modules which received indiscriminate daily irrigation retained significantly less stormwater

and achieved a group mean Cvol 0.5. i.e. Half of all the stormwater was discharged and only half retained.

Within the no irrigation group, the type of planting medium made a significant difference; the

biologically-derived medium had a group mean Cvol of 0.3 and the mineral-based material Cvol 0.4. This

distinction was not statistically significant in the irrigated groups.

Figure 2-5 Regression tree for the runoff coefficients (Cvol) determined on twenty-three extensive green roofs over 12 summer months encompassing May-October 2013, and May-October 2014 .

The relative insignificance of vegetation type and media depth is supported by findings from previous

studies. Two papers examining both depth and planting selection, found no significance in either factor

(Nardini et al., 2012; VanWoert et al., 2005), whilst others have also reported that increased depth alone

makes no significant difference to net stormwater retention (Graceson et al., 2013; Kelly, 2008). It has

been reported that the vegetation makes a large difference in stormwater retention, when studied in

isolation (Berghage et al., 2009; Bousselot et al., 2011; Lundholm et al., 2010). With a holistic green roof

design perspective, this provides leeway for the depth of planting medium to be influenced by loading

summer Cvol

0.4

(23)

None

0.3

(8)

Biological

0.3

(4)

Mineral

0.4

(4)Sensor

0.3

(8)

Daily

0.5

(7)

33

capacity of the roof and plant selection to accommodate other design priorities such as aesthetics,

biodiversity and shading.

The amount of water retained per month was also examined to identify any seasonal trends in the effects

of each design variable (Figure 2-7). Overall, across all modules runoff volumes were slightly higher in

the late season when evapotranspiration rates are lower (Figure 2-6).

Figure 2-6 Monthly group mean volumetric runoff coefficients for 23 green roof modules during the periods between May-October 2013 and May-October 2014.

0.0

0.5

1.0

Cvol

34

Figure 2-7 Monthly group mean volumetric runoff coefficients for four design factors during the during the periods between May-October 2013 and May-October 2014.

0.0

0.5

1.0C

vol

Irrigation

None

Sensor

Daily

0.0

0.5

1.0

Cvol

Medium

Mineral

Biological

0.0

0.5

1.0

Cvol

Depth

10 cm

15 cm

0.0

0.5

1.0

Cvol

Planting

Sedum

Native

35

The relative impact of the irrigation programs is evident in the monthly group mean Cvol data (Figure 2-7).

The impact of irrigation in September and October is attributed to excess saturation of the green roof

media as evapotranspiration reduces. The observed difference between the irrigation programs is in

keeping with previous studies that concluded when irrigation is provided at controlled levels, it is

detrimental only to the retention of the deepest storm events (Schroll et al., 2011), but when provided in

excess it can reduce net retention by producing runoff of its own (Spolek, 2008). Through most of the

summer months, the biological planting medium retained more water and reduced Cvol compared to the

mineral based alternative (Figure 2-7). Also shown in Figure 2-7, there was no clear overall influence of

planting medium depth on the total water discharged and any influence of the vegetation palette on

seasonal Cvol was undiscernible compared to the other design factors. It was notable that there are no

seasonal trends in Cvol, between the two types of vegetation. Through the spring time, the coverage and

canopy structure of the meadow plants develops quickly and more dramatically than in the Sedum

(MacIvor et al., 2013). The Sedum mix was expected to achieve much lower potential transpiration rates

through the summer (Blanusa et al., 2013), which could manifest as higher discharge if the medium was

not being dried so swiftly. A mechanism possibly countering this effect, is canopy capture in the Sedum

modules. Two of the dominant species found in the Sedum mixture during the study period, S.

kamtschaticum and S. spurium (MacIvor et al., 2013), have cup-like leaves and waxy cuticles; foliage

characteristics ideal for intercepting rainfall, which would then be available for direct evaporation from

the leaf surfaces.

The relationship between individual event volumetric discharge and θant.is presented in Figure 2-8. The

choice of planting medium was the most influential design variable; the increase in discharge volume

from moist mineral based modules was much greater than in the biologically based counterparts. The

group mean volumetric runoff coefficients for the 0.25 – 0.35 v/v condition were 0.17 and 0.71 for the

biological and mineral media, respectively. This is a highly significant four-fold improvement of the Cvol

in biological media compared to the mineral media, when both are storing a similar amount of water. The

additional porosity of the biological medium also permits a greater range of θant.; in Figure 2-8 it is

apparent that the Cvol of the biological medium reaches a similar value once it is similarly saturated (θant. =

0.45 – 0.55). This analysis supports the findings above, regarding the annual volumetric runoff

coefficients; there is a compounded detrimental effect in using a very free-draining, mineral-based

planting medium when irrigation is also being used to continuously boost the volumetric water content.

36

Figure 2-8 Box plot of volumetric runoff coefficients over rainstorm events in 2013 and 2014, grouped according to medium type and antecedent volumetric water content over the range 0 – 0.55 v/v. Group means indicated with ‘X’ and connected within the medium type.

Event-based Analysis

To disaggregate storm water retention data to an event level, NRCS curve numbers were calculated (R2

between 0.4 - 0.8, see Appendix B) for all available modules and have also undergone regression tree

analysis to determine influential factors (Figure 2-9)(R2 = 0.984). After two summers, the overall mean

curve number across all twenty-three modules was 90 (σ = 3.5). This curve number is similar to the value

of 92, recently calculated from data collected previously in Toronto (Van Seters et al., 2009), and within

the 90-96 range determined for extensive green roofs in the Dfa/Dfb region (Fassman-Beck et al., 2015).

37

Figure 2-9 Regression tree for the NRCS Curve Numbers determined on twenty three extensive green roofs over 12 summer months encompassing May-October 2013, and May-October 2014.

In most of the fitted curves λ = 0. This is unusually low, even compared to the recently revised value of

0.05 (ASCE/EWRI Curve Number Hydrology Task Committee, 2009), but reflects how green roofs differ

from terrestrial catchments. Events of all size were included, and even many of the lightest storms

resulted in measurable runoff. This approach encompassing light rainfall and the resultant discharge

differs from the approach used by Elizabeth Fassman-Beck et al., (2015) who included only events for

which P > 0.46S, and fixed λ = 0.2. The results from fitting curves whilst maintaining λ = 0.2 are

presented in Appendix B and are, on average, 4 points higher.

Again, the provision of daily irrigation had a significant impact on the curve number raising the value by

5 over the sensor controlled or zero irrigation conditions. The secondary predictor of planting medium

depth in the sensor controlled group is notable, but explained by examining the data contained within the

terminal leaf of “sensor” AND “10 cm” (leaf R2 = 0.50); one datum had a particularly low CN = 82 (CN

curve fit R2 = 0.44). Within the entire data set, this datum had not been identified as a statistical outlier (Z

= 2.0).

Peak Flow

The highest intensity rainfall was recorded at 1.6 mm/min (8 mm/5 min) in August 2013, this value is

similar to the predicted ipeak for a 2-year return period storm in Toronto (ipeak = 1.5 mm/min, where time of

concentration = 10 min (City of Toronto, 2006)). Regression of the paired rainfall intensity and peak flow

NRCS Curve Number

90

(23)

None

89

(8)

Biological

87

(4)

Mineral

91

(4)

Sensor

89

(8)

10 cm

87

(4)

15 cm

90

(4)Daily

94

(7)

38

data over both summers was undertaken for each individual module; R2 values were low (0.3- 0.5) owing

to the simplicity of this model and the wide variation of storm characteristics and antecedent conditions

encompassed. The resultant mean peak runoff coefficient (Cpeak) across all combinations was 0.12 (σ =

0.02). A similar value for Cpeak (0.11) has been reported previously, after Carpenter and Kaluvakolanu

(2011), made observations on 21 storms on an extensive green roof. Regression tree analysis did not find

any of the four design factors significantly influential on this parameter.

To check the validity of the peak runoff coefficient, twelve individual storms from late summer 2015

were modelled and compared to observational data (Figure 2-10). These storms occurred between 22nd

June and 10th August, and varied in peak intensity between 0.04 – 1.16 mm/min. For maximum

robustness, these events were chosen without concern for antecedent conditions, total depth or duration of

the storms.

Figure 2-10 Validation of peak based runoff coefficients using Rational method to calculate peak flow (Qp) and compared to experimental data for twelve, 2015 rainstorm events. Error bars represent the standard deviation of all twenty-four module’s peak flows per event.

Most of the observed peak flow values were slightly overestimated using Cpeak of 0.12, which

demonstrates that a runoff coefficient of 0.12 could be relatively conservative for application of the

Rational method to calculate peak flow from an extensive green roof for up to 2-year return period

rainstorms. Although this parameter does not necessarily extrapolate to higher return period events (e.g.

10 yr = 2.7 mm/min; 100 yr = 4.2 mm/min), field data from these events are naturally more difficult to

obtain. Whilst the results do not point to any single important design parameter to mitigate peak flow,

these are notably low values of Cpeak compared to the current local practice of applying a coefficient of

0.5 (Aster, 2012; e.V., 2008); which appears to be based on the application of ‘coefficients of discharge’

in the referenced FLL guidelines (e.V., 2008) rather than an empirically determined Cpeak.

0.0

0.2

0.4

0.6

0.0 0.2 0.4 0.6 0.8 1.0 1.2

Qp

(L/m

in)

ipeak (mm/min)

Mean measured Qp

Cpeak = 0.12

39

Conclusions

The coefficients determined would be suitable for use in modelling rainstorm responses of extensive

green roofs in a Dfa/Dfb climatic region. Mean net summer seasonal rainfall retention of ~70%,

alternatively described as a volumetric runoff coefficient of 0.3, is consistent with previous reports and

reflects the recommendations of local policy (City of Toronto, 2006). The relative impact of four simple,

easily altered design variables were determined to inform hydrological performance directed design of

extensive green roofs. The provision of daily timed irrigation was detrimental to the net stormwater

retention (down to just 50% seasonal retention), but the use of tensiometer control significantly reduces

excess discharge down to that of a system without irrigation. This type of responsive irrigation would be

particularly well indicated when design teams are balancing priorities including the aesthetic and

biodiversity benefits of reliable vegetation survival. The biologically-derived planting medium,

comprising mostly of aged wood compost, made a supporting contribution, maintaining a lower

volumetric runoff coefficient compared to the mineral-based, in the systems without irrigation. The

biologically-derived material particularly improved individual event retention in wetter antecedent

conditions.

Processing the precipitation and discharge data in event-based pairs yields a similar pattern of influence in

the design factors. The group mean NRCS curve number of 90 was reduced to 89 in the sensor controlled

and no irrigation groups, but raised to 94 for the daily timed irrigation systems. Secondary influences

were noted from both the type of planting medium and the construction depth in the no irrigation and

sensor controlled systems, respectively. The notably low λ < 0.05 adds to a growing body of evidence for

Ia/S < 0.2 calibrations, but can only be directly used in similarly calibrated forecast models. To apply

these CN in a more common Ia = 0.2S model would require some form of transformation (Woodward et

al., 2003).

Compared to the irrigation program and type of planting medium, the depth of medium played relatively

little role in any of the hydrological characteristics analyzed. As the 50% variation in planting medium

depth, between 10 and 15 cm, played no significant role in the measured performance of the roofs, this

design variable remains a discussion point for vegetation planting depth requirements versus roof dead

load capacity. The lack of vegetation influence on Cvol illustrates that transpiration rates are not the only

important factor in the performance of this biotic component. It is suggested that the canopy capture

potential of green roof vegetation warrants further investigation with respect to different foliage

geometries and cuticle properties.

40

The peak runoff coefficient (0.12) was not sensitive to the four design variables tested. This indicates a

robustness making this number suitable for use in modelling any extensive green roofs within the bounds

explored here.

41

: Influences of Four Extensive Green Roof Design Variables on Annual Water Balance

Abstract

This study describes the effect of design decisions on the year-round hydrology of extensive green roofs,

through periods of sub-zero temperatures and warmer periods of drought and high irrigation demand. The

relative impacts of vegetation, type and depth of planting medium and irrigation programming have been

assessed on an annual basis; the annual mean volumetric runoff coefficient of 0.5 across all design

combinations was most significantly affected by the irrigation (daily Cvol = 0.6; sensor controlled Cvol =

0.5; none Cvol = 0.4). This trend was notable throughout the winter as irrigation was correlated with

increased snow accumulation throughout the first of two winter study periods, and with increased

discharge volumes upon thawing after both winters. Over a one-year period green roof modules receiving

daily irrigation or sensor controlled irrigation retained a mean additional 126 mm or 54 mm over those

without, offering an opportunity to evapotranspire excess stored stormwater. The use of a biologically-

derived planting medium was also associated with increase snow accumulation over a mineral based

alternative across both winters, but this did not result in significantly higher discharge volumes. This is

attributed to the higher water retention capacity of the biological medium.

42

Introduction

In Chapter 2, the impact of four categorical variables on event-based summer rainstorm hydrology was

presented. Whilst the resultant parameters are the preferred values for modelling event-based

performance, two aspects of annual water balance were not included: the amount of water supplied and

discharged through the two irrigation programs, and the accumulation and melting of snow during the

winter months. Considering and accounting for both of these is necessary to determine an annual

volumetric runoff coefficient (Cvol) and retention for extensive green roofs. Compared to individual event

analysis, aggregated water balances are a rapid and easy way to compare systems use for stormwater

control. In green roof studies the overall stormwater retention is often stated although these depend on

both the design configuration(s) of the green roof, and the climate in which the study has occurred.

Differences in rainfall distribution and evapotranspiration make comparison between regional studies

difficult, although a recent publication has used statistical analyses to collate existing study data, a

produce a more global assessment of green roof performance (Fassman-Beck et al., 2015).

The provision of irrigation to extensive green roofs remains a topic of much debate in both industry

practices and academic research (Breuning, 2013; Senjug, 2016). LEED certification discourages the

consumption of potable water for irrigation (Haselbach, 2008), however any moisture retained within the

green roof also reduces its capacity to retain stormwater. A lack of irrigation reduces vegetation survival,

diminishes cooling though shading and urban habitat potential, and irrigation water can contribute directly

to evaporative cooling in the summer season (MacIvor et al., 2016). The effect of spray irrigation on an

extensive green roof was assessed by Schroll et al. (2011), who adjusted irrigation either according to a

visual inspection (average 2.97 mm every two days), or in accordance with Portland municipal policy on

irrigation provision (average 3.2 mm every five days). These schedules were compared to control

modules which received no irrigation. The authors recorded just 6 rain events during the irrigated period,

according to a 12-hour inter-event period. Of these events, the irrigation significantly reduced the

retention only during the largest storm of 28.96 mm. An earlier qualitative observation about irrigation

was made by Spolek (2008). Significant discharge was recorded during a period without rainfall; the

authors concluded that the net retention would have been higher without this effect of ‘unused’ irrigation.

From a purely hydrological perspective, irrigation has been shown to be detrimental to green roof

performance. Hardin et al. (2012) used mass balance principles to model a system which couples a cistern

with an extensive green roof irrigated with the stored water. In a Florida climate, they predicted that the

addition of cistern with equivalent of 12.7 cm storage could achieve an annual retention of over 80%

compared to 43% without.

43

In temperate regions, which experience distinct seasonal changes in weather, most studies assessing

water retention by green roofs demonstrate seasonal fluctuations which reflect the seasonal changes in the

rate of evapotranspiration (Berghage et al., 2007; Graceson et al., 2013; Schroll et al., 2011; Spolek,

2008; Stovin et al., 2013). Two such studies from Ontario were published in 2009, each commented on

seasonal variation, but neither included periods of sub-zero conditions. Van Seters et al. (2009) reported

seasonal influences with individual event retention between 78 – 85 % in the summer and 39-64 % in the

spring and fall. Linden and Stone (2009) found a marked change between 81 % retention in August and -

25.5 % in October; the negative value was attributed: “evapotranspiration rates were low and the

vegetated roof was saturated from previous rain events”.

Negative retention values are caused by a green roof system discharging more water than received

precipitation over a given time period. One mechanism that can cause this phenomena, is the melting of

accumulated snow or ice. Globally, less research has been conducted into the behavior of green roofs

during frozen periods. Once the roof is frozen, precipitation patterns are temporally disconnected from

discharge hydrographs, and specialized equipment or methods are required to collect data. Despite these

considerations, a number of research papers do include data including ice and snow. In Connecticut,

Gregoire and Clausen (2011), reported that their green roof study (annual retention 51 %) received

approximately two months of snowfall as part of their annual budget, but did not present monthly or

seasonal values to assess the impact of sub-zero conditions. Berghage et al. (2009), working in

Pennsylvania, attributed cold weather, such as “snow or freezing conditions” as the cause for reduced net

retention during the winter months; their retention rate fell to below 20% in January, compared to 95% in

May, resulting an annual retention of 50%. In the UK climate, Graceson et al. (2013b) noted that just one

particularly large discharge event in January was due in part, to snow and ice melt, but also co-incident

rainfall. Teemusk and Mander (2007) include a more detailed analysis, reporting a two stage discharge

pattern in the thawing of a green roof. They speculate that firstly, the surface snow melts and percolates

through the planting medium, followed by the release of water held frozen within the pore spaces.

44

Methods

Green Roof Innovation Testing laboratory

The Green Roof Innovations Testing laboratory (GRITlab) experiment comprises twenty-four extensive

green roofs, each 2.86 m2 in area. The modules represent every combination of two levels of three

experimental variables and a fourth variable at three levels (2x2x2x3 experimental design). Two types of

planting media were included; one consisting of mostly mineral materials and the other contains mostly

biologically-derived material, composted bark fines. Test modules were constructed at two depths, 10 cm

and 15 cm, and in combination with two types of vegetation, Sedum. and native meadow plants. Irrigation

was delivered at three levels: none, sensor controlled or daily throughout the summer. Each module is

equipped with a 5TE sensor (Decagon Devices), which reports volumetric water content of the planting

media. The flow of discharged water which has percolated through the green roof layers is monitored

beneath each module using an adapted tipping bucket rain gauge (J Hill et al., 2015). Due to the

installation of these gauges, the modules are not in contact with the thermal mass of the building (see

Figure 3-1).

Figure 3-1 GRITlab modules raised above the roof deck to accommodated monitoring equipment.

Irrigation was provided to the modules via drip lines, with 300 mm spacing of the emitters. The flow rate

was fixed, but varied slightly during operation and was monitored (Seametrics SPX-050). The total

volume of irrigation water supplied to each module was controlled by altering the timing. The daily

modules received irrigation every morning, in series starting at 0900 hrs. The sensor controlled modules

each had a custom built, fluid-filled tensiometer installed (Irrometer) which was set to open the irrigation

valve (make demand) when media moisture tension < -25 kPa. In 2013, this irrigation program cycled

through the sensor controlled modules every three hours from 0000 hrs onwards (excluding 0900 hrs

when the daily modules were watered). Irrigation was only received by sensor modules if the valve was

open due to dryness of the media.

45

In 2013 the irrigation system was deployed between the first week of May and the last week of October;

the daily program delivered 5 minutes of irrigation each day, and the sensor program delivered 10

minutes of irrigation for each triggering event, according to the criteria above. During this first season,

both the daily and sensor driven programs resulted in significant outflow during irrigation events. This

rapid discharge over-topped the collection rain gauges prior to their adaptation and were not recorded. In

2014, both irrigation programs were reduced to just 2 minutes of irrigation per module. The sensor

controlled program was reduced to cycle just once every 24 hours, and the irrigation was employed only

between the first week of June and the last week of September. Reducing to 2 minutes of flow meant that

the temporal resolution of the system did not permit direct measurement of the individual module

irrigation amounts, so 2014 daily irrigation values are the mean of the whole daily irrigation period. The

irrigation system is drained and disconnected throughout the winter periods, November to April.

Rainfall was measured on site using a tipping bucket rain gauge (TE525M Texas Electronics) and

parameters used for the automated calculation of reference evapotranspiration were measured using an

adjacent weather station (Allen et al., 2005): Wind monitor (RM Young), CMP 11 pyranometer (Kipp

and Zonen), HMP45C relative humidity and temperature probe (Campbell Scientific). The weather station

is not equipped with a snow gauge, so during the months November – April, the daily precipitation record

from the national network has been used for water balance calculations. The Environment Canada

weather station at Toronto City (Station ID: 6158355) is located approximately 1 km north of the

laboratory (Canada, 2015). All precipitation data is presented in millimeters, where “the water equivalent

of snowfall is computed by dividing the measured amount by ten.” (Environment Canada, 2011). Further

details regarding the construction and layout of the GRITlab are available in Chapter 2. Snow

accumulation on the green roof modules was measured manually at four or five points using a metal ruler,

in accordance with the Environment Canada (2011) procedure “measured at several points that appear

representative of the immediate area and then averaged”.

Theory and Calculations

Monthly water balance calculations were made according to Equation 3-1, without including the change

in storage from one period to the next. The inputs are precipitation (P, mm) and irrigation (Irr., mm) and

the outputs discharge (Q, mm) and evapotranspiration (ET, mm). The change in storage (ΔS, mm) has not

been included in calculations, as the values are very small compared to the components over the monthly

timescale:

𝑃 + 𝐼𝑟𝑟. = 𝑄 + 𝐸𝑇 + ∆𝑆

Equation 3-1

46

In the summer months, the aggregated (monthly and seasonal) volumetric runoff coefficients (Cvol) have

been calculated as the sum of the individual event discharge volumes (Q, mm), as a proportion of the sum

of the event precipitation depth (P, mm) as presented in Chapter 2:

𝑪𝒗𝒐𝒍 = ∑ 𝑸

∑ 𝑷

Equation 3-2

In effect this excluded rainfall events which occurred shortly after 0900 hrs, and in 2013, any other

rainfall events which coincided with the sensor controlled program irrigating. This approach was applied

to all summer months including May to October. During the winter months of November to April, when

the national weather record was being used, precipitation depth data is available on a daily interval, there

is no irrigation in the water balance, and in freezing conditions there is a temporal disconnect between

precipitation and discharge. So for this period Cvol was calculated by summing all Q and P within each

month or across the whole season. Discharge due to irrigation is calculated by subtraction of the summed

event –based discharge from the total discharge recorded over the four months of June – September 2014.

ANOVA analyses were performed using NCSS 10 (NCSS, 2015), and regression trees using Orange Data

Mining (Demšar et al., 2013).


Irrigation and Water Retention

As the irrigation programming in 2013 had resulted in incalculable water balance values, the analysed

data in this study arise from the 12-month period October 2013 to September 2014. The total water

retained by an extensive green roof encompasses losses to evapotranspiration and changes in storage (ET

+ ΔS). In Figure 3-2, the total water retained over the whole year is grouped by irrigation program type,

which is the most consistently significant factor in summer season Cvol and curve number (Chapter 2). At

the end of September 2014, the modules contained an average 80% of the volumetric water content in the

media, compared to October 2013. From the group mean system storage of 28 mm (from CN 90, Chapter

2), this equates to approximately 6 mm change in storage between the start and end of the calculations.

This is a relatively insignificant compared to the group means of the total: None = 351 mm, Sensor = 405

mm, and Daily = 477 mm; which can therefore be largely attributed to evapotranspiration.

47

Figure 3-2 Extensive green roof annual total water retention for months October 2013 –September 2014, grouped by irrigation program.

The difference in storage is evident in the monthly water balance data shown in Figure 3-3. In the winter

months of November to February, all of the green roofs retain significantly more water than can be lost to

evapotranspiration, or can be accounted for by saturation of the green roof media. System storage in these

months includes significant snow depth accumulated on top of the planting layer. In April the

accumulated snow thawed and discharged leading to the negative water balance for modules which had

received either type of irrigation.

Figure 3-3 Monthly water retained group means for three levels of irrigation between October 2013 to September 2014. Reference ET from the GRITlab weather station.

-20

0

20

40

60

80

100

120

Tota

l w

ater

ret

ained

(m

m)

Reference ET

None

Sensor

Daily

48

Although the daily irrigation program delivered significantly more water (57 mm/month), than the sensor

controlled program (27 mm) in the months June-September, both discharged a similar excess (11 and 9

mm) (Figure 3-4). A regression tree analysis found that none of the three design variables; planting,

medium type, and depth, significantly affected the irrigation demand from sensor controlled systems in

the months over which irrigation was provided.

.

Figure 3-4 Input and output volumes associated with two irrigation programs

The provision of irrigation through a drip-line rather than spray may cause the development of localized

macropores and preferential flow paths, and the lack of capillary forces moving water laterally through

the granular media may also contribute to these discharge volumes (Rowe et al., 2014). This may have

resulted in the sensor controlled system making demand for additional irrigation owing to lateral

heterogeneity of the applied water.

Winter Climate and Snow Accumulation

Throughout the winter months of November to April, Toronto normally has a climate with periods of

thaw and relatively short periods below -10 °C (Environment Canada, 2013). As shown in Figure 3-5, the

coldest month, January, normally experiences nine days with minimum temperatures < -10 °C, but up to

five days with minimum temperature over 0 °C. Consequently, snow cover is intermittent even in the

coldest periods, with measurable snow (≥ 1 cm) normally being observed on just 20 days of the month

through January and February. Overall, the depths of precipitation events are evenly distributed between

the months, with a very slight tendency for heavier events in November (three days with > 10 mm).

49

Figure 3-5 Winter months climate normal snow cover, daily minimum temperatures and daily precipitation depth from 1981-2010 data in Toronto, Ontario (Environment Canada, 2013).

The mean daily temperature and precipitation data for the two winters in this study are presented in

Figure 3-6. In February 2015 the mean daily air temperature never rose above 0 °C, which is unusual

compared to the normal conditions (Environment Canada, 2013). A total of 331 mm precipitation fell

between November 2013 - April 2014, and 244 mm in November 2014 – April 2015. The total normal

0

5

10

15

20

25

30

Nov Dec Jan Feb Mar Apr

Nu

mb

er o

f d

ays

Normal snow cover - Toronto, Canada

>= 1 cm

>= 5 cm

>= 10 cm

>= 20 cm

0

5

10

15

20

25

30


Num

ber

of

day

s

Normal days with minimum temperature - Toronto, Canada

<= 0 °C

< -2 °C

< -10 °C

< -20 °C

0

5

10

15

20

25

30


Num

ber

of

day

s

Normal days with precipitation - Toronto, Canada

>= 0.2 mm

>= 5 mm

>= 10 mm

>= 25 mm

50

precipitation for those months is 384 mm, so both periods were also relatively dry. A single day of > 25

mm precipitation occurred in both Aprils, which is also slightly unusual, as November is a more common

month for large rainstorms. In both November and April, the mean daily temperature only occasionally

dips below zero and most precipitation is received as rainfall.

Figure 3-6 Mean daily air temperature (dashed line) from GRITab and precipitation record (bars) from Toronto City weather station for the periods encompassing November 2013 to April 2014, and November 2014 to April 2015.

Forty-seven observations of snow depth across the twenty-four green roof modules were made over the

two winters: Twenty-nine between November 2013 and April 2014, and eighteen between November

2014 and April 2015. Once snow fell in December 2013, cover was continuous with only partial thaws in

every month until late March and early April 2014 when the remaining snow melted. Some of the green

roofs had accumulated deeper snow than the official record in December 2013. As these elevated green

roof modules would cool in response to air temperature faster than the terrestrial ground temperature, this

0

5

10

15

20

25

30-20

0

20

Nov-13 Dec-13 Jan-14 Feb-14 Mar-14 Apr-14

Pre

cip

itat

ion

(m

m)

Dai

ly m

ean

air

tem

per

atu

re (

°C)

Winter 1: 2013-14

0

5

10

15

20

25

30-20

0

20

Nov-14 Dec-14 Jan-15 Feb-15 Mar-15 Apr-15

Pre

cipit

atio

n (

mm

)

Dai

ly m

ean a

ir t

emper

ature

(°C

)

Winter 2: 2014-15

51

may account for the disparity in early winter snow depth. By late winter, all recorded snow depths were

lower than on the ground: by late February rooftop measurements were over 10 cm lower than the at the

weather station. This may be due to losses from wind scour or due to freeze/thaw compaction of the snow

without discharge from the modules.

In December 2014 there was only one measurable snowfall, and no prolonged period of cover until

February 2015. As shown in Figure 3-7, snow depth on many of the green roof modules exceeded the

official record of snow on the ground after the first fall, but by March this had been reversed, with all

modules accumulating less snow than on the ground.

Figure 3-7 Twenty-four modules accumulated snow depth throughout winters 2013-14 and 2014-15, plotted over ground level data.

Spatial autocorrelation was performed on the snow depth data from Nov 2013 – April 2014, and the

distribution of Moran’s I is presented in Figure 3-8. Moran’s I test results can range between 1 and -1,

where higher values indicate more clustered data and low/negative values indicate an even distribution of

the parameter across space. The Moran’s I data had a mean of 0.03 (σ = 0.09). This small value indicates

that the snow depth was not even, but was randomly distributed across the experimental area; not

influenced by the roof shape or aspect.

0

5

10

15

20

25

30

Snow

dep

th (

cm)

Snow on Ground (cm)

0

5

10

15

20

25

30

Snow

Dep

th (

cm)

Snow on Grnd (cm)

52

Figure 3-8 Moran’s I from winter 2013-14 centred about zero, indicate no significant geospatial clustering or trends in the snow depth across the GRIT lab experiment.

The accumulation of snow data was not aggregated in any fashion and all observations were treated as

independent, although co-variance and correlation naturally occurred between some consecutive

measurements. One-way MANOVA analysis (p < 0.05) of the four design factors, on all 47 independent

observations over both winters found the type of medium to be a significant indicator snow depth most

frequently (see Appendix C) i.e. The difference between the two types of medium was statistically

significant on 38 separate observations across both winters, as shown in the group means in Figure 3-9.

The biologically derived, compost planting medium was significantly associated with increased snow

accumulation. The occasions on which the medium was significant were not clustered particularly early

or late in the seasons, or after significant thaw events. The provision of irrigation was also significant on

15 of these dates, but only throughout the first winter, shown in Figure 3-9. The pattern in the first year

irrigation indicates that the modules without irrigation retained significantly less snow; despite the fact

that there was no irrigation applied past October in 2013 of that year. The type of planting was a

significant factor on the first three observations dates in 2013, with the meadow planting retaining more

snow.

53

Figure 3-9 Mean snow depth, grouped by medium type (top), and irrigation (bottom) throughout winters 2013 and 2014.

It has been reported elsewhere (MacIvor et al., 2013) that the growth of the meadow plants on GRITlab

modules was statistically sensitive to the medium and provision of irrigation. Whilst the vegetation was

not a direct indicator for snow accumulation, as shown in Figure 3-10, the density of the vegetation was

reduced on the mineral medium, particularly without irrigation. This vegetation was left standing

throughout the winter months and may be factor in the retention of fresh snowfall, which may otherwise

blow away. Lundholm et al. (2014) found that canopy biomass contributed to greater snow accumulation

on green roof test plot.

0

5

10

15

20

25

16-Dec-13 16-Jan-14 16-Feb-14

Sn

ow

Dep

th (

cm)

Medium 2013-14

Biological Mineral

0

5

10

15

20

25

11-Dec-14 11-Jan-15 11-Feb-15

Sn

ow

Dep

th (

cm)

Medium 2014-15

Biological Mineral

0

5

10

15

20

25

16-Dec-13 16-Jan-14 16-Feb-14

Snow

Dep

th (

cm)

Irrigation 2013-14

None Sensor Daily

0

5

10

15

20

25

11-Dec-14 11-Jan-15 11-Feb-15

Snow

Dep

th (

cm)

Irrigation 2014-15

None Sensor Daily

54

Figure 3-10 Native meadow vegetation mix grown on: a) biological medium with daily irrigation, b) mineral medium with daily irrigation, and c) mineral medium without irrigation. Photographs taken 20 September 2013 (University of Toronto, 2013).

Winter Cvol

As stated above, the net seasonal volumetric runoff coefficient (Cvol) for each module during the summer

months were calculated by summing individual event-based rainfall and discharge volumes, and the

winter Cvol by summing daily data. The seasonal Cvol were similarly calculated for May-October and

November - April, and the ranges represented by all available design configurations are presented in

Figure 3-11. The mean, summer season Cvol, was 0.4 (σ = 0.10), whilst the mean winter season Cvol is 0.6

(σ = 0.10).

