Skip to main content
NCBI home page
Annals of Botany logoLink to Annals of Botany
. 2011 Feb 2;107(4):671–679. doi: 10.1093/aob/mcr007

Performance of dryland and wetland plant species on extensive green roofs

J Scott MacIvor 1,*, Melissa A Ranalli 2, Jeremy T Lundholm 1
PMCID: PMC3064539  PMID: 21292676

Abstract

Background and Aims

Green roofs are constructed ecosystems where plants perform valuable services, ameliorating the urban environment through roof temperature reductions and stormwater interception. Plant species differ in functional characteristics that alter ecosystem properties. Plant performance research on extensive green roofs has so far indicated that species adapted to dry conditions perform optimally. However, in moist, humid climates, species typical of wetter soils might have advantages over dryland species. In this study, survival, growth and the performance of thermal and stormwater capture functions of three pairs of dryland and wetland plant species were quantified using an extensive modular green roof system.

Methods

Seedlings of all six species were germinated in a greenhouse and planted into green roof modules with 6 cm of growing medium. There were 34 treatments consisting of each species in monoculture and all combinations of wet- and dryland species in a randomized block design. Performance measures were survival, vegetation cover and roof surface temperature recorded for each module over two growing seasons, water loss (an estimate of evapotranspiration) in 2007, and albedo and water capture in 2008.

Key Results

Over two seasons, dryland plants performed better than wetland plants, and increasing the number of dryland species in mixtures tended to improve functioning, although there was no clear effect of species or habitat group diversity. All species had survival rates >75 % after the first winter; however, dryland species had much greater cover, an important indicator of green roof performance. Sibbaldiopsis tridentata was the top performing species in monoculture, and was included in the best treatments.

Conclusions

Although dryland species outperformed wetland species, planting extensive green roofs with both groups decreased performance only slightly, while increasing diversity and possibly habitat value. This study provides further evidence that plant composition and diversity can influence green roof functions.

Keywords: Green roofs, ecosystem function, biodiversity, Sibbaldiopsis tridentata, Danthonia spicata, Empetrum nigrum, Kalmia polifolia, Scirpus cespitosus, Vaccinium macrocarpon

INTRODUCTION

Green roofs are being installed worldwide to mitigate the negative environmental and economic impacts of impervious rooftops in cities, which exacerbate problems associated with stormwater runoff, energy requirements for interior cooling and poor wildlife habitat (Getter and Rowe, 2006; Oberndorfer et al., 2007). Green roofs are comprised of vegetation and growing medium on top of root barrier and water retention membranes. Two main types of green roofs are recognized: intensive and extensive green roofs. Intensive systems have deep growing medium (>20 cm) permitting most plants adapted to local climatic conditions to thrive. Conversely, extensive green roofs (<20 cm growing medium) are lightweight, cheaper and require less material input, and thus permit many opportunities to retrofit existing buildings (Peck et al., 1999). Extensive green roofs are typically installed to provide specific services, with summer temperature reductions and stormwater capture being the most widely studied (Oberndorfer et al., 2007). Vegetation, growing medium and engineered components all contribute to the performance of these functions (Dunnett and Kingsbury, 2010), yet comparatively little work has been done to quantify the role of different plant species in green roof performance (Oberndorfer et al., 2007).

Due to shallow soils, the survival and growth of plants are key performance indicators on extensive green roofs (Boivin et al., 2001; Emilsson and Rolf, 2005; Monterusso et al., 2005; Rowe et al., 2005). Plant survival and coverage of the roof surface are important for aesthetics and for public acceptance of green roof technology (Dunnett et al., 2008). Further, canopy biomass is correlated with reflectivity, which reduces roof temperatures, and evapotranspiration, which is related to the amount of water the roof can capture during storms (Lundholm et al., 2010). Roof surface temperature is an important indicator of thermal performance (Leonard and Leonard, 2005; Liu and Minor, 2005; Lundholm et al., 2010), as lower roof temperatures (in the summer) lead to less heat flux into a building interior, resulting in energy savings from lower air conditioning costs (Liu and Minor, 2005). Albedo (reflectivity) is a key determinant of roof temperature, with more reflective vegetation correlated with lower surface temperatures (Lundholm et al., 2010). Water capture during rain events provides a direct measure of stormwater retention (Mentens et al., 2006; Carter and Jackson, 2007). Lastly, the rate of water loss from the growing medium after a rain event is related to both the ability of the roof to retain water in subsequent storms (Lundholm et al., 2010) and the potential for evaporative cooling of the roof surface due to combined transpiration and evaporation from the vegetation and soil surfaces (Del Barrio, 1998; Niachou et al., 2001; Lazzarin et al., 2005).

Extensive green roofs have shallow growing media, and so fluctuating soil moisture and drought are key constraints to plant survival and growth. Sedum species (Crassulaceae) are the most frequently used plants on extensive green roofs. They are low-growing succulents, usually with Crassulacean acid metabolism (CAM) capabilities, that thrive in moderate to extremely dry conditions (Koehler, 2003; Mentens et al., 2005; Snodgrass and Snodgrass, 2006). However, since both roof cooling and stormwater capture are related to plant transpiration rates (Theodosiou, 2003; Wolf and Lundholm, 2008; Lundholm et al., 2010) there is a potential trade-off between optimizing survival and growth, and high transpiration rates: plant species adapted to drought typically have conservative water use strategies, including low stomatal conductance rates (Korner et al., 1979), and overall less water uptake from the soil.

Research on green roof plant selection tends to focus solely on plants that prosper under dry conditions, and few studies have tested wetland plants on roofs. In some regions, particularly those along the east and west coasts of North America, rainfall can occur often and so vegetation with high evapotranspiration rates and/or those adapted to fluctuations in soil moisture content may be desirable. Wetland plants are highly plastic in their response to drought stress (Otte, 2001; Touchette et al., 2007), therefore some wetland species may be suitable for green roofs with shallow soils in high rainfall climates.

To optimize water loss in a region that experiences high annual rainfall, Compton and Whitlow (2006) compared water retention of a marshland species (Spartina alternifolia) and an old-field colonizer (Solidago canadensis) and found equivalent performance, though Spartina had higher water consumption rates, and Solidago had higher evapotranspiration rates. The authors suggest that marshland plants would survive better between rain events, and the old-field colonizers can handle storms at a greater frequency, concluding that different plants may be ideal depending on regional microclimate conditions. Thus, in regions with adequate rainfall, species from wetland and dryland habitats might survive and coexist on green roofs. In a greenhouse study, Wolf and Lundholm (2008) showed that species differed in water uptake rates such that the identity of the species with the ability to extract the most water from the growing medium changed over a gradient of wet to dry soil. They speculated that combining such species on a green roof could result in greater evapotranspiration over the entire growing season, since each species would experience optimal water uptake rates at some point in time. If different species use different resources, or use the same resources at different times, a more species-rich community is expected to use environmental resources most efficiently, leading to increases in the rates of some ecosystem processes (Petchley, 2003; Hooper et al., 2005).

