Content uploaded by Jeremy T. Lundholm
Author content
All content in this area was uploaded by Jeremy T. Lundholm on Mar 05, 2018
Content may be subject to copyright.
1Plant Functional Traits Predict Green Roof Ecosystem Services
2Jeremy Lundholm,*Stephanie Tran, and Luke Gebert
3Biology/Environmental Science, Saint Mary’s University, 923 Robie Street, Halifax, NS B3H 3C3 Canada
4*
SSupporting Information
5ABSTRACT: Plants make important contributions to the
6services provided by engineered ecosystems such as green roofs.
7Ecologists use plant species traits as generic predictors of
8geographical distribution, interactions with other species, and
9ecosystem functioning, but this approach has been little used to
10 optimize engineered ecosystems. Four plant species traits (height,
11 individual leaf area, specific leaf area, and leaf dry matter content)
12 were evaluated as predictors of ecosystem properties and services
13 in a modular green roof system planted with 21 species. Six
14 indicators of ecosystem services, incorporating thermal, hydro-
15 logical, water quality, and carbon sequestration functions, were
16 predicted by the four plant traits directly or indirectly via their
17 effects on aggregate ecosystem properties, including canopy density and albedo. Species average height and specific leaf area were
18 the most useful traits, predicting several services via effects on canopy density or growth rate. This study demonstrates that easily
19 measured plant traits can be used to select species to optimize green roof performance across multiple key services.
20 ■INTRODUCTION
21 Constructed ecosystems such as treatment wetlands, green
22 roofs, and biowalls are engineered to provide ecosystem services.
23 These services depend on the presence of living plants and
24 engineered components such as growing medium, membranes,
25 and subsidies such as irrigation.
1−3
Green roofs can reduce heat
26 transfer through building roofs,
4
retain stormwater,
5
trap
27 airborne particulate matter,
6
sequester carbon,
7
and provide
28 habitat,
8
amenity, and aesthetic values.
3
Plant species and
29 vegetation types differ in their ability to provide these services;
3
30 thus, careful selection of plant species can optimize green roof
31 functioning. Green roofs often feature extreme conditions
32 relative to local natural ecosystems, including shallow soils and
33 high winds; thus, plant selection is also important to ensure
34 survival of green roof vegetation. There are thousands of plant
35 species in each region that can be used on green roofs,
9
yet
36 screening all these would be logistically impossible; hence, there
37 is a need for an efficient way to characterize plants based on
38 general traits that predict their function in a green roof setting.
39 Plant functional traits can be used to categorize species based
40 on their effects on ecosystem processes.
11,12
The plant trait
41 approach highlights ecological function regardless of geographic
42 distribution, taxonomic/phylogenetic relationships, and environ-
43 mental niche and thus represents a general method to screen
44 plant species for various purposes. Traits related to plant size and
45 leaf morphology are relatively easy to measure, and relate to
46 general plant strategies that differentiate species having evolved
47 under different environmental conditions: plants that are shorter,
48 with lower maximum growth rates, are typically found in areas
49 with low soil fertility. These traits, in turn, can predict or
50 influence ecosystem processes such as primary production,
51 nutrient and water uptake, and transpiration rate
10,11
but have
52seldom been used in environmental engineering. For example,
53specific leaf area (SLA), leaf area/dry weight, tends to be higher
54for plants with relatively high growth rates, that inhabit relatively
55fertile areas. SLA can be considered a “soft”trait that does not
56directly drive ecosystem process rates but is correlated with
57variables more closely associated with function,
11
such as net
58photosynthetic capacity and maximum growth rate, which in turn
59can drive ecosystem productivity. Functional traits have been
60used recently in green roof studies to select plants appropriate
61for local climates,
13
to predict growth rates,
14
and to maximize
62survival and stormwater retention.
15
This ecological approach
63uses traits measured in natural populations as indicators of
64generic plant strategies and employs these traits to predict
65ecosystem functions in different situations. The current study is
66the first to undertake a comprehensive analysis of the
67relationships between multiple ecosystem services provided by
68green roofs and generic plant traits.
69
■METHODS AND MATERIALS
70The green roof site was on the Saint Mary’s University campus,
71approximately five meters above ground level and circum-
72scribed by buildings one to three stories higher on all sides
73(Appendix S1, Supporting Information). The climate of Halifax
74is cold, humid, and maritime (Appendix S1). We used
75measurements of ecosystem properties and services from two
76previously established experiments in the same modular green
77roof system, comparing species grown in monoculture in the
Received: November 6, 2014
Revised: January 18, 2015
Accepted: January 19, 2015
Article
pubs.acs.org/est
© XXXX American Chemical Society ADOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
*UNKNOWN *|MPSJCA |JCA10.0.1465/W Unicode |es-2014-05426z.3d (R3.6.i7:4236 |2.0 alpha 39) 2014/12/19 13:33:00 |PROD-JCA1 |rq_3232469 |1/27/2015 05:48:25 |9|JCA-DEFAULT
78 same growing medium on the same site.
14,16−18
Plant species
79 were selected from harsh environments that have similarities to
80 green roofs (Appendix S1). Experiment 1 involved the estab-
81 lishment of 13 species in 2007,
14,16
three modules per species
82 (n= 3) (Table 1) (planting details in Appendix S1). Each
83 replicate module was a black plastic nursery tray (36 cm ×
84 36 cm ×12 cm), with a free-draining base, lined with a
85 composite nonwoven water-retention layer (Huesker Inc.,
86 Charlotte, NC), followed by an Enkamat (Colbond Inc.,
87 Enka, NC) drainage/filter layer which was topped with growing
88 medium (Sopraflor X, Soprema Inc., Drummondville, QC,
89 Canada) to a depth of approximately 6 cm (above the
90 Enkamat).
16
All modules were weeded throughout the experi-
91 ment to remove volunteer species.
92 Canopy density, considered an “ecosystem property”here
93 (number of contacts with live plant parts/0.07 m3), was
94 determined in each module using a three-dimensional pin
95 frame (36 cm ×36 cm ×36 cm) with 16 pins. The total
96 number of contacts between live plant parts and the pins was
97 recorded, and data from the 2010 biomass peak (taken in
98 August). Canopy growth rate incorporated canopy density
99 measurements at the end of the growing season (August) in
100 year 1 (2007) and year 2 (2008) to calculate the change in
101 density as a rate, relative to initial density: ln(density at t2)−
102 ln(density at t1)/number of days between t1and t2. Albedo was
103 quantified by placing modules one-by-one on a gray colored
104 weed barrier fabric (Quest Plastics Ltd., Mississauga, ON,
105 Canada). A single LI-COR pyranometer sensor and LI-250A
106 light meter (LI-COR Biosciences, Lincoln, NE) was affixed to
107 a retort stand, with the sensor and light meter 35 cm above
108 the upper edge of the module. Under clear sky conditions
109 within 1 h of solar noon incoming and reflected radiation were
110 measured in August 2010, within 1 week of canopy density
111 sampling. Incoming radiation (W/m2) was measured by
112 directing the pyranometer sensor toward the sky (180°away
113 from the module), and reflected solar radiation was measured
114 by facing the sensor directly at the module surface. These are
115 relative measurements with validity only within our green roof
116 system and were taken to compare relative reflectivity of the
117 different plant species in monoculture (Appendix S1).