Figure 3-11 Mean volumetric runoff coefficients from 23 modules, over 12 months of summertime events May-Oct in 2013 and 2014 and 12 months of wintertime balance, Nov-April in 2013-2014 and 2014-2015.

55

Throughout the first winter, the proportional volume of discharged water increased month on month,

owing to partial thaws in every month until late March and April 2014 when the remaining snow melted

resulting in the two highest monthly Cvol (1.3 and 1.1 respectively) (Figure 3-12). March had the highest

Cvol in both years owing, in part, to the discharge of meltwater, but also as input precipitation is at the

lowest for the whole year (16 mm and 11 mm in 2014 and 2015 respectively). The high Cvol in April 2014

is attributable to the meltwater discharge coupled with a deep rainstorm later in the month.

Figure 3-12 Group mean runoff coefficients per month through May 2013 to April 2015.

In the second winter a different pattern emerges. Up to 20 cm of snow accumulated in a single event in

early December 2014 (Figure 3-7), followed by two warm periods before the New Year (Figure 3-6); this

accounts for the higher Cvol in December (Figure 3-12). Up to 20 cm of snow persisted through February

2015 as there was no thaw in that month (Cvol = 0); the following Cvol in March 2015 (0.8) was lower than

in March 2014 (1.3). Despite the influences and interactions of various design factors on the accumulation

of snow, the only single factor demonstrating an overall influence on winter season Cvol was the irrigation

F (2,17) = 5.18 (p = 0.02). The group mean Cvol for the entire of both winter periods were: None = 0.5,

sensor = 0.6, and daily = 0.7. In Figure 3-13, some distinction between the irrigation group means is

evident in the late first winter, consistent with the snow accumulation seen in Figure 3-9. Here, a lack of

irrigation reduces the snow accumulation, and in so doing, reduces the Cvol. However, the effect of

irrigation on Cvol is more pronounced throughout the second winter, when irrigation had not influenced

snow depth.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

May

-13

Jun

-13

Jul-

13

Au

g-1

3

Sep

-13

Oct-

13

No

v-1

3

Dec-1

3

Jan

-14

Feb

-14

Mar-

14

Ap

r-1

4

May

-14

Jun

-14

Jul-

14

Au

g-1

4

Sep

-14

Oct-

14

No

v-1

4

Dec-1

4

Jan

-15

Feb

-15

Mar-

15

Ap

r-1

5

Cvo

l

56

Figure 3-13 Group mean runoff coefficients by irrigation program, for months through November – April 2013-14 and 2014-15.

Although the medium type was significant in the accumulation of snow depth, this did not translate into a

significant impact on monthly or winter season Cvol. This may be due to the biologically derived medium

being associated with increase snow accumulation, but also with increased water holding capacity

(Chapter 2, 4, 5), causing any additional meltwater from the snow to be retained within the planting

medium during thaw conditions.

Annual Cvol

The water balances from summer months have been taken from Chapter 2 and still exclude the

quantitative input and outputs of irrigation, as: the impact of the irrigation is reflected in the individual

rainfall event retention, and in the daily irrigation program the additional 46 mm of applied irrigation,

which is not discharged (57 mm in – 11 mm out), would make this design decision appear optimal,

erroneously. The ability of green roof modules to absorb maximum irrigation water is not the research

question. When the winter months Cvol are combined with the summer data, the net mean annual Cvol

across all systems is 0.5, as shown in Figure 3-14 (R2 = 0.954). This is a 0.1 increase over the summer

season Cvol reported previously (Chapter 2), reflecting the decrease in performance of the green roof

systems during freezing conditions and reduced evapotranspiration potential.

0.0

0.5

1.0

1.5

Cvol

None Sensor Daily

57

Figure 3-14 Annual volumetric runoff coefficients for extensive green roofs, calculated from 24 months of data spanning May 2013- April 2015. Each cell contains: Design factor ‘level’, group mean value, and (# modules).

The influential design factors remain ranked as in the summer season, with irrigation having the greatest

influence. Using a green roof system with no irrigation can retain up to 60% of the annual precipitation,

whilst the use of a daily program during just 4 to 6 months of the year can reduce the retention to just

40%. The use of a biological planting medium increases retention from 50% to 60%, but only in modules

with no irrigation.

Conclusions

With a moderate irrigation program is applied as in the summer of 2014, with 2 minutes of water being

supplied daily, or only upon demand from the sensor controlled system, the amount of additional

discharge being created from each of these systems was very similar (3 or 4 mm per month). The daily

irrigation consumed around five times as much water (20 mm per month compared to 4 mm for sensor

controlled). This water consumption could be a net cost to the custodians of the roof if a municipal supply

is being used, but many systems are installed on LEED accredited projects where potable irrigation water

is prohibited (Haselbach, 2008), or as part of a site-wide stormwater management strategy where water

from a cistern is being used to meet an annual retention target (Cheung, 2016). In these cases, where there

is an abundance of irrigation water available, the additional 16 mm water retained in the daily irrigated

beds represents an opportunity to capitalise upon the summer microclimatic effect of evaporative cooling

from the green roof.

As anticipated, cooler weather results in higher runoff volumes and the mean Cvol value for months

November-April is 0.2 higher than in the other six months of the year (0.6 compared to 0.4 for May –

October). Irrigation in summer months was associated with increased snow accumulation on extensive

Annual Cvol

0.5

(23)

None

0.4

(8)

Biological

0.4

(4)

Mineral

0.5

(4)Sensor

0.5

(8)

Daily

0.6

(7)

58

green roofs, resulting in increased meltwater discharge during thawing periods and higher seasonal Cvol

(0.7 for daily irrigation, compared to 0.5 without). A possible mechanism for this is that irrigation

improves vegetation 3D structure, protecting accumulated snow from wind scour. Biologically derived

planting medium also increases snow accumulation, possibly through the same planting. However, this is

not correlated with increased discharge owing to the improved water holding capacity of the medium.

The annual Cvol of 0.5 is increased over summer Cvol (0.4), by inclusion of months which have lower

evapotranspiration rates, and the rapid influx of meltwater during thaw events. However, whilst the

annual water balance values are useful for assessing and comparing the annual performance of green roof

systems, they are unsuited for modelling rainstorm responses of a green roof system.

59

: Physicochemical Properties of Extensive Green Roof Planting Media

Abstract

Increasing the biological component of the planting media has a positive effect of increasing water

holding capacity for extensive green roofs, which is beneficial for stormwater control and vegetation

health. However, the biological component in planting media is a source of phosphorus, which can leach

out over time, polluting the discharged water from a site, and contributing to nutrient excess in natural

water bodies. Based on a short sampling period of discharged water from five-year-old green roof

modules, the combined influences of discharge volumes, and total phosphorous (TP) concentrations are

considered over three design factors: biological (bark compost) and mineral based planting medium, 10

cm and 15 cm planting medium depth, and whether daily irrigation has been provided during preceding

summers. The sampling period occurred in March and April in 2016, and yielded a mean TP of 1.1 mg/L

in mineral based medium and mean of 5.5 mg/L from a biologically-derived medium. Within the

biological medium, a secondary influence on the TP concentrations was the previous provision of

irrigation, and finally, the 15 cm design increased TP slightly over the 10 cm planting medium depth.

Two contextual findings resulted; discharge TP concentrations were not correlated with previous testing

of the planting medium for water extractable TP, and that if all green roofs comprised matured compost,

the annual loading from the current extent of installations in the City of Toronto could be up to 900 kg.

The electrical conductivity in discharged water was compared with the electrical conductivity of the pore

water within two types of soilless planting medium. The pore water electrical conductivity varied

according to prior irrigation; the electrical conductivity of the discharge water varied with medium type.

Significant salinity was not found in either the discharged or pore water.

60

Introduction

This chapter presents two studies into the physicochemical properties of the experimental green roof

modules discussed from a hydrological perspective in chapters 2 and 3:

1. Almost uniquely amongst low impact development technologies, green roofs are often a source of

phosphorous in storm water, albeit the discharge concentrations are subject to a number of

factors. This study determines the concentration of total phosphorous (TP) associated with

matured green roofs (~5 years) of varying designs. In addition to medium type, the effect of

irrigation leaching, the additional initial reservoir of 50% extra depth of growing medium, and the

nutrient uptake of two types of planting will be encompassed by the analysis of both media and

discharged water samples. Using data from March-April 2106, the TP concentrations in discharge

water are used to estimate the potential phosphorous loading that green roofs in the City of

Toronto could contribute to Lake Ontario and subsidiary receiving water bodies.

2. Long-term management of extensive green roofs often encompasses some irrigation to maintain

vegetation health through droughty periods. Consecutive cycles of irrigation and

evapotranspiration are associated with increased soil salinity in the long term. Where green roofs

are specifically designed and used to cycle water in this fashion, an accumulation of salts in the

growing medium could present problems for the vegetation health. In this study the bulk

electrical conductivity of the media will be used to determine the volumetric water content (θ),

the permittivity of the planting medium types (ɛ0), and so the conductivity of the pore water (σp).

Background

Phosphorus

Whilst green roofs provide retention of surplus stormwater in urban settings, as a low impact development

strategy, they often raise concerns as they are well documented as contributors, rather than mitigators, of

environmental macronutrients, particularly phosphorous (Czemiel Berndtsson, 2010; Hashemi et al.,

2015). As green roof planting mixtures are usually a blended product, rather than a natural soil, there is

some degree to which this can be designed for or against. By creating their own planting media blends,

varying only the proportion of compost, Moran et al., (2004) and Toland (2010) concluded that the

composted materials where the primary course of phosphorous in the discharged waters from the green

roofs studied. However, regardless of the initial composition, without the use of additional fertilizer, as

green roofs mature and develop, TP concentrations in the discharged water will reduce as mobile

phosphorus is leached or flushed from the media (Harper et al., 2015; Köhler et al., 2002; Van Seters et

al., 2009). The dominance of initial phosphorous composition over the effects of maturation is evident in

the distribution of findings from previous studies shown in Figure 4-1. Amongst the studies illustrated

61

are the four annual flow weighted concentrations from (study B, Toland, 2010) which reported ten times

increased TP in a blend with 15% additional compost, and the reduction seen after the first nine months

aging in a proprietary product (study F, Harper et al., 2015).

Figure 4-1 Summary of previous studies assessing the total phosphorous discharge from extensive green roofs: A (Gregoire and Clausen, 2011); B (Toland, 2010); C (Berndtsson et al., 2006); D (Teemusk and Mander, 2007); E (Van Seters et al., 2009); F (Harper et al., 2015); G (Beck et al., 2011). Many of the mixtures contain lightweight expanded aggregate (LEA).

The addition of high-organic, composted materials is strongly correlated with improved stormwater

retention (Nagase and Dunnett, 2011), creating a driver for their incorporation into green roof planting

mixtures. In green roofs constructed from biologically derived planting media such as wood, bark or yard

waste compost (Chapter 6), excess phosphorous can be of particular concern, as these materials are

naturally much higher in organic matter and associated macronutrients at the time of installation,

compared to the recommendations for mineral based green roof media (e.V., 2008). Many studies into

the composition of green roof media do not include information about the source of the compost

component. The initial nutrient content of composted materials is fundamentally affected by the source

material or feedstock. Materials such as wood, and paper products typically contain much lower nitrogen

< 1 yr

1 yr

1 yr

7 yrs

3 yrs

1 yrs

1 yrs

2 yrs

1 yr

1 yr

< 1 yr

0 yr

0 yr

0 yr

0.01 0.1 1 10 100

75% LEA, 15% compost, 10% perlite

coarse LEA

LEA fines

lava rock/peat/LEA

Tartu™

lava/peat

lava rock/compost/peat/LEA/sand

Sedum mat

85% coarse LEA/15% compost

85% LEA fines/15% compost

GAF Gardenscapes™

Pro-grow™ mix with 7% biochar

Pro-grow™ extensive mix

GAF Gardenscapes™

AB

BC

DC

EC

BB

FG

GF

TP in green roof discharge water (mg/L)

62

and phosphorous content compared to manure or yard waste (Chatterjee et al., 2013). Using a poultry

manure based compost in their green roof blend, Rowe et al. (2006) recommended that any use of

compost be avoided in green roof media, citing the potential for excess nitrogen and phosphorus

concentrations in the discharge water and possible physical degradation of the media. Instead the authors

recommended the addition of a slow-release fertilizer granule. However, a more nutrient rich compost

was recommended for green roof blends (20 % v/v) by Young et al. (2014), who recognised that potential

for nutrient leaching, but observed significant improvement of plant establishment in using a manure

compost, over a wood compost. When the compost forms the bulk of the product though (> 50 %)

nutrient rich sources of compost are avoided by commercial manufacturer in favour of matured wood or

bark based composts (Schumacher, 2015; Stroud, 2016).

Electrical Conductivity

Electrical conductivity in water (σw) is often employed a surrogate for total dissolved solids (TDS), as an

indicator of overall water quality (Government of Canada, 2009). Conversion between the two parameters

requires specific calibration to account for differences in the molar conductance of different ions (Walton,

1989). This type of measurement still provides useful information about the discharged water quality for

environmental and/or reuse applications (Miguntanna et al., 2010; Raviv et al., 2002). Most notably if

discharged water is to be reused for landscape irrigation, high levels of TDS (described as salinity) can

inhibit root water uptake by plant by increasing the osmotic potential of the water within the planting

medium. The Food and Agriculture Organization of the United Nations (FAO) recommend irrigation

water to have σw < 700 μS/cm, and advise caution in the range 700 < σw < 3000 μS/cm (Ayers et al.,

1985).

Dissolved solids in pore-water are accrued by leaching from the planting medium or soil, but also by

deposition from irrigation and rainwater, which then evaporates rather than being flushed from the

system. As green roofs are designed to capitalise on the water balance and thermal benefits of

evapotranspiration, it is theoretically possible that these systems may increase in salinity as they go

through subsequent cycles of wetting and drying, which can harm vegetation health (Moritani et al.,

2013). However, none of the green roof studies summarized in Figure 4-2, found particularly elevated σw

in the discharged water.

63

Figure 4-2 Summary of previous studies which state the electrical conductivity of discharge from extensive green roofs: A (Beecham and Razzaghmanesh, 2015); B (Gnecco et al., 2013); C (Göbel et al., 2007); D (Buffam et al., 2016); E (Van Seters et al., 2009); F (Buccola and Spolek, 2011)

In situ measurements of electrical conductivity can also be used to determine the volumetric water content

of unsaturated porous media, the composition of the solid particles and even the tortuosity of the pore

network between them (Friedman, 2005).

< 1 yr

3 yrs

< 1 yr

1-3 yrs

1 yr

0 yr

0 50 100 150 200 250 300

Mineral

Vulcaflor™

Aggregated figure (n < 5)

Organic mix (with 50% compost)

Tremco™

lava rock/sand/compost

lava rock/sand/compost

AB

CA

DE

F

Electrical Condcutivity of green roof discharge (μS/cm)

64

Methods

Green Roof Experimental Set up

The Green Roof Innovation Testing laboratory (GRITlab) is located on roof of John H. Daniels faculty of

Architecture, Landscape and Design at the University of Toronto, Ontario, Canada. This multivariate

experiment, constructed in May 2011, comprises 24 green roof modules (2.86 m2), randomly arranged, to

assess the impact of design changes to four parameters on stormwater hydrology and water chemistry.

The choice of vegetation, planting medium type and depth are considered at two levels, whilst irrigation

provision is trialed at three levels. The details of each level within the experimental variables are given

below in Table 4-1.

Table 4-1 Levels of four experimental variables being considered at the GRITlab

Variable Level a Level b Level c Vegetation Mix of 34 Sedum species Mix of 16 native species. - Planting medium type Mineral based Biological - Planting medium depth 10 cm 15 cm - Irrigation None Sensor controlled Daily

Details regarding the planting have been published previously (MacIvor et al., 2013) and manufacturers

specifications for the commercial planting media are available (Bioroof Systems, 2011); the mineral

based product contains a mixture of porous aggregates and the ‘biological’ largely comprises composts

pine bark fines. Drip line irrigation was used to reduce potential losses of water associated with spray; the

flow rate was monitored (Seametrics SPX-050). Irrigation programs were employed throughout the

summer months, typically from May/June until September/October. Volumetric water content, electrical

conductivity and temperature within the planting medium were monitored using 5TE sensors embedded at

half height within the modules (Decagon Devices). A modified tipping bucket rain gauge was used to

measure the discharged water from each individual module (J Hill et al., 2015), and an unaltered tipping

bucket rain gauge on site collected precipitation data (TB6, Hydrological Services).

Phosphorous

In October 2014, when the installation had aged 3.4 years, a media sample was collected from each

module and tested for water extractable total phosphorous (WETP). Slurries of sieved (1.7 mm), air-dried

planting substrate, 1:10 (w/w), were prepared after 10 minutes shaking (150 rpm) in analytical grade

water (Fuhrman et al., 2005). Samples were filtered (VWR 494) and tested for total phosphorus using a

persulfate digestion with ascorbic acid reduction method, and a SMART 3 colorimeter, in accordance

with the Low-range total Phosphorous, La Motte Standard Operating procedures (SOP) (La Motte, 2012).

Values are expressed in ppm, the equivalent mg/kg of medium from which the extract was prepared. In

65

house calibration of this method had been undertaken for these samples which would otherwise fall

outside of the normal specifications of the manufacturers SOP (Ogura, 2014).

In March and April of 2016, when the installation had aged a total of 4.8 years, samples of discharged

water were collected from six modules with the design combinations shown in Table 4-2. During this

period, between four and six samples were collected according to availability; on some occasions, frozen

conditions had prevented discharge from just some modules. Collection occurred as follows: Meltwater

discharge (7/8 March), discharge resulting from rainfall (14/15/29 March), and a combination of both (11

April).

Table 4-2 Subset of green roof modules tested for TP in discharged water

ID Planting medium Depth (cm) Irrigation A Mineral 10 Daily B Mineral 10 None C Biological 10 Daily D Biological 10 None E Biological 15 Daily F Biological 15 None

Again the analyses were performed using persulfate digestion with ascorbic acid reduction method, and

the SMART 3 colorimeter (Motte, 2012). The vegetation variable was excluded from this experiment

after demonstrating no significant influence on the volume of stormwater discharged (Chapters 2 and 3),

or the TP within the medium samples tested previously. As the discharge water samples had originated

from both meltwater and rainfall, which may have different residence times in the green roof media, the

results were tested in time-series Wald-Wolfowitz Runs Tests for Randomness using the median as a

reference value, and correlated with mean daily temperature for the sampling days. Neither test

demonstrated any significance in the results.

Water balance calculations from a 24 month period between May 2013 and April 2015 have been

undertaken and are presented in Chapters 2 and 3. With reference to these data, summarized in Table 4-3,

the discharge TP concentrations, and mean precipitation statistics (Environment Canada, 2013) an

estimate of the overall annual load of TP per condo area will be calculated. As shown in Equation 4-1, it

is the product of the average new condo size (Acondo, m2), annual mean precipitation (pann, mm.), the

volumetric runoff coefficient (Cvol) and the concentration of total phosphorous in the discharged water

(TPgr, mg/L):

𝐺𝑟𝑒𝑒𝑛 𝑟𝑜𝑜𝑓 𝑇𝑃 𝑙𝑜𝑎𝑑 𝑝𝑒𝑟 𝑐𝑜𝑛𝑑𝑜 = 𝐴𝑐𝑜𝑛𝑑𝑜 × 𝑝𝑎𝑛𝑛. × 𝐶𝑣𝑜𝑙 × 𝑇𝑃𝑔𝑟

Equation 4-1

66

The comparison figure of annual load TP per condo was calculated in Equation 4-2 as the product of the

daily waste water output per day, per capita (Vww, L) and the mean TP concentration in municipal

wastewater (TPww, mg/L).

𝑇𝑃 𝑙𝑜𝑎𝑑 𝑝𝑒𝑟 𝑠𝑖𝑛𝑔𝑙𝑒 𝑜𝑐𝑐𝑢𝑝𝑎𝑛𝑐𝑦 𝑐𝑜𝑛𝑑𝑜 = 𝑉𝑤𝑤 × 365 × 𝑇𝑃𝑤𝑤

Equation 4-2


On 9th July 2013, when the installation had aged 2.2 years, grab samples of discharged stormwater were

collected from the draining modules the morning after a rainstorm exceeding a 100-year return period (Di

Gironimo et al., 2013). On 8th July 96.8 mm of rainfall was recorded at Toronto City weather station, the

majority of which fell in the two hours between 4:30 and 6:30 pm (Environment and Climate Change

Canada, 2013). A storm of this magnitude meant that discharge water samples could be manually

collected from all 24 modules, which were still draining the following morning. The samples were then

immediately tested for pH (SenTix 22 pH probe, CanLab 607 pH meter), turbidity (66120-200, VWR

scientific) and electrical conductivity (EC 110, FieldScout).

Calibration of the bulk dielectric permittivity (ɛb) to volumetric water content (θ) was undertaken in the

laboratory in accordance with the instrument manufacturers guideline for the two type of planting media

(Cobos and Chambers, 11AD). The media were air dried before using the 5TE sensor to record the raw

dielectric permittivity in triplicate. Two samples of this material were then weighed and dried (90 °C ± 5

°C, to constant weight) to determine the corresponding water content. A portion of water was mixed into

the bulk volume of the planting substrate and permitted to equilibrate throughout, before conducting the

testing again. The sequence was repeated until the medium became fully saturated.

Three weeks of field data (~60,000 data points) spanning 23rd October to 3rd November 2014 were used to

perform calibration of ε0 for each material by linear regression. This period was chosen as θ was generally

> 0.2 and the media temperatures were around 10 °C. Calibration is unreliable in dry media (Decagon

Devices, 2016) and ice has very different permittivity to water. Two modules were excluded owing to

equipment failure, so that a total of 11 modules of each medium type were used. ε0 is an offset which

conceptually equates to the dielectric permittivity of the media in the absence of any bulk electrical

conductivity and is required to calculate the electrical conductivity of the pore water, σp (dS/m) from the

bulk electrical conductivity, σb (which includes contributions from media solids, air and water) and the

bulk dielectric permittivity, εb (Equation 4-3). The pore water electrical conductivity is also a function of

the dielectric permittivity of the water, εp which is itself a function of soil temperature, Tsoil (Equation

4-4):

67

𝜎𝑝 =휀𝑝𝜎𝑏

휀𝑏 − 휀0

Equation 4-3

휀𝑝 = 80.3 − 0.37(𝑇𝑠𝑜𝑖𝑙 − 20)

Equation 4-4

The electrical conductivity of the pore water can then be employed to estimate the osmotic potential of

the water (ψ0) within the media (Encyclopedia of Soil Science, 2007), see Equation 4-5:

𝜓𝑜(𝑘𝑃𝑎) = −36 × 𝜎𝑝(𝑑𝑆 𝑚⁄ )

Equation 4-5


Phosphorous

Water extracts of the planting media samples found that only the type of medium, mineral or biological,

was of any significance as an predictor for water extractable total phosphorous (WETP), F (1,2) = 74.17

(p < 0.05). The mean WETP from the biological media was 90 mg/kg (σ = 33 mg/kg), compared to the

mean 46 mg/kg (σ = 25 mg/kg) extracted from the mineral media samples (Figure 4-3). The other three

variables of irrigation (3 levels), medium depth (2 levels) and planting type (2 levels) were not significant

in relation to the extractable phosphorous. No distinction between modules which have received daily

irrigation over five summers and those that had not, suggests that all of the most rapidly and readily

leachable phosphorus has left the systems at this time.

68

Figure 4-3 The water extractable total phosphorous in twenty-four, 3.4-year-old green roof modules is distinguished only by the type of planting medium.

Between four and six samples were analysed from each of the six green roof modules during March and

April 2016 (Figure 4-4).

Figure 4-4 TP in discharge water from six green roof modules.

With conflicting reports from previous literature on the relationship between TP discharge and seasonal

conditions, it is difficult to place these observations into an annual context. Although their study had the

lowest mean TP value in the summary above (Figure 4-1), Gregoire and Clausen (2011) found that their

highest value was recorded in meltwater from accumulated snow. Toland (2010) reported a significant

drop in the TP loadings from their green roof modules between the first season Sept-Nov and the final

season of their study Jun-Sept. However, as acknowledged by the author, much of this change may be

attributable to the rapid leaching of the phosphorous from the new installations. During a winter-long

study in Missouri, Harper et al. (2015) noted elevated TP in late January and early February, which had

been attributed to the dormancy of the plants reducing their physical stabilisation of the media at this time

(Harper, 2013). An in-depth study was undertaken in Ohio, by Buffam et al. (2016), who found that TP in

green roof discharge increased with warmer temperatures, 41% of the variance in their measurements was

accounted for by the mean temperature in the previous week. This, the authors hypothesized, may be due

to microbial mineralization of organic matter within the media, being increased in warmer temperatures.

However, comparisons can be drawn between the difference design combinations represented by these

modules. A regression tree approach has been employed to rank the impacts and is presented in Figure

4-5 (R2 = 0.991). The medium type had the greatest effect on the TP concentration being discharged from

0

1

2

3

4

5

6

7

8

9

TP

(m

g/L

)

Sampling date

A

B

C

D

E

F

69

modules of this age. Within the biological media systems, daily irrigation reduced TP. As all of the

installations were constructed at the same time and from the same materials, it can be concluded that the

impact of irrigation in this data illustrates the flushing of more TP prior to the sampling.

Figure 4-5 Regression tree illustrating the relative influence of three design factors on the TP concentrations in samples taken March/April 2016.

Although both the WETP and the later TP measurements were most strongly influenced by the medium

type; with the biological, compost based, medium producing more phosphorous, there was no significant

correlation found between the two data. This suggests that laboratory testing of planting media may not

result in accurate predictions regarding leaching behaviour. This would be exacerbated by many routine

laboratory testing methods which employ dilute acids to extract the nutrient more fully from the medium

(Legg, 2016).

Accounting for the stormwater retention, and the TP concentrations discharged from the materials, the

annual loading values have been calculated using a normal annual precipitation of 786 mm (Environment

Canada, 2013) (Table 4-3).

Table 4-3 Group mean volumetric runoff coefficients for six extensive green roof design combinations. *n=2, encompassing both types of vegetation, apart from case E, where n = 1 (meadow planting only).

ID Group mean*

Cvol Runoff (mm)

TP conc. (mg/L)

Annual TP (g/m2 roof)

A 0.53 417 1.1 0.5

B 0.49 385 1.1 0.4

C 0.57 448 4.1 1.8

D 0.34 267 5.5 1.5

E 0.54 424 7.8 3.3

F 0.40 314 4.8 1.5

Group mean

4.1 mg/L

(6)

Biological

5.5 mg/L

(4)

None

6.6 mg/L

(2)

Daily

4.4 mg/L

(2)Mineral

1.1 mg/L

(2)

70

The theoretical annual TP employs the key assumption that the limited TP measurements made were

representative of the annual TP concentrations being discharged from the green roofs throughout all

seasons. The highest annual load of TP would come from a green roof 15 cm deep and containing

biologically derived media which had received no supplementary irrigation in five years.To place this into

context, the largest wastewater treatment plant in Toronto received mean influent TP concentration of 7.5

mg/L in 2015 (Toronto, 2016), approximately 250 L wastewater is produced per capita daily

(Environment and Climate Change Canada, 2016) and the average new condo size is approximately 74 m2

(Perkins, 2014). The highest loading green roof described above, covering just a single storey condo, with

single occupancy would produce less than half the annual TP load of the occupant (0.24 kg compared to

0.69 kg). As many new residential towers in urban centres comprise 50 or more storeys, the annual

phosphorous contribution from the green roof becomes increasingly insignificant. Furthermore, the TP

load arising from the occupants is underestimated owing to the assumption of single occupancy, and as

the municipal wastewater TP concentration used comes from a watershed encompassing some combined

sewer systems.

Within the wider urban landscape context, green roofs are not the only significant contributor of TP

loading, as it is common practice to apply fertilizer both to commercially maintained green space and

over 50% of residents in Toronto apply fertilizer to their lawns and gardens (Government of Canada,

2015). Many other elements of green infrastructure are also net contributors to phosphorous in urban

runoff; as they comprise living biological components actively undertaking nutrient cycling. Lawns have

been identified as significant sources of TP on an urban catchment scale, and street tree canopy coverage

is positively correlated with increased TP in boulevard runoff (Waschbusch et al., 1999).


The grab samples for physicochemical testing were collected several hours following an intense storm

event, (detailed above). First flush of macro nutrients, including phosphorous, has been observed during

laboratory simulated rainfall events on an extensive green roof of mineral planting material by Czemiel

Berndtsson et al. (2008). But, whilst Gnecco et al. (2013) were focussing on the deposition of heavy

metals, their study found no evidence of a first-flush phenomenon occurring in their green roof study. The

mean turbidity in the grab samples was low, with a mean value of 1.26 NTU (σ = 0.6). None of the design

factors demonstrated any influence on the turbidity of the discharged water. Turbidity in the discharge

water may be controlled due to the pore size of the geotextile used to retain the planting medium above

the drainage layer, or as the sampling took place sometime after the initial outflow from this event, so that

small particulate matter had been flushed from the modules already.

71

Overall, the pH of the discharged water was slightly basic, with a mean value of 8.0 (σ = 0.2), which is

the upper recommended limit for green roof media (e.V., 2008) or horticultural soils (Jones and Jr., 2012).

As shown in Figure 4-6, pH was significantly lowered in discharge from the biological medium, F (1,18)

= 19.41 (p < 0.05). In the mineral based medium, which had a group mean pH greater than 8, it is possible

that phosphate may be complexed and not available for plant uptake (White and Hammond, 2008), which

could prompt excess additional fertilization of this type of green roof. Also shown in Figure 4-6, the mean

electrical conductivity (σw ) of the discharged water in July 2013 was 310 µS/cm (σ = 70 µS/cm).

Although these values are low from the perspective of horticultural concern (Jones and Jr., 2012), they are

higher than those reported in previous studies. The very high organic matter content in the biological

material suggests that macronutrients such as nitrogen and phosphorous may be the primary contributors.

But, σw was significantly higher in the discharge from the mineral medium, F (1,18) = 43.4 (p < 0.05);

this experimental material contained a fractions of particles believed to leach heavy metals. The type of

planting and the flushing with irrigation over five summers had not influenced the pH or σw.

Vijayaraghavan and Raja (2015) had observed that vegetation made a difference to electrical conductivity

in discharge water from extensive green roofs, but this was in comparison to an unplanted plot, and their

observations were made on ‘relatively new’ installations.

Figure 4-6 The influence of green roof medium type on the physicochemical parameters, pH and electrical conductivity.

In April 2016, the σw values from the sub-sample of modules described above as A-F, spanned a similar

range to 2013 (170 – 300 and 180 – 330 μS/cm), although there was no observable trend or significant

correlation between the repeated measures from the same modules.

The use of a combined sensor system can confound the measurement of both volumetric water content

and the salinity of this water. To account for the physical characteristics of the media, the calibration for

water content was undertaken first. The quadratic curve fitted from the calibration of the biologically

derived planting medium is shown in Figure 4-7 and Equation 4-6. It is similar to those derived for peaty

72

and organic soils by Oleszczuk et al. (2004), and particularly that of Myllys and Simojoki (1996) who

fitted near identical parameters in Sphagnum and Carex. peat deposits. In comparison, the mineral based

medium behaved quite differently (see Figure 4-7). It has a lower water holding capacity, which

corresponds with being more poorly graded, having a higher proportion of coarse particles, lower porosity

and much lower organic matter content (see Chapter 5). The S –shaped cubic curve (Figure 4-8) is the

more usual form of expression fitted for natural mineral soils (Oleszczuk et al., 2004).