A study on green roof water capture showed no advantage of species mixtures (Dunnett et al., 2008), but a later study with different species showed performance advantages in roof cooling, stormwater capture and evapotranspiration for combinations of plant life forms (Lundholm et al., 2010). Butler and Orians (2009) showed that Sedum species, which typically have low water uptake rates, facilitate the growth of grasses during dry conditions, possibly due to the ability of Sedum to reduce soil temperatures, or to prevent evaporation from the soil surface. While greater plant biodiversity can increase productivity and functioning within terrestrial ecosystems (Tilman and Lehman, 2002; Cardinale et al., 2007), there has been little examination of plant species, life form and functional diversity as determinants of ecosystem performance in constructed ecosystems such as green roofs (Ranalli and Lundholm, 2008). The objective of this study was to compare the performance of dryland and wetland species of broadly analogous life forms on extensive green roofs, and to determine whether combining species from each habitat type enables improved roof cooling and stormwater retention. Six species, three from each habitat type, were planted in monoculture and in combinations of one, two and three species from each of the two habitats to explore the effect of combinations of plants from different life forms and original habitats on green roof functioning.

MATERIALS AND METHODS

Study system

The study was conducted over two growing seasons (2007–2008) on top of the Patrick Power Library (Saint Mary's University) in Halifax, Nova Scotia (44 °39'N, 63 °35'W). The green roof site is described in detail in Lundholm et al. (2010). During the growing season (May–October), Halifax is characterized by daily maximum temperatures between 13 and 23 °C, daily minimum temperatures between 6 and 15 °C, and monthly precipitation values of 98–135 mm (Environment Canada, 2010). Halifax vegetation endures winters characterized by intermittent snow cover and, throughout November to April, the city typically experiences temperatures between –8·6 and 1·2 °C, and monthly precipitation values of 113·8–160·0 mm (Environment Canada, 2010).

A total of 190 free-draining plastic, square modules (Botanicals Nursery LLC, Wayland, MA, USA) were employed, each measuring 36 × 36 cm and representing a single sampling unit (Supplementary Data Fig. S1, available online). Modules were lined with conventional green roof root barrier and water retention layers (see Lundholm et al., 2010), topped with 6 cm of growing medium and then planted. A common green roof growing medium was used (‘Sopraflor X’, Soprema Inc., Drummondville, QC, Canada) which has a pH of 6·0–7·0 and consisted of crushed brick, blond peat, perlite, sand and vegetable compost. Weed barrier fabric (Quest Plastics Ltd, Mississauga, ON, Canada) was laid over the pre-existing rooftop grass (under the modules) to minimize any influence the grass might have had on the functions examined. These modules are used commercially as a green roofing system in the eastern USA.

Plant selection and planting

Coastal barrens in Nova Scotia are a shrub-dominated habitat characterized by high winds, areas of shallow substrate depth, variability in soil moisture content and an absence of tree cover (Oberndorfer and Lundholm, 2009), from which several species have been successfully grown on extensive green roofs (Ranalli, 2009; Lundholm et al., 2010). Both exposed bedrock and moist bog habitat exist throughout the barrens, each supporting a variety of plant species (Oberndorfer and Lundholm, 2009). Pairs of plants representing three life form groups (rhizomatous dwarf shrub, stoloniferous dwarf shrub and cespitose graminoid), with each pair comprised of one species found in wet areas and one found in dry areas (Oberndorfer and Lundholm, 2009), were examined in monoculture and in combination in extensive green roof conditions over two growing seasons (2007–2008). Dryland plants used were: Empetrum nigrum, a creeping stoloniferous shrub; Danthonia spicata, a cespitose graminoid; and Sibbaldiopsis tridentata, a creeping, rhizomatous shrub. Their wetland counterparts were: Vaccinium macrocarpon, a creeping stoloniferous shrub; Scirpus cespitosus, as the cespitose graminoid; and Kalmia polifolia, a rhizomatous shrub (Table 1). These wetland species are typically found in bogs within the coastal barrens (Oberndorfer, 2006). Empetrum nigrum has a fairly broad habitat tolerance, and is sometimes found in bogs as well, but dominates dry, rocky heathlands in Nova Scotia. Likewise, V. macrocarpon, while considered to be a wetland species, is often found in dry, rocky areas within the coastal barrens system. Our classification of these species is based on canonical correspondence analysis (CCA) showing that the three wetland species are more closely associated with wet substrate conditions than the other three species (Oberndorfer, 2006).

Table 1.

Coastal barren plant species used in this study

Habitat Species Family Common name Life form
Dry Sibbaldiopsis tridentata (Ait.) Rydb. Rosaceae Three-toothed cinquefoil Rhizomatous shrub
Danthonia spicata (L.) P. Beauv. ex Roem. & Schult. Poaceae Poverty grass Cespitose graminoid
Empetrum nigrum L. Empetraceae Black crowberry Stoloniferous shrub
Wet Kalmia polifolia Wangenh. Ericaceae Bog laurel Rhizomatous shrub
Scirpus cespitosus L. Cyperaceae Deer grass Cespitose graminoid
Vaccinium macrocarpon Aiton Ericaceae Large cranberry Stoloniferous shrub
Open in a new tab

Seeds were germinated indoors after being stored dry (dryland species) or in tap water (wetland species) at 4 °C for at least 12 weeks on a bed of Pro-Mix potting soil (Premier Horticulture, Riviere-du-Loup, QC, Canada). After germination, seedlings were transferred to plugs and grown in the Saint Mary's University greenhouse between summer 2006 and spring 2007. Some species had low germination rates, so plants with roots were transplanted from the field into plugs with the same potting soil and allowed to establish for at least 2 weeks prior to planting, which was at least 8 weeks prior to the collection of data. To control for differences in plant size within species, we planted a mix of both relatively large and small plants in all treatments with that particular species. Twenty-one plants were planted per module, which was consistent among all modules regardless of the number of species or habitat groups included in any given module. The module planting arrangement involved staggering four rows of four plants (on approx. 4·5 cm centres) and a centre row of five plants (on approx. 3·5 cm centres; Supplementary Data Fig. S2, available online). The planting sequence involved alternating species from different habitats (dryland and wetland), with life form grouping (stoloniferous and rhizomatous shrubs, and cespitose graminoids) and species being randomly chosen (without replacement) until all species to be included had been selected once, after which the same pattern was repeated throughout the module. By repeating the initial randomly chosen sequence, all species had an equal chance of interacting with, or being exposed to the other species included. Modules were planted between 5 and 19 June 2007 and watered by hand 3–6 times per week until 18 July 2007, after which modules received water exclusively through natural rain events, except 750 mL added during experimentation on 15 August 2008.