118 An index of water loss rate was derived from water addition
119 experiments during the growing season of 2010: modules were
120 weighed using a PX-Series Checkweighing bench scale (ATRON
121 Systems Inc., West Caldwell, NJ), 1.3 L of water was added to
122 the substrate (representing a medium-sized rain event for the
123 region, equivalent to 10 mm of rain), by slowly pouring the water
124 at the base of the vegetation layer, slowly moving the container
125 so that the water would end up equally distributed across the
126 surface. Modules were reweighed 10 min later to determine
127 “stormwater capture”as the difference between pre- and
128 postwatering weights, twice between June−August. Modules
129 were reweighed 24 and 48 h later with the difference between
130 postwatering weights and weights 24 or 48 h later used as an
131 index of water loss (Appendix S1). These experiments were
132 carried out on sunny days within 1 h of solar noon. Ten
133 unplanted control modules were established and integrated with
134 the planted modules at the start of the experiment in 2007
135 (Appendix S1). Indices of relative water capture and loss were
136 created for each planted module each time the experiment was
137 performed by dividing the planted module value by the average
138 of the control (unplanted) modules in the same block.
139 Indicators of ecosystem services were quantified during the
140 2010 growing season. The effect of vegetation on substrate
141
temperatures during summer indicates a cooling service: lower
142
substrate temperatures are linearly correlated with lower net
143
heat flux into the building.
16
To create an index that can be
144
used to compare vegetation at different times with variation in
145
ambient air temperatures and insolation, we calculated an index
146of relative cooling by dividing the substrate temperatures in
147
planted modules within 1 h of solar noon on a sunny day
16
148
(once in May, once in July) by the average substrate
149
temperature of the control modules in the same block at the
150
same time (measurement details in Appendix S1). Three
151
variables related to soil chemistry services were calculated for
152
experiment 1. Growing medium samples (250 mL per module,
153
Appendix S1) were collected at the end of the growing season
154
(September 2010) and assessed for organic matter content
155(%)(loss on ignition), nitrate-N (ion-specific electrode), and
156
phosphate (P2O5) (Mehlich 3 extraction, inductively coupled
157
argon plasma), and were again converted to indices of relative
158
content by dividing planted module values by the average
159
value of control modules in the same block. Organic matter
160
content was included as an index of the ability of green roof
161
vegetation to increase carbon storage over time;
7
soil nutrient
162
concentrations were included, as greater nutrient uptake by the
163
vegetation may be associated with higher runoffwater quality,
19
164
so lower nutrient concentrations were assessed as indicating
165higher performance of nutrient removal services.
166
In 2009 we set up experiment 2, involving 10 replicates each
167
of 12 species and an unplanted control using the same modular
168
system as the previous experiment
17,18
(four of these species
169
were also in experiment 1) (Appendix S1). Canopy growth
170rate was determined for the first two years of growth as in the
171
other experiment. Substrate temperature was measured in the
172
same way as in experiment 1 during the same growing season
173
(2010) in May, July, and August. The stormwater capture
174
trials were performed three times between June−August 2010.
175
Snow depth was quantified as it impacts substrate temperature
176
and heat flux in winter and is affected by vegetation.
18
As an
177
ecosystem property, we used average snow depths when
178
there was snow coverage from January 7−March 7, 2011
179
(Appendix S1). As experiment 1 was harvested and soil
180nutrients quantified at the end of the growing season in 2010,
181
we could only examine winter performance for experiment 2.
182
Snow depths were used to represent the general effect of the
183
species monocultures on the differential accumulation of snow
184
on the substrate surface. In winter, net heat flux out of the
185
building is negatively correlated with substrate temperature,
18
186
so we used the minimum substrate temperature recorded be-
187
tween November 5, 2010, and March 31, 2011 (Appendix S1)
188
and divided by the average minimum of the control modules in
189the same block as an index of “heat trapping”in winter (higher
190
minima represent less heat lost in the winter). The experiment
191
2 modules were used in a subsequent experiment so substrate
192was not extracted and analyzed for nutrients and organic matter.
193
For indices that were measured more than once, we used the
194temporal average for each module. We then took an average of
195all modules of that species for each variable representing an
196
ecosystem property or ecosystem service for use in multiple
197
regression analyses. Indicators of ecosystem services for
198
which lower values indicate greater performance (e.g., a lower
199
substrate temperature in summer indicates greater cooling
200
ability of the vegetation, so replicates with cooler substrates
201
registered low indices relative to controls) were first reflected
202
(multiplied by −1) so that higher performance would register
203as a higher value. This was performed for the “substrate
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
B
Table 1. Measured Plant Traits, Ecosystem Properties, and Ecosystem Service Indicators (means ±standard error)
species
a
growth form
height
(cm)
b
leaf area (mm2)
specific leaf
area
(mm2/mg)
leaf dry
matter
content
(mg/
gfresh weight)
canopy growth rate
((contactsT2 −contactsT1)/
growing season days)
albedo
index
canopy
density
(leaf
contacts/
0.07m3)
water loss
index
stormwater
capture
index
substrate
cooling
index
c
organic
matter
index
d
phosphate
index
nitrate
index
substrate
winter
temperature
increase
index
snow depth
index
Arctostaphylos
uva-ursi
(expt 1)
creeping shrub 6. ±1 128.5 ±16.6 5.9 ±0.5 444.7 ±7.3 6.5 ×10−4±1.1 ×10−40.17 ±0 22 ±12 0.95 ±0.0 1.03 ±0.0 1.02 ±0.0 0.97 ±0.0 0.96 ±0.1
Symphotrichum
novi-belgii
(expt 2)
tall forb 54 ±5 1062.5 ±220.8 23 ±1.4 195.6 ±6 3.8 ×10−3±4.4 ×10−40.2 ±0.0 55 ±9 1.22 ±0.1 1.04 ±0.0 0.88 ±0.0 1.12 ±0.0 0.91 ±0.11
