R

Ig

Aa

b

ARRA

KUGSPPS

1

trt2tOsDreGib

n

0h

Ecological Engineering 58 (2013) 156– 164

Contents lists available at SciVerse ScienceDirect

Ecological Engineering

j ourna l ho me pa g e: www.elsev ier .com/ locate /eco leng

esearch paper

nvestigation of weed phenology in an establishing semi-extensivereen roof

yako Nagasea,∗, Nigel Dunnettb,1, Min-Sung Choib,1

Chiba University, Graduate school of Engineering, Division of Design Science, 1-33 Yayoicho, Inage-ku, Chiba-shi, Chiba 263-8522, JapanUniversity of Sheffield, Department of Landscape, Arts Tower, Western Bank, Sheffield S10 2TN, UK

a r t i c l e i n f o

rticle history:eceived 24 November 2012eceived in revised form 7 April 2013ccepted 8 June 2013

eywords:rban landscaperavel mulchubstrate depthlant species diversitylanting density

a b s t r a c t

Although weeding is one of the most time consuming tasks in green roof maintenance, there have been fewstudies of weed phenology and it is not clear how planting design affects weed colonisation on green roofs.This study investigated weeds including self-sowing planted species during the establishment period ina semi-extensive green roof in Rotherham, UK. This green roof was installed in the summer of 2005,and 54 plant species were planted in 10 cm (areas with gravel mulch) and 20 cm (areas without mulch)of the substrate. The planting density was 18–22 plants/m2. Thirty-two quadrats (50 × 50 cm) were setup through the combinations of plant species diversity (high and low), planting density (high and low),four aspect and covering 2.5 cm of gravel mulch (with and without). Drip irrigation was installed forsupplementary watering in dry seasons. All weeds and self-sowing in each quadrat were not removed.The remainder of the roof was weeded six times in this period. Nine weed species were found on the green

elf-sowing roof. They were all native species and could have value of biodiversity. High planting density reducedweeds effectively whereas plant diversity did not affect weed colonisation significantly. Moreover, theuse of gravel mulch significantly reduced the number of weeds. Knowing phenology of expected weedsallows targeting maintenance to remove them before they set seeds. 29 species planted on this green roofwere self-sowing, Allium schoenoprasum, Campanula rotundifolia, Festuca spp. and Petrorhagia saxifragashowed a very high number of seedlings.

2othbatSeistc

. Introduction

Recently, the use of green roofs has received considerable atten-ion and the number of papers related to green roofs is increasingapidly. At present, the environmental benefits of green roofs seemo be most frequently studied (Berndtsson, 2010; Sailor and Hagos,011; Tonietto et al., 2011). Other examples of research topics arehe adaptation of plants to the green roof environment (Butler andrians, 2011; MacIvor and Lundholm, 2011; Rowe et al., 2012),

ubstrate composition for green roofs (Emilsson, 2008; Nagase andunnett, 2011; Molineux et al., 2009.) and social aspects of green

oofs (Francis and Lorimer, 2011; Kosareo and Ries, 2007). How-ver, there has been little research on green roof maintenance.

reen roof maintenance tends to rely on people’s experience, and

nformation on green roof maintenance is mainly obtained fromooks (Dunnett and Kingsbury, 2008; Weiler and Scholz-Barth,

∗ Corresponding author. Tel.: +81 043 290 3113; fax: +81 043 290 3121.E-mail addresses: a-nagase@faculty.chiba-u.jp (A. Nagase),

.dunnett@shef.ac.uk (N. Dunnett), mschoi2012@gmail.com (M.-S. Choi).1 Tel: +44 0114 222 0611; fax: +44 0114 275 4176.

ntm

rpsra

925-8574/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ecoleng.2013.06.007

© 2013 Elsevier B.V. All rights reserved.