Figure 4-7 Calibration of 5TE sensor in biological planting medium (top), and mineral based medium (bottom)

𝜃 = −0.0008휀𝑏2 + 0.0473휀𝑏 − 0.0762

R² = 0.93

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 5 10 15 20 25 30 35

θ(v

ol/

vol)

bulk dielectric permitivity, εb

Biological

R² = 0.84

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 5 10 15 20 25 30 35

θ (

vol/

vol)

bulk dielectric permitivity, εb

Mineral

73

Equation 4-6

𝜃 = 0.0004휀𝑏3 − 0.0123휀𝑏

2 + 0.143휀𝑏 − 0.2947

Equation 4-7

To make use of the 5TE sensors to monitor or calculate the electrical conductivity of the water held

within the planting medium, it is necessary to determine the dielectric permittivity of the solids within the

planting materials. To do this, regressions of εb against σb were performed, see Appendix D: Data relating

to Chapter 4. The range of ɛ0 in each of the planting media are presented in Figure 4-8. There was a much

greater spread of values obtained in the biologically derived media, indicating a higher level of

homogeneity in the green roof systems containing mineral based planting media. The group mean ɛ0 for

the biologically derived material was 7.7, and the group mean for the mineral media was 6.4, which is

closer to the manufacturers recommended value for natural soils of 6 (Decagon Devices, 2016). Hilhorst

(2000) calculated ε0 values between 1.9 to 7.6 in a variety of porous media (not including any with > 5%

organic matter) and recommended the use of an average value of 4.1.

Figure 4-8 Range of ε0 in eleven green roof modules containing bioloigcally derived planting medium (left), and mineral based green roof planting medium (right).

Using the group mean ɛ0 values, σp on 11th April 2016 were calculated in eighteen of the green roof

modules. The mean volumetric water content (θ) of the media at this time was 0.45 in the biological

media and 0.29 in the mineral media, which represents the near saturated conditions under which

discharge had been occurring. The relative influences of the four design characteristics were assessed

using a regression tree approach (Figure 4-9)(R2 = 0.991). As the impact of medium type has been

controlled for in the calculation of σp, this did not emerge as a significant indicator of pore-water salinity,

74

the provision of irrigation was the only determining factor, where increased irrigation over a number of

years was associated with lower pore-water salinity.

Figure 4-9 Irrigation makes a more significant impact on pore water electrical conductivity in April 2016, than any other design factor: planting medium type, depth or planting type.

Pore-water conductivities (σp) were not well correlated with the conductivity of discharge water samples

(σw) collected on the same day, again due to the complexity if the water potential within the media. For

discharge and reuse purposes, monitoring the electrical conductivity of the free discharge water is

valuable, but this may not be the most suitable indicator of vegetation sustainability if large volumes of

reused water were being applied to a green roof for irrigation purposes.

Conclusions

Mean WETP was twice as high in the biological medium samples compared to the mineral materials, with

no significant influence from differential uptake by plating selection, leaching due to levels of irrigation

or additional retention in a deeper 5 cm reservoir of material. Five years of summer irrigation had made

significant impact on the remaining TP concentrations found in discharge water, as had the initial

inclusion of an extra 5 cm of depth. Although, the choice of media was still the larger determinant; TP

values in water discharged from biological medium green roofs (group mean 5.5. mg/L) were similar to

untreated wastewater. On a regional scale, even under ‘worst case’ scenario, with highest discharge TP,

green roofs would still make a relatively small contribution to net annual urban TP release.

Electrical conductivity has been used to calibrate for both volumetric water content (θ) and the dielectric

of two type of planting media (ɛ0). The volumetric water content values have been employed in

hydrological analyses (Chapters 2 and 3) and the ɛ0 used to determine pore water conductivity and from

σp

784 μS/cm

(18)

None

1246 μS/cm

(6)

Sensor

659 μS/cm

(5)

Daily

479 μS/cm

(7)

75

that an osmotic potential of around 44 kPa in relatively saturated media of both types. The electrical

conductivity values of up to 1220 dS/m within the systems and discharged at up to 450 μS/cm are not

high enough to raise any concerns for the horticultural wellbeing of the system, or potential pollution to

the storm water discharge.

76

: The influence of depth and porosity on the hydraulic properties of green roof planting media

Abstract

An essential component of a building integrated vegetation system, such as an extensive green roof, is the

layer of lightweight planting medium which supports rooting and stores water. How the water is held

within the planting medium is important for the health of the plants, but also determines the storage of the

system to capture rainwater and contribute to stormwater management goals in urban environments.

Predicting and describing the storm water management performance of green roofs requires reliable data

regarding the water retention properties of the planting medium layer.

Ten materials proposed for use on green roofs, including four mineral components, three biological

components and three commercial blends have been characterized through measurement of their water

retention curves (WRC). In combination with the particle size distributions, the resultant data demonstrate

that some of the materials contain measurable intra-particle pore networks, in addition to the inter-particle

void spaces described in classical soil hydrology. The WRC are also used to model the maximum water

storage under static equilibrium conditions, throughout a 15 cm profile of each material.

In freely draining, unsaturated green roof systems, the role of the intra-particle pores may be limited to

increasing microscale roughness of particle surfaces, thereby reducing film flow under drier conditions.

The highly organic, biologically derived materials: screened compost, bark fines and shredded wood all

demonstrated some hydrophobicity when air dried, but wetting occurred within less than 30 minutes on

all occasions, which would be within the time frame of many rainstorms. Saturated hydraulic conductivity

was lower in soils with a higher proportion of fines (< 106 μm). The proportion of fines was also the

dominant factor determining the storage within the modelled green roof profiles.

77

Introduction

Extensive green roofs comprise a layer of lightweight, (typically soilless), planting medium, up to a

maximum depth of 15 cm, that is used to support the growth of plants (Czemiel Berndtsson, 2010). They

are designed and installed in new and retrofit scenarios, to provide a wide array of benefits to both their

host building and at an infrastructural scale across urban environments (Yocca and Sale, 2012). One such

benefit is the retention of stormwater, where green roofs can help to reduce peak storm flows and volumes

discharged to storm and combined sewer systems (Nawaz et al., 2015). On an annual basis extensive

green roofs can capture, store and evapotranspire 40-70% of the total precipitation (Fassman-Beck et al.,

2015). In most climates, the total annual retention is constrained by the inherently limited capacity of the

green roof and the distribution of storm depths (Adams and Papa, 2000). This means that a proportion of

the received rainstorms will be expected to exceed the storage of the green roof system, producing

discharge water or runoff (Stovin et al., 2015).

The vast majority of extensive green roofs contain a drainage layer immediately beneath the planting

medium, separated with a geotextile to maintain the free space beneath. The purpose of this layer is to

prevent ponded water from penetrating a leaky roof membrane and/or damaging the plant root system.

Drainage layers are often preformed plastic, but may be constructed with a coarse, freely draining

aggregate layer (e.V., 2008; Optigrün international, 2008; Roofmeadow, 2013). As illustrated in the

conceptual Figure 5-1, a green roof system at maximum water storage capacity would cease to drain when

there is complete saturation of the planting medium at the lower boundary with the geotextile (Handreck

and Black, 2002; Iwata, 1994). A meniscus forms at every void along the geotextile and matric pressure

increases upwards through the profile, analogous to the capillary zone in a subsurface environment.

Comparable conditions have been observed and described for potting containers at plant nurseries

(Cassel and Neilsen, 1986; Handreck and Black, 2002; Heller et al., 2015) and for the maintenance of golf

courses and sports fields (McCoy, 1998). How this translates into the maximum water storage of the

medium depends on the slope and shape of the water retention curve (WRC); the dashed line in Figure

6-1.

78

Figure 5-1 Green roof matric pressure as a function of medium depth under static equilibrium with maximum water storage. Where θ = volumetric water content, and θs = saturated volumetric water content

This hydraulic condition is not incorporated into many green roof models (Li and Babcock, 2014).

Programs commonly used by design and engineering consultants such as SWMM (Burszta-Adamiak and

Mrowiec, 2013) or simple models provided by municipalities such as City of Calgary (Struck et al.,

2014), treat the water holding capacity of a green roof medium as a fixed property throughout the profile.

Similarly, Kasmin et al. (2010) defined the maximum water storage as the product of medium depth and

the maximum water holding capacity of the material. Zhang and Guo (2012) also use this type of linear

relationship in their analytical probabilistic model. Metselaar (2012) used the Soil Water Atmosphere and

Plant (SWAP) model to simulate the performance of four planting media including fine sand and peat

moss, over 46 years, using a seepage face boundary. The SWAP model demonstrated that with this lower

boundary condition, there was a non-linear relationship between medium depth and the fraction of rainfall

retained. The SWMS-2D model employed by Palla et al. (2009) included a geotextile between the

planting medium (Vulcaflor), and the underlying coarse drainage layer (lapillus). The vertical profile of

water content during a simulated storm event illustrated that the hydraulic conductivity of the geotextile

boundary was most influential in maintaining a horizontal wetting front. She and Pang (2010) instead

reasoned that the wide range of pore sizes found in their planting medium meant that gravity drainage

would occur from their green roof system through the network of macro pores before saturation was

reached.

In natural soils the particle size distribution (PSD) can be used to predict both the maximum water

capacity (MWC) (Wang et al., 2008) and saturated hydraulic conductivity (KSat) (Chapuis, 2012) through

statistical processes or programs termed ‘pedotransfer functions’ (Pachepsky and Rawls, 2004; van

Genuchten et al., 1991). Although these relationships are usually derived from empirical evidence, the

79

physical principles that determine the water behavior are based on the broad assumption that the porous

medium is comprised of regularly shaped, inert particles. The PSD of the relatively coarse green roof

planting media is readily measured through sieving (Carbone et al., 2014; Poë et al., 2011; Yuristy, 2013).

However, for green roof media, which often comprise a high proportion of materials with internal

porosity, such as expanded aggregates or compost, it may be erroneous to infer the pore size distribution

from the more readily measured PSD owing to the unusual physical properties of the constituent particles.

Research regarding the storage of stormwater in extensive green roof media has considered a great many

materials and combinations. These most often include some form of naturally porous mineral with added

biological materials (Burszta-Adamiak, 2012; Ma et al., 2012; Schroll et al., 2011; Uhl and Schiedt, 2008;

Voyde et al., 2010) and/or man-made expanded aggregates with added biological materials (Carter and

Rasmussen, 2006; Cronk, 2012; Gregoire and Clausen, 2011; Hathaway et al., 2008; Prowell, 2006;

VanWoert et al., 2005). Biologically derived materials with high organic matter content are usually

incorporated in small quantities and then the bulk properties are assessed (Molineux et al., 2009). The use

of large quantities of biological material in green roof media has been associated with the leaching of

excess nutrients in runoff water (Czemiel Berndtsson, 2010; Van Seters et al., 2009), and these media

have been observed to shrink (Schindler et al., 2015) and become hydrophobic (Raviv et al., 2008b) when

dried.

The distribution of water within the green roof profile has implications both for stormwater capture

(Carter and Jackson, 2007; Hilten et al., 2008; Qin et al., 2012; Uhl and Schiedt, 2008) and for the healthy

growth of the plants (Breuning, 2013; MacIvor et al., 2013; Rowe et al., 2014). Green roof plantings are

essential, not only to maintaining stakeholder engagement with the system (McGlade and Hill, 2014), but

also in preventing wind scour and associated media loss (J. Hill et al., 2015). Within the growing media

higher plant available water storage is essential to keep a plant’s tissues turgid and growing, whilst a free

air space (FAS) of ≥ 10 % is recommended in the rooting zone to prevent waterlogging and rotting (Raviv

et al., 2008b). There have been many laboratory studies describing the hydraulics of planting media for

green roof applications (Babilis and Londra, 2011; Dal Ferro et al., 2014). There remains a paucity of data

regarding the hydraulic functions of coarse aggregates with intra-particle porous networks (Raviv et al.,

2008b) although data is emerging from a few nursery horticulture (Schindler et al., 2016) and green roof

(Dal Ferro et al., 2014; Latshaw et al., 2009) research communities. This study examines the properties of

bulk mineral and biological green roof materials, including the WRC and considers the implications for

their use in extensive green roof systems up to 15 cm in depth.

80

Methods

Ten coarse soilless materials were selected for analysis and comparison: graded sand (Sample A, serving

as a control), six bulk materials (Samples B-G) and three commercial blended products, designed for use

on extensive green roofs (Samples H-J). Their descriptions and identifying key are presented in Table 5-1.

Table 5-1 Identity and shared sources of ten sample materials for analysis and comparison

Sample ID Product A Graded silica sand B Lightweight expanded aggregate (LEA) - Supplier 1 C Lightweight expanded aggregate (LEA) - Supplier 2 D Crushed brick E ¼” screened composted wood F Bark fines G Shredded Pine H Commercially blended biologically derived medium – Manufacturer A I Commercially blended mineral based medium – Manufacturer A J Commercially blended mineral based medium – Manufacturer B

Medium property measurements

In all cases, sample moisture determinations from a given condition were made by oven drying to

constant weight mass (85 C, ±5 C) and measuring the change in weight. Air-dry moisture content and

related experiments were conducted in a laboratory with normal indoor climate control (temp: 20-22 °C,

RH: 11- 29 %). Sessile droplet adsorption was determined from dynamic contact angle data measured on

samples with an FTA200 goniometer (First Ten Angstroms). Five replicates were conducted on each

material, using individual particles or peds taken from air-dried subsamples, placed on glass slides and

droplets of water (15 μL) deposited from < 1 cm height using a computer controlled syringe pump. The

bulk shrink/swell properties were determined by packing (with shaking) 300 cm3 of test material into a

glass cylinder. Water was added until no trapped air pockets were visible, but only to the surface of the

test material to prevent floating the dry particles. The samples were soaked for 24 hours and the depth of

the material read on the side of the cylinder to the nearest 5 cm3 increment. The samples were then dried

to constant weight and the particles settled by tapping the base of the cylinder by hand before assessing

the final volume.

Solid particle densities (ρs) were calculated using gas pycnometry, determined with a Stereopycnometer

(Quantachrome) supplied with compressed air. Porosity (𝜙) was calculated from dried bulk (ρd) and

particle densities, Equation 5-1:

𝜙 = 1 −𝜌𝑑

𝜌𝑠

81

Equation 5-1

Organic matter content (OM %) was determined by loss on ignition (550 C, 2 hours) using a muffle

furnace (Matthiessen et al., 2005). Particle size distribution was performed by sieve analysis on air dry

samples. Screen number and sizes were No. - (9.51 mm), No. 6 (3.36 mm), No. 12 (1.7 mm), No. 16

(1.18 mm), No. 30 (0.60 mm), No. - (0.300 mm) and No. 104 (0.106 mm). Gradation quality was

assessed by calculation of the coefficients of uniformity (CU) and curvature (CC), using the diameters of

the percentage of particles passed at 10, 30 and 60 % (dx), Equation 5-2 and Equation 5-3:

𝐶𝑈 =𝑑60

𝑑10

Equation 5-2

𝐶𝑐 = (𝑑30)2

𝑑60𝑑10

Equation 5-3

Samples of these non-cohesive test materials (250 cm3) were packed, with shaking by hand, into stainless

steel rings (SZ 250, UMS GmbH; 5 cm height) for hydraulic conductivity and drying curve

measurements. Saturated hydraulic conductivity was measured ten times, in five repeats of two replicates,

using a 1 cm constant head pressure and 1 second measurement intervals on a KSAT apparatus (UMS

GmbH). Rates were normalized to 10 °C by accounting for the temperature dependence of water

viscosity. X-ray computed tomography was performed on an individual fragment of lava rock and porous

material recovered from sample I, using v|tome|x’s CT system (GE Instruments). Resolution was 16.77

micron voxels; geometric analyses (circularity and fractal dimension) were performed with FracLac for

ImageJ (Karperien, 2013; Schneider et al., 2012).

Water Retention Parameters

Drying curves were acquired on single samples saturated for 24 – 30 hours, using an automated

evaporation method with a HYPROP instrument with integrated balance (UMS GmbH)(Schindler et al.,

2016). The resulting water retention curves (WRC) were fitted to a bimodal, constrained van Genuchten

expression, with the addition of an adsorptive water retention function (Sad) which forces the volumetric

water content to 0 at pF 6.8 (= 618,758 kPa) (HYPROP-FIT code 1211 (Peters and Durner, 2015)). This

extrapolation was selected as these materials would experience evaporative drying beyond the irreducible

saturation point in use. In Equation 5-4 the volumetric water content (θ, vol/vol), at a capillary pressure

head (h, cmH2O) is calculated as a function of the saturated water content (θs), the irreducible water

82

saturation (θr), with the curve fitting parameters corresponding the air entry pressure, (αi, 1/cm) and the

unitless pore size distribution (ni). The bimodal distribution is weighted in the first two terms with (wi)

𝜃(ℎ) =𝑤1(𝜃𝑠 − 𝜃𝑟)

[1 + (𝛼1ℎ)𝑛1](1−

1𝑛1

)+

𝑤2(𝜃𝑠 − 𝜃𝑟)

[1 + (𝛼2ℎ)𝑛2](1−

1𝑛2

)+ 𝜃𝑟𝑆𝑎𝑑

Equation 5-4

Integrating Equation 5-4 also permits comparison of the maximum theoretical water storage (S) of these

systems at different design depths up to 15 cm (Equation 5-5):

𝑆𝑖−𝑗 = ∫ (𝑤1(𝜃𝑠 − 𝜃𝑟)

[1 + (𝑎1ℎ)𝑛1](1−

1𝑛1

)+

𝑤2(𝜃𝑠 − 𝜃𝑟)

[1 + (𝑎2ℎ)𝑛2](1−

1𝑛2

)) 𝑑ℎ

𝑗

𝑖

Equation 5-5

The adsorption term Sad does not contribute in these relatively high saturation conditions, and has been

omitted from Equation 5-5 for simplicity. Horticulturally important parameters were also derived from the

WRC. The free air space in the material at 15 cm head (FAS15) was calculated from:

𝐹𝐴𝑆15 = 𝜙 − 𝜃(ℎ15)

Equation 5-6

and the plant available water (PAW), the difference between the water held under static tension (container

capacity) and the permanent wilting point (ψm = -1.5 MPa = 15296 cm (Tolk, 2003)) from:

𝑃𝐴𝑊15 = 𝜃(ℎ15) − 𝜃(ℎ15296)

Equation 5-7

As the evaporative drying occurred in the absence of a flowing liquid phase, the Young-Laplace Equation

was used to determine the distribution of pore radiuses within each material:

𝑟 =2𝛾 cos 𝛿

−𝜓𝑚

Equation 5-8

83

This relates emptying pore radius (r, m) to the matric potential (-ψm, Pa) as a function of the interfacial

tension between water and air in the pore spaces (γ, 0.072 N/m at 20 °C) and the receding contact angle

between the materials (δ, °). The assumption is made that if medium wetting is complete prior to the

drying measurements, the receding contact angle of water/particle interface is 0° (Dal Ferro et al., 2014;

Ravi et al., 2015). Statistical comparisons, including Pearson correlation coefficients (PCC) were

calculated using NCSS 10 (NCSS, 2015), and regression trees using Orange Data Mining (Demšar et al.,

2013).


Density and porosity

As the samples originated from very different materials, there was a large variation in the bulk density

(ρb), particle density (ρs) and porosity (𝝓); the mean values were 0.78 g/cm3, 2.11 g/cm3 and 0.66

respectively (see

84

Table 5-2). The graded sand (A) had the highest ρb, as all of the other test materials were selected as being

lighter-than-soil alternatives for supporting building integrated vegetation. Not only are the silica solids

relatively dense (Chandra and Berntsson, 2002; Shmulsky and Jones, 2011), but the packing (D. C. Beard,

1973; Sezer and Göktepe, 2010) results in a much lower 𝜙 compared to the other materials. The ρb and ρs

of the mineral based materials (B-D and the blends I and J) were all significantly higher (≥ 0.7 g/cm3)

than those of the biologically derived media (≤ 0.42 g/cm3) (E-G and blend H). The crushed brick sample,

D, has a similar ρs to the sand, but the packing of the larger irregular fragments led to a lighter-weight

bulk material with a higher 𝜙. Sample D had a 𝜙 of 0.58 similar to a crushed brick sample (B5, 𝜙 = 0.51)

used in a study by Graceson et al. (2013), but coarser and with a higher 𝜙 than that used in the green roof

study (𝜙 = 0.47) by Molineux et al. (2009). The crushed brick analyzed in this study is particularly poorly

graded, with low coefficients of uniformity (CU = 2.9) and curvature (CC = 0.9). The narrow grading of

samples A and C are also evident in low CU (2.0 and 1.7 respectively). None of the bulk mineral products

(A-D) held any measurable water content after open storage in the laboratory for several weeks (Air-dry

θ) whereas the biological derived products (E – G) did retain some moisture; sample E had the highest

volumetric water content (Air-dry θ = 0.11) after this period of air-drying. In all cases the observed water

content upon air drying was lower than the irreducible water content values shown in Table 5-3, justifying

the extrapolation into the adsorbed proportion of water in the samples under drier conditions.

85

Table 5-2 Density, porosity and organic matter content of ten porous test materials.

ρd

(g/cm3) ρs

(g/cm3) 𝝓 Air-dry θ OM (%) CU CC

A 1.58 2.65* 0.40 0 0 2.0 1.0

B 1.08 2.42 0.56 0 0 20 1.8

C 0.77 2.31 0.67 0 0 1.7 1.0

D 1.13 2.67 0.58 0 1 2.9 0.9

E 0.42 1.81 0.77 0.11 48 6.3 0.9

F 0.22 1.41 0.84 0.08 97 4.8 1.2

G 0.24 1.35 0.83 0.05 87 5.0 1.1

H 0.39 1.87 0.79 0.03 47 6.7 0.9

I 0.91 2.16 0.58 0.01 7 23 1.9

J 1.07 2.41 0.56 0.01 5 14 1.9

*Value assumed (Zihms et al., 2013)

The raw PSD data are presented as a histogram for each material in Figure 5-3, Figure 5-4, and Figure

5-6. For gravels to be considered to be ‘well-graded’ CU ≥ 4, and 1 ≤ CC ≤ 3, and for sand CU ≥ 6 (ASTM

D2487 -11, 2011). Apart from one of the LEA (B) and the compost (E), most of the bulk products were

poorly-graded as reflected by low CU; materials D and E also had a low CC. The blended materials (H-J)

were all more well-graded (i.e. CU > 6), although sample H had a low CC (< 1).

Naturally, the biologically derived materials contained much higher proportions of combustible carbon.

The screened compost, E, and the similar blended product, H, each contained almost 50% organic matter

by weight, whilst the bark fines, F, was almost entirely decomposed and was 97% organic matter. Higher

OM was strongly and positively correlated with higher porosity (PCC = 0.86), both of which are

associated with particulate wood products (Shmulsky and Jones, 2011). Similarly, bark porosity has been

reported at 0.86, various composted green waste materials having porosity > 0.7, and wood fibre having

porosity up to 0.95 depending on settling and compaction treatments (Raviv et al., 2008a).

WRC parameters

Drying van Genuchten parameters for all materials were fitted to the WRC data according to Equation

5-4. The raw data and the fitted WRCs are presented in Figure 5-2, and the parameters in Table 5-3.

Materials A, C, E and G showed significant dominance of the primary weighted sub-factor (w1),

indicating an essentially unimodal pore distribution, and little influence of any secondary or intra-particle

porosity.

86

Figure 5-2 Drying curve data from the analysis of ten samples. Grey circles are raw data, lines are the fitted curves: Bulk materials A, C, E, and G are grouped as having significant (w1 > 0.9) weighting on the inter-particle voids (top); bulk materials B, D, and F are grouped as having distinctly separate and more evenly weighted van Genuchten parameters (middle); blended materials H, I, and J (bottom)

0

20

40

60

80

100

0.01 1 100 10000 1000000

θ(v

ol/

ol)

ψm(-kPa)

A: Graded sand

C: LEA supplier 2

E: Composted wood

G: Shredded pine

0

20

40

60

80

100

0.01 1 100 10000 1000000

θ(v

ol/

vol)

ψm (-kPa)

B: LEA supplier 1

D: Crushed brick

F: Bark fines

0

20

40

60

80

100

0.01 1 100 10000 1000000

θ(o

l/vol)

ψm (-kPa)

H: Biological blend a

I: Mineral blend a

J: Mineral blend b

87

Table 5-3 the van Genuchten parameters from the fitted curves arising from the evaporative drying of ten test materials

θr

(vol/vol) θs

(vol/vol)

Inter-particle voids Intra-particle pores

w1 α1 (cm-1) n1 w2 α2 (cm-1) n2 A 0.06 0.40 0.91 0.079 15 0.09 0.048 15

B 0.00 0.45 0.37 0.304 2.3 0.63 0.011 1.4

C 0.16 0.31 0.93 0.650 3.1 0.07 0.000 1.3

D 0.00 0.37 0.57 0.355 2.9 0.43 0.001 1.5

E 0.32 0.81 0.97 0.118 1.9 0.03 0.010 7.7

F 0.29 0.76 0.65 0.129 3.8 0.35 0.006 3.8

G 0.28 0.75 0.94 0.300 1.9 0.06 0.012 7.8

H 0.28 0.79 0.34 0.080 4.5 0.66 0.041 1.7

I 0.23 0.55 0.78 0.110 1.4 0.22 0.053 4.4

J 0.18 0.50 0.85 0.098 1.7 0.15 0.003 3.8

As detailed above, the materials were soaked > 24 hrs for determination of θs, and 𝜙 was calculated from

particle density measurements taken on oven dried samples. The θs were higher (≥ 0.75) in the biological

samples, than in the mineral samples (≤ 0.55), corresponding with the higher 𝜙 in the biological

materials. For most samples 𝜙 ≈ θs, indicating little discrepancy between the measurement methods. An

exception was sample C, 𝜙 = 0.67 and θs= 0.31; a difference which could be caused by pore spaces within

the particles, accessible by the compressed air during pycnometry, but not filled with water by soaking in

the 5 cm deep HYPROP apparatus. The physical characteristics of material C are also evident in Figure

5-2. The rapid drainage of the inter-particle pores in this coarse, poorly-graded light weight expanded

aggregate is seen in the high α2 value (0.650 cm-1) related to low air entry pressure. There is a very low

weighting (w2 = 0.07) of the secondary intra-particle pores in sample C. Expanded aggregates have been

reported to have relatively little intra-particle porosity compared to other anthropogenic or naturally

porous granular materials (Hemmings et al., 2009).

Figure 5-3 illustrates the very narrow distribution of inter-particle voids in the uniformly graded sand,

corresponding with the high and matched n values of 15 in sample A. This fits with the classical models

relating particle size distributions to pore size distributions and to matric potential. The WRC of material

E demonstrated a very consistent drying rate (Figure 5-2), dominated by a relatively narrow pore size

distribution with a low n1 (1.9), which matches the well-graded particle distribution seen in Figure 5-3.

Any intra-particle pore networks, which may be present, are obscured by the fraction of finer fragments,

which would also create smaller voids within the bulk matrix. Sample G follows a similar trend, although

an inflection in the curve suggests that there may be some influence of particularly small intra-particle

pores at higher matric potential.

88

Figure 5-3 The largely unimodal pore size distributions (line) plotted over the particle size distributions (bars) found in: A: Sand, C: Poorly-graded LEA, E: ¼” Screened composted wood, and G: Shredded Pine.

Mineral materials B and D, and F (bark fines) all demonstrated significant bimodality in their pore size

distributions (Figure 5-4). Naasz et al. (2008) has also described a largely bi-modal pore distribution in

pine bark, with the following parameters: w1 = 0.33, α1 = 3.56 kPa-1, and w2 = 0.62, α2 = 2.07 kPa-1.

Material B (LEA) was well-graded in comparison to material C (LEA). The secondary pore network in B

may be due to smaller particles below those characterized by sieve analysis, this is suggested by a

relatively high proportion of particles in the lowest fraction. Also, as B did not have the same discrepancy

between ϕ and θs as found in C, the smaller spaces appear to be well-connected as inter-particle voids,

preventing trapped air. Both sample B and F were more well-graded than Sample D (Figure 5-4) however

0%

10%

20%

30%

40%

50%

60%

0

1

2

3

4

5

6

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re V

olu

me

(v/v

)

Size (μm)

A: Graded silica sand

0%

20%

40%

60%

80%

100%

0

0.1

0.2

0.3

0.4

0.5

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

C: LEA - Supplier 2

0%

5%

10%

15%

20%

25%

30%

35%

0

0.2

0.4

0.6

0.8

1

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

E: ¼” screened composted wood

0%

10%

20%

30%

40%

0

0.2

0.4

0.6

0.8

1

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

G: Shredded Pine

89

none of these were significantly gap-graded, i.e. none were missing a mid-sized particle fraction, which

could create bi-modal distribution of inter-particle void sizes. Neither D nor F had pores visible to the

naked eye within the particles however, intra-particle pores with mode around 10 μm in both materials

were visible under optical microscopy, as were those in material B (

Figure 5-5).

Figure 5-4 The largely bimodal pore size distributions (line) plotted over the particle size distributions (bars) found in: B: Well-graded LEA, D: Crushed brick, and F: Bark fines.

0%

5%

10%

15%

20%

25%

30%

35%

0

0.1

0.2

0.3

0.4

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

B: LEA - Supplier 1

0%

10%

20%

30%

40%

50%

60%

70%

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

D: Crushed brick

0%

5%

10%

15%

20%

25%

30%

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

F: Bark fines

90

Figure 5-5 Surface detail visible under 100x magnification: left) B: LEA, centre) D: Brick particle, right) F: Bark fragment

Commercial green roof blends I and J, shown in Figure 5-6, both exhibit a more balanced weighting

between the inter-particle voids and intra-particle pore networks, indicative of these materials having been

blended from a number of constituents such as expanded shale, brick fragments and compost fines.

Materials I and J had relatively low air entry pressure, i.e. high α1 (0.110, 0.098 cm-1) and contained

correspondingly higher proportions of coarse fragments compared to the other commercial mixture (H).

Green roof blend H, was similarly graded and of similar appearance to the screened compost, E. Although

H had a balanced weighting between two pore ranges (w2 = 0.66) compared to E, both shared a wide pore

size distribution, and similar values for saturated water content θs and maximum adsorbed water content

θr (approximately 0.8 and 0.3 respectively). As the α1 value for Sample H was of the same order of

magnitude as α2 (0.080, 0.041 cm-1) the pore distribution is not clearly separated into two distinct

domains, as can be seen in the gradually changing slope of the curve in Figure 5-6. As in Sample B, there

was a relatively high proportion of particles in the lowest fraction following sieve analysis of Sample H,

and no evidence of trapped air in comparison of ϕ and θs. These observations suggest again, that a

network of smaller inter-particle voids, rather then intra-particle pores are within the blend H.

91

Figure 5-6 The pore size distributions (line) plotted over the particle size distributions (bars) found in commercial green roof planting media blends: H: Compost based - Manufacturer A, I: Mineral based - Manufacturer A, and, J: Mineral based - Manufacturer B.

0%

5%

10%

15%

20%

25%

0

0.2

0.4

0.6

0.8

1

1.2

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

H: Commercially blended biological medium

– Manufacturer a

0%

5%

10%

15%

20%

25%

30%

35%

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

I: Commercially blended mineral based

medium – Manufacturer a

0%

5%

10%

15%

20%

25%

30%

35%

0

0.1

0.2

0.3

0.4

0.5

0.001 0.1 10 1000

Par

ticl

es b

y w

eight

Po

re v

olu

me

(v/v

)

Size (μm)

J: Commercially blended mineral based

medium – Manufacturer b

92

System water storage

By integration of the WRC, the maximum theoretical storage capacities of the first three incremental 5 cm

depths of green roof depth have been calculated and are presented in Figure 5-7. Whilst sample A

maintains a similar capacity between the first and second 5 cm increments, the other bulk materials B – G

all have markedly decreasing water storage capacity per unit increased depth compared to the blended

products H – J, owing to low air-entry pressures and poor grading. The greatest difference was seen in

sample C, which could retain just 17% as much water in the uppermost 10-15 cm increment (0.6 mm),

compared to the lowest 0-5 cm section of the profile (3.6 mm). The highest performing material overall

was product H, which stored over 23 mm/5 cm unit depth; H had a notably high air entry pressure, was

well-graded and had a high proportion of fines.

Figure 5-7 Modelled water storage in three 5 cm increments of green roof profile depth, for seven bulk materials (A-G) and three commercial blended materials (H-J).