Experimental design

To examine differences in performance of plant species from dryland and wetland habitats, we created five replicates of each species in monoculture (one per block); ten replicates of all dryland or all wetland species combined; five replicates (one per block) of each combination of individual dryland and wetland species and each combination of two dryland and two wetland species; and ten replicates of all species in combination (Table 2). Additionally, ten growing medium-only modules (two per block) with only potting soil plugs (no plants) inserted into the substrate layer served as controls. To maintain the species composition of each treatment, all colonizing species not originally planted were removed twice a month during inspections of each module over each of the two growing seasons.

Table 2.

Species included in each treatment, and treatment performance (mean ± s.e.) for each function

Treatment Combination Replicates No. of species Plants Survival (%) Cover (%) Temp 2007 (°C) Temp 2008 (°C) Albedo (%) Water capture (%) Water loss (%)
C − 10 0 None − − 34·08 ± 0·46 31·22 ± 0·91 16·26 ± 0·36 67·0 ± 2·4 1·82 ± 0·04
D-1 a 5 1 ST 85·7 ± 3·5 66·2 ± 5·5* 33·48 ± 0·74 27·24 ± 0·99* 17·75 ± 0·45* 56·7 ± 4·1* 1·87 ± 0·05
D-1 b 5 1 DS 98·1 ± 3·5 96·2 ± 5·5* 32·16 ± 0·74 27·20 ± 0·99* 16·66 ± 0·45 61·1 ± 4·1 1·74 ± 0·05
D-1 c 5 1 EN 95·2 ± 4·0 25·0 ± 6·1 32·54 ± 0·80 29·59 ± 1·07 15·81 ± 0·49 60·0 ± 4·4 1·86 ± 0·06
D-2 a 5 2 ST, DS 98·1 ± 3·5 100·0 ± 5·5* 31·50 ± 0·74* 25·10 ± 0·99* 18·49 ± 0·45* 70·1 ± 4·1 1·80 ± 0·05
D-2 b 5 2 ST, EN 95·2 ± 3·5 73·7 ± 5·5* 31·86 ± 0·74* 26·10 ± 0·99* 18·01 ± 0·45* 64·8 ± 4·1 1·85 ± 0·05
D-2 c 5 2 DS, EN 99·0 ± 3·5 98·7 ± 5·5* 32·06 ± 0·74* 27·34 ± 0·99* 17·28 ± 0·45* 67·2 ± 4·1 1·74 ± 0·04
D-3 a 10 3 ST, DS, EN 90·5 ± 2·5 96·9 ± 3·9* 32·07 ± 0·60* 25·97 ± 0·81* 18·38 ± 0·37* 68·1 ± 3·4 1·82 ± 0·05
W-1 a 5 1 KP 95·2 ± 3·5 15·0 ± 5·5 34·14 ± 0·74 29·96 ± 0·99 16·31 ± 0·45 59·3 ± 4·1 1·97 ± 0·05*
W-1 b 5 1 SC 86·7 ± 3·5 41·2 ± 5·5* 33·54 ± 0·74 30·00 ± 0·99 16·58 ± 0·45 59·1 ± 4·1 1·83 ± 0·05
W-1 c 5 1 VM 92·3 ± 3·5 33·7 ± 5·5 33·12 ±0 .74 28·36 ± 0·99* 16·23 ± 0·45 59·2 ± 4·1 1·86 ± 0·06
W-2 a 5 2 SC, VM 85·7 ± 4·0 37·5 ± 6·1* 32·41 ± 0·80* 29·74 ± 1·07 16·42 ± 0·49 60·3 ± 4·4 1·84 ± 0·05
W-2 b 5 2 KP, VM 88·6 ± 3·5 22·5 ± 5·5 32·84 ± 0·74 29·22 ± 0·99* 16·36 ± 0·45 61·5 ± 4·1 1·83 ± 0·05
W-2 c 5 2 KP, SC 81·0 ± 3·5 30·0 ± 5·5 33·48 ± 0·74 30·18 ± 0·99 15·98 ± 0·45 49·3 ± 4·1* 1·85 ± 0·04
W-3 a 10 3 KP, SC, VM 77·1 ± 2·5 25·0 ± 3·9 34·03 ± 0·60 30·14 ± 0·81 16·19 ± 0·37 60·7 ± 3·4 1·79 ± 0·05
W + D-1 a 5 2 ST, KP 92·4 ± 3·5 67·5 ± 5·5* 32·26 ± 0·74* 27·48 ± 0·99* 19·12 ± 0·45* 67·2 ± 4·1 1·88 ± 0·05
W + D-1 b 5 2 DS, KP 91·4 ± 3·5 96·2 ± 5·5* 32·94 ± 0·74 28·10 ± 0·99* 17·40 ± 0·45* 65·7 ± 4·1 1·87 ± 0·06
W + D-1 c 5 2 EN, KP 90·5 ± 4·0 12·5 ± 6·1 34·32 ± 0·80 28·61 ± 1·07* 16·54 ± 0·49 54·5 ± 4·4* 1·86 ± 0·05
W + D-1 d 5 2 ST, SC 79·0 ± 3·5 67·5 ± 5·5* 33·16 ± 0·74 27·72 ± 0·99* 18·01 ± 0·45* 63·0 ± 4·1 1·87 ± 0·06
W + D-1 e 5 2 DS, SC 83·3 ± 4·0 100·0 ± 6·1* 32·71 ± 0·80 27·14 ± 1·07* 17·39 ± 0·49* 63·3 ± 4·4 1·77 ± 0·05
W + D-1 f 5 2 EN, SC 92·4 ± 3·5 28·8 ± 5·5 32·52 ± 0·74* 28·46 ± 0·99* 16·63 ± 0·45 62·0 ± 4·1 1·79 ± 0·05
W + D-1 g 5 2 ST, VM 94·3 ± 3·5 81·2 ± 5·5* 33·66 ± 0·74 26·64 ± 0·99* 18·41 ± 0·45* 62·0 ± 4·1 1·77 ± 0·05
W + D-1 h 5 2 DS, VM 95·2 ± 3·5 98·7 ± 5·5* 33·24 ± 0·74 26·86 ± 0·99* 17·87 ± 0·45* 73·5 ± 4·1 1·74 ± 0·05
W + D-1 i 5 2 EN, VM 99·0 ± 3·5 23·8 ± 5·5 33·04 ± 0·74 29·66 ± 0·99 16·12 ± 0·45 61·6 ± 4·1 1·81 ± 0·05
W + D-2 a 5 4 DS, EN, SC, VM 88·6 ± 3·5 71·2 ± 5·5* 32·88 ± 0·74 27·12 ± 0·99* 16·85 ± 0·45 59·5 ± 4·1 1·78 ± 0·05
W + D-2 b 5 4 DS, ST, KP, SC 94·3 ± 3·5 90·0 ± 5·5* 32·30 ± 0·74* 24·30 ± 0·99* 18·68 ± 0·45* 72·1 ± 4·1 1·82 ± 0·05
W + D-2 c 5 4 DS, ST, SC, VM 89·3 ± 4·0 92·2 ± 6·1* 32·79 ± 0·80 25·22 ± 1·07* 18·54 ± 0·49* 63·2 ± 4·4 1·75 ± 0·06
W + D-2 d 5 4 DS, ST, VM, KP 93·3 ± 3·5 95·0 ± 5·5* 32·60 ± 0·74* 25·66 ± 0·99* 18·15 ± 0·45* 64·9 ± 4·1 1·85 ± 0·05
W + D-2 e 5 4 DS, EN, KP, VM 90·4 ± 3·5 87·5 ± 5·5* 32·16 ± 0·74* 27·20 ± 0·99* 17·16 ± 0·45* 63·9 ± 4·1 1·76 ± 0·05
W + D-2 f 5 4 ST, EN, KP, VM 87·6 ± 3·5 60·0 ± 5·5* 33·62 ± 0·74 27·60 ± 0·99* 18·27 ± 0·45* 71·0 ± 4·1 1·80 ± 0·05
W + D-2 g 5 4 ST, EN, SC, VM 83·8 ± 3·5 76·2 ± 5·5* 33·04 ± 0·74 25·78 ± 0·99* 18·78 ± 0·45* 65·2 ± 4·1 1·82 ± 0·05
W + D-2 h 5 4 DS, EN, KP, SC 87·6 ± 3·5 88·7 ± 5·5* 34·20 ± 0·74 26·06 ± 0·99* 17·44 ± 0·45* 60·7 ± 4·1 1·82 ± 0·05
W + D-2 i 5 4 ST, EN, KP, SC 75·0 ± 4·0 60·9 ± 6·1* 33·38 ± 0·74 25·86 ± 0·99* 17·51 ± 0·45* 59·8 ± 4·1 1·83 ± 0·05
W + D-3 a 10 6 ST, DS, EN, KP, SC, VM 86·2 ± 2·5 85·0 ± 3·9* 32·76 ± 0·60* 26·91 ± 0·81* 17·67 ± 0·37* 59·4 ± 3·4* 1·83 ± 0·04
Open in a new tab