Campanula
rotundifolia
(expt 1)
tall forb 16 ±3 102.2 ±23.9 16.3 ±1.8 262 ±22.1 3.3 ×10−3±1.3 ×10−30.19 ±0 110 ±5 1.04 ±0.1 0.96 ±0.0 0.8 ±0.0 1.09 ±0.0 0.87 ±0.1 0.2 ±0.1
Carex
argyrantha
(expt 2)
sod forming
graminoid
63 ±0 1451.1 ±137 35.3 ±4.8 580.8 ±27.4 8.2 ×10−4±8.7 ×10−40.2 ±0.01 163 ±25 1.14 ±0.0 1.04 ±0.0 0.8 ±0.0 0.81 ±0.1 1.39 ±0.2
Carex nigra
(expt 2)
sod forming
graminoid
67 ±3 1116.1 ±141.4 10.5 ±0.9 401.2 ±23.5 9.4 ×10−4±1.4 ×10−40.2 ±0.0 154 ±20 1.10 ±0.0 1 ±0.0 0.82 ±0.0 0.7 ±0.0 1.34 ±0.1
Danthonia
spicata
(expts1and2)
bunch forming
graminoid
36 ±2 90.2 ±10.8 9.5 ±1 402.8 ±46.6 1.1 ×10−3±3.8 ×10−40.18 ±0 121 ±13 0.87 ±0.0 0.99 ±0.0 0.85 ±0.0 0.95 ±0.1 1.11 ±0.2 0.48 ±0.2 0.67 ±0.0 1.16 ±0.1
Deschampsia
flexuosa
(expts1and2)
bunch forming
graminoid
57 ±2 81.8 ±13.2 13.3 ±2.2 368.8 ±45.9 1. Seven ×10−3±4.4 ×10−40.17 ±0 98 ±15 0.95 ±0.0 1.01 ±0.0 0.9 ±0.0 1.07 ±0.1 0.84 ±0.1 0.85 ±0.2 0.9 ±0.1 0.98 ±0.1
Empetrum
nigrum
(expts1and2)
creeping shrub 8 ±1 4.6 ±0.4 40.1 ±3.1 90.2 ±20.8 1.7 ×10−4±4.5 ×10−40.17 ±0.01 56 ±10 0.91 ±0.0 1.07 ±0.0 0.94 ±0.0 0.95 ±0.1 1.13 ±0.1 0.46 ±0.3 0.92 ±0.0 0.94 ±0.1
Festuca rubra
(expt 2)
sod forming
graminoid
55 ±6 141.1 ±26.6 16.1 ±4 303.5 ±38.8 1.3 ×10−3±1.5 ×10−30.18 ±0.0 243 ±38 1.02 ±0.0 1.11 ±0.0 0.85 ±0.0 0.68 ±0.0 1.46 ±0.1
Gaultheria
procumbens
(expt 1)
creeping shrub 10 ±1 357.8 ±38.4 7.3 ±0.5 382.9 ±10.1 −2.7 ×10−3±1.3 ×10−30.16 ±05±4 1.11 ±0.1 1.04 ±0.0 0.95 ±0.0 0.99 ±0.2 1.15 ±0.2 0.89 ±0.3
Plantago
maritima
(expt 1)
tall forb 14 ±2 228.7 ±81.2 7.3 ±0.6 112.6 ±6.5 1.3 ×10−3±4.3 ×10−40.17 ±0 47 ±7 0.87 ±0.1 1.01 ±0.0 0.88 ±0.0 1.24 ±0.2 0.92 ±0.2 0.32 ±0.1
Poa compressa
(expt 1)
sod forming
graminoid
26 ±6 1662.0 ±465.0 41.8 ±6.9 244.8 ±57.3 3.4 ×10−3±6.8 ×10−40.19 ±0 209 ±60 0.92 ±0.1 1.01 ±0.0 0.74 ±0.0 0.99 ±0.1 1.09 ±0.2 0.1 ±0.0
Sagina
procumbens
(expt 1)
creeping forb 2 ±0 16.5 ±4.0 92.4 ±23.3 167 ±19.5 −2.7 ×10−3±5.1 ×10−40.18 ±0 101 ±8 1.10 ±0.1 0.95 ±0.0 0.86 ±0.1 1.01 ±0.1 1.17 ±0.2 0.83 ±0.4
Sedum acre
(expt 1)
succulent 4 ±0 18.5 ±2.8 37.0 ±8.4 124.7 ±13.6 1.4 ×10−3±5.1 ×10−40.26 ±0.0 124 ±30 1.07 ±0.1 0.87 ±0.0 0.74 ±0.0 1.13 ±0.2 0.78 ±0.1 0.87 ±0.3
Rhodiola rosea
(expt 1)
succulent 18 ±0 152.8 ±36.0 16.9 ±1.1 85.4 ±4.9 −2.7 ×10−3±1.0 ×10−30.16 ±0 22 ±6 0.96 ±0.0 0.96 ±0.0 0.94 ±0.0 0.92 ±0.1 1.18 ±0.2 1.25 ±0.9
Sedum spurium
(expt 1)
succulent 5 ±1 162.9 ±11.8 30.0 ±3.2 86.7 ±5.5 −8.5 ×10−4±1.5 ×10−30.21 ±0 87 ±39 0.84 ±0.1 0.93 ±0.0 0.8 ±0.0 0.93 ±0.1 1.11 ±0.2 0.35 ±0.1
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
C
204cooling”,“substrate winter temperature increase”, phosphate,
205and nitrate “removal”indices (Appendix S1).
206Plant traits are usually measured on plants growing in typical
207environmental conditions in field settings, so that trait values
208are representative of natural populations.
20
To determine plant
209traits for each species used in these experiments, five leaf samples
210per species were obtained from different individuals growing
211in their natural habitats (Appendix S1). Following established
212protocols,
11,20
healthy, fully expanded leaves from well developed
213plants were selected (Appendix S1). Within 2 h of collection
214fresh weight was measured and the five leaves per species
215scanned. Leaves were dried at 50 °C for 48 h and dry weight was
216measured. From these measurements, the following traits were
217obtained for the 21 test species: plant height, leaf area (LA),
218specific leaf area (SLA), and leaf dry matter content (LDMC).
219Plant height was determined as the average height of the five
220plants sampled in the field (to the nearest mm). LA was
221determined from the average of the five leaves, by measuring
222the one-sided surface area of the scanned image of each
223individual leaf using ImageJ software (version 1.47; NIH, USA)
224(Appendix S1). SLA was calculated as one-sided leaf area divided
225by its oven-dried weight, in mm2/mg. LDMC is the measure of
226dry leaf weight (mg) divided by fresh leaf weight (g).
227We evaluated the relationships between plant species traits,
228monoculture ecosystem properties, and ecosystem service
229indicators using multiple linear regression in order to construct
230a path diagram. We first evaluated the relationships between
231ecosystem properties and service indicators (model results in
232Table S1, Supporting Information) and then the relationships
233between plant traits and ecosystem properties. As we
234hypothesized that plant traits affect ecosystem services via
235differences in ecosystem properties, we only examined the direct
236effects of plant traits on ecosystem services if there were
237no adequate models linking ecosystem properties directly to
238services. Variables were transformed to improve homogeneity of
239variance, and standardized (Z-scores)
21
(Table S1). Model fit was
240evaluated using the AICc criterion.
22
All models were compared
241using delta-weights, and when multiple models had delta-
242weights lower than 7,
23
model averaging was used to generate
243standardized regression coefficients (Table S1). The path
244diagram was constructed using only those coefficients whose
24595% confidence limits did not overlap zero. Model selection and
246averaging procedures used the MuMIn package in R.
24
247
■RESULTS AND DISCUSSION
248Measured traits differed greatly among species with leaf
249area spanning more than 3 orders of magnitude (Table 1).
250Leaf area was highest in sod-forming graminoids but highly
251variable within that group, relatively high in bunch-forming
252graminoids, and lower but highly variable within the other
253growth forms (Table 1). Height was consistently low in
254creeping forbs and shrubs, relatively low but highly variable
255within the succulents, and high but variable within a growth
256form for the grasses and tall forbs. Leaf dry matter content
257(LDMC) showed the least variation with an almost 7-fold
258difference between the lowest and highest species values
259(Table 1). Low LDMC species included all the succulents and
260Empetrum nigrum,Plantago maritima,andSagina procumbens
261(each from a different growth form group). Specificleaf
262area (SLA) tended to be low for creeping shrubs, except for
263E. nigrum. SLA for other growth forms varied greatly within a
264group. S. procumbens, the only creeping forb included, had the
265highest SLA (Table 1).