Table 5-4 presents the measured Ksat values, and the calculated parameters of maximum system water

storage (mm/15 cm), θ(h15), θ(h15296), FAS15 and PAW. The maximum water storage was correlated

against the physical parameters of the media: d10, ϕ, proportion of fines passing a No. 104 sieve, CU, CC,

and OM, α1 and α2. Higher storage was most strongly and negatively correlated with increasing size of the

d10 fraction (PCC = -0.86), and positively correlated with the proportion of fines (PCC = 0.52). These

relationships indicate the influence of the smallest inter-particle voids, by increasing the retention of

capillary water at higher potential. A strong, negative correlation between theoretical maximum water

storage and α1 (PCC = -0.80) confirms that the air entry pressure is also highly significant, controlling the

onset of gravity water drainage under low tension. All of the tested materials demonstrated acceptable (>

10%) free air space in the upper part of the green roof profiles modelled (FAS15). The compost, E and the

0

5

10

15

20

25

30

A B C D E F G H I J

Wat

er s

tora

ge

(mm

)

0 - 5 cm

5 - 10 cm

10 - 15 cm

93

blended material, H each provided the greatest plant available water (PAW), at 0.43 and 0.50

respectively. The poorly-graded materials, A and C, provided almost no PAW at just 0.09 and 0.08.

Table 5-4 System static and dynamic air and water properties for ten samples

Maximum system

storage (mm/15

cm)

Ksat (m/hr) (σ) FAS15

θ(h15)

container

capacity

θ(h15296)

wilting

point

PAW15

A 44 2.7 (0.6) 0.29 0.11 0.03 0.09

B 53 8.3 (6.3) 0.26 0.29 0.04 0.26

C 5 13.6 (4.6) 0.52 0.15 0.07 0.08

D 34 13.2 (3.9) 0.41 0.17 0.03 0.13

E 57 1.6 (1.0) 0.20 0.57 0.14 0.43

F 51 1.8 (1.4) 0.36 0.48 0.13 0.36

G 38 8.3 (3.5) 0.43 0.39 0.12 0.28

H 74 0.9 (0.02) 0.15 0.64 0.13 0.50

I 42 0.4 (0.01) 0.13 0.45 0.13 0.34

J 41 0.3 (0.06) 0.17 0.39 0.08 0.31

Some stormwater management models require input figures for field (container) capacity, θ(h15) and

wilting point, θ(h15296) for planting media in green roofs (Westhoff Engineering Resources Inc., 2011;

Zhang and Guo, 2012). Routine testing of green roof media does not include these parameters (ASTM

E2777 -14, 2014; e.V., 2008) so regression tree analysis is applied as a pedotransfer function to predict

these values from the frequently tested parameters, ρd, ϕ, and OM (Figure 5-8)(θ(h15) R2 = 0.999, θ(h15296)

R2 = 0.952). OM was the primary predictor for both of these θ values, with higher OM again being

associated with higher θ. The secondary predictor was ρd, and ϕ did not feature as a useful predictor in

this data set. An additional sample, comprising a binary mixture (1:1) of C and E was prepared and used

to validate the performance of these trees. The input parameters for this test sample were ρd = 0.64 g/cm3

and OM = 24 %. The predicted values for θ(h15) and θ(h15296) were 0.39 and 0.12, respectively; the

measured values were 0.34 and 0.11. The discrepancy in the modelled and measured values is very small

compared to the near two-fold difference from example guidelines for generic green roof material based

on sand texture (θ(h15) = 0.132 and θ(h15296) = 0.057, (Westhoff Engineering Resources Inc., 2011)).

94

Figure 5-8 Regresison trees for prediction of container capacity (θ(h15)) and wilting point (θ(h15296), from predictors ρd, ϕ, and OM.

The system water storage capacity and Ksat were negatively correlated (-0.69). The Ksat values of the bulk

materials A-G have high standard deviations; the variation occurring between replicates rather then

repeated measurements, reflecting the heterogeneity of the materials rather than internal erosion processes

(Chapuis, 2012). The blended products H - J had the lowest values of saturated hydraulic conductivity

(each < 0.5 m/hr). The highest values were observed in the coarsest bulk materials C and D (each > 13

m/hr). Despite the use of a low and constant head pressure during the testing, the rapidity of flow in these

materials was outside of the specifications of the KSAT instrument (i.e. > 8.3 m/hr). Higher Ksat was

associated with lower α1 (PCC = 0.92) and larger d10 (PCC = 0.79).

Under non-saturated conditions, hydraulic conductivity may be reduced by the limited areas of contact

between the large particles; as the intra-particle pores increase the surface roughness, these influence the

θ(h15)

0.36

(11)

OM > 3 %

0.47

(7)

OM > 35 %

0.52

(4)

OM ≤ 35 %

0.39

(3)

OM ≤ 3 %

0.18

(4)

ρd ≤ 1.105 g/cm3

0.22

(5)

ρd > 1.105 g/cm3

0.14

(2)

θ(h15296)

0.09

(11)

OM > 6 %

0.12

(6)

OM ≤ 6 %

0.05(5)

ρd ≤ 1.075 g/cm3

0.08

(2)

ρd > 1.105 g/cm3

0.03

(3)

95

processes of film flow. This has been described as Regime III flow, where the contact area between

particles, containing capillary water is constrained by the proportion of non-pore surface on the individual

fragments (Carminati et al., 2008). Figure 5-9 shows a single image slice from an X-ray scan of a typical

particle taken from the largest fraction of green roof blend I. Analysis of 849 slices through this same

particle shows a low degree of circularity (0.40), but also a low mean fractal dimension of the surface (D

= 1.1) which indicates that the surface has low roughness. Although the particle has around 60% internal

porosity, few of these pores connect to the outside surface. Those that do are dead-end pores rather than

extensive networks with high connectivity and fluid entry into such pores is not favoured under the

normal atmospheric pressure conditions encountered on green roofs (Hemmings et al., 2009).

Figure 5-9 Binary image from x-ray of material I particle (left), results of surface fractal analysis to show the network of connected pores (right)

Hydrophobicity, wetting and shrink/swell characteristics

Materials with high organic content can demonstrate hydrophobicity and shrink/swell characteristics

which can impact the bulk properties (Petrell and Gumulia, 2013). Hydraulic conductivity can vary with θ

in biologically derived materials. For example, Khoshand and Fall (2014) reported a ‘U’ shaped

relationship between the two parameters; a two-fold decrease in the hydraulic conductivity of their

compost occurred between θ = 0.2 and 0.4 which corresponded with the maximum dry density at θ = 0.4,

and then an increase in hydraulic conductivity was observed between θ = 0.4 and 0.6 owing to higher

connectivity of the wetted regions and a trend towards Ksat.

96

The shrink-swell properties of the four biologically derived materials, E – H, were measured in a single

cycle of wetting and drying (Table 5-5). All materials had presumably been stored in external aggregate

yards prior to arrival at the laboratory and have therefore been through innumerable cycles of wetting and

drying previously. The shredded pine, G, showed the greatest change in depth (L) compared to the air-

dried initial condition, attributed to the expansion of pore spaces between the long fibrous fragments as

they saturated. The three other materials, comprising smaller and more regularly shaped particles showed

less expansion, but did settle lower in the vessel after drying.

Table 5-5 Dynamic contact angle data from the analysis of the biologically derived materials E-G

Sample Mean initial

contact angle (σ) Mean wetting time (secs) (σ)

Mean wetting time (min)

Soaked L

Dried L

Compost: E 98 (7) 357 (364) 6 100% 87% Bark Fines: F 90 (10) 468 (410) 8 102% 87%

Shredded Pine: G 109 (10) 1724 (413) 29 108% 100% Blend: H 111 (8) 1532 (222) 26 103% 90%

The initial advancing contact angles ≥ 90 seen in Table 5-5 indicate that all three biologically derived

bulk materials and the blended material, H, were hydrophobic when air dry. The variance in all of these

measurements reflects the heterogeneity in the surface properties of the fragments at the scale of the

testing. Rewetting of the fragments in E and F was relatively rapid, taking place in just 6 or 8 minutes, the

pine and the blended materials (G and H) took over twenty minutes. The long rewetting time of G and H

could inhibit adsorption of draining water during more short, high intensity rainstorms, particularly in

summer months when the materials would be drier at onset. Dry bark has been reported as hydrophobic as

a bulk characteristic (Beardsell and Nichols, 1982) and partially hydrophobic by static contact angle (60°

- 80°)(Naasz et al., 2008). The initial contact angles measured in this study were at much higher tensions

(~13 MPa) than those observed by Naasz et al (up to ~0.3 MPa).

Conclusions

Whilst some of these coarse materials, such as the lightweight expanded aggregate, crushed brick and

bark fines, contained measurable intra-particle porosity, the impact of these networks on bulk material

properties was minimal. The poorly-graded bulk materials (A – G) demonstrated high mean Ksat, with

larger variability, compared to the < 1 m/hr observed in all three commercially blended products H – J.

Across the whole data set, Ksat was correlated to the d10, which is in agreement with many observations

and calculations made in natural soils (Chapuis, 2012). The pores observed in some particles may

influence hydraulic conductivity under drier conditions by increasing surface roughness and reducing the

area of contact between adjacent fragments.

97

Modelling the materials in systems of up to 15 cm indicated that both the largest and finest inter-particle

void spaces were influential in determining the maximum water storage capacity. Overall, the blended

products maintained a higher proportion of their storage per unit depth better than the bulk materials. As

the coarse bulk materials had low air entry pressures (large van Genuchten α values), the water storage

capacity of many of these materials dropped significantly within the upper parts of the 15 cm maximum

specification for extensive green roofs. The implication is that the addition of a few extra centimeters of

planting material is not necessarily the most effective way to increase the storm water capture of an

extensive system. Instead, design specifications should focus on using a well graded medium with a high

capacity for capillary retention of water allowing higher storage per unit depth through the entire design

bed depth. Higher water storage in the upper portion of a green roof may also increase the evaporation

potential of the system. The drainage conditions would also depend upon the permeability of the lower

boundary geotextile, a factor which warrants further exploration in extensive green roof systems.

98

: Comparisons of Extensive Green Roof Media in Southern Ontario

Abstract

Thirty-three extensive green roofs in southern Ontario were surveyed and samples of planting media

recovered for hydrological laboratory analyses. The resulting data demonstrate a significant dominance of

the role of organic matter in the physical and chemical properties of the media. This is very apparent

owing to the local practice of basing some green roof media on composted materials rather than on the

European recommendations of lightweight minerals or aggregates. The desirable characteristics of

increased maximum water holding capacity and lower bulk density are both well correlated to the

percentage of organic matter found in the media samples. Within the roofs surveyed, the percentage of

organic matter, even in association with increasing maturation, was not a significant indicator for loss of

green roof planting depth. Particle size distribution (PSD) parameters were also measured and compared

to the water transport properties of the media (permeability and maximum water holding capacity).

Introduction

To date, much of the research into the stormwater management performance of green roof planting media

has followed one of a few formats: detailed laboratory testing of bulk material (Ouldboukhitine et al.,

2012; Yio et al., 2013), computer simulation or modelling (Stovin et al., 2013; Tabares-Valasco and

Srebric, 2012; Vesuviano et al., 2014), and/or monitoring of a full-scale installation(s) (Carson et al.,

2013; Schroll et al., 2011). As Germany has historically led the development of green roof design, it is

fitting that one of the most comprehensive reviews of multiple maturing installations has come from

Berlin (Köhler and Poll, 2010). There, Köhler and Poll found that significant changes in construction

practices had occurred over time, resulting in interesting comparisons and lessons for future designs.

In North America, where the green roof industry is younger, Toronto, ON, is home to a large and growing

number of green roofs (Keesmaat, 2013), some of the most mature extensive green roofs (Matuk, 1999)

and earliest North American research on performance (Liu and Minor, 2005; Van Seters et al., 2009),

coupled with pioneering municipal policy (Kwik, 2001) and legislation.

In 2009 the City of Toronto passed the Toronto Green Roof Bylaw, which stipulates that all new

developments above particular size thresholds must have some form of green roof covering between 20 -

60% of the roof area. The size thresholds and the proportion of the roof to be covered vary according to

the purpose of the development, e.g. residential development below six storeys are exempted, but

industrial units below 6 storeys are required to comply (Toronto, 2009). The bylaw includes few

99

technical specifications regarding the construction materials and methods, and does not require the green

roofs to meet any specific storm water management performance objectives. Instead, the bylaw

emphasises the survival of the ‘selected plants’ at three years, rather than overtly aligning the bylaw with

other planning documents which address the seasonal retention of stormwater (City of Toronto, 2014).

But in doing so, the properties of the overall green roof systems must be such that they retain sufficient

water to support vegetation, and so capitalizing upon rainwater retention, through the use of well-

designed porous medium meets both objectives (Czemiel Berndtsson, 2010).

Owing to the early development of two locally distinct approaches regarding organic matter in green roof

media (Buist and Friedrich, 2008), the Greater Toronto Area is an ideal location in which to examine the

development and maturation of various extensive green roof types. Extensive green roofs (those ≤ 15 cm

deep) are popular as they are relatively lightweight and cost effective for meeting the requirements of the

bylaw. Lightweight planting media, used on green roofs, typically comprise an engineered mixture of

graded materials including composted organic matter and granular mineral components. Inspired by the

German FLL guidelines for green roof construction, many manufacturers produce a default green roof

planting mix which is a freely draining combination of coarse aggregates, with a low total organic content

(e.V., 2008). The compositions of such commercial substrates are considered proprietary information, but

even branded products have mixtures that vary according to materials available locally and at the required

time (Rugh, 2013). In Southern Ontario, more than one local green roof manufacturer has advocated for,

and constructed extensive green roofs planted in media comprised of blends with a majority of mature

compost. Such blends can be specified to include up to 50% initial organic matter content (Bioroof

Systems, 2011), far above the values investigated in most other studies (Getter et al., 2007; Graceson et

al., 2013; Nagase and Dunnett, 2011; Rowe et al., 2006; Yio et al., 2013). The use and specification of

composted or harvested materials in green roof media can sometimes cause confusion, owing to

widespread colloquial use of the term ‘organic matter’ to describe the bulk materials used in the blended

mixtures. In the current study the percentage of organic matter (OM) is used to mean the amount of

combustible carbon in the resultant mixture. Different biologically sourced materials, such as compost,

peat or coir, can have very widely varying properties including actual carbon content (Abad et al., 2002;

Zmora-Nahum et al., 2007); this ultimately affects their contribution to a planting medium blend.

The objective of this paper is to examine and report the in-situ, physical and chemical properties of

extensive green roof planting media collected from Toronto green roofs. As new green roof construction

materials continue to be developed, and this region now has a number of roofs over ten years old, it is

timely to perform a review of some of these installations.

100

Methods

Between 2013 and early 2015, thirty-three green roofs across the Greater Toronto Area were surveyed,

and samples of the green roof media recovered for laboratory analyses. This region of southern Ontario is

within humid continental climate (Dfa/Dfb) region (Kottek et al., 2006). Climate normals include sub-

zero degree Celsius mean temperatures for three months of the year and between 54 and 84 mm

precipitation each month (Canada, 2013).

The projects were selected to represent the broadest range of ages and the widest variety of planting

medium types. Although efforts were made to encompass green roofs of different ages and roofs up to 17

years old were included; nearly half of the roofs surveyed were three years old or less at the time of

surveying, Figure 6-1. This is due to the relatively recent adoption of green roofing in the region,

followed by rapid expansion of the market.

Figure 6-1: Age of thirty-three green roofs at the time of surveying and sampling.

101

At the time of the sampling, thirty-one of the green roofs were classed as ‘extensive’ construction (≤ 15

cm mean depth of planting material).

System Properties

On each roof the depth of the planting medium was measured in several locations using a graduated metal

probe (1 cm precision). Access to installations was often restricted spatially due to safety concerns, or

restricted to a short time owing to normal building operations. Some sampling visits were made at

relatively short notice or without any prior knowledge about the roof location, layout or construction. For

these reasons, a grab sampling strategy was employed, including both edge and center areas, for both the

depth measurement and medium sampling. Measurements of infiltration rate were made using a custom

built infiltrometer which is shown in Figure 6-2. The internal diameter of the instrument is 41 mm, a

suction head of 1 cm was applied, and a few millimeters depth of coarse sand were employed to provide a

contact surface between the disk and the planting media. Full depth cores of the green roof planting media

were extracted in sections of clear PVC tube (i.d. 45 mm) that were capped for transportation to the

laboratory.

102

Figure 6-2 Schematic (not to scale) and photograph of the infiltrometer used for in situ measurements

Physical Properties

Whilst retaining the sample with minimal disturbance within the PVC sampling tube, the bottom cap was

removed and replaced with a single layer of polypropylene geotextile (Permittivity, ASTM D4491: 2.2 s-1

from manufacturer’s documentation). The sample was then entirely submerged in water for 24 hours to

fully saturate all of the pore spaces. After this time Kf was measured (six replicates) using a falling head

of water within the sample tube, and the sample was subsequently left to drain for a period of 120

minutes. The sample was then discharged from the tube into a pre-weighed crucible and weighed before

being placed in a drying oven (95 ± 5 °C) for 8-12 hours. After the dry sample was weighed, the PSD was

determined using a set of brass sieves (9510 μm plus US sizes 6/12/16/30/50/140). Finally, the particle

density of each sample was calculated using water displacement, following a further 24 hours of

submersion. The equipment differed from that prescribed in other laboratory tests (ASTM E2399 -11,

103

2011; e.V., 2008) in order to preserve the initial structure of the core samples. The equations used to

derive additional parameters are presented in Table 6-1.

Table 6-1 Equations used to summarize physical characteristics of the porous media.

Parameter Calculation Reference

Depth ratio surveyed depth/design depth -

Permeability

from falling head test Kf =

L

(t1 − t0)ln (

h0

h1) (Al-Khafaji and Andersland, 1992)

Maximum Water

Capacity MWC = 100 × (ρm − ρd) (ASTM E2399 -11, 2011)

Porosity from

particle density ϕ = 1 −

ρd

ρs (ASTM E2399 -11, 2011)

Free Air Space FAS = (1 −ρmρd

ρs) − (1 − ρd)ρm (Annan and White, 1998)

Coefficient of

Uniformity CU =

d60

d10 (Das and Sivakugan, 2015)

Coefficient of

Curvature Cc =

(d30)2

d60d10 (Das and Sivakugan, 2015)

Chemical Properties

Samples for chemical analyses were taken from blended, full profile cores that had been air-dried at the

laboratory prior to testing. Percentage organic matter was determined by loss on ignition, with a

decomposition temperature of 550 °C held for two hours (Matthiessen et al., 2005). Saturated slurries of

the planting media 1:5 (w/w) were tested for pH (SenTix 22 pH probe, CanLab 607 pH meter) after 30

minutes shaking at 150 rpm in analytical grade water. After further dilution to 1:10 (w/w) and another 10

minutes shaking, the samples were centrifuged at 4g for 10 minutes to remove suspended solids. The

supernatant was then tested for electrical conductivity, σw (FieldScout EC 110 Meter), colour, water-

extractable ortho-phosphate and WETP using a SMART 3 colorimeter (Fuhrman et al., 2005), in

accordance with La Motte Standard Operating procedures (Motte, 2012).

The relationships between the various physical and chemical factors were determined using statistical

methods. These included: Pearson product-moment correlation to identify simple linear relationships,

scatter plot matrices and least square curve fitting to identify and define non-linear relationships, and

regressions trees to rank and perform regression with multiple independent variables. The regression tree

was generated using Orange (Demšar et al., 2013). All other interpretations of numerical data were

performed using NCSS statistical software (NCSS, 2015).

104


The data arising from this observational study have been classed as either input/design choices

(independent variables) or outputs/performance indicators (dependant variables). These designations are

listed in Table 6-2.

Table 6-2 Six independent and fourteen dependant variable measured on the surveyed roofs.

Independent Variables Dependent Variables

Age of installation

Particle composition parameters:

Organic matter content (%)

Mean particle density

Particle size parameters:

d10

Coefficient of Uniformity

Coefficient of Curvature

System properties:

Depth ratio

Bulk media physical properties:

Maximum media density

Dry bulk density

Porosity

Free air space

Water interaction properties:

Maximum water capacity

Infiltration rate

Permeability

Chemical properties:

pH

Water extractable colour

Water extractable electrical conductivity

Water extractable phosphorous

Age of installation

During the period of seventeen years, no significant trends were noted in the choice of planting medium

type, i.e. none of the physical properties of the media were correlated with age. This reduces confounding

influences and supports an independent assessment of the physical properties.

One of the most widely hypothesized changes in matured green roofs concerns collapse or compaction of

the planting medium due to settling of mineral particles and decomposition of organic matter (Buist and

Friedrich, 2008; Yuristy, 2013). To assess the extent of any changes in depth, ratios were calculated using

survey data and original design depth information for thirty of the roofs (architectsAlliance, 2004; Boivin,

n.d.; Guo and Mrosovsky, 2014; Inc, n.d.; Liu and Minor, 2005; LiveRoof, 2015; Ltd, 2010; McGlade,

2015; McGlade and Hill, 2014; Toronto, 2010; Van Seters et al., 2009; Vonk, n.d.; Zoo, n.d.). Three roofs

were excluded from this analysis; two because they were designed to include a variety of depths; and one

because original design information was not available. Overall, the survey measurements had a high

standard deviation, owing to the tolerance of the original construction (typically ± 10%) and changes at

the site due to wind scour and particle settling. This property was tested against the age of the roofs,

resulting in poor correlation (PCC 0.16, p > 0.05); rejecting the hypothesis that maturation up to 17 years,

105

results in losses of media depth. One possible reason for this would be pedogenesis mechanisms adding

decomposed bulk biological material to the profile; a process that has been reported in the context of

matured green roofs elsewhere (Köhler and Poll, 2010).

Another possible mechanism upon aging, is the degradation of particles to create more fines within the

media, which could reduce permeability (Gaches et al., 2013). As stated above, the d10 of the samples

recovered was not associated with age, due to the wide variety of media blends encountered. As such this

study is unable to make a logical assessment of this type of change. Similarly, green roofs have

previously been associated with increased phosphorous in the outflow waters (Czemiel Berndtsson,

2010), with concentration reportedly diminishing as the roof matures (Harper, 2013; Köhler et al., 2002;

Van Seters et al., 2009). In this study, the wide variety of materials encountered meant that none of the

chemical properties of the planting media were significantly statistically associated with the age of the

roofs.

Particle composition

The two current design approaches regarding OM content are clearly identifiable in Figure 6-3; Twenty-

two of the roofs had low organic matter content, up to around 8%, as recommended by FLL (albeit, some

have up to 11%). The other third of the media mixtures appear to have been formulated without any

regard for the FLL guidelines, resulting in OM ≥ 30 %. The OM of the planting media demonstrated the

most widespread and often most significant effect on many of the dependant variables listed in Table 6-2.

106

Figure 6-3 Organic matter content of plating media recovered from thirty-three extensive green roofs; roofs are alphabetical from oldest to most recently constructed, the dashed line crosses at 8 %

Under laboratory test conditions, composted materials have demonstrated high compressibility (Khoshand

and Fall, 2014); this is one reason why it has been theorized that the depth of compost-based green roofs

would diminish over time. It has been argued that the decomposition of excessive organic matter over

time would lead to compaction of the media, increased surface runoff and root damage (Nagase and

Dunnett, 2011). Loss of depth has been attributed to other potential causes of media loss including wind

scour of the lightweight particles prior to establishment of vegetation (J. Hill et al., 2015) or biological

degradation of the material (Emilsson and Rolf, 2005). When considered alone, the depth ratio was not

well correlated with OM (PCC 0.23, slope p > 0.05), and in combination with age, through the use of

multiple regression, none of the terms were deemed significant (p > 0.05). These survey data do not

support the hypothesis that the use of high OM planting media leads to any significant loss in media depth

in up to 17 years.

The media samples were grouped according to OM content (< 30% ‘low’ and ≥ 30% ‘high’) and the

density properties compared. The low and high OM samples had ρm 1.2 and 1.0 g/cm3, ρd 0.8 and 0.3

g/cm3, and ρs 2.0 and 1.7 g/cm3 respectively, as shown in Figure 6-4. As green roofs are constructed with

a consideration for maximum dead load, so it is notable that both classes (high and low OM) have a

similar mean ρm and neither type of material is strongly preferable for this property. The two-fold

107

difference in the class means for ρd indicates that the lighter, high OM types of materials will typically

have higher MWC, as this is defined as the difference between ρm and ρd.

Figure 6-4 Multifactor box plot of bulk and solid particle densities, divided between low (< 30%) and high (≥ 30%) OM content.

Planting medium/Water Interactions

Increased OM has been previously associated with increased plant health and MWC in green roof media

(Buist and Friedrich, 2008; Getter et al., 2007; Nagase and Dunnett, 2011; Rowe et al., 2006; Yio et al.,

2013; Young et al., 2014), albeit many previous studies have focused on materials with relatively low

OM. In our total set of observational data, encompassing both low and high OM content samples, the

relationship between MWC and OM was established using curve fitting (Figure 6-5) such that R2 = 0.80:

108

𝑀𝑊𝐶 = 28.8 + 10.6 ln(𝑂𝑀)

Equation 6-1

Figure 6-5 Relationship between maximum water content and organic matter content of green roof media

Across all the entire data set, it can be observed that some samples, particularly with higher OM, had the

potential to hold over three times the water (by volume) compared to samples with low OM. This can

have significant implications for plant available water, by improving vegetation drought resilience, and

facilitating the evaporative cooling provided by green roofs to mitigate urban heat island effects (Jim and

Peng, 2013; Wanphen and Nagano, 2009). This increased capacity of high OM materials also has a

significant potential for increasing the net stormwater retention over a season (Molineux et al., 2009),

particularly in temperate conditions, which have a high proportion of low intensity storms (Simmons,

2015).

Viewed in isolation, the logarithmic expression in Equation 6-1, accommodates a wide range of MWC

within a low OM group, and a relatively small range of MWC within the high OM group of materials. So,

109

other physical parameters must also contribute to the overall MWC of each blend; an example of which

would be the void size distribution, which is often linked to PSD for ease of measurement. The overall

PSD of porous media can be summarized using CU and CC.

These parameters have been calculated for the planting media recovered from all of the green roofs. Well

graded mixtures of particles have CU > 6; in this entire sample set CU values were normally distributed

between 5.5 and 20.7, indicating that the sample materials were generally well-graded. Samples with

higher OM were significantly correlated with lower CU, the distributions of CU in high and low OM

materials are presented in Figure 6-6. As many mineral based planting materials contain intra-particle

pores, through the use of naturally porous rocks or lightweight expanded aggregate, a possible mechanism

for the regrading of matured green roof planting media has been explored by Gaches et al. (2013). They

found that repeated cycling of freeze-thaw conditions in saturated porous aggregate causes the individual

fragments to shatter and reshapes the particle size distribution.

A smaller particle size in the lowest decile fraction (d10) was associated with higher OM values (PCC -

0.38, slope p < 0.05), i.e. a greater proportion of fines in the medium was associated with higher OM. The

relative size of the smallest 10% of the particles (d10) was linearly correlated with many physical

properties such as porosity and bulk media densities. But the overall pattern of influence strongly

indicated that all of these relationships were as a result of the connection between increased OM and

smaller d10.

110

Figure 6-6 Multifactor box plot of particle size distribution coefficients, divided between low (< 30%) and high (≥ 30%) organic matter content.

Another parameter used to summarize the shape of the PSD curve is the CC which describes whether the

PSD curve bulges towards a dominance of finer or coarser particles in the midrange. Preferable values lie

in the range 1 ≤ CC ≤ 3, which indicates an even distribution of particle sizes. Most of the media samples

had ‘good’ or low CC. As seen in Figure 6-6, lower values were significantly correlated with higher OM

(PCC -0.36, slope p < 0.05). One roof had a notably high value of CC, 4.9; in Figure 6-7, a comparison is

presented between this PSD and another sample that shares a CU value, but has a CC of 0.5. In the less

well graded mixture (solid line), the sharp upturn between 3.4 and 9.5 mm indicates a missing size

fraction of particles that could contribute towards a relatively high Kf and low MWC.

111

Figure 6-7 Particle size distirbution curves from green roof planting media recovered from green roofs with CU ≈ 16. Dashed line media CC = 0.5; solid line media CC = 5.5.

The MWC of the samples was linearly correlated with CU (PCC -0.46, slope p < 0.05) and CC (PCC -0.48,

slope p < 0.05). The relative influence of these two particle size parameters, in combination with the OM

was considered using a regression tree approach. In Figure 6-8, it is apparent that the OM has a large

influence, with three break points at 10.5% and then 3.5% and 39%. In the low OM samples, the CC is

more influential on the MWC, whereas in the high range CU is the relevant factor (R2 = 0.59 on random

sampling).

112

Figure 6-8 Regression tree of MWC demonstrating the relative importance of OM and interaction with particle size parameters CU and CC.

The media textures ranged from a tight clayey topsoil type media (low φ), to some extremely lightweight

fluffy textured potting compost type materials (high φ). Overall the values of φ spanned 0.6 units, centred

about a mean of 0.6; the data were not normally distributed; OM was a significant predictor for φ (PCC

0.54, slope p < 0.05). An inverse relationship was also identified between FAS and OM (Equation 6-2 and

Figure 6-9) and defined using curve fitting, such that R2 =0.52:

𝐹𝐴𝑆 =0.36

1 + 0.06(𝑂𝑀)

Equation 6-2

113

Figure 6-9 Relationship between free air space and organic matter content in green roof planting media

Incorporating additional independent variables with further regression did not improve the prediction of

FAS. During survey visits to eighteen of the green roofs, infiltration rates were measured in situ; these

yielded high standard deviations owing to the relative heterogeneity of the planting media compared to

the instrument size. The green roof media samples were also very coarse compared to natural soils, with

large inter-particle pore sizes and associated high Kf values, as seen in Figure 6-10. It is worth noting that

laboratory measurements of Kf can be confounded by a number of factors that are particularly relevant to

green roof planting media. Material inhomogeneity in relation to sample size (particularly where sample

size is limited), contributes to a wider standard deviation between replicates. The poorly graded mixtures

also make the samples prone to internal erosion during testing, whereby fines migrate through the coarse

fragments (Chapuis, 2012; Li, 2008). During these tests, this manifests as a reduction of Kf between

replicates due to the clogging of the retaining geotextile layer. By the same rationale, it could by

anticipated that during a standard test using a sample column with perforated base, repeated analyses on

114

the same sample would show increased Kf as the fine materials are washed out of the sample entirely

(ASTM E2399 -11, 2011).

The infiltration results were compared to the laboratory data for Kf from these roofs and the two were

significantly correlated as shown in Figure 6-10 (PCC 0.52, slope p < 0.05). The infiltration rates were

not significantly influenced by OM (PCC -0.42, slope p > 0.05), but were significantly influenced by d10.

As anticipated, media with larger minimum particle sizes have a more open pore network and increased

infiltration rates (PCC 0.41, slope p < 0.05).

Kf values were significantly influenced by OM (PCC -0.39, slope p < 0.05). The difference in OM

influence on water flow properties between the field and laboratory measurements may be due to

shrink/swell characteristic of the biologically sources materials (Manel et al., 2011). The in-situ

measurements were usually taken whilst the green roofs were well below saturation levels of water

content, whilst the Kf were measured on saturated and soaked samples, which could cause biologically

sourced materials to expand differentially compared to mineral aggregate particles. High OM materials

then had typically lower Kf values owing to the reduction of connected pore spaces and channels within

the samples.

Many predictive models relating Kf to a particle size parameter, do so in the form 𝐾 ∝ (𝑑10)2 (Chapuis,

2012; Chapuis et al., 1989). Using data from thirty-one samples, a curve with the exponent 2 resulted in

an R2 = 0.09. As other models have used other exponents this was also tried (Chapuis, 2012; Shahabi et

al., 1984) (𝑦 = 𝑎 × 𝑥𝑏) and also demonstrated no discernable relationship (R2 = 0.13). The CU is another

parameter frequently used to predict Kf; using the same curve fitting as above also resulted in a poor fit

(R2 = 0.05).