Survival and cover were recorded on May 29 2009. Temp 2007 represents July 2007 data, and Temp 2008 represents June 2008 data. Albedo and water capture were recorded in 2008, and water loss is the sum of the two (July and August) measurements from 2007. Species are coded as: ST, Sibbaldiopsis tridentata; DS, Danthonia spicata; EN, Empetrum nigrum; KP, Kalmia polifolia; SC, Scirpus cespitosus; VM, Vaccinium macrocarpon. The lettering in the Combination column indicates different within treatment plant combinations. The lettering is arbitrary, and repeated letters in different treatments are independent of one another. An asterisk (*) indicates a significant difference from the growing medium-only control (α = 0·05). The top ten treatments for each roof function are in bold.

A one-way, randomized complete block design was used with modules organized in five long, narrow blocks, each block being two modules wide. Blocks were oriented approximately North to South since the dominant sunlight and shadow gradient (from surrounding buildings) occurred along a West to East orientation across the site. To control for the effect of environmental variation within blocks on measured green roof functions, modules were randomly ordered within blocks and were also rotated within a block six times throughout the study.

Green roof performance measures

The percentage survival for each module was estimated at the end of May 2008, capturing which species survived over winter. Average survival was recorded for each treatment type. Final vegetation cover, used as an index of above-ground biomass, was estimated in each growing season, first during the week of 13–21 August 2007 and then on 25 June 2008, using the point interception method (Floyd and Anderson, 1987) with a three-dimensional metal pin frame (Domenico Ranalli, Regina, SK, Canada). The pin frame was custom designed to fit an individual module (36 × 36 cm), and was 30 cm high. The frame divides the module into 25 sub-plots using 16 equally spaced rods. Plant cover was estimated as the frequency of any vegetation contact with any of the pins (to a maximum of 16 out of 16).

Hand-held Taylor 9878 Slim-Line Pocket Thermometers (Commercial Solutions Inc., Edmonton, AB, Canada) were used to measure substrate surface temperature. Temperature readings were taken near the centre of modules, while in sunlight [between 1030 h and 1330 h Atlantic Standard Time (AST)] on 24 July, 29 August, 17 September 2007 and 25 June 2008.

The incoming and reflected solar radiation for each module was measured on 25 June 2008 under clear-sky conditions prior to solar noon (between 1030 h and 1230 h AST) when the sun is highest in the sky and the variability of incoming solar radiation is minimal (Sailor and Vasireddy, 2006). When measured, each module was moved away from the block layout and placed on top of grey coloured weed barrier fabric (Quest Plastics Ltd) to ensure that the grass on our study roof, as well as adjacent modules, did not influence the reflected values recorded. Measurements were made with a pair of fixed position LI-200SL LI-COR pyranometer sensors (LI-COR Biosciences, Lincoln, NE, USA), with the lowermost pyranometer located 25 cm from the ground. Albedo is expressed as: (reflected radiation/incoming radiation) × 100 %.

A PX-Series Checkweighing bench scale (ATRON Systems Inc., West Caldwell, NJ, USA) was used to weigh each module to determine the substrate water content on 15 August 2008. Each module was first weighed and then 750 mL of water was added to the centre using a watering jug to simulate a 5 mm rain event. Ten minutes after the addition of water, each module was weighed again. Within those first 10 min some water passed through modules and any remaining was considered to represent the amount of water captured (i.e. when the substrate was at field capacity).

Water loss is an indirect estimate of evapotranspiration and was calculated for each module between 23 and 25 July 2007 and 10 and 12 August 2007 as the difference between the initial module mass immediately following a rain event and the final module mass 72 h later. If it had rained again within the 72 h period, the data would have been discarded. The amount of time between the end of a rainfall and the beginning of a weighing event was not the same at each rain event (i.e. modules were weighed the morning following a rain, regardless of whether the rain had stopped the previous day or overnight). All modules were weighed within 1·5 h of each other for a single weighing event to minimize differences in weight (between modules) resulting from differences in time of measurement. Unlike the instantaneous nature of water capture estimates, measured water loss reflected the evaporation and transpiration that occurs over the 72 h period between weighing events.