Table 1. continued
species
a
growth form
height
(cm)
b
leaf area (mm2)
specific leaf
area
(mm2/mg)
leaf dry
matter
content
(mg/
gfresh weight)
canopy growth rate
((contactsT2 −contactsT1)/
growing season days)
albedo
index
canopy
density
(leaf
contacts/
0.07m3)
water loss
index
stormwater
capture
index
substrate
cooling
index
c
organic
matter
index
d
phosphate
index
nitrate
index
substrate
winter
temperature
increase
index
snow depth
index
Sibbaldiopsis
tridentata
(expt 2)
creeping shrub 9 ±3 395.6 ±108.3 8.1 ±0.3 380.9 ±6.8 −3.4 ×10−3±1.8 ×10−30.17 ±0.0 40 ±15 1.02 ±0.0 1.07 ±0.0 0.92 ±0.0 0.92 ±0.0 1.12 ±0.1
Solidago bicolor
(expts1and2)
tall forb 29 ±3 653.8 ±222.2 8.0 ±0.6 267.8 ±12.9 2.5 ×10−3±5.7 ×10−40.2 ±0.0 58 ±18 0.95 ±0.0 1.03 ±0.0 0.89 ±0.0 1.16 ±0.1 0.97 ±0.1 0.14 ±0.0 1.04 ±0.0 1.05 ±0.1
Solidago
puberula
(expt 2)
tall forb 41 ±3 705.5 ±75.1 28.4 ±2.9 448.4 ±24.5 2.2 ×10−3±7.1 ×10−40.2 ±0.0 55 ±15 0.94 ±0.0 1.06 ±0.0 0.93 ±0.0 0.99 ±0.0 1.08 ±0.1
Vaccinium
macrocarpon
(expt 2)
creeping shrub 8 ±1 26.0 ±2.3 7.3 ±0.6 582.6 ±17.2 −1.4 ×10−3±8.5 ×10−40.18 ±0.01 14 ±4 0.9 ±0.0 1.04 ±0.0 0.99 ±0.0 0.96 ±0.1 0.99 ±0.1
Vaccinium
vitis-idaea
(expt 1)
creeping shrub 4 ±0 56.2 ±7.6 6.2 ±0.4 465.4 ±9.1 −3.1 ×10−3±8.9 ×10−40.16 ±0.01 0 ±0 1.11 ±0.1 0.97 ±0.0 0.97 ±0.0 0.85 ±0.1 1.29 ±0.2 1.12 ±0.4
a
Nomenclature follows.
25
b
Trait data (height, SLA, leaf area, and LDMC) for experiment 1 are from the literature.
14
c
Several variables were reflected in analyses (cooling, phosphate, nitrate, winter
temperature increase); hence, low values in this table represent high performance of desired functions.
d
Soil chemistry data were only collected from experiment 1; winter data (snow depths and substrate
winter temperatures) were only collected for experiment 2.
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
D
266 Measured ecosystem properties also differed across growth
267 forms and species. Growth rates were slightly negative for
268 S. procumbens,Gaultheria procumbens,Sibbaldiopsis tridentata,
269 Vaccinium macrocarpon,V. vitis-idaea,Rhodiola rosea, and Sedum
270 spurium (Table 1). These are species that generally inhabit
271 extremely shallow and low-nutrient soils
26
and are expected to
272 have low growth rates. These results also indicate that the
273 canopies produced by these species reached their approximate
274 maximum densities in the green roof system by the end of the
275 first growing seasons. Tall forbs had consistently high canopy
276 growth rates, as did both groups of grasses (Table 1). The
277 succulent S. acre also had a relatively high growth rate. Creeping
278 shrubs and forbs had low positive or negative canopy growth
279 rates. The index of albedo showed the greatest range of
280 variation within the succulent group with both the lowest and
281 highest values (∼16% reflectivity for R. rosea and ∼26% for
282 Sedum acre). Species in the other groups varied much less
283 (between 16% and 19% reflectivity). Canopy density ranged
284 from zero (for the extremely short ground cover V. vitis-idaea)
285 to 243 contacts/0.07m3for the sod-forming graminoid Festuca
286 rubra. Grasses tended to have consistently high canopy density,
287 whereas tall forbs had consistently intermediate values (Table 1).
288 While all creeping shrubs were at the low end of canopy
289 densities, there was considerable variation among species in
290 the group, as well as in the succulent group. Water loss rates
291 ranged from ∼16% lower than unplanted controls (S. spurium),
292 implying that the vegetation blocked evaporation from the
293 bare substrate (relative to controls) but also had very low
294 transpiration rates, to 22% higher (S. novi-belgii) than controls.
295 Values substantially lower than controls were found in other
296 species (Danthonia spicata, E. nigrum, and V. macrocarpon) that
297 tended to concentrate their leaf biomass close to the substrate
298 surface, likely presenting a barrier to evaporation from the
299 substrate surface. There was otherwise little consistency in
300 water loss within a growth form group. The effect of the
301 planted species on snow depths ranged from 46% higher than
302 unvegetated control modules in the dense-canopied F. rubra
303 and other sod-forming graminoids, to close to no difference
304 from controls for sparse-canopied tall forb Symphotrichum
305 qnovi-belgii, and some creeping shrubs (Table 1).
306 Consistent with other green roof studies,
27
vegetation had
307 a relatively small effect on stormwater capture, with the greatest
308
increase relative to controls around 10% for F. rubra (Table 1).
309
Creeping shrubs tended to be at the high end for stormwater
310
capture, along with tall forbs, with succulents at the low end.
311
Several species had lower capture than controls (S. acre had one
312
of the lowest). Vegetation reduced summer substrate temper-
313
atures up to 26% lower than controls (Poa compressa and S. acre
314
had the coolest substrate). Some species were very similar in
315
substrate temperature to controls (Arctostaphylos uva-ursi,
316
V. macrocarpon, V. vitis-idaea). In general, creeping shrubs
317
tended to cool the substrate the least, followed by tall forbs,
318
with inconsistent results among grass species and highly
319
variable performance among the succulents (Table 1). Organic
320
matter showed little difference from controls, with high
321
variability within a species and no consistent patterns between
322
growth forms, save that the three tall forb species tended to have
323
relatively high amounts of organic matter. Most species
324
contained equivalent phosphate in their substrate to controls;
325
the only species standing out with relatively low phosphate were
326
S. acre and Deschampsia flexuosa. Substrate nitrate varied greatly
327
among species with P. compressa showing very low levels (10% of
328
the control values), considerable range among the three grass
329
species, consistently low values among the three tall forbs, and
330
no clear pattern among the other species (Table 1). Most species
331
increased minimum winter substrate temperatures (registering as
332
lower values in Table 1) relative to controls. Two grasses with
333
relatively dense canopies close to the ground (D. spicata and
334
F. rubra) raised winter minimum temperatures by approximately
335
30% relative to controls (Table 1). Grasses tended to have the
336
highest winter minimum temperatures, whereas tall forbs and
337creeping shrubs had similar values to controls.
338
Most of the ecosystem service indicators were predicted by
339
vegetation properties (Figure 1). Stormwater capture was
340
positively correlated with plant height but not any of the
341
vegetation properties. The index of water loss rate was not
342
predicted by plant traits and did not predict any of the services.
343
The three services related to substrate chemistry (organic
344
matter, phosphate, and nitrate removal) were all positively
345
predicted by canopy growth rate, which was positively
346
predicted by plant height (Figure 1; Table S1). Substrate
347
cooling was positively related to albedo and canopy density.