High permeability is associated with lower air entry pressures, which could result in non-linear MWC

change with green roof depth. When tested, no correlation was found between MWC and surveyed depth

of the samples (PCC 0.26, slope p > 0.05).

115

Figure 6-10 Relationship between infiltration and permeability rates in eighteen green roof media samples

116

Chemistry

The chemistry of water extracts taken from the planting media are presented in Table 6-3, all of the

measurements were strongly impacted by the proportion of OM. Significant positive linear correlations

were found with the colour, σw, phosphate and TP content of water extract, whereas pH was significantly,

negatively, linearly correlated (PCC in Table 6-3, all slopes p < 0.05). The pH values ranged between

6.20-8.45, which represents a slightly wider range than the 6.5 – 8.0 recommendation from the FLL (e.V.,

2008) and is encompassed by the terms ‘slightly acidic’ to ‘moderately alkaline’ by the USDA and

considered adequately neutral for growth of soil microbiota including actinomycetes, bacteria, fungi and

protozoa (USDA Natural Resources Conservation Service, 2011).

All of the samples yielded visible coloration of the water extracts. A similar result to the reported in

Pennsylvania, indicating a color range for individual green roof runoff events of roughly 100 to 1,000

Pt/Co color equivalents (Berghage et al., 2010). As these were concentrated preparations, rather than

runoff water collection samples, some particularly compost-rich installations exceeded the normal range

of the instrument. Electrical conductivity measurements were taken of the solutions prepared for the

phosphorous analyses and are presented as an overall indicator of ‘salinity’ for comparison between

samples. A strong correlation between the σw and reactive phosphate indicates that there is no

independently variable source of ions in the media (PCC 0.59, slope p < 0.05). In addition to the

relationship with OM, the correlation between reactive phosphate and TP suggests that synthetic

fertilizers are not a significant source of phosphate in this sample set, despite some samples having visible

colored particles or granules. As the σw and colour of the sample extracts were highly correlated with

phosphorous (and OM), it may be possible for future studies to make use of these rapidly measured

parameters as surrogates for phosphorous concentrations in field studies involving green roof planting

media.

Table 6-3 Chemistry of water extracts prepared from thirty three green roof media samples.

Mean Standard Deviation Min Max PCC with OM

Colour (Pt/Co) 708.9 579.5 162 2712 0.79

Reactive phosphate (ppm) 2.5 2.1 0.2 9.7 0.79

WETP (mg/kg) 4.5 4.1 0.5 16.1 0.86

σw (dS/m) 0.13 0.13 0.04 0.65 0.47

pH 7.2 0.5 6.2 8.2 -0.49

117

Conclusions

Overall, the mean depths of the individual green roofs were not significantly changed from their original

designs. In particular, the combination of higher organic matter and up to seventeen year of maturation

has not caused there to be significant loss in the planting medium. This may be due in part, to early

processes of pedogenesis, i.e. breakdown products of the mosses and other plants contributing to the

overall media profiles. This observation supports the hypothesis that biologically derived planting media

can be employed for sustainable, long term extensive green roof installations in this region.

The samples containing higher organic matter had lower density, compared to the mineral alternative

when both saturated and dried (ρm 1.2 and 1.0 g/cm3, ρd 0.8 and 0.3 g/cm3 respectively). This is a

significant benefit for roof dead loads; a particular concern for retrofit roofs. In some circumstances

employing a lighter-weight medium may mean a greater planting depth can be used and with that, a

greater range of vegetation may be supported.

Higher OM planting media was associated with greater water retention under laboratory test conditions.

This extra water retention would improve green roof performance in a stormwater hydrology context,

provide more plant available water, and increase the potential for evaporative cooling. However, higher

OM was correlated with more yellow/brown color, dissolved salts and, phosphorous in water discharging

from the media. In making green roof design decisions for stormwater management, there is a balance to

be struck between reducing discharged water, versus a potential increase in nutrient runoff.

Samples with low organic matter content had typically higher permeability. To date, this has been

promoted as a desired property, to reduce loading and ponding on green roofs. But there is also some

benefit to be derived from slowing the passage of water slightly, and potentially increasing peak lag times

in situ.

118

: Conclusions

Throughout the four preceding chapters a number of regression trees have been presented to help inform

decision making processes in the design of green roofs with a focus on stormwater management. Even

within the confines of this objective, the recommendations and findings depend upon the purpose of the

extensive green roof: to trap water, to slow the flow of water, or to remove water from an integrated

system. The major field studies described from GRIT were subject to the local climate and the constraints

of the physical operation of the system; such that the distribution of storm characteristics observed fell

largely within the expected range for a period of up to two years (albeit that a >100 yr event was observed

without data collection). These parameters include the depth of storms recorded and the peak intensity of

rainfall.

With the expectation of climate change bringing more extreme rainfall events, the metrics calculated and

presented may not hold true decades into the future. But by including statistical analyses from green roof

installations in the field and laboratory analyses of the most important components of green roof

construction, this body of work improves our understanding of the changes that may occur in maturing

extensive green roofs, and provides a basis for modeling the performance of such systems under more

dramatic weather conditions.

The Extensive Green Roof as a Reservoir

Irrigation

On average, the extensive green roofs studied at the GRITlab retained 50 % (Cvol = 0.5) of the total annual

precipitation, including all rain and snow. The irrigation programming made a significant difference to

this first decimal place; the accuracy to which this figure is commonly employed. Daily irrigation raises

Cvol up to 0.6, and removing all irrigation drops it down to 0.4. The FLL recommend the use of a

‘coefficient of discharge’ as 0.4 for green roofs between 10-15 cm depth (and <15 °) (e.V., 2008); this

document is currently cited as a reference for consulting engineers locally (Water, 2006). Which means in

practice, when green roofs are modelled, they may be underrepresenting the annual discharge volumes, if

irrigation practices are not considered.

The Toronto Green Standard equates annual runoff coefficients to storm depth (note that the definition of

‘each rainfall’ is daily precipitation):

“WQ 2.2 Stormwater retention & reuse: Retain at least the first 5 mm from each rainfall through

rainwater reuse, on-site infiltration and evapotranspiration

119

OR

Ensure that the maximum allowable annual runoff volume from the development site is no more than 50%

of the total average annual rainfall depth.”

(City of Toronto, 2014)

This connection also comes from the Wet Weather Flow Management Guidelines (Water, 2006), which

references “almost identical” results in an earlier report issued by the MOE (MMM Ltd., 1994). However,

this relationship must be applied with caution if the objective of analysis is to determine the storage

potential of an extensive green roof system during a rainstorm event. Whilst in fall months Cvol rises due

to reduced evapotranspiration potential, spring time Cvol are very much higher owing to meltwater runoff

(see Chapter 3). Snow accumulation in the winter months of November to April was positively associated

with irrigation and the type of planting medium; but only the irrigation resulted in significantly higher

meltwater discharge. Incorporating this thaw data for use to predict the storage potential of green roof

systems during rainstorms appears inappropriate. For this reason, summer Cvol values, incorporating only

rainfall events have been determined and presented separately. The group mean summer Cvol across all

GRITlab modules was 0.4, and again most significantly affected by irrigation programming. Daily

irrigation increases the Cvol to 0.5, but the sensor controlled programming Cvol was indistinguishable from

the none irrigated systems, both having Cvol of 0.3.

A second important disconnect between the per event and annual net rainwater retention is evident though

the modeling of NRCS curve numbers. In all of the data sets, from every module, there were many events

< 5 mm which resulted in discharged water despite the net seasonal retention being 50% or greater

(Appendix B: Data relating to Chapter 2). This is due to the use of a particularly short inter-event time

definition (1 hour). This definition creates a larger number of smaller rainfall events which can and do

occur quickly one after another, creating wetter antecedent conditions more frequently. Overall, this

manifest as a very low λ, which has not been reported previously in the literature and demonstrates a

potential weakness in the application of curve numbers to model the performance of green roof systems.

However, the production of curve numbers does permit comparison of the different designs. The impact

of irrigation was evident through individual event analysis, where daily irrigation adds 5 points to the

NRCS curve number (94) over the value of 89 for roofs without irrigation or using a sensor controlled

system. The NRCS method relates the curve numbers to a theoretical system potential storage value, (S,

mm); which worked out to be 17 mm for roofs with daily irrigation and 32 mm for those with sensor

controlled or none.

120

Planting Medium

In systems without irrigation, the use of a mineral based planting medium increased summer season Cvol

by 0.1 over the biologically-derived material. In this case, without irrigation and with the biological

medium, Cvol is 0.3. Under laboratory test conditions the compost based planting medium, identified as

material H, was predicted to retain up to an additional 70% water over the mineral based planting

medium, identified as material I (50 mm/30 mm, at 10 cm). However, the testing and subsequent

modelling of a static equilibrium condition created by completely saturating the materials, a scenario

which rarely occurs during infiltration and percolation of rainwater through green roof materials.

Both blended planting materials (H and I, in Chapter 5) outperformed bulk components, which were

generally poorly-graded regardless of their material composition. Across all of the biologically derived

and mineral based materials, the maximum potential stormwater storage was positively associated with a

higher proportion of fines, whilst the amount of plant available water was better correlated with OM.

When modelled from the water retention curves, all of the porous materials had the potential to retain less

water in the upper part of a 15 cm medium profile compared to the saturated zone at the bottom.

Hydraulic conductivity was negatively correlated with stormwater storage, as it is reduced by a higher

proportion of fines. For this reason, guidelines for the preparation of green roof planting medium caution

against a higher proportion of fines, citing potential waterlogging problems (e.V., 2008). The container

capacity and wilting point of the bulk components and planting media could be successfully predicted

from the commonly measured parameters, dried bulk density and organic matter by loss on ignition,

within this limited sample set. Whilst the margins of error are large, the values determined are

dramatically different to some guidelines which use currently use sand as a representative material in

modelling green roofs as reservoirs (Westhoff Engineering Resources Inc., 2011).

Around one third of the surveyed installations in Chapter 6 appear to have been constructed using some

form of biologically derived planting medium with OM ≥ 30. The practice of using a higher OM planting

medium did not appear to have become more or less popular over the last 18 years, as this type of

medium was equally found in some of the oldest and newest installations. On average, all of the

installations had lost around 10% depth over their design specifications, but this was not associated with

age or the use of higher OM materials. In the laboratory, the biologically derived media produced water

extracts which were more coloured, had lower pH and higher total phosphorous (TP) concentration.

Depth

In the GRITlab experiment the depth of the extensive green roofs was of relatively little significance in

the hydrology in either the summer or winter season. A small contribution was made to annual Cvol in the

121

irrigated conditions, but these changes in the second significant figure are not impactful in practice.

Laboratory testing and modelling of bulk materials and blended mixtures found that improved water

retention was associated with particle gradation properties and that increased air entry pressure was

associated with maintaining the storage capacity per unit depth through a green roof profile up to 15 cm.

Planting type

The type of vegetation made no significant contributions to the statistical data. In the GRITlab experiment

it was hypothesised that taller meadow species, under improved growing conditions may provide

structural support, maintaining snow depth against wind erosion in the winter months. In the appendices

the importance of the development of vegetation in green roofs is discussed qualitatively in respect to: the

processes of succession that occur, with the influx of volunteer species over time, and as provision of

cover to reduce the impact of wind scour eroding the panting media.

The Extensive Green Roof as an Orifice

The peak runoff coefficient, Cpeak, for use in the Rational method, was determined for the experimental

extensive green roofs on GRITlab. For rainfall events of intensity up to 1.6 mm/min the Cpeak was 0.12

and was unchanged by any of the design parameters tested. This reduction in peak flow is very

comparable to previous studies (Carpenter and Kaluvakolanu, 2011; Moran et al., 2004; Voyde et al.,

2010) and is markedly lower than the often employed value of 0.5 (Aster, 2012; e.V., 2008). This

significant reduction of peak flow may be coupled with longer residence time of the rainwater within the

green roof medium. Ponding of rainwater was not observed on any of the green roof modules during the

two-year study at the GRIT laboratory. This may be due to the occurrence of a maximum peak rainfall

intensity matching this return period (~1.5 mm/min). To assess the theoretical risk of ponding on the

green roofs, the Ksat (Chapter 5) and Kf (Chapter 6) can also be compared against this value. The materials

(A-J) tested in Chapter 5 all had Ksat greater than the 100-year return period ipeak of 4.2 mm/min (City of

Toronto, 2006). From these results, no ponding would be anticipated on these materials.

The testing of Kf on the cored materials taken from existing green roof installations in Chapter 6 were

conducted with the media supported by geotextile rather than a porous plastic frit (as is used in the KSAT

apparatus). The geotextile emulates field conditions more closely by including the interaction effect of the

finest media particles within the pores of the fabric. Of the thirty-two samples tested; twenty-six had Kf <

4.2 mm/min and may cause water to pond during a particularly intense 100-year storm, nineteen had Kf <

2.7 mm/min and may cause water to pond during a particularly intense 10-year storm, and eight had Kf <

1.5 mm/min, corresponding to a 2-year return period ipeak.

122

It is understood from industry information that a number of these installations (Chapter 6) were

constructed from the green roof media identified as samples H and J in Chapter 5 (Bioroof Systems,

2014; LiveRoof, 2015, 2009a, 2009b; McGlade, 2015). This experimental evidence supports the

hypothesis that the geotextile plays a significant role in the rate of discharge flow from green roofs. It is

often specified that the geotextile must be a free-flowing as possible to reduce saturated conditions within

the planting medium, citing plant health as a concern (e.V., 2008). However, it may be possible to

decrease the permittivity of the geotextile, permitting short-term periods of water-logging within the

medium and reducing Qp in higher intensity events, without harming long-term vegetation survival.

Surface flow on a green roof has been reported in just one published study by Voyde et al., (2010),

although the authors did note that sustained ponded water was not observed. The pore size of the

geotextile also influences the drainage characteristics and saturation profile throughout the depth of the

green roof systems modelled in Chapter 6. By decreasing pore size, reducing drainage and increasing θ at

the upper boundary of the planting media, evaporation from the planting medium could be maximized.

The Extensive Green Roof as an Evaporation pan

The extreme extension of the idea to optimize evaporation is the ‘blue roof’, where free water is permitted

to pond on the roof during a rainstorm event and then released in a controlled fashion at a later time, or

retained until fully evaporated. Whilst this is being explored in New York ((NYC Environmental

Protection, 2016)) and in other less high-profile jurisdictions (Crawford, 2013), it requires a high degree

of confidence in the waterproofing, which not all manufacturers are prepared to warranty. However,

almost all proprietary green roof systems incorporate some form of retention in the drainage trays, and

some manufacturers are working specifically increasing the capacity of this reservoir component

(LiveRoof, 2014), some of which incorporate wicking components (Zinco, 2011), or combine blue roof

and green roof components (Duncan, 2015). Combining a green roof surface with a separate rooftop

cistern is not a novel idea (Mann, 2004; McGlade, 2015), although the modelling of such systems has not

been conducted in a climate which experiences large seasonal swings in precipitation and temperature.

The purpose of the green roof also serves to empty out the cistern quickly as a recipient of irrigation with

the harvested water, to maximize the available storage space before the next storm event. An initial

evaluation of the extensive green roofs tested found that through June – September up to 16 mm (L/m2)

per month could be evapotranspired through this method. Using a closed loop system to capture up to

100% of annual rainfall between the green roof and cistern can entirely negate concerns over leaching of

excess nutrients.

Compost based media has the positive benefit of increased stormwater retention, but the negative trade-

off of increased TP in the discharged waters. The group mean value of 5.5 mg/L in the discharge from the

123

compost based, ‘biologically derived’ modules is very similar to that in untreated wastewater. There was

around a three-fold increase in net phosphorous loading from compost green roofs compared to the low

OM alternative, marking this type of system out in a Low Impact Development toolkit of systems

designed to reduce stormwater flow and improve water quality. The relative co-benefits of each

individual green roof would have to be weighed to determine if the additional phosphorous discharge was

of relative concern. Green roofs comprise a relatively small area of green space in the City of Toronto

overall, so that the net effect at a watershed scale is relatively minor, particularly as there is no restriction

to the liberal application of fertilizer in any other green space as an owner or stakeholder sees fit. In

neighbouring jurisdictions such as in the Lake Simcoe watershed, this may not be acceptable (Ontario

MOECC, 2010).

Recycling harvested rainwater and potentially supplementing with selected greywater for irrigation may

yet prove the best way for green roofs to manage stormwater, by removing up to all of annual volume,

flow and chemical loading. Irrigation supports a wider selection of vegetation which may not be so

drought tolerant but support more native insect and animal species and maximises summer season

evaporative cooling. This type of integrated system presents the most exciting opportunities to optimise a

number of infrastructural and ecosystem co-benefits and reduce the compromises which otherwise exist

between excess discharge water and vegetation productivity. Although from a commercial perspective,

the problem remains that extensive green roofs are a relatively expensive mechanism by which to

discharge harvested rainwater.

124

Further Work

This body of work has raised questions, not only regarding policy, but also on technical and operational

matters. Some of the content alluded to in this section has not been fully explored or included in previous

chapters, and has been incorporated for reference:

Irrigation

In these unusually freely draining materials with high hydraulic conductivity and low air entry pressures,

there is a significant gap in the literature in modelling, observing and explaining the application of

irrigation. Excess water from just five minutes of drip irrigation onto GRITlab modules was in excess of

100 L per module per month, whilst areas in between the lines were barely wetted. The critical ratio

between the radius (x, y) and penetration (z) of the wetted zone over time could easily be modelled using

the Van Genuchten parameters already derived.

The most significant mechanism by which irrigation impacts rainwater retention is by increasing the

antecedent moisture condition of the planting medium in advance of the event. There are increasingly

economic and accessible irrigation systems with mobile applications which schedule irrigation in

anticipation of incoming rain rather than the type of sensors investigated here which are simply

responsive to previous rainfall. These types of ‘smart’ programme systems should improve the

stormwater retention of extensive green roofs and field trials of various combinations could be

considered.

Other non-biotic components

There is a dearth of published studies exploring the effect of changing the properties of the geotextile

used in green roof construction. The assumption has always been that the most desirable outcome is to

drain the medium as quickly as possible for the health of the vegetation and to reduce loading on the roof.

However, slowing the rate of percolating water by a number of hours would not necessarily damage the

planting and the loading would be comparable to winter time snow loading which is accounted for in new

building design.

Similarly, there is growing interest in storing more water in the preformed ‘drainage’ boards, and there

are many questions asked by industry practitioners, about the fate of the water that is contained within the

individual cups of some existing pre-formed boards. Completely redesigning the reservoir/drainage

component to create green roof/ blue roof hybrid systems which permit more free water to flow beneath

the planting medium is an emerging trend which has not been researched.

125

Cisterns

Green roofs can be employed as a convenient repository for excess stormwater which has been retained in

an underground vault or cistern since the last precipitation event. There remain significant opportunities

to research how best to maximise the efficiency of a green roof system explicitly designed to

evapotranspire excess stored stormwater on site. If stormwater is only ever recirculated until evaporated

to dryness, rather begin permitted drain out of the system, there may come a time at which salinity rises to

unacceptable levels. Is this an actual risk? How long would it take? Can it be modelled or only measured?

If field measurements are required to undertake this, a good estimate of the bulk dielectric of green roof

media would be required (see Chapter 4).

Nutrition versus pollution

Much green roof chemistry research to date has focussed on the common, relatively low OM type

materials. There is much work to be done in profiling and understanding the chemical species involved in

the macro-nutrient elements in different type of composted products. Different compounds of nitrogen

and phosphorous have different plant availability and differing propensity to elute into discharged water,

and different feedstocks create very different composts.

Development of Organic Matter

It has been observed that mineral based modules on GRITlab have the potential to increase in OM, it

would be interesting to see if this were widespread and which of the other design factors were influential

in this development. Ideally this study would be replicated in future years as pedogenesis is a slow

process, compared to an academic program.

Final Comments: The ‘Best’ Extensive Green Roof?

Not only is this a difficult question to answer with due consideration to other co-benefits of green roofs

such as biodiversity, thermal, aesthetic etc. not even discussed in this research; but, it is a difficult

question to answer just with stormwater management as a priority:

Firstly, an increased depth of planting medium profile had negligible impact compared to the other design

factors, so 10 cm would suffice in order to reduce roof loading. Modelling saturation from the base,

indicates that this would also have the greatest maximum potential θ at the surface, particularly in

combination with the additional water retention potential through the use of a compost-based or

biologically-derived media. This type of materials also reduces the overall weight of the system, even

when saturated.

126

The use of Sedum mats prevents loss of medium through wind erosion and on GRITlab, it is readily

observed that the Sedum are more resilient to fluctuations in water availability. However, from experience

meeting dozens of green roof owners and designers, and surveying their installations, I believe that all

green roofs eventually end up neglected (Appendix F: Success and Succession). For this reason, a

‘ceremonial’ top dressing of wildflower seeds would be recommended once the installation is completed

and the ribbon being cut. The idea is to communicate an understanding of the green roof system as an

evolving piece of landscape, rather than another building material that will simply ‘degrade’ over time.

On GRITlab, the nearest combination is the: 10 cm, no irrigation, compost based, Sedum. module is

shown on 23 June 2015 in Figure 7-1. Of the installations surveyed, the glorious Toronto Botanical

Garden roof, referred to as ‘roof H’ in Chapter 6. comes closest to the ‘best’ extensive green roof, see

Figure 7-2.

Figure 7-1 GRITlab module 6E, 23 June 2015

127

Figure 7-2 Toronto Botanical Garden Extensive Green Roof, 28 May 2014

128

: References

Abad, M., Noguera, P., Puchades, R., Maquieira, A., Noguera, V., 2002. Physico-chemical and chemical

properties of some coconut coir dusts for use as a peat substitute for containerised ornamental

plants. Bioresour. Technol. 82, 241–245. doi:10.1016/S0960-8524(01)00189-4

Adams, B.J., Papa, F., 2000. Urban Stormwater Management Planning with Analytical Probabilistic

Models. John Wiley Sons Inc.

Al-Khafaji, A.W., Andersland, O.B., 1992. 5.2 Permeability measurment - laboratory, in: Geotechnical

Engineering & Soil Testing. Saunders College, USA, pp. 136–139.

Allen, R.G., Walter, I.A., Elliott, R., Howell, T., Itenfisu, D., Jensen, M., 2005. THE ASCE

STANDARDIZED REFERENCE EVAPOTRANSPIRATION EQUATION (RPRT). ASCE

Publications.

Almssad, A., Almusaed, A., 2015. Environmental reply to vernacular habitat conformation from a vast

areas of Scandinavia. Renew. Sustain. Energy Rev. 48, 825–834. doi:10.1016/j.rser.2015.04.013

Annan, J.S., White, R.K., 1998. Evaluation of techniques for measuring air filled- porosity in composts of

municipal biosolids and wood chips, in: Conferences of Composting the South East . South

Carolina, USA, pp. 88–96.

Anselin, L., Syabri, I., Kho, Y., 2006. GeoDa: An Introduction to Spatial Data Analysis. Geogr. Anal. 38,

5–22. doi:10.1111/j.0016-7363.2005.00671.x

architectsAlliance, 2004. Specifications for green roof at Pond Road residence, York University.

ASCE/EWRI Curve Number Hydrology Task Committee, 2009. Findings and Developments, in:

Hawkins, R.H., (U.S.)., E. and W.R.I., Committee., A.C.N.H.T. (Eds.), Curve Number Hydrology :

State of the Practice. American Society of Civil Engineers, Reston, Va.

Aster, D., 2012. Toronto Green Roof Construction Standard: Supplementary Guidelines.

ASTM D2487 -11, 2011. Standard Practice for Classification of Soils for Engineering Purposes (Unified

Soil Classification System). doi:10.1520/D2487-11

ASTM E2399 -11, 2011. Standard Test Method for Maximum Media Density for Dead Load Analysis of

Vegitative (Green) Roof Systems. doi:10.1520/D6836-02R08E02

129

ASTM E2777 -14, 2014. Standard Guide for Vegetative (Green) Roof Systems. doi:10.1520/E2777-14

Ayers, R.S., Westcot, D.W., Food and Agriculture Organization of the United Nations., 1985. Water

quality for agriculture. Food and Agriculture Organization of the United Nations.

Babilis, D.J., Londra, P.A., 2011. Assessment of various methods for the determination of hydraulic

conductivities of two green roof substrates by steady state infiltration experiments. Irrig. Drain.

doi:10.1002/ird.543

Bates, B., 2008. Analysing regional aspects of climate change and water resources: North America, in:

Bates, B.C., Kundzewicz, Z.W., Wu, S., Palutikof, J.P. (Eds.), Climate Change and Water, IPCC

Technical Paper VI. IPCC Secretariat, Geneva, pp. 102–108.

Beardsell, D.V., Nichols, D.G., 1982. Wetting properties of dried-out nursery container media. Sci.

Hortic. (Amsterdam). 17, 49–59. doi:10.1016/0304-4238(82)90061-9

Beck, D.A., Johnson, G.R., Spolek, G.A., 2011. Amending greenroof soil with biochar to affect runoff

water quantity and quality. Environ. Pollut. 159, 2111–2118. doi:10.1016/j.envpol.2011.01.022

Beecham, S., Razzaghmanesh, M., 2015. Water quality and quantity investigation of green roofs in a dry

climate. Water Res. 70, 370–384.

Berghage, R., Jarrettm, A., Beattie, D., Kelley, K., Husain, S., Rezai, F., Long, B., Negassi, A., Cameron,

R., Hunt, W., 2007. Quantifying Evaporation and Transpirational Water Losses from Green Roofs

and Greeen Roof Media Capacity for Neutralizing Acid Rain (RPRT). National Decentralized

Water Resources Capacity Development Project (NDWRCDP).

Berghage, R., Miller, C., Bass, B., Moseley, D., Weeks, K., 2010. Stormwater Runoff from a Large

Commerical Roof in Chicago, in: CitiesAlive!: Eighth Annual Green Roof and Wall Conference.

Green Roofs for Healthy Cities, Vancouver.

Berghage, R.D., Beattie, D., Jarrett, A.R., Thuring, C., Razaei, F., 2009. Green Roofs for Stormwater

Runoff Control (RPRT). National Risk Management Research Laboratory.

Berndtsson, J.C., Emilsson, T., Bengtsson, L., 2006. The influence of extensive vegetated roofs on runoff

water quality. Sci. Total Environ. 355, 48–63. doi:10.1016/j.scitotenv.2005.02.035

Bioroof Systems, 2014. Native Child and Family Services Center [WWW Document]. URL

http://www.bioroof.com/BR_projects10.html (accessed 8.8.16).

130

Bioroof Systems, 2011. Bio-mix Eco-blend BIO-010 [WWW Document]. URL

http://www.bioroof.com/_links/TD010_BioMixEcoBlend.pdf (accessed 6.6.16).

Blanusa, T., Monteiro, M.M., Fantozzi, F., Vysini, E., Li, Y., Cameron, R.W.F., 2013. Alternatives to

Sedum on green roofs: Can broad leaf perennial plants offer better “cooling service”? Build.

Environ. 59, 99–106.

Boivin, M.-A., n.d. Older green roofs.

Bousselot, J.M., Klett, J.E., and Koski, R.D., 2011. Moisture Content of Extensive Green Roof Substrate

and Growth Response of 15 Temperate Plant Species during Dry Down. HortScience 46, 518–522.

Breuning, J., 2013. Irrigation on Extensive Green Roofs - Facts Study [WWW Document]. URL

http://www.greenroofs.com/content/Irrigation-on-Extensive-Green-Roofs-Facts-Study.htm

Buccola, N., Spolek, G., 2011. A pilot-scale evaluation of greenroof runoff retention, detention, and

quality. Water. Air. Soil Pollut. 216, 83–92. doi:10.1007/s11270-010-0516-8

Buffam, I., Mitchell, M.E., Durtsche, R.D., 2016. Environmental drivers of seasonal variation in green

roof runoff water quality. Ecol. Eng. 91, 506–514. doi:10.1016/j.ecoleng.2016.02.044

Buist, R., Friedrich, C., 2008. The Organic Question. Living Archit. Monit. 17–20.

Burszta-Adamiak, E., 2012. Analysis of the retention capacity of green roofs. J. Water L. Dev. 16, 3.

Burszta-Adamiak, E., Mrowiec, M., 2013. Modelling of Green roofs’ hydrologic performance using

EPA’s SWMM. Water Sci. Technol. 68, 36–42. doi:10.2166/wst.2013.219

Canada, E., 2013. Toronto 1971 to 2000 Canadian Climate Normals station data [WWW Document].

URL

http://climate.weather.gc.ca/climate_normals/results_e.html?stnID=5051&lang=e&dCode=1&Statio

nName=TORONTO&SearchType=Contains&province=ALL&provBut=&month1=0&month2=12

(accessed 3.9.16).

Canada, G. of, 2015. Historical Climate Data [WWW Document]. URL

http://climate.weather.gc.ca/index_e.html#access (accessed 10.26.15).

Carbone, M., Garofalo, G., Nigro, G., Piro, P., 2014. A Conceptual Model for Predicting Hydraulic

Behaviour of a Green Roof. Procedia Eng. 70, 266–274.

131

Carlisle, S., Piana, M., 2013. Growing resilience: Long-term plant dynamics and green roof

performance , in: Cities Alive - 11th Annual Green Roof and Wall Conference. Green Roofs for

Health Cities, San Francisco.

Carminati, A., Kaestner, A., Lehmann, P., Fl??hler, H., 2008. Unsaturated water flow across soil

aggregate contacts. Adv. Water Resour. 31, 1221–1232. doi:10.1016/j.advwatres.2008.01.008

Carpenter, D.D., Kaluvakolanu, P., 2011. Effect of Roof Surface Type on Storm-Water Runoff from Full-

Scale Roofs in a Temperate Climate. J. Irrig. Drain. Eng. 137, 161–169.

doi:10.1061/(ASCE)IR.1943-4774.0000185

Carson, T.B., Marasco, D.E., Culligan, P.J., McGillis, W.R., 2013. Hydrological performance of

extensive green roofs in New York City: observations and multi-year modeling of three full-scale

systems. Environ. Res. Lett. 8, 24036. doi:10.1088/1748-9326/8/2/024036

Carter, T., Jackson, C.R., 2007. Vegetated roofs for stormwater management at multiple spatial scales.

Landsc. Urban Plan. 80, 84–94. doi:10.1016/j.landurbplan.2006.06.005

Carter, T.L., Rasmussen, T.C., 2006. HYDROLOGIC BEHAVIOR OF VEGETATED ROOFS. JAWRA

J. Am. Water Resour. Assoc. 42, 1261–1274.

Cassel, D.K., Neilsen, D.R., 1986. Field Capacity and Available Water Capacity, in: Klute, A. (Ed.),

Methods of Soil Analysis - Part 1. Physical and Mineralogical Methods. pp. 901–924.

doi:10.2136/sssabookser5.1.2ed.c28

Castleton, H.F., Stovin, V., Beck, S.B.M., Davison, J.B., 2010. Green roofs; Building energy savings and

the potential for retrofit. Energy Build. 42, 1582–1591. doi:10.1016/j.enbuild.2010.05.004

Central Lake Ontario Conservation, 2010. Technical Guidelines for Stormwater Mnagement Submissions

(RPRT). CLOCA.

Chandra, S., Berntsson, L., 2002. Lightweight Aggregate Concrete, Lightweight Aggregate Concrete.

Elsevier. doi:10.1016/B978-081551486-2.50005-5

Chao-Hsien, L., En-Hao, H., Yie-Ru, C., 2015. Designing a rainwater harvesting system for urban green

roof irrigation. Water Sci. Technol. Water Supply 15, 271.

Chapuis, R., 2012. Predicting the saturated hydraulic conductivity of soils: a review. Bull. Eng. Geol.

Environ. 71, 401–434. doi:10.1007/s10064-012-0418-7

132

Chapuis, R.P., Gill, D.E., Baass, K., 1989. Laboratory permeability tests on sand: influence of the

compaction method on anisotropy. Can. Geotech. J. 26, 614–622.

Chatterjee, N., Flury, M., Hinman, C., Cogger, C.G., 2013. Chemical and Physical Charactersitcs of

Compost Leachates: A Review.

Cheung, P., 2016. A quick request.