Statistical analyses

In order to compare the performance between planted treatments containing different combinations of the wetland and dryland species with unplanted, growing medium-only controls, separate linear mixed effects models in R (v. 2·1·7, R Foundation for Statistical Computing, Vienna, Austria) were fit to surface temperature and water loss recorded in 2007, and albedo, surface temperature and water capture recorded in 2008. An analysis of variance (α = 0·05) for unbalanced designs was used to detect differences between treatments and the growing medium-only control, with the block variable treated as a random factor. The analysis was repeated on planting combinations grouped by functional type [wetland (W), dryland (D), wetland + dryland (W + D)] and the number of species planted (see Table 2). Treatment combinations were grouped to detect generalized differences or patterns in performance among wetland, dryland and combination treatments, as well as among planted diversity levels. All analyses were repeated once more with cover included as a covariate, since vegetation cover increases shading (Niachou et al., 2001) and trapping of low temperature air pockets below the canopy (Dimoudi and Nikolopolou, 2003), both of which impact green roof cooling and water retention through evapotranspiration.

RESULTS

Plant survival and cover

All treatments had >75 % survival after the first growing season, and there was no evident trend in survival related to functional type (wetland or dryland species) or diversity (number of species planted; Table 2). All modules containing dryland species had significantly higher cover than controls, except E. nigrum in monoculture or when combined with wetland species, while some treatments containing only wetland species were not different from the controls (with zero cover; Table 2).

Surface temperature

In 2007, mean surface temperature over the growing season was not significantly different when treatment was grouped by functional type and species combination (F9,180 = 0·66, P = 0·77) but cover was a significant effect when included as a covariate (F1,180 = 4·41, P < 0·05), with greater cover associated with lower surface temperatures. Interestingly, surface temperature recorded in July 2007 was significantly different by treatment (F9,180 = 2·76, P < 0·05), but the cover covariate had no effect, probably because most plants were still small, and had only been growing outdoors for a month. Block also had a significant influence on surface temperature in July 2007 (F4,185 = 5·87, P < 0·05). The greatest improvement over the growing medium control in 2007 occurred in July when combinations of two dryland species decreased surface temperature by an average of 2·27 °C, with the best mixture of two dryland species [D-2-a (S. tridentata, D. spicata)] decreasing surface temperature by 2·58 °C (Table 2). The best wetland-only treatment combination [W-2-a (S. cespitosus, V. macrocarpon)] decreased surface temperature by 1·67 °C and the best treatment containing both wet- and dryland species [W + D-1-a (S. tridentata, K. polifolia)] decreased surface temperature by 1·82 °C.

At the end of July 2008, surface temperature was not significantly influenced by treatment grouped by functional type and species combination (F9,180 = 1·51, P > 0·05), but cover had a very strong effect F1,180 = 23·53, P < 0·05), again with a negative influence on temperature. As found in Lundholm et al. (2010), increases in cover tend to result in lower roof surface temperatures, which might explain why dryland species kept the roof surface much cooler than wetland species. The block effect was also significant (F4,185 = 22·81, P < 0·05). Eight of the top ten performing treatments for this function contained S. tridentata (Table 2). The top performing treatment was 6·92 °C cooler than the growing medium-only controls and contained both dryland and wetland species [W + D-2-b (S. tridentata, D. spicata, K. polifolia, S. cespitosus); Table 2). Comparing July 2007 with June 2008, surface temperature varied more among treatments in the second season, but decreased relative to the controls, probably as a result of plants surviving the winter and having greater cover. In the second season, the most diverse dryland treatments [D-3 (S. tridentata, D. spicata, E. nigrum)] only decreased surface temperature by 1·23 °C over the best performing dryland species in monoculture (D. spicata), whereas the best performing dryland monoculture decreased surface temperature by 2·39 °C over the worst dryland monoculture (E. nigrum) (Table 2). Each wetland monoculture performed better than the most diverse wetland-only treatment. Eight of the top ten performing treatments were combinations of dryland and wetland species.

Albedo

Albedo differed significantly among treatments grouped by functional type and species combination (F9,180 = 2·17, P = 0·04), but even more so in relation to cover (F1,180 = 17·07, P < 0·05) when included as a covariate (albedo was positively correlated with cover). Overall, albedo tended to increase with increasing diversity of dryland plants, but decreased with increasing numbers of wetland species (Table 2). The block effect was also slightly significant (F4,185 = 4·74, P < 0·05). Combinations of all three dryland species (D-3) had the highest albedo values and combinations of two wetland species (W-2) had the lowest recorded albedo values. When the effect of cover was ignored, the D-3 combination increased albedo by 13 % compared with controls and by 10 % compared with modules containing a single dryland species (D-1). In modules containing both wetland and dryland species, performance was good overall, but modules containing two species from each habitat (four species total) outperformed modules containing a single species, or three species from each habitat (six species total). The best performing treatment [W + D-1-a (S. tridentata, K. polifolia)] reflected 15 % more than the growing medium-only control (Table 2). The top ten performing treatments all contained S. tridentata, which has glossy leaves, and was one of the most reflective species in MacIvor and Lundholm (2011), a study conducted at the same facility as that described here.

Water capture

In 2008, water capture was significantly different between treatments grouped by functional type and species combination (F9,180 = 2·32, P = 0·03) and the cover covariate was significant (F1,180 = 6·50, P = 0·01), which had a positive effect on water capture. The block effect was not a significant factor (F4,185 = 0·68, P = 0·62). Top performers were combinations of two and three dryland species (D-2, D-3), which were the only treatments to have greater water capture than the growing medium-only controls, possibly because dryland species generally exhibit higher evapotranspiration rates, as suggested by Compton and Whitlow (2006). Furthermore, evapotranspiration might have been higher in dryland species because of their greater biomass, as estimated by cover, which was a significant covariate and would increase surface area for stomatal conductance. Modules containing wetland species captured the least amount of water, and all modules that had significantly lower water capture compared with controls contained K. polifolia (Table 2).

Water loss

Mean water loss recorded in the first growing season did not differ significantly among treatments grouped by functional type and species combination (F9,180 = 1·17, P = 0·34), but was significantly affected by cover (F1,180 = 4·43, P > 0·05), which had a negative effect on module water loss. The block effect was significant (F4,185 = 11·58, P < 0·05). The average amount of water lost by all treatments in the July measurement period was 0·81 kg and this was 1·00 kg during the August measurement period. It is interesting to note that although some treatments with high cover values performed best in terms of maximizing water loss, many with little or no cover (controls) performed just as well or better than other treatments. This demonstrates the importance of growing medium in capturing and storing water, and that the effect of cover increases over time, with cover actually reducing water loss through evapotranspiration for some species. In both July and August 2007, modules with wetland species in monoculture (W-1) lost the most water of all treatment types over a 72 h period. While D. spicata in monoculture captured relatively little water, the treatments in which D. spicata was combined with other species tended to have the highest levels of water capture (Table 2).