348
Albedo was also related to canopy density but not any of the
349plant traits. Canopy density was positively related to both SLA
Figure 1. Relationships between plant traits (left), green roof ecosystem properties (center), and indicators of ecosystem services (right).
Standardized regression coefficients determined by model averaging (showing only relationships with coefficients whose 95% confidence intervals
did not overlap zero).
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
E
350 and plant height. Substrate winter temperature increase was
351 positively correlated with snow depth and canopy density.
352 Deeper snow was associated with denser canopies, lower
353 canopy growth rates, and higher albedo. Leaf area and LDMC
354 were also positive predictors of snow depth.
355 Plant traits represent general characteristics of plant species,
356 but their effects on ecosystem processes are mediated through
357 canopy characteristics and the effects of plant canopies on
358 microclimates.
28−30
In this study, “ecosystem properties”
359 represent empirically measured canopy variables from mono-
360 culture species populations grown in the green roof setting.
361 Traits related to overall plant size and leaf characteristics
362 measured under natural conditions predicted canopy density,
363 canopy growth rate, and snow depth, which in turn predicted
364 the ecosystem service indicators. The plant traits used as
365 predictors here represent easy-to-measure “soft”
20
traits that are
366 correlated with process rates but not necessarily closely related
367 mechanistically to ecological processes.
368 Plant height is a general indicator of high growth rates
369 and/or evolution in resource-rich environments where com-
370 petition for light is important, although there are many
371 trade-offs involved in growing tall.
31
The tallest species in
372 this study were primarily graminoids and tall forbs (Table 1),
373 which are associated with deeper soils and more fertile
374 conditions, and had higher growth rates in the green roof
375 system. It is important to note that “height”and the other
376 leaf traits are general indicators of the fertility of the habitats
377 where the species tend to grow naturally. While taller species
378 performed some ecosystem services more effectively than
379 shorter species in this system, there are important caveats
380 relevant to interpreting these results. First, the species selected
381 for these experiments all grow in relatively harsh or resource-
382 limited environments (Appendix S1) and tend to be relatively
383 short compared with species from more fertile habitats. Our
384 finding implies that the species that are tallest in a set of short
385 species perform these services better than the shorter ones.
386 Second, while the species we grew on the green roof generally
387 grew shorter than under natural conditions, there may be dis-
388 advantages for relatively tall species in green roof environments.
389 Relatively tall species may be more susceptible to drought,
32
390 and tall species from relatively fertile environments may have
391 higher resource requirements leading to population crashes
392 if fertilization or supplemental irrigation are not applied. This
393 study ran for a relatively short period of time so longer studies
394 are required to determine how height and long-term survival
395 may be related.
396 Stormwater capture was positively related to plant species
397 height, possibly due to an overall greater resource demand and
398 water uptake, possibly mediated through greater root biomass,
399 which supports what has been found in other studies.
33,34
400 In this study, species height also predicted canopy growth rate,
401 and density, which were positive correlates of all the ecosystem
402 service indicators except stormwater capture.
403 Canopy growth rate relates to the overall primary production
404 of a population of a given species and is expected to be correlated
405 with the organic matter content of soils,
35
and rates of nutrient
406 uptake.
36
Here canopy growth rate was a relatively strong
407 predictor of all three services related to soil properties: organic
408 matter, phosphate, and nitrate contents. Species from relatively
409 fertile environments (indicated by taller heights) grew faster in
410 the green roof system and resulted in higher substrate organic
411 matter, and lower phosphate and nitrate contents, likely as a
412 result of overall higher rates of multiple metabolic processes.
413
Canopy density represents the density of live leaf and stem
414
parts in aboveground vegetation and was higher in the green
415
roof system for species that were generally taller in their natural
416
environments and had higher SLA values. These species, in
417
turn, gave rise to lower summer substrate temperatures and
418higher winter minima. SLA represents the amount of invest-
419
ment in light-intercepting area per unit of dry mass and is
420
expected to be higher in species from more resource-rich
421
environments, resulting in a higher ability to harvest light per
422
unit biomass allocation.
37
Again it is important to emphasize
423
that the high SLA species in this study are generally from the
424
most fertile areas in low productivity areas (coastal barrens,
425
rock outcrops) that have some similarities with the green roof
426
environment (shallow substrates, exposure to wind and sun).
427SLA is positively correlated with net photosynthetic capacity,
38
428
potential growth rate,
39
and evapotranspiration rate.
40
High
429
SLA species in this study belong to a greater diversity of growth
430
forms than tall species: forbs, graminoids, shrubs, and
431succulents were all among the highest SLA species.
432High LDMC species included two dwarf shrubs in the same
433genus (Vaccinium); low LDMC species were the succulents,
434
E. nigrum and Plantago maritima, which have relatively thick
435
leaves and likely high water storage similar to the true succulents.
436
Species with large individual leaf area included some graminoids
437
and tall forbs. Both leaf area and LDMC were weak positive
438
predictors of snow depth, although canopy density had the
439
strongest effect of any of the individual variables. This shows
440
that dense canopies led to greater trapping of windborne snow,
441
accumulation of less dense snow, and/or decreases in melt rate,
442as predicted in previous work.
18
When canopy density is held
443
constant there were also independent effects of leaf area and
444
LDMC so species with larger individual leaves and/or greater dry
445
matter content also resulted in deeper snow layers which have a
446
moderating effect on winter substrate temperatures (Figure 1).
447
While we did not measure canopy density in the winter (i.e.,
448
we used August canopy densities to predict snow depths in the
449
following winter) the strong predictive power of summer canopy
450
density implies that species with dense canopies in summer
451
retain dense canopies of dead stems and leaves or woody tissue
452in winter. Some of the high canopy density species tested in the
453
snow study were relatively slow growing; thus, there was a
454
negative relationship between growth rate and snow depth.
455
The relationships between snow accumulation and plant traits
456
are complex, as snow accumulation represents the outcome of
457
several distinct processes.
41
Canopy density had an effect on
458
raising minimum winter substrate temperatures independent of
459
snow depth, suggesting that plant residues, leaves from evergreen
460
species, or other structures helped retain heat in the substrate.
461One possible mechanism is the reduction of convective heat
462
losses due to reduction of air movement within the canopy
42
and
463
lower wind speeds at ground level
43
even when canopy material
464is largely dead.
465
Here we used an ecological approach to predict green roof
466services: plant traits measured on plants growing in their natural
467habitats were linked to ecosystem services via statistical models,
468
using empirical data from replicated green roof modules. The
469
approach taken by environmental engineers usually involves
470
using numerical models of physical processes to predict key
471
variables, and altering parameter values to explore sensitivity
472
to environmental variability due to climate or vegetation type.
4
473
The finding that substrate cooling has a positive relationship with
474
both albedo and canopy density is congruent with the findings
475from numerical heat balance models, although our temperature
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
F
476 index includes several sources of variation from averaging across
477 replicates within a species and sampling dates. Parametric
478 variations in simulation studies have found leaf area index (LAI,
479 representing single-sided total leaf area in the canopy) to be one
480 of the most important parameters in determining the reduction
481 of substrate surface temperatures, with LAI inversely related to
482 substrate surface temperature.