Cirillo, C., Podolsky, L., 2012. Health, Prosperity and Sustainability: The Case for Green Infrastructure in

Ontario (RPRT). Green Infrastructure Ontario Coalition; Ecojustice, Toronto.

City of Toronto, 2016a. Toronto Beach Map - Stormwater Management - Water.

City of Toronto, 2016b. Community Council Area Profiles - Demographics [WWW Document]. URL

http://www1.toronto.ca/wps/portal/contentonly?vgnextoid=91a998b613412410VgnVCM10000071d

60f89RCRD&vgnextchannel=1e68f40f9aae0410VgnVCM10000071d60f89RCRD (accessed

8.4.16).

City of Toronto, 2010. Cleaning up Our Waterways: The Don River and Central Waterfront Project

Frequently Asked Questions. Toronto, Canada.

City of Toronto, 2006. Wet Weather Flow Management Guidelines (RPRT). City of Toronto.

City of Toronto, C.P., 2014. Water Quality, Quantity and Efficiency. Toronto Green Stand. Mid to High-

Rise Resid. Non-resid. Dev. v 2.0.

Cobos, D.R., Chambers, C., 11AD. Calibrating ECH2O Soil Moisture Sensors (RPRT). Decagon

Devices.

Contributors, 2016. List of North American cities by population [WWW Document]. Wikipedia. URL

https://en.wikipedia.org/w/index.php?title=List_of_North_American_cities_by_population&oldid=7

18495163 (accessed 5.13.16).

Crawford, D., 2013. Blue roofs.

Cronk, E., 2012. Stormwater retention capabilities of steep sloped, extensive green roof systems (THES).

ProQuest Diss. Theses. Michigan State University, United States -- Michigan.

Czemiel Berndtsson, J., 2010. Green roof performance towards management of runoff water quantity and

quality: A review. Ecol. Eng. 36, 351–360.

133

Czemiel Berndtsson, J., Bengtsson, L., Jinno, K., miljöteknologi, I. för bygg-och, Faculty of Engineering,

L.T.H., Technology, D. of B. and E., universitet, L., vattenresurslära, A. för T., Engineering, D. of

W.R., LTH, I. vid, Högskola, L.T., University, L., LTH, D. at, 2008. First flush effect from

vegetated roofs during simulated rain events. Hydrol. Res. 39, 171.

Czuchajowska, Z., Paszczynska, B., Pomeranz, Y., 1992. Functional properties of psyllium in wheat-

based products. Cereal Chem. 69, 516.

D. C. Beard, P.K.W., 1973. Influence of Texture on Porosity and Permeability of Unconsolidated Sand.

Am. Assoc. Pet. Geol. Bull. 57, 349–369.

Dal Ferro, N., Pagliarin, C., Morari, F., 2014. Pore network and water retention characteristics of volcanic

porous media. Eur. J. Soil Sci. 65, 672–683. doi:10.1111/ejss.12163

Das, B., Sivakugan, S., 2015. 2.6 Grain Size Distribution Curve, in: Introduction to Geotechnical

Engineering, 2nd Ed. Cengage Learning, pp. 12–19.

Decagon Devices, 2016. 5TE Water Content, EC, and Temperature Sensor.

Demšar, J., Curk, T., Erjavec, A., Gorup, Č., Hočevar, T., Milutinovič, M., Možina, M., Polajnar, M.,

Toplak, M., Starič, A., 2013. Orange: data mining toolbox in Python. J. Mach. Learn. Res. 14,

2349–2353.

Dhalla, S., Zimmer, C., 2010. Low impact development stormwater management planning and design

guide (RPRT). Credit Valley Conservation Authority; Toronto and Region Conservation Authority,

[Toronto].

Di Gironimo, L., Bowering, T., Kellershohn, D., 2013. Impact of July 8, 2013 storm on the City’s Sewer

and Stormwater Systems. Toronto.

Doerr, S.H., Shakesby, R.A., Walsh, R.P.D., 2000. Soil water repellency: its causes, characteristics and

hydro-geomorphological significance. Earth Sci. Rev. 51, 33–65.

Drokin, A., 2010. Easyfit - Distribution fitting software.

Duncan, B., 2015. Blue roof.

e.V., F.L.L., 2008. Guidelines for the Planning, Construction and Maintenance of Green Roofing – Green

Roofing Guideline. Bonn.

134

Emilsson, T., Rolf, K., 2005. Comparison of establishment methods for extensive green roofs in southern

Sweden. Urban For. Urban Green. 3, 103–111. doi:10.1016/j.ufug.2004.07.001

Encyclopedia of Soil Science, 2007. . Springer Science & Business Media.

Environment and Climate Change Canada, 2016. Residential Water Use in Canada Indicator Data 1991-

2011 [WWW Document]. Environ. Indic. URL https://www.ec.gc.ca/indicateurs-

indicators/default.asp?lang=en&n=553CC57B-1 (accessed 9.23.16).

Environment and Climate Change Canada, 2013. Canada’s Top Ten Weather Stories for 2013 [WWW

Document]. URL https://ec.gc.ca/meteo-weather/default.asp?lang=En&n=5BA5EAFC-

1&offset=3&toc=hide (accessed 5.3.16).

Environment Canada, 2013. Canadian Climate Normals 1981-2010 Station Data.

Environment Canada, 2011. Glossary [WWW Document]. URL

http://climate.weather.gc.ca/glossary_e.html (accessed 5.6.16).

Farrell, C., Mitchell, R.E., Szota, C., Rayner, J.P., Williams, N.S.G., 2012. Green roofs for hot and dry

climates: Interacting effects of plant water use, succulence and substrate. Ecol. Eng. 49, 270.

Fassman-Beck, E., Carpenter, D., Hunt, W., Kurtz, T., Berghage, R., Stovin, V., Wadzuk, B., Colwell, S.,

2015. Curve Number and Runoff Coefficients for Extensive Living Roofs Across Climate Zones.

Low Impact Dev. Conf.

Fassman-Beck, E., Hunt, W., Berghage, R., Carpenter, D., Kurtz, T., Stovin, V., Wadzuk, B., 2015. Curve

Number and Runoff Coefficients for Extensive Living Roofs. J. Hydrol. Eng.

Fassman-Beck, E., Miller, C., 2016. Citation: peak lag.

Federal Interagency Stream Restoration Working Group (FISRWG), 1998. Stream Corridor Restoration:

Principles, Processes, and Practices. 10/98. Fig. 3-21.

Friedman, S.P., 2005. Soil properties influencing apparent electrical conductivity: A review. Comput.

Electron. Agric. doi:10.1016/j.compag.2004.11.001

Fuhrman, J.K., Zhang, H., Schroder, J.L., Davis, R.L., Payton, M.E., 2005. Water-Soluble Phosphorous

as Affected by Soil to Extractant Ratios, Extracction Times, and Electrolyte. Commun. Soil Sci.

Plant Anal. 36, 925–935. doi:10.1081/CSS-200049482

135

Gaches, W., Lea-Cox, J., Cohan, S., Ristvey, A., Sullivan, J., Davis, A., 2013. Substrate Particle Size

Distribution of Mature Mid-Atlantic Green Roofs, in: 2013 International Low Impact Development

Symposium. Saint Paul, Minnesota, p. 31.

Gerrits, A.M.J., 2010. The role of interception in the hydrological cycle (THES). Delft University of

Technology.

Getter, K.L., Rowe, D.B., Andresen, J.A., 2007. Quantifying the effect of slope on extensive green roof

stormwater retention. Ecol. Eng. 31, 225–231. doi:10.1016/j.ecoleng.2007.06.004

Gnecco, I., Palla, A., Lanza, L.G., Barbera, P., 2013. The Role of Green Roofs as a Source/sink of

Pollutants in Storm Water Outflows. Water Resour. Manag. 27, 4715–4730. doi:10.1007/s11269-

013-0414-0

Göbel, P., Dierkes, C., Coldewey, W.G., 2007. Storm water runoff concentration matrix for urban areas.

J. Contam. Hydrol. 91, 26–42.

Google Trends, 2016. Explore "Green Roofs" [WWW Document]. URL

https://www.google.ca/trends/explore?date=all&q=green roofs (accessed 8.8.16).

Government of Canada, 2015. Households and the environment survey, use of fertilizer and pesticides,

Canada, provinces and census metropolitan areas (CMA) [WWW Document]. Stat. Canada. URL

http://www5.statcan.gc.ca/cansim/a26#F3 (accessed 6.13.16).

Government of Canada, 2009. Guidelines for Canadian Drinking Water Quality: Guideline Technical

Document - Total Dissolved Solids (TDS) [WWW Document]. URL

http://healthycanadians.gc.ca/publications/healthy-living-vie-saine/water-dissolved-solids-matieres-

dissoutes-eau/index-eng.php#a1 (accessed 6.9.16).

Graceson, A., Hare, M., Monaghan, J., Hall, N., 2013. The water retention capabilities of growing media

for green roofs. Ecol. Eng. 61, 328–334. doi:10.1016/j.ecoleng.2013.09.030

Green Roofs for Healthy Cities, 2016. GRHC [WWW Document]. URL http://www.greenroofs.org/

(accessed 8.8.16).

Greenland International Consulting Inc., 2002. Command List, in: Visual OTTHYMOTM v2.0 Reference

Manual. Schaeffer & Associates Ltd., pp. 5–16.

Gregoire, B.G., Clausen, J.C., 2011. Effect of a modular extensive green roof on stormwater runoff and

136

water quality. Ecol. Eng. 37, 963–969. doi:10.1016/j.ecoleng.2011.02.004

Guo, T., Mrosovsky, L., 2014. AccessPoint on Danforth Green Roof Profile (RPRT). Access Alliance,

Toronto, Canada.

Hahn, E., 2009. Urban green roof vegetation assemblage demography, classification and design

recommendations (THES). Ryerson University.

Handreck, K.A., Black, N.D., 2002. Growing Media for Ornamental Plants and Turf. UNSW Press.

Hardin, M., Wanielista, M., Chopra, M., 2012. A Mass Balance Model for Designing Green Roof

Systems that Incorporate a Cistern for Re-Use. Water 4, 914–931.

Harper, G.E., 2013. Green roof water quality impacts and physicochemical stability (JOUR). Missouri

University of Science and Technology.

Harper, G.E., Limmer, M.A., Showalter, W.E., Burken, J.G., 2015. Nine-month evaluation of runoff

quality and quantity from an experiential green roof in Missouri, USA. Ecol. Eng. 78, 127–133.

doi:http://dx.doi.org/10.1016/j.ecoleng.2014.06.004

Haselbach, L., 2008. The engineering guide to LEED: new construction : sustainable construction for

engineers. McGraw-Hill, New York.

Hashemi, S.S.G., Mahmud, H. Bin, Ashraf, M.A., 2015. Performance of green roofs with respect to water

quality and reduction of energy consumption in tropics: A review. Renew. Sustain. Energy Rev. 52,

669–679. doi:http://dx.doi.org/10.1016/j.rser.2015.07.163

Hathaway, A.M., Hunt, W.F., Jennings, G.D., 2008. A Field Study of Green Roof Hydrologic and Water

Quality Performance. Trans. Am. Soc. Agric. Biol. Eng. 51, 37–44.

Heller, H., Bar-Tal, A., Assouline, S., Narkis, K., Suryano, S., de la Forge, A., Barak, M., Alon, H.,

Bruner, M., Cohen, S., Tsohar, D., 2015. The effects of container geometry on water and heat

regimes in soilless culture: lettuce as a case study. Irrig. Sci. 33, 53–65. doi:10.1007/s00271-014-

0448-y

Hemmings, R.T., Cornelius, B.J., Yuran, P., Wu, M., 2009. Comparative Study of Lightweight

Aggregates, in: World of Coal Ash (WOCA) Conference. Laxington, Kentucky, pp. 1–15.

Hilhorst, M.A., 2000. A Pore Water Conductivity Sensor. Soil Sci. Soc. Am. J.

137

doi:10.2136/sssaj2000.6461922x

Hill, J., Drake, J., Sleep, B., 2016. Comparisons of extensive green roof media in Southern Ontario. Ecol.

Eng. 94, 418–426. doi:10.1016/j.ecoleng.2016.05.045

Hill, J., Kuszkowski, H., Sleep, B., Drake, J., 2015. Gone with the wind?, in: Environmental Connection

Conference 2015. International Erosion Control Association.

Hill, J., Perotto, M., Yoon, C., 2015. Processes of Quantifying the Hydrological Performance of

Extensive Green Roofs, in: RCI International Convention Proceeedings.

Hilten, R.N., Lawrence, T.M., Tollner, E.W., 2008. Modeling stormwater runoff from green roofs with

HYDRUS-1D. J. Hydrol. 358, 288–293. doi:10.1016/j.jhydrol.2008.06.010

Inc, N.D., n.d. Central YMCA [WWW Document]. URL

http://www.roofgarden.ca/index.php?view=pageView&pageid=255

International Joint Commission, 2016. International Joint Commission [WWW Document]. URL

http://www.ijc.org/en_/ (accessed 5.13.16).

Iwata, 1994. Capillarity, in: Soil-Water Interactions: Mechanisms Applications, Second Edition, Revised

and Expanded. CRC Press, pp. 229–256.

Jim, C.Y., Peng, L.L.H., 2013. Green-Roof Effects on Neighborhood Microclimate and Human Thermal

Sensation. Energies 6, 598–618.

Jones, J.B., Jr., 2012. Plant Nutrition and Soil Fertility Manual, Second Edition. CRC Press.

Karperien, A., 2013. FracLac for ImageJ [WWW Document]. URL

http://rsb.info.nih.gov/ij/plugins/fraclac/FLHelp/Introduction.htm (accessed 6.1.16).

Kasmin, H., Stovin, V.R., Hathway, E.A., 2010. Towards a generic rainfall-runoff model for green roofs.

Water Sci. Technol. 62, 898–905.

Keesmaat, J., 2013. Within five years we’ll have more “green roofs” than any other city in the world

#ourpoliciesareworking . #TOpoli [WWW Document]. URL

https://twitter.com/jen_keesmaat/status/402477653330247680 (accessed 3.7.16).

Kelly, M., 2008. A comparative analysis of green roof designs including depth of media, drainage layer

materials, and pollution control media (THES). University of Central Florida, ctx_ver=Z39.88-

138

2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-

8&rfr_id=info:sid/summon.serialssolutions.com&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&rft.

genre=dissertation&rft.title=A+comparative+analysis+of+green+roof+designs+including+depth+of

+media%2C+draina.

Khoshand, A., Fall, M., 2014. Geotechnical Characterization of Compost Based Biocover Materials.

Geotech. Geol. Eng. 32, 489–503.

Köhler, M., Poll, P.H., 2010. Long-term performance of selected old Berlin greenroofs in comparison to

younger extensive greenroofs in Berlin. Ecol. Eng. 36, 722–729.

Köhler, M., Schmidt, M., Friedrich, W.G., Laar, M., Vera Lúcia, de A., Tavares, S., 2002. Green roofs in

temperate climates and in the hot‐humid tropics – far beyond the aesthetics. Env Mgtt Heal. 13,

382–391. doi:10.1108/09566160210439297

Kottek, M., Grieser, J., Beck, C., Rudolf, B., Rubel, F., 2006. World Map of the Köppen-Geiger climate

classification updated. Meteorol. Zeitschrift 15, 259–263. doi:10.1127/0941-2948/2006/0130

Krzeminski, J., 2013. The Next Step for Mandate Markets. Toronto Green Roof GatherinThe Next Step.

Kwik, J., 2001. Plotting over city hall. Altern. J. 27, 3.

Lake Simcoe Region Conservation Authority, 2013. LSRCA Technical Guidelines for Stormwater

Management Submissions (RPRT). LSRCA.

Latshaw, K., Fitzgerald, J., Sutton, R., 2009. Analysis of Green Roof Growing Media Porosity. Rural.

Rev. Undergrad. Res. Agric. Life Sci. 4.

Legg, J., 2016. Compost Specs.

Li, M., 2008. Seepage induced instability in widely graded soils (THES). The Univesity of British

Columbia.

Li, Y., Babcock, R.W., 2014. Green roof hydrologic performance and modeling: A review. Water Sci.

Technol. doi:10.2166/wst.2013.770

Linden, K. Vander, Stone, M., 2009. Treatment performance of an extensive vegetated roof in Waterloo,

Ontario. Water Qual. Res. J. Canada 44, 26–32.

Liu, K., Minor, J., 2005. Performance evaluation of an extensive green roof, in: Greening Rooftops for

139

Sustainable Communities. National Research Council of Canada / City of Toronto, Washing D.C.,

pp. 1–11.

LiveRoof, 2015. University of Toronto Mississauga (UTM) Instructional Centre [WWW Document].

URL http://liveroof.com/showcase-view/?show=488&ordernum=11 (accessed 8.8.16).

LiveRoof, 2014. Hybrid Green Roofs System Overview [WWW Document]. URL

http://liveroof.com/system-overview/ (accessed 8.8.14).

LiveRoof, 2009a. Toronto City Hall Podium Roof, Toronto, Ontario [WWW Document]. URL

http://www.liveroof.com/showcase-view/?project=223_1_e_toronto-city-hall-podium-

roof_toronto_ontario (accessed 8.8.16).

LiveRoof, 2009b. Esri Canada, Toronto, Ontario [WWW Document]. URL

http://www.liveroof.com/showcase-view/?project=119_1_e_esri-canada_toronto_ontario (accessed

8.8.16).

Lottie, 2016. Overhead and Underfoot [WWW Document]. URL

http://thecurlyjen.wix.com/lottie#!overhead-and-underfoot/c1mqf (accessed 8.8.16).

Lstiburek, J.W., 2011. Seeing red over green roofs. ASHRAE J. 53, 68.

Ltd, F.C., 2010. Project Work Order for green roof at Life Sciences Building, York U.

Lundholm, J., Macivor, J.S., Macdougall, Z., Ranalli, M., 2010. Plant species and functional group

combinations affect green roof ecosystem functions. PLoS One 5, e9677.

Lundholm, J.T., Weddle, B.M., Macivor, J.S., 2014. Snow depth and vegetation type affect green roof

thermal performance in winter. Energy Build. 84, 299–307. doi:10.1016/j.enbuild.2014.07.093

Ma, L., Qin, B., Chang, Q.Z., 2012. Performance of Urban Rainwater Retention by Green Roof: A Case

Study of Jinan. Appl. Mech. Mater. 178–181, 295.

MacIvor, J.S., Margolis, L., Perotto, M., Drake, J.A.P., 2016. Air temperature cooling by extensive green

roofs in Toronto Canada. Ecol. Eng. 95, 36–42. doi:10.1016/j.ecoleng.2016.06.050

MacIvor, J.S., Margolis, L., Puncher, C.L., Carver Matthews J., B., 2013. Decoupling factors affecting

plant diversity and cover on extensive green roofs. J. Environ. Manage. 130, 297–305.

Manel, K.R., Michel, J.-C., Medhioub, K., Morel, P., 2011. Evaluation of Compost Maturity,

140

Hydrophysical and Physicochemical Properties: Indicators for Use as a Component of Growing

Media. Compost Sci. Util. 19, 226.

Mann, S., 2004. Blue Roof [WWW Document]. Exp. Des. Lab. URL

http://www.experientialdesignlab.com/projects/auxiliary-projects/blue-roof (accessed 8.8.16).

Margolis, L., 2013. gritlab [WWW Document]. URL http://grit.daniels.utoronto.ca/

Matthiessen, M.K., Larney, F.J., Brent Selinger, L., Olson, A.F., 2005. Influence of Loss‐on‐Ignition

Temperature and Heating Time on Ash Content of Compost and Manure. Commun. Soil Sci. Plant

Anal. 36, 2561–2573.

Matuk, N., 1999. Green building: Mountain Equipment Co-op [Toronto]. Can. Archit. 44, 22–23.

McCoy, E., 1998. Hydrology of Modern Putting Green Root Zones, in: Golf Course Soil and Water

Science. Ohio State University.

McGlade, T., 2015. Green roof construction documents archive.

McGlade, T., Hill, J., 2014. Success and Succession, in: Cities Alive - 12th Annual Green Roof and Wall

Conference. Green Roofs for Health Cities, Nashville.

Mentens, J., Raes, D., Hermy, M., 2006. Green roofs as a tool for solving the rainwater runoff problem in

the urbanized 21st century? Landsc. Urban Plan. 77, 217–226.

Metselaar, K., 2012. Water retention and evapotranspiration of green roofs and possible natural

vegetation types. Resour. Conserv. Recycl. 64, 49–55. doi:10.1016/j.resconrec.2011.12.009

Miguntanna, N.S., Egodawatta, P., Kokot, S., Goonetilleke, A., 2010. Determination of a set of surrogate

parameters to assess urban stormwater quality. Sci. Total Environ. 408, 6251–6259.

Mishra, S.K., 2003. Soil conservation service curve number (SCS-CN) methodology. Kluwer Academic

Publishers, Boston.

Molineux, C.J., Fentiman, C.H., Gange, A.C., 2009. Characterising alternative recycled waste materials

for use as green roof growing media in the U.K. Ecol. Eng. 35, 1507–1513.

doi:10.1016/j.ecoleng.2009.06.010

Moran, A., Hunt, B., Jennings, G., 2004. A North Carolina field study to evaluate greenroof runoff

quantity, runoff quality, and plant growth., in: Green Roofs for Healthy Cities 2004 Conference.

141

Green Roofs for Healthy Cities, Portland, OR.

Moritani, S., Yamamoto, T., Andry, H., Inoue, M., Kato, K., Saito, H., 2013. Effect of combined water

and salinity stress factors on evapotranspiration of Sedum kamtschaticum Fischer in relation to

green roof irrigation. Urban For. Urban Green. 12, 338–343.

Motte, L., 2012. SMART 3 Colorimeter Operator’s Manual 1–294.

Myllys, M., Simojoki, A., 1996. Calibration of time domain reflectometry (TDR) for soil moisture

measurements in cultivated peat soils. Suo 47, 1–6.

Naasz, R., Michel, J.-C., Charpentier, S., 2008. Water repellency of organic growing media related to

hysteretic water retention properties. Eur. J. Soil Sci. 59, 156–165.

Nagase, A., Dunnett, N., 2011. The relationship between percentage of organic matter in substrate and

plant growth in extensive green roofs. Landsc. Urban Plan. 103, 230–236.

Nardini, A., Andri, S., Crasso, M., 2012. Influence of substrate depth and vegetation type on temperature

and water runoff mitigation by extensive green roofs: shrubs versus herbaceous plants. Urban

Ecosyst. 15, 697–708.

Nawaz, R., McDonald, A., Postoyko, S., 2015. Hydrological performance of a full-scale extensive green

roof located in a temperate climate. Ecol. Eng. 82, 66–80. doi:10.1016/j.ecoleng.2014.11.061

NCSS, L.L.C., 2015. NCSS 10 Statistical Software.

NYC Environmental Protection, 2016. Blue Roof and Green Roof [WWW Document]. URL

http://www.nyc.gov/html/dep/html/stormwater/green_pilot_project_ps118.shtml (accessed 8.7.16).

Oberndorfer, E., Lundholm, J., Bass, B., Coffman, R.R., Doshi, H., Dunnett, N., Gaffin, S., Köhler, M.,

Liu, K.K.Y., Rowe, B., 2007. Green Roofs as Urban Ecosystems: Ecological Structures, Functions,

and Services. Bioscience 57, 823–833.

Ogura, A.P., 2014. Method Development of Phosphate Analysis for Green Roof Soil Media. University

of Toronto.

Oleszczuk, R., Brandyk, T., Gnatowski, T., Szatylowicz, J., 2004. Calibration of TDR for moisture

determination in peat deposits. Int. Agrophysics 18, 145–151.

Optigrün international, 2008. Green Roof Systems with Granular Drainage [WWW Document]. URL

142

http://www.conservationtechnology.com/greenroof_systems_granular.html (accessed 5.8.16).

Ouldboukhitine, S.-E., Belarbi, R., Djedjig, R., 2012. Characterization of green roof components:

Measurements of thermal and hydrological properties. Build. Environ. 56, 78.

Pachepsky, Y., Rawls, W.J., 2004. Development of pedotransfer functions in soil hydrology. Elsevier,

New York.

Palla, A., Gnecco, I., Lanza, L.G., 2012. Compared performance of a conceptual and a mechanistic

hydrologic models of a green roof. Hydrol. Process. 26, 73–84.

Palla, A., Gnecco, I., Lanza, L.G., 2009. Unsaturated 2D modelling of subsurface water flow in the

coarse-grained porous matrix of a green roof. J. Hydrol. 379, 193–204.

doi:10.1016/j.jhydrol.2009.10.008

Perkins, T., 2014. The new way Toronto condo developers are saving space [WWW Document]. Globe

Mail. URL http://www.theglobeandmail.com/real-estate/the-market/the-new-way-toronto-condo-

developers-are-saving-space/article18134051/ (accessed 9.23.16).

Peters, A., Durner, W., 2015. SHYPFIT 2.0 User’s Manual. Berlin, Germany.

Petrell, R.., Gumulia, A., 2013. Saturated and unsaturated flow through sloped compost filter beds of

different particle sizes. Water Sci. Technol. 67, 2406–2411.

Philippi, P.M., 2005. Introduction to the German FLL-Guideline for the Planning, Execution and Upkeep

of Green-Roof Sites., in: Greening Rooftops for Sustainable Communities. Washington, DC, p. 7.

Podolsky, L., 2013. The Great Lakes Sewage Report Card [WWW Document]. URL

https://www.ecojustice.ca/wp-content/uploads/2014/08/FINAL-The-Great-Lakes-Sewage-Report-

Card-2013.pdf (accessed 8.1.16).

Poë, S., Stovin, V., Dunsiger, Z., 2011. The Impact of Green Roof Configuration on Hydrological

Performance, in: 12th International Conference on Urban Drainage. Porto Alegre, Brazil.

Prowell, E.S., 2006. An Analysis of Stormwater Retention and Detention of Modular Green Roof Blocks

(THES). The University of Georgia.

Qin, H., Peng, Y., Tang, Q., Yu, S.-L., 2016. A HYDRUS model for irrigation management of green

roofs with a water storage layer. Ecol. Eng. 95, 399–408.

143

Qin, X., Wu, X., Chiew, Y.-M., Li, Y., 2012. A Green Roof Test Bed for Stormwater Management and

Reduction of Urban Heat Island Effect in Singapore. Br. J. Environ. Clim. Chang. 2, 410–420.

Ravi, S., Dharmarajan, R., Moghaddam, S., 2015. Measurement of Capillary Radius and Contact Angle

within Porous Media. Langmuir 31, 12954–9. doi:10.1021/acs.langmuir.5b03113

Raviv, M., Lieth, J.H., MAHER, M., PRASAD, M., RAVIV, M., 2008a. Soilless Culture, Soilless

Culture. Elsevier. doi:10.1016/B978-044452975-6.50013-7

Raviv, M., Lieth, J.H., WALLACH, R., 2008b. Soilless Culture, Soilless Culture. Elsevier.

doi:10.1016/B978-044452975-6.50005-8

Raviv, M., Wallach, R., Silber, A., Bar-Tal, A., 2002. Chapter 2: Substrates and their Analysis, in:

Passam, H., Savvas, D. (Eds.), Hydroponic Production of Vegetables and Ornamentals. Embryo

Publications, pp. 25–77.

Roehr, D., 2010. Runoff Reduction Effects of Green Roofs in Vancouver, BC, Kelowna, BC, and

Shanghai, P.R. China 35, 53–68.

Roehr, D., Kong, Y., 2010. Stormwater Runoff Reduction Achieved by Green Roofs: Comparing SWMM

Method to TR-55 Method, in: Low Impact Development 2010. American Society of Civil

Engineers, pp. 1012–1021. doi:doi:10.1061/41099(367)88

Roofmeadow, 2013. Type III Green roofs [WWW Document]. URL

http://www.roofmeadow.com/details-specs-services/details-specs/type-iii/ (accessed 5.8.16).

Rowe, D.B., Getter, K.L., Durhman, A.K., 2011. Effect of green roof media depth on Crassulacean plant

succession over seven years. Landsc. Urban Plan. 104, 310.

Rowe, D.B., Kolp, M.R., Greer, S.E., Getter, K.L., 2014. Comparison of irrigation efficiency and plant

health of overhead, drip, and sub-irrigation for extensive green roofs. Ecol. Eng. 64, 306–313.

doi:10.1016/j.ecoleng.2013.12.052

Rowe, D.B., Monterusso, M.A., Rugh, C.L., 2006. Assessment of Heat-expanded Slate and Fertility

Requirements in Green Roof Substrates. Hortic. Technol. 16, 471–477.

Rugh, C.L., 2013. Green Roof Workshop.

Schindler, U., Müller, L., Eulenstein, F., 2016. Measurement and evaluation of the hydraulic properties of

144

horticultural substrates. Arch. Agron. Soil Sci. 62, 806–818. doi:10.1080/03650340.2015.1083982

Schneider, C.A., Rasband, W.S., Eliceiri, K.W., 2012. NIH Image to ImageJ: 25 years of image analysis.

Nat. Methods 9, 671–675.

Schroll, E., Lambrinos, J., Righetti, T., Sandrock, D., 2011. The role of vegetation in regulating

stormwater runoff from green roofs in a winter rainfall climate. Ecol. Eng. 37, 595–600.

doi:10.1016/j.ecoleng.2010.12.020

Schumacher, K., 2015. “Soil” Science.

SENES Consultants Ltd., 2011. Toronto’s Future Weather and Climate Driver Study - Volume 1 (RPRT).

City of Toronto, Toronto.

Senjug, C., 2016. Irrigation.

Sezer, a., Göktepe, a. B., 2010. Effect of particle shape on density and permeability of sands. Proc. ICE -

Geotech. Eng. 163, 307–320. doi:10.1680/geng.2010.163.6.307

Shahabi, A.A., Das, B.M., Tarquin, A.J., 1984. An empirical relation for coefficient of permeability of

sand.

She, N., Liu, J., 2013. Using Preferential Flow to Model Green Roofs, in: World Environmental and

Water Resources Congress 2013. American Society of Civil Engineers, pp. 377–390.

doi:doi:10.1061/9780784412947.036

She, N., Pang, J., 2010. Physically Based Green Roof Model. J. Hydrol. Eng. 15, 458.

Shmulsky, R., Jones, P.D., 2011. Density and Specific Gravity, in: Forest Products and Wood Science An

Introduction. Wiley-Blackwell, pp. 175–195. doi:10.1002/9780470960035.ch8

Simmon, R., 2012. NASA Earth Observatory Image of City Lights of the United States 2012: Suomi NPP

VIIRS data.

Simmons, M.T., 2015. Climates and Microclimates: Challenges for Extensive Green Roof Design in Hot

Climates, in: Sutton, R.K., service)., S. (Online (Eds.), Green Roof Ecosystems. Imprint: Springer,

Cham, pp. 63–80.

Simmons, M.T., Gardiner, B., Windhager, S., Tinsley, J., 2008. Green roofs are not created equal: the

hydrologic and thermal performance of six different extensive green roofs and reflective and non-

145

reflective roofs in a sub-tropical climate. Urban Ecosyst. 11, 339–348.

Snodgrass, E., 2005. 100 extensive green roofs: lessons learned, in: Greening Rooftops for Sustainable

Communities. Green Roofs for Health Cities, Washington.

Spolek, G., 2008. Performance monitoring of three ecoroofs in Portland, Oregon. Urban Ecosyst. 11,

349–359.

Starry, O., 2013. The comparative Effects of Three Sedum Species on Green Roof Stormwater Retention

(THES). University of Maryland.

Steinfeld, C., Del Porto, D., 2008. Green Roof Alternative Substrate Pilot Study (RPRT). University of

Massachusetts.

Stovin, V., Poë, S., Berretta, C., 2013. A modelling study of long term green roof retention performance.

J. Environ. Manage. 131, 206.

Stovin, V., Vesuviano, G., De-Ville, S., 2015. Defining green roof detention performance. Urban Water J.

1–15. doi:10.1080/1573062X.2015.1049279

Stroud, C., 2016. Phosphorous.

Struck, S., Ross, K., Sagi, J., 2014. Low Impact Development Guidelines: Module 3 - Green Roofs.