DISCUSSION

After two growing seasons, dryland species performed green roof functions better than wetland species in extensive green roof modules. In general, monocultures and mixtures of dryland species performed best, followed by mixtures of dryland and wetland species, with monocultures and mixtures of only wetland species performing the poorest. In this study, there was no consistent effect of combining more species. Plant cover was important in reducing surface temperature and increasing reflectivity; however, it was less useful in determining which treatments were the best for water capture and loss. Adding both wetland species and dryland species tended to reduce overall roof performance slightly, thus it can be concluded that planting species adapted to dryland conditions will result in improved performance. Differences in performance between species and between life forms were much more obvious in the second growing season, probably due to increased growth and above-ground biomass, as described in Lundholm et al. (2010). A strong block effect was observed, with the easternmost blocks experiencing higher surface temperatures and greater water loss, both suspected to be as a result of longer daily exposure to direct sunlight due to shading from surrounding buildings, and indicative of the microclimatic influence on roof performance. It is expected that there was no block effect for water capture, and only a slight effect for albedo, because these variables were measured by removing each module from the block layout.

In the second season, the number of S. tridentata plants in each module increased substantially through rhizomatous growth. Danthonia spicata had a high growth rate in the first growing season, and in the second year a large amount of litter from the first season remained; however, only living plant material was included in cover assessments. Excess biomass production is undesirable on extensive green roofs because it increases potential fire risks, thus D. spicata is best suited to roofs that include a maintenance strategy. Empetrum nigrum had a slower growth rate than the other dryland creeping shrub (S. tridentata), and behaved more like the wetland species it was paired with, V. macrocarpon. Interestingly, in Oberndorfer (2006) these two species were most closely associated in the CCA classification.

It is unknown whether planting a variety of species that occupy different niches, rather than monocultures where plants are vying for the same resources, improved performance, or whether the inclusion of S. tridentata in extensive green roof plantings increased performance by some advantageous function. For example, through facilitation species such as S. tridentata might provide shade for more sensitive plant species (Pugnaire and Haase, 1996; Huston, 1997). Butler and Orians (2009) examined species' abilities to facilitate water uptake in other species on green roofs, and, as suggested in Ranalli (2009), certain species are thought possibly to play a significant role in sustaining biodiverse green roofs by retaining water in the growing medium longer when it is limiting. They advocate that these plant species are those with conservative water use strategies (such as Sedum species and other CAM species). Others suggest that plants with low mat-forming habits can prevent evaporation of water from the substrate (Wolf and Lundholm, 2008). Interspecies facilitation may provide an easy and low cost method to reduce the abiotic stress of a green roof system, which could expand the range of plants able to live in this habitat and, consequently, increase habitat value for insects and other invertebrates.

In the dryland species, the significant effect of cover on albedo and surface temperature provides evidence that stand architecture and plant biomass play an influential role in vegetation reflectivity and cooling functions (Oke, 1978; Diaz et al., 2005). For example, leaf orientation can affect albedo, with horizontal leaves being better at reducing radiation penetration compared with vertical leaves (Oke, 1978). This partially explains why monocultures of certain species such as S. tridentata and D. spicata had some of the highest albedo values. As found in Lundholm et al. (2010), exposed growing medium increased roof temperature and evaporation, while modules with high productivity had the coolest roof surface, probably through increased reflectivity, absorption and transpiration. Although a rooftop consisting of growing medium only may optimize water retention, planting species that have high evapotranspiration rates would also contribute to increased roof cooling while reducing soil moisture content, thereby increasing water retention on the roof surface at the next rain event. Clearly, plant canopy cover has an inconsistent influence on green roof water relations, and further research is required to tease apart the various features of the canopy that act in retaining water, and shedding it.

In this study, despite high survival rates, reduced cover values were documented for the wetland species, which resulted in a greater area of the growing medium being exposed and, therefore, low albedo and high surface temperature measures. In these treatments, more of the red-coloured brick that comprised a significant portion of the growing medium was exposed. Niachou et al. (2001) recorded higher green roof temperatures in areas of red vegetation or bare soil compared with green vegetation. Similarly, although mulch is not equivalent to plant material or brick, Decoteau et al. (1989) found that common red-coloured gardening mulch on top of a soil surface can decrease albedo and increase surface temperature. Therefore, evidence exists for why growing medium controls, modules containing wetland species in monoculture and mixtures of wetland species only performed equivalently in terms of albedo and surface temperature. With respect to water retention in the wetland species, Touchette et al. (2007) examined the response of five wetland plants to water withdrawal and found that all species had drought-tolerant strategies. However, because wetlands often experience only short periods of drought, many wetland species initially close their stomata in response to soil flooding to avoid uptake of phytotoxic compounds, such as reduced forms of Fe and Mn, that accumulate under flooded conditions and can be injurious to plants (Pezeshki, 2001). Thus wetland plants are less likely to take up large amounts of water during or immediately following a substantial rain event.

Overall, there were greater differences between dryland and wetland species groups than within groups. Interestingly, Wolf and Lundholm (2008) also found large differences in overall water loss among species within the same growth form, and suggested that generalizations about water loss in green roof systems should not be made based on growth form or cover values alone. In our study, growing medium-only controls were one of the top performing treatments for water capture and loss, and most modules performed similarly for water loss. Planting wetland species, particularly K. polifolia, led to higher water loss. Whereas a previous study using only dryland species found positive correlations between water loss and water capture (Lundholm et al., 2010), presumably because drier soils can hold more water in subsequent rain events, in this study wetland species had poor performance for water capture. Moreover, here we are comparing capture and loss from two different seasons. This requires further investigation into below-ground factors such as root biomass, which may have been higher in wetland species exposed to periodic drought (Miller and Zedler, 2003) and in shrubs such as K. polifolia (Crow and Wieder, 2005), reducing the amount of pore space in soil for water capture.

While this experiment was only conducted over two growing seasons, a different experiment that included two of the species here (E. nigrum and D. spicata) in the third growing season showed similar results for water capture and loss (Empetrum had higher loss but lower capture than other species, while Danthonia had both relatively low capture and loss) (Lundholm et al., 2010). The finding that D. spicata performed poorly for water capture when in monoculture but mixtures containing this species tended to perform well suggests some kind of interaction between this grass and the other species, or some kind of amelioration. In other studies, D. spicata shows lower water loss rates than growing medium-only controls (Wolf and Lundholm, 2008; Lundholm et al., 2010), suggesting that it prevents evaporation from the soil by shading the surface. In this study, water loss rates for this species were very low, although not significantly lower than controls.