44−46
This result is due to LAI
483 being a factor in the calculation of convective heat transfer,
484 evapotranspiration, and shading. These heat fluxes are generally
485 calculated according to the “big leaf”approach, i.e., the Penman−
486 Monteith equation for evapotranspiration, whereby the heat
487 fluxes of a single, average leaf are extrapolated to the canopy level
488 via the LAI. While we did not measure LAI directly, within a
489 species, LAI and canopy density should be linearly and positively
490 related. Further empirical work to quantify LAI for different
491 green roof vegetation types is the next step to link screening of
492 species using ecological traits with prediction of green roof
493 functioning.
494 Sensitivity analyses suggest that plant height is one of the
495 primary variables determining thermal gain through green
496 roofs.
47
In this study, average plant height for a given species in
497 its natural habitats is correlated with substrate cooling via its
498 role in canopy density (Figure 1), possibly via greater reflection
499 and/or absorption of solar radiation, represented in numerical
500 models by the extinction coefficient. Plants that were relatively
501 tall in natural habitats produced a lot of canopy density on
502 the green roof, due to high productivity and a generally taller
503 stature than the other species. However, the role of plant height
504 in reducing ground heat flux is also likely due to its role in the
505 aerodynamic resistance of convective and latent heat transfer.
506 According to the logarithmic wind profile approach adopted
507 in numerous green roof heat balance models,
4,48,49
vertical
508 wind shear within the canopy increases the convective and
509 evapotranspiration heat losses from the green roof system for
510 taller plants by reducing the aerodynamic resistance and thus
511 reducing the heat flux into the substrate. Thus, taller plants
512 likely influence substrate temperatures via several mechanisms
513 simultaneously.
514 Additional cooling mechanisms reduce the importance of
515 albedo in the surface energy balance of a green roof compared
516 to a conventional roof;
50
however, the results of this study
517 showed canopy albedo still had a positive relationship with
518 substrate cooling. To simulate the effects of albedo on heating
519 loads, one study compared Sedum tomentosum in monoculture
520 to a mixture of Sedum species,
51
which had measured canopy
521 reflection coefficients of 0.23 and 0.11, respectively, using a
522 previously validated green roof model.
52
The average peak net
523 radiation difference and average net radiation between the two
524 plant conditions was as high as ∼16% and ∼20%, respectively,
525 suggesting that plant screening for canopy albedo may result in
526 improvements to green roof thermal performance. None of the
527 plant traits we used predicted albedo, so additional traits related
528 to leaf optical properties would need to be incorporated to
529 allow this kind of screening.
530 Several empirical studies show that stormwater retention in
531 green roofs is related to antecedent soil moisture content;
5,27,53
532 thus, high transpiration rates in the vegetation should be able to
533 increase the water holding capacity of soil. While some green
534 roof studies show a relationship between water loss rate and
535 stormwater capture,
16
the current study did not show this. The
536 index of water loss used here represented water lost within
537 the first 48 h of water addition, when conditions were wet
538 and water loss would be a function of both evaporation from
539
substrate surface and transpiration from the canopy, as well as
540
possible drainage after the first 10 min of runoff. A longer
541
drying period, leading to a drier substrate surface, might have
542
produced a clearer signal of differential transpiration rates
543
among species and a stronger effect on subsequent stormwater
544retention, as evaporation from the substrate surface would
545
have been minimized after it dried out completely. Stormwater
546
capture should be a function of water loss over the entire interval
547
between rain events; thus, the experimental method employed
548
here likely overlooked the differences among plant species in ET
549rates over the longer term.
550Since leachate from green roofs can be high in nutrients,
53,54
551
the uptake and retention of nutrients by vegetation could
552
mitigate nutrient pollution in runoff. Here we measured
553
substrate concentrations of nitrate and phosphate after four
554
years of plant growth; the index of removal was relative to
555
control modules in which leaching was presumed to be the
556
primary process driving nutrient loss. The low nutrient content
557
relative to controls in some plant species treatments (Table 1)
558
suggests that plant uptake and cycling within the plant−
559substrate system is responsible for the reductions, rather than
560greatly increased leaching rates for vegetated modules. We did
561
not quantify leachate concentrations directly, and it is clear
562
that further work needs to be done to examine the effects of
563
plant species on nutrient dynamics and water quality in green
564
roof systems. Likewise, our measurement of organic matter
565
is intended to be an index of the differential ability of plant
566
species to contribute toward carbon capture in the system,
567
although it omits key variables such as inorganic carbon that are
568
essential to understand the entire impact of green roof vege-
569tation on carbon sequestration.
7
These indices only partially
570capture the important ecosystem services of green roofs, but
571
they clearly respond to differences in plant species, suggesting
572
that plant traits can be used to optimize ecosystem service
573provisioning.
574
While shallow-substrate (extensive) green roofs are usually
575planted with succulents due to their superior drought tolerance,
576
the plants with the highest trait values tended not to be
577
succulents in this study. The most common vegetation for
578
shallow-substrate green roofs is mixtures of Sedum species or
579
other succulents.
9
While here we only quantified the effects of
580
plant traits on monoculture provisioning of ecosystem services,
581
a past study on the same system showed that a Sedum mixture
582
and S. acre monocultures performed similarly for summer sub-
583
strate temperature reductions, albedo, and stormwater capture
584
(other services were not quantified).
16
This suggests that Sedum
585mixtures should be effective at summer cooling but relatively
586
poor in stormwater capture, relative to other species we
587
examined. As several other studies have shown,
16,33,35,55,56
588
succulents are not always the best choice if particular ecosystem
589
services are to optimized besides roof cooling, but less drought
590
tolerant species sharing the traits identified here could be
591
included on green roofs depending on regional climate,
592availability of irrigation, and other factors.
593
In addition to Sedum mixtures, shallow-substrate green roofs
594
often employ more diverse species and growth form mixtures.
8
595
Some designs feature multiple species planted as monocultural
596
beds, and our results are relevant to single-species patches within
597
diverse roofs of this kind. However, predicting ecosystem
598
services from species-diverse roofs featuring multispecies
599
communities requires a different approach. Future studies should
600
examine functional trait diversity in species mixtures to
601determine the effects on ecosystem services. The approach we
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
G
602 adopted allows prediction of monoculture performance as a first
603 step toward screening lists of species for possible inclusion in
604 mixed plantings.
605 This study also shows that broad growth form groupings can
606 exhibit considerable variation in traits among species,
607 suggesting that functional traits, rather than growth form
608 groups might be used for green roof plant selection. Height and
609 SLA were the best predictors, and while height was consistent
610 for only some of the growth form groups, SLA was highly
611 variable both within and between groups. Out of the species
612 tested here, we could recommend designing green roofs with
613 species that have relatively high SLA and height values, such as
614 F. rubra. It is also possible that mixtures of species with
615 opposite trait values (e.g., a short species with high SLA such as
616 S. procumbens and a tall species such as Carex nigra with low
617 SLA) might be combined to yield optimal functioning, if the
618 diversity in morphology leads to complementary resource use
619 or other synergistic effects.
16
However, the design criteria for a
620 particular roof, including ecosystem services to be maximized,
621 may require attention to other traits not measured here such
622 as flowering period or other variables related to habitat use by
623 animals or human aesthetic criteria.