Calgary.

Tabares-Valasco, P.C., Srebric, J., 2012. A heat transfer model for assessment of plant based roofing

systems in summer conditions. Build. Environ. 49, 310–323.

Taylor, P., 2011. The wetting of leaf surfaces. Curr. Opin. Colloid Interface Sci. 16, 326–334.

Teemusk, A., Mander, U., 2007. Rainwater runoff quantity and quality performance from a greenroof:

The effects of short-term events. Ecol. Eng. 30, 271–277. doi:10.1016/j.ecoleng.2007.01.009

Toland, D.C., 2010. Stormwater Runoff and Plant Survival on Mock Green Roofs at the Watershed

Research and Education. University of Arkansas.

Tolk, J.A., 2003. Soils, Permanent Wilting Points, in: Encyclopedia of Water Science. Marcel Dekker.

Toronto, C. of, 2016. Annual Wastewater Treatment Plant Reports [WWW Document]. URL

http://www1.toronto.ca/wps/portal/contentonly?vgnextoid=da8807ceb6f8e310VgnVCM10000071d

146

60f89RCRD&vgnextchannel=0db7f75e4f18f310VgnVCM10000071d60f89RCRD (accessed

5.3.16).

Toronto, C. of, 2010. Toronto Zoo Penguin Exhibit Green Roof Specifications. RFT #54 (2010 - 09).

Toronto, C. of the C. of, 2009. Green Roofs. Adopt. by Counc. City Toronto 2009-05-27 by By-law No.

583- 2009. .

Toronto Region Conservation Authority, 2016. Watershed Management [WWW Document]. URL

https://trca.ca/conservation/watershed-management/ (accessed 5.13.16).

Toronto Region Conservation Authority, 2014. Flood Plain Management [WWW Document]. URL

http://www.trca.on.ca/the-living-city/water-flood-management/flood-plain-management.dot

Uhl, M., Schiedt, L., 2008. Green Roof Storm Water Retention - Monitoring Results, in: 11th

International Conference on Urban Drainage. Edinburgh.

United Nations, Department of Economic and Social Affairs, P.D., 2014. World Urbanization Prospects:

The 2014 Revision.

United States Environmental Protection Agency, 2014. Green Infrastructure [WWW Document]. URL

http://water.epa.gov/polwaste/green/#fact

Univeristy of Toronto Libraries, 2016. Search Journal Articles “green roof” [WWW Document]. URL

https://query.library.utoronto.ca/index.php/search/q?kw=green+roof&facet[0]=addFacetValueFilters

(ContentType,Journal+Article)&ps=20&facet[1]=addRangeFilter(PublicationDate,2016:2019)

(accessed 8.8.16).

University of Toronto, J.H.D.F. of A.L. and D., 2013. gritlab [WWW Document]. URL

http://grit.daniels.utoronto.ca/

USDA Natural Resources Conservation Service, 2011. Soil Quality Indicators: soil pH.

van Genuchten, M.T., Leij, F.J., Yates, S.R.R., 1991. The RETC Code for Quantifying the Hydraulic

Functions of Unsaturated Soils (No. 6.xx), United States Environmental Reseach Laboratory. PC-

Progress. doi:10.1002/9781118616871

van Hoof, J., van Dijken, F., 2008. The historical turf farms of Iceland: Architecture, building technology

and the indoor environment. Build. Environ. 43, 1023–1030.

147

Van Seters, T., Rocha, L., Smith, D., MacMillan, G., 2009. Evaluation of Green Roofs for Runoff

Retention, Runoff Quality, and Leachability. Water Qual.Res.J.Can. 44, 33–47.

VanWoert, N.D., Rowe, D.B., Andresen, J.A., Rugh, C.L., al, et, 2005. Green Roof Stormwater

Retention: Effects of Roof Surface, Slope, and Media Depth. J. Environ. Qual. 34, 1036–1044.

Vergroesen, T., Man Joshi, U., van de Giesen C., N., van de Ven M., F.H., 2010. High resolution rainfall

– runoff measurement setup for green roof experiments in a tropical environment. Hydrol. Earth

Syst. Sci. Discuss. 7, 9367–9410.

Vesuviano, G., Sonnenwald, F., Stovin, V., 2014. A two-stage storage routing model for green roof runoff

detention. Water Sci. Technol. 69, 1191.

Vijayaraghavan, K., Raja, F.D., 2015. Pilot-scale evaluation of green roofs with Sargassum biomass as an

additive to improve runoff quality. Ecol. Eng. 75, 70–78.

doi:http://dx.doi.org/10.1016/j.ecoleng.2014.11.029

Vonk, J., n.d. Academic Research into Green Roofs.

Voyde, E., Fassman, E., Simcock, R., 2010. Hydrology of an extensive living roof under sub-tropical

climate conditions in Auckland, New Zealand. J. Hydrol. doi:10.1016/j.jhydrol.2010.09.013

Walpole, R.E. (Ed.), 2007. Probability & statistics for engineers & scientists. Prentice Hall, Upper Saddle

River, NJ.

Walton, N.R.G., 1989. Electrical Conductivity and Total Dissolved Solids—What is Their Precise

Relationship? Desalination 72, 275–292. doi:10.1016/0011-9164(89)80012-8

Wang, Y., Grove, S.M., Anderson, M.G., 2008. A physical–chemical model for the static water retention

characteristic of unsaturated porous media. Adv. Water Resour. 31, 701–713.

doi:10.1016/j.advwatres.2008.01.005

Wanphen, S., Nagano, K., 2009. Experimental study of the performance of porous materials to moderate

the roof surface temperature by its evaporative cooling effect. Build. Environ. 44, 338–351.

doi:10.1016/j.buildenv.2008.03.012

Waschbusch, R.J., Selbig, W.R., Bannerman, R.T., 1999. Sources of Phosphorus in Stormwater and

Street Dirt from Two Urban Residential Basins In Madison, Wisconsin, 1994–95. Middleton,

Wisconsin.

148

Westhoff Engineering Resources Inc., 2011. User Manual for Water Balance Spreadsheetv1.2. Calgary.

White, P.J., Hammond, J., 2008. The ecophysiology of plant-phosphorus interactions.

Woodward, D.E., Hawkins, R., Jiang, R., Hjelmfelt A., J., Van Mullem, J., Quan, Q.D., Hydrologist, N.,

Maryland, D., Quan, Q.D., Engineer, H., Nrcs, U., Dc, W., 2003. Runoff Curve Number Method:

Examination of the Initial Abstraction Ratio, in: World Water & Environmental Resources Congress

2003. pp. 1–10. doi:doi:10.1061/40685(2003)308

Yio, M.H.N., Stovin, V., Werdin, J., Vesuviano, G., 2013. Experimental analysis of green roof substrate

detention characteristics. Water Sci. Technol. 68, 1477.

Yocca, D., Sale, L. V, 2012. Green Roof and Wall Performance Standards. Living Archit. Monit. 16–17.

Young, C.B., McEnroe, B.M., Rome, A.C., 2009. Empirical Determination of Rational Method Runoff

Coefficients. J. Hydrol. Eng. 14, 1283–1289. doi:10.1061/(ASCE)HE.1943-5584.0000114

Young, T., Cameron, D.D., Sorrill, J., Edwards, T., Phoenix, G.K., 2014. Importance of different

components of green roof substrate on plant growth and physiological performance. Urban For.

Urban Green. 13, 507–516. doi:10.1016/j.ufug.2014.04.007

Yuristy, G., 2013. Considering a Green Roof Substrate for Northern Climates (THES). The University of

Guelph.

Zhang, S., Guo, Y., 2012. An Analytical Probabilistic Model for Evaluating the Hydrologic Performance

of Green Roofs. J. Hydrol. Eng. 399.

Zihms, S., Switzer, C., Irvine, J., Karstunen, M., 2013. Effects of high temperature processes on physical

properties of silica sand. Eng. Geol. 164.

Zinco, 2011. Semi-intensive Green Roofs [WWW Document]. URL http://www.zinco.ca/semi-

intensive.html (accessed 8.8.16).

Zmora-Nahum, S., Hadar, Y., Chen, Y., 2007. Physico-chemical properties of commercial composts

varying in their source materials and country of origin. Soil Biol. Biochem. 39, 1263–1276.

doi:10.1016/j.soilbio.2006.12.017

Zoo, T., n.d. Green Roofs [WWW Document]. URL

http://www.torontozoo.com/conservation/greenroof.asp

150

Appendix A: Glossary

Dielectric permittivity – ɛ, is the resistance a material presents to the application of an electric field. In the

context of this research, the measurement frequency is 70 MHz (Decagon Devices, 2016)

Green Roof - The term “green roof” is used to describe a wide range of engineered, building-integrated,

vegetated systems. A number of terms that are broadly agreed upon for the types of building based

vegetation systems include:

Eco-roofs – These systems receive no-irrigation and little maintenance despite plant death or

dormancy during drought periods. Media must resist shrinkage and hydrophobicity to retain the

maximum amount of water during subsequent rainstorm events (Spolek, 2008). This terminology

and policy is particularly found in Washington State, USA.

Vegetated roofs – Optimized for plant health, aesthetic enjoyment and biodiversity. This term

may also be broadly used to describe any type of greenery on a horizontal building surface.

Roof Gardens – Any combination of container planting, and/or layered construction at any depth;

may also include vegetable gardening. Roof gardens are usually defined by the amenity benefit,

and so the availability of access onto the space.

Intensive green roofs – Use a variety of herbaceous and woody plants that require deeper planting

substrate (>15 cm). Through the use of larger plants, the landscapes created are often comparable

to ground level schemes. As there is no commonly agreed upper limit, at the most extreme, this

could include city parks or gardens created over parking garages or subway tunnels.

Extensive green roofs (as illustrated overleaf) – The shallowest type of green roof, typically

below 15 cm deep, often planted with drought tolerant species and may be irrigated, despite being

the type specified most frequently for stormwater management (Czemiel Berndtsson, 2010; She

and Pang, 2010).

151

Composition of typical ‘built up’ extensive green roof system.

From the roof deck, extensive green roof systems typically comprise: a root penetration barrier,

a drainage/water retention layer, a layer of planting substrate, and the vegetation. Many

installations also include an irrigation system, either upon the surface of the planting substrate

or below. The drainage/retention layer may be formed from a molded plastic board, an

absorbent material such as industrial felt, or, where loading permits, a layer of coarse aggregate.

GRITlab – The Green Roof Innovation Testing laboratory, henceforth refereed to as GRITlab, is located

on the fifth floor roof of The John H. Daniels Faculty of Architecture, Landscape, and Design, at 230

College Street, Toronto, Ontario, Canada. (University of Toronto, 2013).

Meadow – In this thesis, this refers to a native wildflower, forb and grasses seed mix used to provide

vegetation cover as one level of the ‘planting’ variable on GRITlab modules. For details of the

species included please refer to (MacIvor et al., 2013).

Medium –Syn. Substrate, planting medium or porous medium. Is a soilless matrix of particles of

biological or mineral original, so that water can infiltrate, percolate and drain. Typically employed in

the construction of extensive green roofs to support the growth of the vegetation and retain water.

Permeability – (kf) the rate at which water passes through the media under field conditions, which may be

near to, but are not explicitly saturated conditions.

Planting – syn. Vegetation. The uppermost living layer in a healthy extensive green roof.

Potential – Here, the total water potential (ψt, kPa) is a sum of the matric potential (ψm, kPa), the

gravitational potential (ψz, kPa) and the osmotic potential (ψo, kPa):

Vegetation

Planting medium

Geotextile

Drainage board

Roof membrane

152

𝜓𝑡 = 𝜓𝑚+𝜓𝑧+𝜓𝑜

For water storage and discharge related matters, only the first two terms make a practical

contribution. As green roofs are living systems typifying the SPAC (soil-plant-air continuum), the

plant availability of water is also under consideration, hence inclusion of the osmotic potential. In

subterranean engineering applications where the osmotic potential is omitted, this expression is

commonly made in length units of cm or m, where hydraulic head (H, m) is the sum of the pressure

head (h, m) and the elevation head (z, m):

𝐻 = ℎ + 𝑧

The conversion of units performed as a function of the density of water (ρw) and acceleration

due to gravity (g):

ℎ =𝜓𝑚

𝜌𝑤𝑔

In the unsaturated environment of a freely draining green roof, the hydraulic head equals 0 and

so the pressure head (and matric potential) will be < 0.

Osmotic potential is inversely related to the medium volumetric water content (θ), owing to the

increased in concentration of solute within the system during evaporative drying. The relationship

uses the gas constant, R, and the temperature of the system, T, to connect osmotic potential, ψo, to

the molality of the medium solution, b:

𝜓𝑜 = −𝑏𝑅𝑇

Regression tree – A data mining technique producing a global model of partitioned data, through

recursive partitioning. In this work, regression trees have been generated in Orange Data Mining (for

reference, the map is shown below). Pre-pruning set to n = 3 and post-pruning m = 5. Levels set to 3,

validation performed on training data set.

153

Saturated hydraulic conductivity – (Ksat) the rate at which water passes through media under saturated

conditions.

Storage – The amount of water a specified depth of porous medium can retain under static equilibrium,

expressed in mm.

Water Holding Capacity – The amount of water a volume of medium can retain under static equilibrium,

expressed in %.

154

Appendix B: Data relating to Chapter 2

Key to the modules studied at GRITlab

ID Medium Depth (cm) Vegetation Irrigation

1C Biological 10 Meadow None

2C Biological 15 Meadow Sensor

2E Mineral 15 Sedum None

3W Biological 15 Sedum None

3C Mineral 15 Meadow Daily

3E Mineral 10 Meadow Sensor

4W Mineral 15 Sedum Sensor

4C Biological 15 Meadow None

4E Biological 15 Meadow Daily

5W Biological 10 Meadow Daily

5C Mineral 10 Sedum Sensor

6C Mineral 10 Meadow Daily

6E Biological 10 Sedum None

7W Mineral 10 Sedum None

7C Biological 10 Meadow Sensor

7E Biological 15 Sedum Daily

8W Biological 10 Sedum Sensor

8C Mineral 15 Sedum Daily

8E Mineral 15 Meadow Sensor

9W Mineral 15 Meadow None

9E Mineral 10 Sedum Daily

10C Biological 10 Sedum Daily

11C Biological 15 Sedum Sensor

11E Mineral 10 Meadow None

155

Fitted curves imported directly from NCSS 11: x-axis = precipitation depth (mm), y-axis = total discharge volume (mm). As all modules experienced the same rainfall events, the x-axes are identical, variation exists in the scales of the y-axes.

Model: 𝒚 =(𝒙−(𝝀𝑺))𝟐

(𝒙+((𝟏−𝝀)𝑺))

Upper and lower 90 % confidence intervals shown in shaded region (Bootstrapped: N = 3000, random seed)

1C

2C

2E

3W

156

3C

3E

4W

4C

4E

5W

157

5C

6C

6E

7W

7C

7E

158

8W

8C

8E

9W

9E

10C

159

11C

11E

160

Parameters resulting from the curve number fitting

ID

Fitted to both λ and S Fitted to S with λ = 0.2

R2 λ S (mm) CN R2 S (mm) CN

1C 0.63 0 48 84 0.58 24 91

2C 0.68 0.13 19 93 0.67 17 94

2E 0.68 0.12 28 90 0.66 22 92

3W 0.72 0 28 90 0.71 17 94

3C 0.78 0 17 94 0.76 11 96

3E 0.73 0 29 90 0.71 17 94

4W 0.74 0 35 88 0.71 19 93

4C 0.72 0 44 85 0.67 23 92

4E 0.69 0 16 94 0.66 10 96

5W 0.75 0 18 94 0.73 11 96

5C 0.76 0 29 90 0.72 17 94

6C 0.68 0 14 95 0.65 9 97

6E 0.65 0 31 89 0.61 21 93

7W 0.37 0 35 88 0.28 19 93

7C 0.44 0 56 82 0.3 25 91

7E 0.51 0 39 87 0.49 25 91

8W 0.65 0 34 88 0.63 21 92

8C 0.72 0 14 95 0.7 9 97

8E 0.66 0 32 89 0.61 18 93

9W 0.48 0 20 93 0.44 12 95

9E 0.81 0 21 92 0.79 13 95

10C 0.66 0 16 94 0.63 10 96

11C 0.60 0 26 91 0.57 16 94

11E 0.77 0 21 92 0.74 13 95

161

Linear regressions of peak flow (Qp) and peak rainfall intensity (ipeak)

1C

2C

2E

3W

y = 0.45x

R² = 0.48

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.30x

R² = 0.32

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.29x

R² = 0.49

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.25x

R² = 0.47

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

162

3C

3E

4W

4C

4E

5W

y = 0.45x

R² = 0.44

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.41x

R² = 0.41

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.30x

R² = 0.50

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.26x

R² = 0.31

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.33x

R² = 0.38

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.41x

R² = 0.40

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

163

5C

6C

6E

7W

7C

7E

y = 0.37x

R² = 0.48

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.37x

R² = 0.41

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.27x

R² = 0.40

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.29x

R² = 0.44

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.29x

R² = 0.45

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.22x

R² = 0.28

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

164

8W

8C

8E

9W

9E

10C

y = 0.24x

R² = 0.38

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.35x

R² = 0.39

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.34x

R² = 0.51

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.26x

R² = 0.39

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.42x

R² = 0.44

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.36x

R² = 0.40

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

165

11C

11E

y = 0.25x

R² = 0.30

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

y = 0.34x

R² = 0.36

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.5 1.0 1.5 2.0

Qp

(L/m

in)

ipeak (mm/min)

166

Appendix C: Data relating to Chapter 3

Volumetric coefficients: data from the twenty-four months beginning May 2013, and total water retained (precipitation and irrigaiton) during period Oct 13 - Sept 14

ID Medium Depth Vegetation Irrigation

Summer Cvol

(May – Oct)

Winter Cvol

(Nov - April)

Annual

Cvol

Total water

retained Oct 13

- Sept 14 (mm)

1C Biological 10 Meadow None 0.25 0.51 0.36 350

2C Biological 15 Meadow Sensor 0.33 0.60 0.44 433

2E Mineral 15 Sedum None 0.24 0.38 0.30 339

3W Biological 15 Sedum None 0.33 0.63 0.45 388

3C Mineral 15 Meadow Daily 0.48 0.63 0.54 551

3E Mineral 10 Meadow Sensor 0.35 0.58 0.44 391

4W Mineral 15 Sedum Sensor 0.29 0.71 0.46 387

4C Biological 15 Meadow None 0.24 0.52 0.36 351

4E Biological 15 Meadow Daily 0.53 0.56 0.54 604

5W Biological 10 Meadow Daily 0.46 0.70 0.56 496

5C Mineral 10 Sedum Sensor 0.34 0.59 0.44 394

6C Mineral 10 Meadow Daily 0.56 0.59 0.57 467

6E Biological 10 Sedum None 0.30 0.37 0.32 364

7W Mineral 10 Sedum None 0.41 0.62 0.49 335

7C Biological 10 Meadow Sensor 0.29 0.52 0.38 509

7E Biological 15 Sedum Daily 0.32* 0.76* 0.48* -

8W Biological 10 Sedum Sensor 0.27 0.76 0.47 339

8C Mineral 15 Sedum Daily 0.53 0.74 0.61 465

8E Mineral 15 Meadow Sensor 0.33 0.68 0.47 368

9W Mineral 15 Meadow None 0.53 0.55 0.54 369

9E Mineral 10 Sedum Daily 0.40 0.61 0.49 477

10C Biological 10 Sedum Daily 0.51 0.72 0.59 490

11C Biological 15 Sedum Sensor 0.40 0.61 0.48 421

11E Mineral 10 Meadow None 0.43 0.56 0.48 312

*data from the twelve months beginning May 2013

167

F-values for snow depth in relation to four design variables in winter 2013-2014. Significance in the F-value is indicated with bold text and cell shading

ANOVA F values p (< 0.05)

Date Medium

(d.f. 1, 18)

Depth

(d.f. 1, 18)

Irrigation

(d.f. 2, 18)

Planting

(d.f. 1, 18)

2013-12-16 2.07 0.67 0.81 6.02

2013-12-17 2.1 0.65 0.56 4.93

2013-12-19 3.21 1.15 1.18 6.74

2013-12-20 4.85 0.28 1.11 2.54

2014-01-03 15.82 2.16 1.61 0.67

2014-01-06 6.61 0.86 3.71 1.79

2014-01-07 6.61 0.86 3.71 1.79

2014-01-09 6.61 0.86 3.71 1.79

2014-01-10 8.89 0.56 3.13 0

2014-01-16 4.79 1.72 3.64 4.79

2014-01-17 4.79 1.72 3.64 4.79

2014-01-20 7.68 1.92 6.84 0.48

2014-01-21 7.68 1.92 6.84 0.48

2014-01-27 19.9 1.42 9.44 0.09

2014-01-28 11.08 2.29 8.27 0.37

2014-01-30 17.22 0.48 7.21 1.08

2014-01-31 15.5 3.2 5.86 1.15

2014-02-02 8.25 3.16 3.54 0.17

2014-02-04 9.17 1.68 4.51 0.02

2014-02-06 5.21 0.96 1.44 1.3

2014-02-08 15.08 0.1 1.42 2.48

2014-02-10 10.61 0.13 2.12 1.18

2014-02-11 10.42 0.01 2.48 0.69

2014-02-13 5.72 1.63 1.91 1.63

2014-02-18 4.6 2.71 1.59 0.12

2014-03-03 32.27 1.07 10.47 0.27

2014-03-06 42.55 9.06 9.88 1.01

2014-03-14 19.45 0.12 3.79 0.28

168

F-values for snow depth in relation to four design variables in winter 2014-2015. Significance in the F-value is indicated with bold text and cell shading

ANOVA F values p (< 0.05) Date Medium

(d.f. 1, 18) Depth

(d.f. 1, 18) Irrigation (d.f. 2, 18)

Planting (d.f. 1, 18)

2014-12-11 9.37 0.67 0.56 0.14

2014-12-12 9.76 0.31 0.1 0.06

2015-01-03 10.79 1.6 3.64 0.06

2015-01-29 11.31 5.77 1.62 0.23

2015-01-30 9.08 4.88 1.25 0.36

2015-02-02 5.14 1.21 0.24 0.21

2015-02-04 8.67 0.6 0.67 0

2015-02-05 9.34 1.04 1.3 0.03

2015-02-08 5.03 1.43 0.2 0.36

2015-02-09 4.37 1.24 0.15 0.7

2015-02-11 7.37 2.07 0.06 0.92

2015-02-12 6.48 3.57 0.29 0.33

2015-02-23 0.97 0.54 0.04 0.33

2015-02-27 0.87 0.01 0.03 0.12

2015-03-02 1.5 0 0.01 0.17

2015-03-04 0.51 0.21 0.01 0.04

2015-03-07 0.85 0.85 0.03 0.19

169

Appendix D: Data relating to Chapter 4

Water quality parameters tested from the discharge waters off twenty-four green roof modules. Also water extractable total phosphorous (WETP) measurements from media samples taken in 2014.

Medium Depth (cm) Planting Irrigation 9 July 2013

WETP (mg/kg) pH σw (dS/m) Turb. (NTU) Abs 400

Mineral 10 Meadow Daily 8.0 0.18 0.8 0.198 18

Mineral 10 Meadow None 7.9 0.31 3.0 0.227 82

Mineral 10 Meadow Sensor 7.9 0.27 2.1 0.245 14

Mineral 10 Sedum Daily 8.3 0.25 1.7 0.211 12

Mineral 10 Sedum None 8.2 0.28 1.7 0.308 25

Mineral 10 Sedum Sensor 8.2 0.19 1.2 0.264 84

Mineral 15 Meadow Daily 8.3 0.30 1.2 0.272 54

Mineral 15 Meadow None 8.3 0.30 0.8 0.365 35

Mineral 15 Meadow Sensor 8.2 0.29 1.0 0.371 14

Mineral 15 Sedum Daily 8.2 0.25 0.4 0.207 95

Mineral 15 Sedum None 8.1 0.21 0.0 0.382 92

Mineral 15 Sedum Sensor 8.4 0.25 1.3 0.365 32

Biological 10 Meadow Daily 7.9 0.40 0.9 0.303 113

Biological 10 Meadow None 7.5 0.36 1.1 0.364 83

Biological 10 Meadow Sensor 8.0 0.33 1.2 0.508 84

Biological 10 Sedum Daily 7.9 0.43 0.5 0.231 121

Biological 10 Sedum None 8.0 0.32 1.6 0.442 117

Biological 10 Sedum Sensor 7.9 0.33 1.4 0.418 33

Biological 15 Meadow Daily 8.0 0.33 1.3 0.536 94

Biological 15 Meadow None 8.0 0.42 2.2 0.682 76

Biological 15 Meadow Sensor 8.0 0.42 0.5 0.760 84

Biological 15 Sedum Daily 7.9 0.40 1.3 0.562 67

Biological 15 Sedum None 7.9 0.33 1.8 0.855 116

Biological 15 Sedum Sensor 7.9 0.23 1.3 0.343 96

170

Regression to determine ε0 in eleven green modules containing mineral based planting medium.

Regression to determine ε0 in eleven green modules containin bark compost based planting medium.

5

6

7

8

9

10

11

12

13

14

15

0 0.01 0.02 0.03 0.04 0.05 0.06

Bu

lk d

iele

ctri

c p

erm

itti

vit

y

Bulk electrical conducitivity (dS/m)

0

5

10

15

20

25

30

0 0.05 0.1 0.15 0.2

Bulk

die

lect

ric

per

mit

tivit

y

Bulk electrical conductivity, (dS/m)

171

Appendix E: Data relating to Chapter 6

Onsite observations and background information

Roof Age

(years)

Design depth

(cm)

Surveyed depth

(cm)

Depth

ratio n σ

A 17 -- 2 8.3 1.8 -

B 16 12 10 9 2 0.75

C 15 10 14 10.9 16 1.09

D 15 7.5 13 7.5 1 1.00

E 13 14 21 13 2.1 0.93

F 10 10 22 8 1.8 0.80

G 10 15 16 14 2.4 0.93

H 9 15 39 15 2.2 1.00

I 8 15 20 15 1.5 1.00

J 8 15 22 14 1.8 0.93

K 8 10 6 10 0.6 1.00

L 8 12.5 14 13 1.8 1.04

M 5 15 20 11 3.5 0.73

N 5 6.25 43 7 1.4 1.12

O 5 15 13 11 3.9 0.73

P 4 12.5 3 11.7 2.4 0.94

Q 3 10 1 13.5 - 1.35

R 3 12.5 32 13 1.4 1.04

S 4 10.5 5 10 1.3 0.95

T 3 12.5 32 12 2.3 0.96

U 3 Various 36 13 3.8 -

V 3 Various 33 25 10.6 -

W 3 8.75 10 8 1.1 0.91

X 2 15 31 13 2.4 0.87

Y 3 - 4 7 1.1 -

Z 2 17.5 12 14 3.2 0.80

a 2 15 5 15 1 1.00

b 2 - 6 9 0.9 -

c 1 20 33 19 1.5 0.95

d 1 15 29 11 1.9 0.73

e 0 10.5 5 9 1.1 0.86

f 0 10.5 3 9 0.5 0.86

172

Bulk physical properties of green roof media

Roof n

Maximum media density

(g/cm3)

Dry bulk density

(g/cm3) MWC (%)

Particle density (g/cm3) Porosity

Permeability (mm/s)

σ σ σ

A 2 1.1 0.07 0.4 0.19 68 0.9 0.8 - -

B 4 1.0 0.09 0.5 0.05 52 2.0 0.7 0.05 0.01

C 3 0.9 0.08 0.5 0.01 40 1.2 0.6 0.17 0.02

D 3 1.3 0.09 0.7 0.08 54 2.2 0.7 0.03 0.00

E 6 1.2 0.03 0.6 0.02 57 2.1 0.7 0.03 0.01

F 4 1.2 0.03 0.7 0.03 48 2.1 0.7 0.02 0.01

G 4 1.2 0.06 0.8 0.07 36 1.7 0.5 0.11 0.05

H 6 0.9 0.09 0.2 0.03 69 0.6 0.5 0.03 0.03

I 2 1.1 0.31 0.6 0.38 54 2.8 0.7 0.08 0.04

J 4 1.1 0.28 0.5 0.26 63 2.4 0.8 0.04 0.02

K 6 1.0 0.03 0.3 0.03 69 1.1 0.7 0.05 0.02

L 6 1.0 0.04 0.3 0.03 70 3.6 0.9 0.01 0.01

M 4 1.3 0.08 0.8 0.06 53 2.4 0.7 0.03 0.01

N 6 1.0 0.07 0.3 0.05 70 1.3 0.8 0.01 0.01

O 6 1.2 0.04 0.8 0.03 43 2.1 0.6 0.03 0.01

P 2 1.3 0.04 0.8 0.01 51 1.6 0.5 0.06 0.01

Q 1 1.7 - 1.3 - 43 2.4 0.5 0.03 -

R 6 1.0 0.10 0.6 0.11 44 2.2 0.8 0.07 0.03

S 6 1.3 0.09 0.8 0.11 48 1.9 0.6 0.04 0.01

T 5 1.2 0.09 0.7 0.05 48 0.9 0.3 0.07 0.03

U 5 1.3 0.07 0.8 0.07 50 2.2 0.6 0.02 0.02

V 6 1.1 0.05 0.3 0.03 73 1.6 0.8 0.00 0.00

W 2 1.0 0.02 0.3 0.02 77 1.4 0.8 0.04 0.01

X 6 1.2 0.07 1.0 0.05 24 2.2 0.6 0.04 0.02

Y 4 1.1 0.09 0.8 0.08 27 2.0 0.6 0.05 0.03

Z 3 1.3 0.14 0.8 0.15 44 2.2 0.6 0.01 0.01

a 2 1.0 0.02 0.2 0.01 81 2.2 0.9 0.00 0.00

b 2 1.2 0.07 0.9 0.06 25 2.1 0.6 0.13 0.01

c 6 1.1 0.04 0.5 0.01 64 2.0 0.8 0.00 0.04

d 6 1.0 0.08 0.3 0.07 68 1.2 0.6 0.03 0.02

e 2 1.1 0.00 0.7 0.01 40 2.0 0.6 0.11 0.00

f 3 1.3 0.07 0.8 0.04 47 1.9 0.6 0.07 0.01

173

Particle size parameters and organic matter content of green roof media

Roof n CU CC d10 (mm) OM

A 1 6 0.8 0.20 69

B 2 11 0.4 0.31 10

C 3 16 0.5 0.16 1

D 3 17 5.0 0.33 4

E 6 14 0.5 0.15 11

F 1 15 0.7 0.19 7

G 4 16 4.9 0.36 6

H 6 10 1.0 0.11 46

I 2 21 1.6 0.18 6

J 4 12 1.0 0.18 30

K 3 6 0.7 0.13 59

L 3 6 0.9 0.09 51

M 4 14 1.8 0.19 3

N 6 10 0.9 0.13 48

O 3 19 1.5 0.09 3

P 2 10 0.7 0.37 5

Q 1 12 0.8 0.25 2

R 5 10 1.5 0.15 7

S 2 10 1.7 0.15 7

T 3 15 1.7 0.45 5

U 3 15 0.3 0.27 6

V 4 6 1.1 0.14 53

W 1 8 0.8 0.16 65

X 3 17 3.4 0.28 2

Y 4 13 3.2 0.49 3

Z 3 12 0.7 0.13 5

a 2 11 1.0 0.13 49

b 2 6 2.5 0.98 2

c 6 10 1.0 0.19 35

d 5 8 0.8 0.09 43

e 2 12 2.1 0.3 4

f 3 13 1.6 0.23 5

174

Appendix F: Success and Succession

Abstract

A city builds a highway and knows what to expect in terms of maintenance and performance 10 and 20

years from now. It is starting to be recognized that green infrastructure behaves differently, as growing

and evolving systems: some of which are quite well characterized, such as urban forests. Green roofs are

younger, more dynamic and often highly designed. All of which leaves more questions about their

maturation and development largely unanswered in many regions. A retrospective, comprising several of

the oldest roofs in North America will be presented, to provide a rich framework within which to consider

ecological succession as an ultimate end goal of extensive green roofs. Fundamental issues concerning the

post-construction phases of green roof management include the economic and environmental expenses of

weeding, fertilization and irrigation of extensive green roofs:

What happens when the manicured aesthetic ideal is rejected either intentionally or through

neglect?