Although the inclusion of wetland species on extensive green roofs might not necessarily improve stormwater and roof cooling benefits in the short term, survival was high, and perhaps the greater heterogeneity in plant structure and phenology would improve habitat value, as suggested in Dunnett and Kingsbury (2010). Persistence and cover are often more valued characteristics of extensive green roof vegetation than high above-ground biomass productivity, which can result in diminished soil nutrient resources and potential fire hazards (Sutton, 2008). Clearly the relationship between species from different habitats in combination with dryland species and the potential benefits on extensive green roofs should be examined in greater detail.

Interest in the role of biodiversity in improving green roof functioning is gaining momentum as more studies emerge that examine novel species and plant palettes. Dvorak and Volder (2010) even suggest that thousands of plants in North America remain untested, representing many species that might thrive under extensive green roof conditions within their respective ecoregions. It is clear from this study that the testing of plant species for inclusion on extensive green roofs should continue to focus on regional dryland plant communities, even though understanding the requirements necessary for wetland species on extensive green roofs remains valuable information for those involved in green roof design. Continued testing will lead to predicting changes within the plant community and in ecosystem functioning by quantifying plant traits over time within well-designed experimental studies (Lavorel and Garnier, 2002). Huston (1997) suggested that balance among species that have different growth and flowering times or tolerances for stressful conditions increases their permanence within a given habitat. This is highly desirable although much less understood on extensive green roofs. Understanding the biology and habitat requirements of different plant species, as well as the compatibility and stability of specific mixtures, could increase levels of plant richness that might be sustained on extensive green roofs over an extended period, thereby reducing maintenance and cost, while potentially increasing technical performance, habitat value for wild species, and compliance with climate change strategies in urban areas.

SUPPLEMENTARY DATA

Supplementary data are available online at aob.oxfordjournals.org and consist of the following figures. Figure S1. Planting scheme used in each of the extensive green roof modules. Figure S2. The extensive green roof modular system employed in the study.

Supplementary Data
supp_107_4_671__index.html (774B, html)

ACKNOWLEDGEMENTS

We thank Andre Thibault, Sarah Robinson, Zachary MacDougall, Adam Harris and Crystal Hillier for help collecting data and providing insight during the preparation of the manuscript. This work was supported by a National Science and Engineering Research Council Industrial Post-graduate scholarship (NSERC IPS) in collaboration with Elevated Landscape Technologies (first author), a National Science and Engineering Research Council Canadian Graduate Scholarship (NSERC CGS) (second author), and the Saint Mary's University Faculty of Graduate Studies.