624 This study used plant leaf and traits determined by an
625 average of measurements from plants from five naturally
626 occurring populations. The ecosystem properties and service
627 indicators measured on the green roof also represent averages
628 within each species. Despite the considerable variation in traits
629 and response variables within a species (Table 1), there was still
630 a high predictive ability across species. The novelty of this study
631 is to show that simple leaf and canopy traits can predict
632 multiple important green roof functions. While the indices we
633 used cannot be directly converted into precise estimates of
634 ecosystem services for use in cost benefit analyses or use in
635 numerical modeling, traits can be used to screen large numbers
636 of species. Plant traits for many regional floras are now avail-
637 able in databases
57
and can be used to generate plant lists for
638 further empirical testing and for the design of many kinds of
639 constructed ecosystems.
640 ■ASSOCIATED CONTENT
641 *
SSupporting Information
642 Appendix S1: Detailed methods. Table S1: Multiple linear
643 regression models used for construction of path diagram in
644 Figure 1. This material is available free of charge via the
645 Internet at http://pubs.acs.org.
646 ■AUTHOR INFORMATION
647 Corresponding Author
648 *Tel: (902) 420-5506. E-mail: jlundholm@smu.ca.
649 Author Contributions
650 The manuscript was written through contributions of all
651 authors. All authors have given approval to the final version of
652 the manuscript.
653 Notes
654 The authors declare no competing financial interest.
655 ■ACKNOWLEDGMENTS
656 We thank J. Scott MacIvor and Melissa Ranalli for data
657 collection, and Amy Heim and Emily Walker for constructive
658 criticism of the manuscript. This work was funded by NSERC
659 Discovery (311788-2010) and Strategic Project (413116-2011)
660 Grants.
661
■ABBREVIATIONS
662SLA specific leaf area
663LDMC leaf dry matter content
664LAI leaf area index
665ET evapotranspiration
666
667
■REFERENCES
(1) 668Darlington, A. B.; Dat, J. F.; Dixon, M. A. The biofiltration of
669indoor air: Air flux and temperature influences the removal of toluene,
670ethylbenzene, and xylene. Environ. Sci. Technol. 2001,35, 240−6.
(2) 671Stottmeister, U.; Wiener, A.; Kuschk, P.; Kappelmeyer, U.;
672Kastner, M.; Bederski, O.; Müller, R.; Moormann, H. Effects of plants
673and microorganisms in constructed wetlands for wastewater treatment.
674Biotechnol. Adv. 2003,22, 93−117.
(3) 675Lundholm, J. T.; Williams, N. S. G. Effects of vegetation on green
676roof ecosystem services. In Green Roof Ecosystems; Sutton, R. K., Ed.;
677Springer: New York, 2015 (in press).
(4) 678Sailor, D. J. A green roof model for building energy simulation
679programs. Energy Build. 2008,40, 1466−1478.
(5) 680
Villarreal, E. L.; Bengtsson, L. Response of a sedum green-roof to
681individual rain events. Ecol. Eng. 2005,25,1−7 DOI: 10.1016/
682j.ecoleng.2004.11.008.
(6) 683Speak, A. F.; Rothwell, J. J.; Lindley, S. J.; Smith, C. L. Urban
684
particulate pollution reduction by four species of green roof vegetation
685in a UK city. Atmos. Environ. 2012,61, 283−293.
(7) 686Getter, K. L.; Rowe, D. B.; Robertson, G. P.; Cregg, B. M.;
687Andresen, J. A. Carbon sequestration potential of extensive green
688roofs. Environ. Sci. Technol. 2009,43, 7564−7570.
(8) 689Williams, N. S. G.; Lundholm, J. T.; MacIvor, J. S. Can green
690roofs improve biodiversity conservation in cities? J. Appl. Ecol. 2014,
69151, 1643−1649.
(9) 692Snodgrass, E. C.; Snodgrass, L. L. Green Roof Plants: A Resource
693and Planting Guide; Timber Press: Portland, OR, 2006.
(10) 694Lavorel, S.; Garnier, E. Predicting changes in community
695composition and ecosystem functioning from plant traits: Revisiting
696the Holy Grail. Funct. Ecol. 2002,16, 545−556.
(11) 697
Cornelissen, J. H. C.; Lavorel, S.; Garnier, E.; Diaz, S.;
698Buchmann, N.; Gurvich, D. E.; Reich, P. B.; ter Steege, H.; Morgan, H.
699D.; van der Heijden, M. G. A.; Pausas, J. G.; Poorter, H. A handbook
700of protocols for standardised and easy measurement of plant functional
701traits worldwide. Aust. J. Bot. 2003,51 (4), 335−380.
(12) 702Díaz, S.; Cabido, M. Vive la difference: plant functional diversity
703
matters to ecosystem processes. Trends Ecol. Evol. 2001,16 (11), 646−
704655.
(13) 705
Van Mechelen, C.; Dutoit, T.; Kattge, J.; Hermy, M. Plant trait
706analysis delivers an extensive list of potential green roof species for
707Mediterranean France. Ecol. Eng. 2014,67, 48−59.
(14) 708Lundholm, J.; Heim, A.; Tran, S.; Smith, T. Leaf and life history
709traits predict plant growth in a green roof ecosystem. PLoS One 2014,
7109(6), e101395 DOI: 10.1371/journal.pone.0101395.
(15) 711Farrell, C.; Szota, C.; Williams, N. S. G.; Arndt, S. K. High water
712users can be drought tolerant: Using physiological traits for green roof
713plant selection. Plant Soil 2013,372, 177−193.
(16) 714Lundholm, J.; MacIvor, J. S.; MacDougall, Z.; Ranalli, M. Plant
715species and functional group combinations affect green roof ecosystem
716functions. PLoS One 2010,5(3), e9677 http://dx.doi.org/10.
7171371%2Fjournal.pone.0009677.
(17) 718MacIvor, J. S.; Lundholm, J. T. Performance evaluation of native
719plants suited to extensive green roof conditions in a maritime climate.
720Ecol. Eng. 2011,37, 407−417.
(18) 721Lundholm, J. T.; Weddle, B. M.; MacIvor, J. S. Snow depth and
722vegetation type affect green roof thermal performance in winter.
723Energy Build. 2014,84, 299−307.
(19) 724
Berndtsson, J. C.; Bengtsson, L.; Jinno, K. Runoff water quality
725from intensive and extensive vegetated roofs. Ecol. Eng. 2009,35 (3),
726369−380.
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
H
(20)
727 Hodgson, J. G.; Wilson, P. J.; Hunt, R.; Grime, J. P.; Thompson,
728 K. Allocating C- S-R plant functional types: A soft approach to a hard
729 problem. Oikos 1999,85, 282−294.
(21)
730 Maestre, F. T.; Quero, J. L.; Gotelli, N. J.; Escudero, A.; Ochoa,
731 V.; Delgado-Baquerizo, M.; et al. Plant species richness and ecosystem
732 multifunctionality in global drylands. Science 2012,335 (6065), 214−
733 218.
(22)
734 Burnham, K. P.; Anderson, D. R. Model Selection and Multimodel
735 Inference: A Practical Information-theoretic Approach; Springer: New
736 York, 2002.
(23)
737 Burnham, K. P.; Anderson, D. R.; Huyvaert, K. P. AIC model
738 selection and multimodel inference in behavioral ecology: Some
739 background, observations, and comparisons. Behav. Ecol. Sociobiol.
740 2011,65, 23−35.
(24)
741 Bartoń, K. MuMIn: Multi-model inference. R package version
742 1.9.13. 2013. http://CRAN.R-project.org/package=MuMIn.