What it the long term viability of a green roof as a piece of ecological urban infrastructure versus

its survival as a horticultural canvas?

What can we learn from revisiting old projects and how can we design with the evolution of

individual sites in mind?

These six case studies are selected on the basis of age (average age ten years), as well as the diversity in

terms of design characteristics, subsequent maintenance and stakeholder perspectives. They include a

highly sloped Sedum roof, a roof with three different compositions of planting medium and a few roofs

which have received almost no maintenance or attention since installation.

175

Introduction

In the City of Toronto, where green roofs are mandated through municipal law, the measure of success is

defined:

“The plant selection and design shall be such that within three years of the planting date the selected

plants shall cover no less than 80% of the vegetated roof.” (Toronto, 2009)

Beyond these early years many green roofs undergo changes due to either deliberate or neglectful

behavior by their owners and managers (Carlisle and Piana, 2013). In terrestrial environments complete

abandonment would ultimately result in a climax community (Snodgrass, 2005), but even minimally

maintained roofs have trees removed once they reach a noticeable 1.8 m. The transition towards a

naturalized ecosystem is further confounded by changes in the ownership and management of green roofs

and the corresponding changes in the priorities in the use and function. Again this marks green roofs apart

from most land based vegetation and plantings.

In addition to the careful selection of the initial planting material, the subsurface components play an

important role in the continued development of a system. The percentage of organic matter in the planting

substrate has a significant impact in both vegetation health and water holding capacity (Buist and

Friedrich, 2008; Nagase and Dunnett, 2011; Rowe et al., 2006; Yio et al., 2013). It has been suggested

that organic matter ≥ 25% could cause long term failure of the vegetation through overgrowth and

reduced drought resiliency (Nagase and Dunnett, 2011). It has also been previously reported that

‘traditional’, low organic matter substrate can accumulate organic matter over a number of years,

resulting in physic-chemical changes (Getter et al., 2007).

Methods

Three complementary studies have been undertaken on a total of eight green roofs on six sites. The green

roofs are detailed in Table 0-1 below.

176

Table 0-1 Details of study green roofs in Toronto

Ref

Co

nstru

ction

da

te

Ro

of g

eom

etry

Sh

ad

ing

structu

res

Org

an

izatio

n

typ

e

Su

rrou

nd

ing

s

Earth Rangers: South roof ERS 2003 Flat North Non-profit Natural

George Vari Engineering,

Ryerson University GVE 2004 Flat Many University

Dense

Urban

Jackman Avenue School JAS 2005 Flat

North/East/

South Junior School Urban

Toronto Botanical Garden:

Flat roof TBGF 2005 Flat East

Non-profit Park Toronto Botanical Garden:

Sloped roof TBGS 2005

Slopes to

North None

Arts and Administration,

University of Toronto UTSC 2005 Flat North/West University Park

Royal Ontario Museum ROM 2008 Flat North Non-profit

Dense

Urban

Earth Rangers: East roof ERE 2010 Flat West Non-profit Natural

The roofs were selected for study owing to their age, although it is notable that all of the roofs are owned

and managed by organizations with an education mandate. Only the George Vari Engineering Building

roof is accessible to the community, albeit under restricted conditions. The community gardening group,

Rye's HomeGrown CG, have a group of volunteers who regularly farm on the green roof.

Study 1 – Succession in Management

Individuals who own or maintain the case study roofs, were asked to complete a short survey. Two

stakeholders per case study were invited to take part. Question 1 established their relationship to the green

roof. The second question asked them to rank a number of services or functions which their green roof

provides in 2014. The functions included were: Aesthetics, Biodiversity, Community, Identity, Education

or research, and Infrastructure. Finally, the simple question ‘Is your roof successful?’ was asked with

three categories of possible response.

Study 2 – Succession of Plant Communities

Assessments of the vegetation diversity were made during visits to the study roofs in 2013, 2014 or in

some cases both summer seasons. The number of species were recorded on site, or by later reference to

photographs of encountered specimens. In two cases the plant communities had been surveyed in 2008

(Hahn, 2009). A summary of the original planting lists is presented below in Table 0-2.

177

Table 0-2 Selected planting details study 2 Genus only identified where the species is unknown (in a proprietary seed mixture) or where multiple species have been used.

ERS ERE GVE JAS ROM TBGF TBGS UTSC

Achillea millefolium X

Allium schonaprasum X X X X

Aquilegia canadensis X

Aster sp. X

Calendula sp. X

Carex sp. X

Centaurea sp. X

Coreopsis grandiflora ‘Early

Sunrise’

X

Coreopsis lanceolata X X

Coreopsis tripteris X

Cosmos sp. X

Equisetum hymale X

Erysimum sp. X

Geranium maculatum X

Geranium psilostemon ‘Rozanne’ X

Hemerocallis ‘Catherine woodbury’ X

Leucanthemum × superbum X

Liatris spicata X X

Lobelia siphilitica X

Lolium multiflorum X

Lupinus polyphyllus X

Monarda fistulosa X

Monarda punctata X

Panicum virgatum X X

Papaver rhoeas X

Penstemon digitalis X X

Pycnanthemum tenuifolium X

Rudbeckia hirta X X

Rudbeckia nitida 'Herbstsonne' X

Rudbeckia laciniata 'Goldquelle' X

Saponaria officinalis X

Schizachyrium scoparium X

Sedum sp. (#) 8 8 4

Sempervivum sp. X

Silene sp.

Symphyotrichum novae-angliae X X X

Trifolium repens X

Study 3 – Development of Substrate

The depth of the planting substrate was recorded on-site using a graduated metal probe. Samples of the

planting substrate were air-dried and stored at the laboratory prior to testing. Portions of the substrate

(~20 g) were subject to loss-on-ignition (550 °C, 120 minutes) to determine the % organic matter. pH

measurements were made using 1:2 (w/v) preparations in reagent grade water, stood for 60 minutes prior

178

to testing (Radiometer PHM92 with Sentix22 probe). Additional suspensions were prepared at 1:5 (w/v)

and tested for electrical conductivity (FieldScout EC110).


Study 1 – Succession in Management

Question 1 established that three of eight respondents have been with their organization since their roof

was constructed and three have joined their organizations since. Four respondents identified themselves as

responsible for the upkeep and maintenance of their roof.

Question 2 results regarding the importance of various factors, are presented below in Figure 0-1. In each

respondent’s polygon, more important functions project further from the centre. Based on mean rankings

the key priority was education and research, whilst the lowest average ranking was given to aesthetics.

Half of the respondents said that their roof was ‘successful’ and selected ‘somewhat’ to this question.

Frustrations were expressed about limited accessibility and use of roofs, particularly for educational

purposes. This directly reflects the nature and priorities of the organizations. Another comment was made

regarding the Toronto bylaw and questioned the effectiveness of green roofs from a sustainability

perspective. No concerns were expressed about the upkeep or evolution of their green roof design over

time.

179

Figure 0-1 Stakeholder rankings of the importance of green roof functions (n=7).

Study 2 – Succession of Plant Communities

Earth Rangers

This green roof was originally commissioned with a strong interest in the infrastructural benefits in terms

of storm water retention and moderating the temperature within the building envelope. Despite these

primary motivations, the roof is overlooked from the building’s upper storey and is one of the most highly

maintained roofs studied. Monthly weeding and occasional manual irrigation results in the visible

retention of the geometric design over ten years later (Figure 0-2). The Southern green roof (pictured) is

the most ‘traditional’ of the study installations, comprising Sedum species planted into a low organic

medium. Eight Sedum species were planted with Chives (Allium schoenoprasum) and White Clover

(Trifolium repens) providing additional height and texture. Originally this design spanned three roofs;

north, east, and south over the first floor wings of the buildings. After six years, this system had not

developed significantly and lacked the dynamism that the owners wanted to promote green technology

awareness and to integrate the roof into the natural surroundings. So the Eastern roof was replaced in

2010 to support a variety of native and ornamental perennials in a substrate with higher % compost. Due

to the maintenance, this roof has also remained largely unchanged since installation although some areas

of geraniums (Geranium maculatum) and all of the Tickseed (Coreopsis sp.) were lost in the winter

2013/2014. In the winter of 2013/2014 a weir was installed to measure the output flow from the entire

Aesthetics

Biodiversity

Community

Identity

Education or research

Infrastructure

180

green roof (South-East and a North wing). This is the final data stream for the building owners, Earth

Rangers, to track their entire water balance (both indoor and outdoor uses) on a continuous basis.

Figure 0-2 Earth Rangers Southern roof: 2005, after 2 years establishment (left), and 2013 (right).

George Vari Engineering Building, Ryerson University

In contrast, the green roof of the George Vari Engineering building at Ryerson University was originally

conceived as a monoculture of daylilies (Hemerocallis ‘Catherine woodbury’) and was equipped from the

outset with a suite of hydrological and climatological sensors for research purposes. This roof received

minimal maintenance for over eight years, with just woody species being removed to prevent membrane

damage. For many years there was no irrigation system installed. After just four years the number of

species had risen to thirty-nine, as many volunteer species colonized (Hahn, 2009). Some Hemerocallis

remain still in 2014, although, until 2013, this urban ecosystem continued to evolve into more of a prairie

meadow with Aster sp., Asclepias tuberosa, Linaria vulgaris and Vicia cracca prevalent amongst the

grasses. A dramatic change in the management of this roof occurred between 2013 and 2014, as a group

of community gardeners took control of the space and turned much of the area to vegetable production

(Figure 0-3). With this new function the space is accessed twice per week through the summer months

and now receives automatic irrigation and regular soil amendments in addition to weeding care.

181

Figure 0-3 George Vari Engineering Building roof, 2013 (left), and 2014 (right)

Jackman Avenue School

This 90 m2 roof was designed in partnership with the school community who were and still are concerned

primarily with the potential for evaporative cooling to affect air flow coming across the installation and in

through adjacent windows. A palette of eleven native species was selected according to the sheltered and

shaded position. The roof is not accessible under usual circumstances and it is explained to the young

students as being a ‘roof’ rather than a ‘garden’. Weeding maintenance is provided infrequently

(approximately annually) and irrigation is not provided. Since installation the total number of species has

remained similar (11 originally). The original design intent has been largely obscured due to the relative

success of the planting and some replacement with volunteer species.

Toronto Botanical Garden

This roof, of two parts, was intended to fit within the overall botanical garden theme of a curated garden

landscape. The largest part of the roof (60 m2) is sloped to the north on a curve, from 22 to 30°. This area

has battens to retain the planting substrate and was planted with four varieties of Sedum. The slope

restricts maintenance, as access requires the use of harnesses and steel lines to climb the loose surface

(Figure 0-4). Although this is the second Sedum roof examined, the slope, the composition of the original

planting substrate (high compost content), lower maintenance and higher vegetation coverage have led to

a very different outcome. A thick layer of haircap moss (~10 cm) has formed an understory, resulting in a

wetter environment at the substrate surface which supports fruiting fungi in the late summer. The local

seed bank is diverse and ornamental compared to other green roofs, leading to the establishment of

Penstemmon, Coneflower (Ratibida pinnata), Jerusalen Sage (Phlomis tuberosa 'Amazone') in addition to

the more common Golden rod (Solidago sp.) and the seemingly ubiquitous Black medic (Medicago

lupulina).

Figure 0-4 Toronto Botanical Garden sloped section, 2006 (left) and 2014 (right).

182

University of Toronto, Scarborough, Arts and Administration Building

Installed at the time of the building construction, this roof was originally seeded with a wildflower

mixture, with a nurse crop of Annual Rye (Lolium multiflorum). The clients were clear that they required

a minimal maintenance system and did not provide irrigation beyond the first season. The system is still

dominated by the rye grass with other abundant species including yarrow (Achillea millefolium) dog

strangling vine (Cynanchum rossicum) (Figure 0-5).

Figure 0-5 Arts and Administration green roof, University of Toronto, 2005 (left) and 2013 (right).

In addition to the noted herbaceous changes, all of the roofs had saplings of woody species observed or

reported as recently removed. The most common among which were the Chinese elm (Ulmus parvifolia)

and the Manitoba maple (Acer negundo).

Royal Ontario Museum

Although one of the youngest roofs studied, this green roof (519 m2) has a number of remarkable features.

It is located to the South of an overhanging reflective aluminium fascia, which results in a very hot dry

area directly alongside the only observation point, a restaurant window. The majority of the installation is

suspended 1.2 m above the existing roof with deeper recesses to support ornamental trees. The original

planting comprised Sedums – 8 main varieties with about 16 different cultivars, and Hens and Chicks

(Sempivivum sp.), chives and horsetail grass (Equisetum sp.). The spray irrigation system is programmed

to deliver twice a day during the summer due of the low storage volume of the planting substrate depth. In

the first two years there was an attempt to maintain the design intent with weeding and replacement of

dead Sedums. But with the spread of the chives and the horsetail grass into the highly designed Sedum

fields it was seen as an impossible maintenance endeavor and after the two-year warranty period all

routine attention ceased. Over time it has acquired a diverse look from the original highly stratified design

183

of white, yellow and pink Sedum areas. The design intent was fraught with unrealistic expectations from

the start – form beat function to the detriment of good horticultural practice or practicality. One area

which suffered particularly is beneath the overhang, the intense heat and light eradicated the Hens and

Chicks leaving an opening, which was occupied by Quick Weed (Galinsoga quadriradiata) in 2013. This

is an annual weed, which failed to germinate by June 2014, resulting in the exposure and erosion of the

planting substrate, and subsequent damage to the irrigation line, exacerbating the problems (Figure 0-6).

Figure 0-6 Royal Ontario Museum scorched section detail, 2013 (left) and 2014 (right)

184

Study 3 – Development of Substrate

Based on the depth measurements, none of the planting substrates have demonstrated significant bulk

compaction since installation (Figure 0-7).

Figure 0-7 Depth of planting substrate on eight green roofs

The planting substrate used in the Earth Rangers South roof contains ~7% organic matter, which is within

the range on the supplier’s current specification for a similar product (5 - 10%). This material was distinct

from the remaining study roofs which contained a mean of 53% organic matter (σ = 8%). The substrates

had neutral pH, with mean 𝑝𝐻1:2 𝐻2𝑂 = 6.8 (σ = 0.5). The electrical conductivity of the soil solution

indicates the total dissolved solids and is related to the osmotic potential. Electrical conductivity appeared

to reduce with age, although the small sample size resulted in Pearson’s r = 0.75, and r =0.68 with the

exclusion of the Earth Rangers South roof.

Conclusions

Stakeholders who inherit or commission green roofs may expect a zero-maintenance, but the emergence

of tree seedlings on all of the roofs forces the minimum of an annual inspection for even the most

disinterested owners. The sparsest vegetation was observed in the north section of the ROM roof and

across the southern roof at the Earth Rangers site. In both cases the roofs are exceptionally hot and dry

due to their southern aspect and unusual substrate configurations. When applied ≥10 cm deep, the

substrates containing around half organic matter were stable after ten years and successfully supporting a

range of native, ornamental and/or Sedum species.

0

5

10

15

20

Pla

nti

ng s

ub

stra

te d

epth

/cm

Measured Depth Design depth

185

In many of the case studies examined, the original aesthetics had been eroded through abandonment and

the design intent lost to nature. Benign neglect provides a window of opportunity for the development of

alternative, biodiverse urban ecosystems which retain all of their other original functions. As the current

custodians place a relatively low priority on the aesthetics of their green roofs, the mantra ‘form follows

function’ may be good advice for enduring green roof design. In most circumstances where the roof is not

accessed or even overlooked the designers and clients are advised to work towards a shared understanding

of how a roof with minimal maintenance will evolve to ensure the perception of success for the lifetime of

the installation.

186

Appendix G: Gone with the Wind

Abstract

Across all kinds of terrain, wind speeds increase exponentially, and in cities, the forests of tall buildings

can create ‘urban canyons’ and turbulence as well. The forces exerted by the wind can cause erosion of

the planting substrates used in extensive green roofs. In addition to the effect of increased wind speed at

height, the problem is exacerbated by the formulation and use of deliberately low density, soil-less

planting substrates. Such substrates typically comprise a mixture of composted organic material, natural

aggregate and manufactured lightweight expanded aggregate (LEA). Organic matter is known to improve

the texture of the substrate as well as improve the moisture retention properties of the bulk mixture. But

when organic rich mixtures dry out, they are lighter than the mineral aggregate particles. The retained

moisture is determined by the substrate properties, but also by the local precipitation patterns. Finally, and

perhaps most crucially, established vegetation on a green roof will effective prevent loss of substrate

through wind erosion.

Typical extensive green roof vegetation in North America is either a selection of low growing succulents

such as Sedum sp. or grass mixtures with native forbs. Sedum may be sometimes grown on thin layers of

substrate, directly onto geotextiles at a nursery. These ‘mats’ can be cut, rolled and delivered for

immediate effect and usually require little more than staking on high risk sites. However, alternative

methods of planting leave exposed substrate vulnerable to wind scour for at least the first growing season,

necessitating the use of an erosion control product.

As a niche, but rapidly expanding market, extensive green roof construction practices present a

fascinating learning opportunity for the erosion control specialist. In this presentation we will take a tour

of case studies across Canada, and examine the various combinations of erosion control measures and

planting techniques. Products used include: polymer meshes, organic blankets and meshes, 3D cellular

grids and chemical stabilizers. The use of stabilizers in non-soil media is a particularly recent

development; this presentation will showcase some preliminary research from the University of Toronto,

looking at the hydrology of stabilized engineered substrates.

187

Introduction

Green roof construction practices are rapidly evolving across Canada, with alternative plating substrates

being trialled and employed in many full scale installations. The window of time between construction

and complete vegetation coverage can leave new systems vulnerable to wind scour and erosion from

heavy rainfall, necessitating the use of erosion control measures. Some of these control measures are

reviewed, based on the observations of the author and the experiences of colleagues at the construction

company Flynn Canada. Based on the recommendation to use a soil amendment or tackifier in most

circumstances, a suite of laboratory investigations into the effects this type of product has on the

hydraulics of a green roof planting substrate were undertaken.

Vegetated roofs have a long history stemming from a number of European countries and traditions

(Köhler and Poll, 2010). Since the late 20th century they have become commoditized and to some degree,

standardized across the world (e.V., 2008). For over ten years they have grown in popularity in North

America, driven by legislation such as the City of Toronto bylaw, which mandates green roofs on larger

new developments (Toronto, 2009). In Canada, an awareness of urban ecology and concern for

maintaining biodiversity has fostered an interest in native plant species and their use in green roofs. As

these require more nutrition and available water than the Sedum species common in Europe, this has led

to increased adoption of lightweight compost based planting substrates in green roof construction (Buist

and Friedrich, 2008). Extensive green roofs are typically defined as those with only 15 cm planting

substrate depth (Czemiel Berndtsson, 2010). Beneath this would be a root penetration barrier, a layer

which determines a specified water reservoir and drainage, then a geotextile to prevent the planting

substrate from filling the reservoir voids. Many installations also include an irrigation system, either upon

the surface of the planting substrate e.g. spray or below e.g. capillary mats.

In 2014, the American-based ASTM International completed their complete Standard Guide for

Vegetative (Green) Roof Systems (ASTM E2777 -14, 2014). Amongst other design concerns, this

document highlighted the need to resist erosive forces in immature green roofs, whilst the vegetation is

still establishing. Owing to their relative exposure and elevation, extensive green roofs can be subject to

both wind scour and erosions from heavy rainfall.

A popular method of vegetating a green roof with low growing and resilient Sedum species, is to roll out

‘mats’ of pre grown plants onto the planting medium. The mats (typically around 2 m2) are cultivated at

ground level, in nurseries and typically comprise a geotextile of high water permeability, a nylon mesh for

support and a thin layer of growing medium onto which the Sedum are seeded (Rugh, 2013). This system

provides instant greenery and through an instant physical barrier, effectively mitigates all sources of

erosion. Over a number of years the roots of the vegetation penetrate the planting medium, increasing the

188

resiliency and cohesion of the system (see Figure 0-1). In some cases where a building study indicates

that high winds are anticipated, additional anchors can be added to the system using the weight of the

planting Substrate to retain the mats on top (see Figure 0-2).

Figure 0-1. Preparation of pre-grown Sedum ‘mats’ (left), root penetration after two years growth on a green roof (right).

Figure 0-2. Anchors to retain pre grown mats in high wind velocity situations.

In Ontario, the green roof market has been growing for a little over ten years. Since inception there have

been many proponents for adopting alternative planting strategies, focussing on native or ornamental

species rather than the Sedum species, originally recommended by the German pioneers in this field. The

pre-grown mat format is less hospitable to these vegetation types and has led to the development of some

alternative strategies. One such strategy is to establish the desired vegetation into the substrate (0.1 - 0.15

m) contained within a mesh sided crate or module. Owing to their bulk and weight, these have a much

smaller footprint (around 0.2 m2) and may be more expensive, but again provide instant greening and

associated erosion resistance (LiveRoof, 2014). When clients require a more economical or efficient

Anchor constructed

from washers, bolts

and threaded bar

Sedum mat

Planting substrate

Drainage

189

green roof construction method, the composite layers are assembled in situ, sometime referred to as a

‘built up’ installation. Often this requires that the planting substrate is secured in some manner prior to

full establishment of the vegetation and a number of methods have been tried. The advantages and

disadvantages of each erosion control measure depends largely on how the planting will be conducted and

maintained. A summary of some combinations is presented below in Table 0-1.

Table 0-1 Combinations of erosion control measures and planting methods. Red not recommended, yellow may present some difficulty, green represents recommended combinations.

Polymer (e.g. polypropylene) meshes are promoted by some green roof product suppliers and both

bonded and knitted are frequently specified by designers. It has the highest tensile strength of the products

reviewed which is of assistance in retaining aggregate materials on slopes. However, it is not well suited

for green roofs applications for a variety of reasons. It presents a tripping hazard by entangling very easily

with the grips on work boot soles; at grade this may be an annoyance, on a rooftop it can lead to a serious

fall. If rooted stock vegetation is being planted, time and energy is required to cut a hole for every plant;

this quickly adds up to extra man-days when there are hundreds or thousands of plants required. Finally,

when a seeding strategy is employed with broad leaf forbs, tenting of the mesh can easily occur (Figure

0-3). This is only remedied by removal of the mesh, which may be required whilst there is still exposed

planting substrate.

Polypropylene mesh

Biodegradable mesh

Biodegradable blanket

Soil amendment

Seeding broadleaf

Seeding grasses

Nursery plugs

Larger nursery stock

190

Figure 0-3 Tenting of polymer mesh over native wildflower seed mixture on an extensive green roof.

Biodegradable meshes (e.g. jute or coir) have not demonstrated the same level of stumbling hazard.

However, densely woven mesh can still obstruct planting of potted nursery stock and could potentially

tent over broad leaved species seeded below. Some meshes are more loosely woven and can easily be

pulled apart to permit the insertion of plugs, although smaller plugs are still most highly recommended

(72 per 0.27 m2 flat). Biodegradable blankets comprising two lightweight woven layers with an inner

layer of loose material (e.g. straw or coir) provide an additional benefit of significant weed control

decreasing competition with the desired vegetation and reducing maintenance requirements in the first

few years. The open weave and non-woven components are easily torn apart to permit rooted nursery

stock of most sizes to be easily inserted. Depending on construction scheduling the vegetation may be

installed many months after the other components; in these cases, the weathering of the blanket further

decreases the resistance to planting. Blankets also do less to impede the growth of seeded installations,

although on exposed sites, the use of pegs to retain the blanket over an otherwise bare seedbed is

recommended.

The use of soil amendments or tackifiers is a relatively recent development in preventing green roof

erosion. In Southern Ontario, a single product is currently being explicitly marketed for this purpose; it is

a powdered mixture comprising sugar and mucilage containing plant matter. The recommended

application rate of around 1 kg/m2 is made by introduction from a separate reservoir in the blower truck.

It has been success at stabilizing all types of seeded and planted roofs on several projects and does not

have any of the disadvantages of the overlaid erosion control measures. Although research has been

conducted into the use of mucilage containing tackifiers on land based applications with varying degrees

of loam soil, little has been conducted into the unusual rooftop scenarios with compost based substrates.

191

These systems typically have very high infiltration and percolation rates, zero runoff and water balance

very sensitive to changes in storage capacity and evapotranspiration rates.

Hydrological effects of soil amendments in a green roof planting substrate

Methods

The experiments were performed using a proprietary green roof planting substrate, comprising 95%

composted pine bark (Bioroof Systems, 2011). The erosion control ‘soil amendments’, Psyllium husk

powder and polyacrylamide gel fibres (PAM) were tested for their impact on the hydraulic properties of

the planting substrate. Both of the materials are highly hygroscopic, helping to retain water and replace

natural ped formation in the soilless matrix. Psyllium husks can absorb approximately 40 g/g water

(Czuchajowska et al., 1992), the PAM have been specified to absorb 1500 g/g of water. All experiments

were performed with two different dosages; 1g/L and 3 g/L (equal to 1 and 3 kg/m3) with control samples.

Samples were prepared with each amendment, at each dosage. The samples were then saturated so that

approximately 70% of the void spaces contained water (i.e. Si = 0.7). The calculated weight of water was

dosed into the dried substrate, mixed and equilibrated for >12 hours. The prepared samples were placed

into straight sided vessels with similar cross section area. The vessels were stored in the climate

controlled laboratory for just over two weeks and periodically weighed manually to determine the loss of

water. Air dried media samples were settled into tubes of 4.2 cm diameter to a height of 10 cm. The

tubes’ bottom orifices were covered by a geotextile with a water flow rate of 6519 l/min/m². Water was

added from the top of the tubes and the falling head time was measured. This process was done repeatedly

until the rate stabilized. The percolation rate was determined using Equation 0-1, where kf is the rate at

which water moves through the sample (permeability): L is the depth of the sample, Δt is the time taken

for the water level to drop, h0 is the initial water height and Ht is the height of the water at time = t.

𝑘𝑓 =𝐿

∆𝑡𝑙𝑛 (

𝐻0

𝐻𝑡

)

Equation 0-1 Permeability calculated using a ‘falling head’ method.

Media samples were settled into tubes of 4.2 cm diameter till a height of 10 cm. The tubes’ bottom

orifices were covered by a geotextile with a water flow rate of 6519 L/min/m². In an adaptation of ASTM

E2399-11, the samples were immersed in water for a 24-hour period to make them completely saturated

and left to drain for 2 hours. To obtain the retained water mass the samples were ejected from the tubes,

weighed and placed into the drying oven at a temperature of ~100 °C for 24 hours then weighed a second

time.

192

Results

At the 1 g/L dosage, neither amendment appeared to make a significant difference to the rate at which

water was evaporated from the straight sided container. The higher dosage of 3 g/L psyllium husk powder

significantly reduced the amount of water evaporated. This was determined using a paired t- test of time

series lost water values against the control sample (p < 0.05). Polyacrylamide fibres used at this dosage,

3g/L, did not make a significant difference to the rate of evaporation.

Figure 0-4. Evaporation of water from; a) psyllium husk amended compost, b) Polyacrylamide amended compost.

The compost material is relatively heterogeneous leading to a wide variation in the rate at which water

percolates. This is evident in the variation observed between the two control samples seen in Figure 0-5.

The rate at which the water percolates gradually drops and then stabilizes owing to migration of fines into

50%

60%

70%

80%

90%

100%

0 100 200 300 400

Sam

ple

mas

s

Hours

Psyllium Husk

0 g/L

1 g/L

3 g/L

50%

60%

70%

80%

90%

100%

0 100 200 300 400

Sam

ple

mas

s

Hours

Polyacrylamide

0 g/L

1 g/L

3 g/L

193

the interstices of the geotextile and to the swelling characteristic of the compost. The rate at which the

percolation stabilized was defined by 3 or more similar measurements without trend and resulted in rate

of 0.1 and 0.3 mm/s flow rate in the control samples. The amended composts were similarly tested and

the additions did not result in significant change in the final rate. The samples containing psyllium husk

powder did start with a more rapid flow rate, which slowed as the experiment became saturated. The

samples containing PAM did not show this same pattern, but did stabilize after fewer replicates of the

experiment, when lower volume of water had been passed, Figure 0-5 b).

Figure 0-5 Change in percolation rate after replicated measurements in a) psyllium husk, and b) polyacrylamide amended compost.

WRC experiments were conducted in duplicate, permitting comparison of mean results (one sample

containing PAM 3g/L was lost). The small error bars (σ) indicate that this experiment was repeatable and

that the samples were relatively homogenous in this characteristic compared to the hydraulic conductivity.

Each of the experiments resulted in significant change in the water retention capacity, as determined with

0.0

0.2

0.4

0.6

0.8

1.0

0 5 10 15

Per

mea

bil

ity k

f/m

m/s

Replicate

Psyllium husk

Control

Control

1 g/L

3 g/L

0.0

0.2

0.4

0.6

0.8

1.0

0 5 10 15

Per

mea

bil

ity, kf/

mm

/s

Replicate

Polyacrylamide

Control

Control

1 g/L

3 g/L

194

unpaired t-test against the control (p < 0.05). However, this change was not consistent with increasing the

dosage of either product.

Figure 0-6. Water retention as a proportion of the material dry weight in psyllium husk (PH) amended compost and polyacrylamide (PAM) amended compost.

Discussion

The psyllium husk powder was much more easily blended into the planting substrate resulting in a more

even distribution, although this did not appear to have a significant impact on the tests conducted. The

polyacrylamide fibres are more usually distributed by dissolution or aqueous suspension.

It was hypothesised that the amendments would slow the loss of water in the evaporation experiment

owing to the chemical binding of water into the mucilage or gel. This effect was only indicated in the

higher rate of psyllium husk dosage, which suggests that this experiment would benefit from replication

and expansion across a wider range of dosages. Though the polyacrylamide fibres clumped somewhat in

the compost matrix, this did not cause water to be retained more strongly. It had also been anticipated that

the amendments would increase wettability of the dried samples. Although all of the samples dried

sufficiently to shrink from the sides of their containers, the surfaces were still easily rewetted including

the control samples, so that any effect by the amendment was not apparent. In some circumstances, this

type of compost product has been observed to develop ‘hydrophobicity’ where water will bead upon the

surface rather than infiltrate.

The percolation experiment found that the additives did not cause a great variation in the flow rate though

the substrate under constant flow conditions. Such conditions exist only periodically in natural conditions

during particularly heavy rainfall. The migration of fines was emulated by the use of a non-woven

geotextile to retain the substrate sample. A repeat of these trials was conducted after the samples had air

dried for two weeks in the laboratory, resulting faster flow conditions once more. It remains unknown

140

150

160

170

180

190

200

210

Control PH 1g/L PH 3g/L PAM 1g/L PAM 3g/L

Wat

er R

etai

ned

/%

195

what contribution to the slowing flow rate results from the fines and which from the swelling of the

substrate. A clear trend indicating the effect of the amendments to the water retention capacity was not

evident. This may be due to the heterogeneity of the substrate, although this factor would also cause the

errors in the replicate experiments to be larger. In a practical setting the difference between these figures

may not be great enough to influence the overall green roof water retention properties.

Conclusions

In a review of trialled technologies, the soil amendments were preferred to physical barriers for erosion

control owing to the ease of use and the lack of interference with planting and subsequent maintenance

activities. Despite a growing interest within the industry, this type of material has not been widely

explored. Although, several different chemical compounds or natural powder products are employed in

erosion control on the ground, few have made their way onto rooftop applications.

These comparisons between psyllium husk powder and polyacrylamide fibres have indicated that their

use at low application rates may make a significant difference in laboratory settings, but not whether this

would translate to an appreciable difference at the field scale. Much more additional research is required

to determine the optimal application rates required to resist different velocities and wind turbulence, to

widen the range of additives tested and to include the transpiration of water with vegetation.

Designing Green Roofs for Low Impact Development: What ... · Past and current practices in green...

Documents

Transcript of Designing Green Roofs for Low Impact Development: What ... · Past and current practices in green...