LITERATURE CITED

  1. Boivin M, Lamy M, Gosselin A, Dansereau B. Effect of artificial substrate depth on freezing injury of six herbaceous perennials grown in a green roof system. HortTechnology. 2001;11:409–412. [Google Scholar]
  2. Butler C, Orians CM. Proceedings of the 7th North American Green Roof Conference: Greening Rooftops for Sustainable Communities. Atlanta, GA. Toronto: The Cardinal Group; 2009. Sedum facilitates the growth of neighboring plants on a green roof under water limited conditions. [Google Scholar]
  3. Cardinale BJ, Wright JP, Cadotte MW, et al. Impacts of plant diversity on biomass production increase through time because of species complementarity. Proceedings of the National Academy of Sciences, USA. 2007;104:18123–18128. doi: 10.1073/pnas.0709069104. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Carter T, Jackson CR. Vegetated roofs for stormwater management at multiple spatial scales. Landscape and Urban Planning. 2007;80:84–94. [Google Scholar]
  5. Compton JS, Whitlow TH. Proceedings of the 4th North American Green Roof Conference: Greening Rooftops for Sustainable Communities. Boston, MA. Toronto: The Cardinal Group; 2006. A zero discharge green roof system and species selection to optimize evapotranspiration and water retention. [Google Scholar]
  6. Crow SE, Wieder RK. Sources of CO2 emission from a northern peatland: root respiration, exudation, and decomposition. Ecology. 2005;86:1825–1834. [Google Scholar]
  7. Decoteau DR, Kasperbauer MJ, Hunt PG. Mulch surface color affects yield of fresh-market tomatoes. Journal of the American Society for Horticultural Science. 1989;114:216–219. [Google Scholar]
  8. Del Barrio EP. Analysis of the green roofs cooling potential in buildings. Energy and Buildings. 1998;27:179–193. [Google Scholar]
  9. Diaz S, Tilman D, Fargione J, et al. Biodiversity regulation of ecosystem services. In: Hassan R, Scholes R, Ash N, editors. Ecosystems and human well-being: current state and trends. Vol. 1. Washington: Island Press; 2005. pp. 297–329. [Google Scholar]
  10. Dimoudi A, Nikolopoulou N. Vegetation in the urban environment: microclimatic analysis and benefits. Energy and Buildings. 2003;35:69–76. [Google Scholar]
  11. Dunnett N, Kingsbury N. Planting green roofs and living walls. 2nd edn. Portland, OR: Timber Press; 2010. [Google Scholar]
  12. Dunnett N, Nagase A, Booth R, Grime P. Vegetation composition and structure significantly influence green roof performance. Urban Ecosystems. 2008;11:385–398. [Google Scholar]
  13. Dvorak B, Volder A. Green roof vegetation for North American ecoregions: a literature review. Landscape and Urban Planning. 2010;96:197–213. [Google Scholar]
  14. Emilsson T, Rolf K. Comparisons of establishment methods for extensive green roofs in southern Sweden. Urban Forestry and Urban Greening. 2005;3:103–111. [Google Scholar]
  15. Environment Canada. Canadian climate normals 1971–2000: Halifax Citadel, Nova Scotia. 2008 http://climate.weatheroffice.ec.gc.ca/climatenormals/stnselecte.html . (accessed October, 2008) [Google Scholar]
  16. Floyd DA, Anderson JE. A comparison of three methods for estimating plant cover. Journal of Ecology. 1987;75:221–228. [Google Scholar]
  17. Getter KL, Rowe DB. The role of green roofs in sustainable development. Hortscience. 2006;41:1276–1286. [Google Scholar]
  18. Hooper DU, Chapin FS, III, Ewel JJ, et al. Effects of biodiversity on ecosystem functioning: a consensus of current knowledge. Ecological Monographs. 2005;75:3–35. [Google Scholar]
  19. Huston MA. Hidden treatments in ecological experiments: re-evaluating the ecosystem function of biodiversity. Oecologia. 1997;110:449–460. doi: 10.1007/s004420050180. [DOI] [PubMed] [Google Scholar]
  20. Koehler M. Proceedings of the 1st North American Green Roof Conference: Greening rooftops for sustainable communities. Chicago, IL. Toronto: The Cardinal Group; 2003. Plant survival research and biodiversity: lessons from Europe. [Google Scholar]
  21. Korner CH, Scheel JA, Bauer H. Maximum leaf diffusive conductance in vascular plants. Photosynthetica. 1979;13:45–82. [Google Scholar]
  22. Lavorel S, Garnier E. Predicting changes in community composition and ecosystem functioning from plant traits: revisiting the Holy Grail. Functional Ecology. 2002;16:545–556. [Google Scholar]
  23. Lazzarin RM, Castellotti F, Busato F. Experimental measurements and numerical modeling of a green roof. Energy and Buildings. 2005;37:1260–1267. [Google Scholar]
  24. Leonard T, Leonard J. Proceedings of the 3rd North American Green Roof Conference: Greening rooftops for sustainable communities. Washington, DC. Toronto: The Cardinal Group; 2005. The green roof and energy performance – rooftop data analyzed. [Google Scholar]
  25. Liu K, Minor J. Proceedings of the 3rd North American Green Roof Conference: Greening rooftops for sustainable communities. Washington, DC. Toronto: The Cardinal Group; 2005. Performance evaluation of an extensive green roof. [Google Scholar]
  26. Lundholm JT, MacIvor JS, MacDougall JZ, Ranalli MA. Plant species and functional group combination affect green roof ecosystem functions. Plos One. 2010;5:e9677. doi: 10.1371/journal.pone.0009677. doi:10.1371/journal.pone.0009677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. MacIvor JS, Lundholm JT. Performance evaluation of native plants suited to extensive green roof conditions in a maritime climate. Ecological Engineering. 2011 (in press). doi: 10.1016/j.ecoleng.2010.10.004. [Google Scholar]
  28. Mentens J, Raes D, Hermy M. Green roofs as a tool for solving the rainwater runoff problem in the urbanized 21st century? Landscape and Urban Planning. 2006;77:217–226. [Google Scholar]
  29. Miller RC, Zedler JB. Responses of native and invasive wetland plants to hydroperiod and water depth. Plant Ecology. 2003;167:57–69. [Google Scholar]
  30. Monterusso MA, Rowe BD, Rugh CL. Establishment and persistence of Sedum spp. and native taxa for green roof applications. Hortscience. 2005;40:391–396. [Google Scholar]
  31. Niachou A, Papakonstantinou K, Santamouris M, Tsangrassoulis A, Mihalakakou G. Analysis of the green roof thermal properties and the investigation of its energy performance. Energy and Buildings. 2001;33:719–729. [Google Scholar]
  32. Oberndorfer E. Plant, macrolichen and moss community structure and species richness in the coastal barrens of Nova Scotia. Nova Scotia: Saint Mary's University; 2006. MSc Thesis. Halifax. [Google Scholar]
  33. Oberndorfer E, Lundholm JT. Species richness, abundance, rarity and environmental gradients in coastal barren vegetation. Biodiversity and Conservation. 2009;18:1523–1553. [Google Scholar]
  34. Oberndorfer E, Lundholm J, Bass B, et al. Green roofs as urban ecosystems: ecological structures, functions, and services. Bioscience. 2007;57:823–833. [Google Scholar]
  35. Oke TR. Boundary layer climates. London: Methuen & Co. Ltd; 1978. [Google Scholar]
  36. Otte ML. What is stress to a wetland plant? Environmental and Experimental Botany. 2001;46:195–202. [Google Scholar]
  37. Peck SW, Callaghan C, Kuhn ME, Bass B. Greenbacks from green roofs: forging a new industry in Canada. Ottawa: Canada Mortgage and Housing Corporation; 1999. [Google Scholar]
  38. Petchley OL. Integrating methods that investigate how complementarity influences ecosystem functioning. OIKOS. 2003;101:323–330. [Google Scholar]
  39. Pezeshki SR. Wetland plant responses to flooding. Environmental and Experimental Botany. 2001;46:299–312. [Google Scholar]
  40. Platt RH. Regreening the metropolis: pathways to more ecological cities. Annals of the New York Academy of Sciences. 2004;1023:49–61. doi: 10.1196/annals.1319.020. [DOI] [PubMed] [Google Scholar]
  41. Pugnaire FI, Haase P. Comparative physiology and growth of two perennial tussock grass species in a semi-arid environment. Annals of Botany. 1996;77:81–86. [Google Scholar]
  42. Ranalli M. Native plant evaluation and green roof performance: the influence of composition and richness on ecosystem functions. Halifax, Nova Scotia: Saint Mary's University; 2009. MSc thesis. [Google Scholar]
  43. Ranalli M, Lundholm J. Biodiversity and ecosystem function in constructed ecosystems. CAB Reviews: Perspectives in Agriculture, Veterinary Science, Nutrition and Natural Resources. 2008;3(034) http://www.cabi.org/cabreviews/default.aspx?site=167&page=1178 . [Google Scholar]
  44. Rowe B, Monterusso M, Rugh C. Proceedings of the 3rd North American Green Roof Conference: Greening rooftops for sustainable communities. Washington, DC. Toronto: The Cardinal Group; 2005. Evaluation of Sedum species and Michigan native taxa for green roof applications. [Google Scholar]
  45. Sailor DJ, Vasireddy C. Correcting aggregate energy consumption data to account for variability in local weather. Environmental Modelling & Software. 2006;21:733–738. [Google Scholar]
  46. Snodgrass EC, Snodgrass LL. Green roof plants: a resource and planting guide. Oregon: Timber Press; 2006. [Google Scholar]
  47. Sutton RK. Proceedings of the 6th North American Green Roof Conference: Greening rooftops for sustainable communities. Baltimore, MD. Toronto: The Cardinal Group; 2008. Media modification for native plant assemblages on green roofs. [Google Scholar]
  48. Theodosiou TG. Summer period analysis of the performance of a planted roof as a passive cooling technique. Energy and Buildings. 2003;30:271–277. [Google Scholar]
  49. Tilman D, Lehman C. Biodiversity, composition, and ecosystem processes: theory and concepts. In: Kinzig AP, Pacala SW, Tilman D, editors. The functional consequences of biodiversity: empirical progress and theoretical extensions. Princeton: Princeton University Press; 2002. pp. 9–41. [Google Scholar]
  50. Touchette BW, Iannacone LR, Turner GE, Frank AR. Drought tolerance versus drought avoidance: a comparison of plant–water relations in herbaceous wetland plants subjected to water withdrawal and repletion. Wetlands. 2007;27:656–667. [Google Scholar]
  51. Wolf D, Lundholm JT. Water uptake in green roof microcosms: effects of species and water availability. Ecological Engineering. 2008;33:179–186. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Data
supp_107_4_671__index.html (774B, html)
supp_mcr007_mcr007supp_figs.pdf (193.9KB, pdf)

Articles from Annals of Botany are provided here courtesy of Oxford University Press

RESOURCES