(25)
743 Zinck, M. Roland’s Flora of Nova Scotia; Nova Scotia Museum:
744 Halifax, Nova Scotia, 1998.
(26)
745 Gleason, H. A.; Cronquist, A. Manual of Vascular Plants of
746 Northeastern United States and Adjacent Canada; The New York
747 Botanical Garden: Bronx, NY, 1991.
(27)
748 VanWoert, N. D.; Rowe, D. B.; Andresen, J. A.; Rugh, C. L.;
749 Fernandez, R. T.; Xiao, L. Green roof stormwater retention: Effects of
750 roof surface, slope, and media depth. J. Environ. Qual. 2005,34 (3),
751 1036−1044.
(28)
752 Gausman, H. W.; Allen, W. A. Optical parameters of leaves of 30
753 plant species. Plant Physiol. 1973,52 (1), 57−62.
(29)
754 Oke, T. R. Boundary Layer Climates; London: Methuen & Co.
755 Ltd.: London, U.K., 1978.
(30)
756 Gower, S. T.; Kucharik, C. J.; Norman, J. M. Direct and indirect
757 estimation of leaf area index, fAPAR, and net primary production of
758 terrestrial ecosystems. Remote Sens. Environ. 1999,70 (1), 29−51.
(31)
759 Westoby, M.; Falster, D. S.; Moles, A. T.; Vesk, P. A.; Wright, I.
760 J. Plant ecological strategies. some leading dimensions of variation
761 between species. Annu. Rev. Ecol. Syst. 2002,33, 125−159.
(32)
762 Butler, C.; Orians, C. M. Sedum cools soil and can improve
763 neighboring plant performance during water deficit on a green roof.
764 Ecol. Eng. 2011,37, 1796−1803.
(33)
765 Dunnett, N.; Nagase, A.; Booth, R.; Grime, P. Influence of
766 vegetation composition on runoff in two simulated green roof
767 experiments. Urban Ecosyst. 2008,11 (4), 385−398.
(34)
768 Nagase, A.; Dunnett, N. Amount of water runoff from different
769 vegetation types on extensive green roofs: Effects of plant species,
770 diversity and plant structure. Landscape Urban Plan. 2012,104 (3−4),
771 356−363.
(35)
772 De Deyn, G. B.; Cornelissen, J. H.; Bardgett, R. D. Plant
773 functional traits and soil carbon sequestration in contrasting biomes.
774 Ecol. Lett. 2008,11 (5), 516−531.
(36)
775 Hobbie, S. E. Effects of plant species on nutrient cycling. Trends
776 Ecol. Evol. 1992,7(10), 336−339.
(37)
777 Westoby, M. A leaf-height-seed (LHS) plant ecology strategy
778 scheme. Plant Soil 1998,199 (2), 213−227.
(38)
779 Reich, P. B.; Walters, M. B.; Ellsworth, D. S. From tropics to
780 tundra: global convergence in plant functioning. Proc. Natl. Acad. Sci.
781 U. S. A. 1997,94, 13730−13734.
(39)
782 Grime, J. P.; Thompson, K.; Hunt, R.; Hodgson, J. G.;
783 Cornelissen, J. H. C.; Rorison, I. H.; et al. Integrated screening
784 validates primary axes of specialisation in plants. Oikos 1997,79, 259−
785 281.
(40)
786 Reich, P. B.; Ellsworth, D. S.; Walters, M. B.; Vose, J. M.;
787 Gresham, C.; Volin, J. C.; Bowman, W. D. Generality of leaf trait
788 relationships: A test across six biomes. Ecology 1999,80 (6), 1955−
789 1969.
(41)
790 Liston, G.; McFadden, J.; Sturm, M.; Pielke, R., SR. Modelled
791 changes in arctic tundra snow, energy and moisture fluxes due to
792 increased shrubs. Glob. Change Biol. 2002,8, 17−32.
(42)
793 Jones, H. G. Plants and Microclimate: A Quantitative Approach to
794 Environmental Plant Physiology; Cambridge University Press: Cam-
795 bridge, UK, 2013.
(43) 796Aase, J. K.; Siddoway, F. H. Stubble height effects on seasonal
797microclimate, water balance, and plant development of no-till winter
798wheat. Agric. Meteorol. 1980,21 (1), 1−20.
(44) 799Del Barrio, E. P. Analysis of the green roofs cooling potential in
800buildings. Energy Build. 1998,27, 179−193.
(45) 801Kumar, R.; Kaushik, S. C. Performance evaluation of green roof
802and shading for thermal protection of buildings. Build. Environ. 2005,
80340 (11), 1505−1511.
(46) 804Jaffal, I.; Ouldboukhitine, S. E.; Belarbi, R. A comprehensive
805study of the impact of green roofs on building energy performance.
806Renew. Energy 2012,43, 157−164.
(47) 807Olivieri, F.; Di Perna, C.; D’Orazio, M.; Olivieri, L.; Neila, J.
808Experimental measurements and numerical model for the summer
809performance assessment of extensive green roofs in a Mediterranean
810coastal climate. Energy Build. 2013,63,1−14.
(48) 811Alexandri, E.; Jones, P. Temperature decreases in an urban
812canyon due to green walls and green roofs in diverse climates. Build.
813Environ. 2008,43 (4), 480−493.
(49) 814Djedjig, R.; Ouldboukhitine, S. E.; Belarbi, R.; Bozonnet, E.
815Development and validation of a coupled heat and mass transfer
816model for green roofs. Int. Commun. 7Heat Mass Transfer 2012,39 (6),
817752−761.
(50) 818Ouldboukhitine, S. E.; Belarbi, R.; Sailor, D. J. Experimental and
819numerical investigation of urban street canyons to evaluate the impact
820of green roof inside and outside buildings. Appl. Energy 2014,114,
821273−282.
(51) 822Zhao, M.; Tabares-Velasco, P. C.; Srebric, J.; Komarneni, S.;
823Berghage, R. Effects of plant and substrate selection on thermal
824performance of green roofs during the summer. Build. Environ. 2014,
82578, 199−211.
(52) 826Tabares-Velasco, P. C.; Srebric, J. A heat transfer model for
827assessment of plant based roofing systems in summer conditions.
828Build. Environ. 2012,49, 310−323.
(53) 829Stovin, V. The potential of green roofs to manage urban
830stormwater. Water Environ. J. 2010,24 (3), 192−199.
(54) 831Rowe, D. B. Green roofs as a means of pollution abatement.
832Environ. Pollut. 2011,159 (8−9), 2100−2110.
(55) 833Monterusso, M. A.; Rowe, D. B.; Rugh, C. L.; Russell, D. K.
834Runoff water quantity and quality from green roof systems. Acta
835Hortic. 2004,639, 369−376.
(56) 836Blanusa, T.; Vaz Monteiro, M. M.; Fantozzi, F.; Vysini, E.; Li, Y.;
837Cameron, R. W. Alternatives to Sedum on green roofs: Can broad leaf
838perennial plants offer better ‘cooling service’?Build. Environ. 2013,59,
83999−106.
(57) 840Kattge, J.; Diaz, S.; Lavorel, S.; Prentice, I. C.; Leadley, P.;
841Bönisch, G.; et al. TRY−a global database of plant traits. Glob. Change
842Biol. 2011,17 (9), 2905−2935.
Environmental Science & Technology Article
DOI: 10.1021/es505426z
Environ. Sci. Technol. XXXX, XXX, XXX−XXX
I