009; Snodgrass and McIntyre, 2010). Lack of scientific evidencen green roof maintenance may result in inappropriate plant selec-ion and planting design and lead to a high maintenance cost. Itas been reported that the increased maintenance cost could be aarrier to the implementation of green roofs (Zhang et al., 2012)nd reduction of the maintenance cost may be an important initia-ive for installing green roofs. In some countries, such as Germany,witzerland, the United States of America and Japan, governmentsncourage the installation of green roofs through policy, direct andndirect regulation, and financial incentives, and funding of demon-tration or research projects (Carter and Fowler, 2008). However,hey tend to be limited to the encouragement of new green roofonstruction and not green roof maintenance. Moreover, mainte-ance may be a continuing concern for green roof owners becausehe maintenance cost may not be guaranteed for a long period in

any cases.Different types of green roofs may have different maintenance

equirement. Usually, a set of simple annual tasks such as plant

rotection, checking drainage and weeding is carried out for exten-ive and semi-extensive green roofs which are light-weight greenoof systems in which mainly herbaceous plants are used (Dunnettnd Kingsbury, 2008). Intensive green roofs, which have a thicker

Engineering 58 (2013) 156– 164 157

soiscHncamla

mwgnamtMaaadamvwwsd(

usaulaiasp2bvidaN

bqptt2apgacib

cBWptt

shiyfnotivigast

2

NMigobo

Tpawlh

A. Nagase et al. / Ecological

ubstrate and various types of plant species, require maintenanceperations such as the weeding of undesirable plants, fertilisation,nfilling bare spots (with cuttings, plugs or seeds), replacing erodedubstrate, pruning vegetation back from building structures andleaning plant debris from roof drains (Getter and Rowe, 2006).owever, it is important to note that the requirement of mainte-ance may depend on the planting design intent. For example, lawnover extensive roofs may require high maintenance to keep tidynd aesthetic. On the contrary, naturalistic intensive green roofsay allow weed colonisation to get more diversity benefits such as

ike a green roof in Ministry of land, infrastructure transportationnd tourism in Japan.

Weeding is one of the most important and time-consumingaintenance tasks for any type of green roof. In this paper,eeds are defined as spontaneously colonising plants without the

rower’s intention. Weeds may grow aggressively, compete forutrients and water and often shade out desirable plants (Weilernd Scholz-Barth, 2009). Weeds can be brought in with the growingedium, with the wind, by birds or through the shoes, clothing and

ools of people installing or maintaining the roof (Snodgrass andcIntyre, 2010). Weeds are also brought onto roofs with the plants,

nd as contaminants in any seed-mixes used. However, heightbove the ground in green roofs may exclude heavier weed seedsnd some herbivores and dispersers. Weeding is important for gar-en type of green roofs for healthy growth of planted plants and foresthetic. However, extensive green roofs, particularly those of theeadow type, may allow plant colonisation to have a fully covered

egetation layer. It is required to remove vigorous weeds such asillow, birch and buddleia regularly (Grant, 2006). These vigorouseeds tend to compete with desired plants for nutrients, water,

unlight, and other resources (Allaby, 2006). Their roots can alsoamage roof components such as the waterproofing membraneLuckett, 2009).

Weed control can be more efficient if urban ecosystems arenderstood. Although some studies have reported plant coloni-ation in green roofs (Dunnett et al., 2007; Köhler, 2006; Köhlernd Poll, 2010), there has been little research on the efficacy ofsing planting design to reduce weeds on green roofs. Seed estab-

ishment may be influenced by space, light, nutrient and moisturevailability. Three methods are typically used to reduce weed-ng in urban landscapes with the methods being (1) exclude lightt ground surface using sufficient height and density of plants,elf-mulching plants or mulch (Hitchmough, 1994), (2) Use highlant diversity (Cook-Patton and Bauerle, 2012; Lundholm et al.,010; Nagase and Dunnett, 2010, 2012), (3) remove parent plantsefore seed is physiologically capable of germination. Many pre-ious ecological studies have suggested that resident biodiversitys an important determinant of invasion success, arguing that highiversity increases the competitive environment of communitiesnd makes invasion more difficult (Funk et al., 2008; Levine, 2000;aeem et al., 2000; Zimdahl, 2004).

Some plant species, commonly planted for green roofs, coulde self-sowing; they disperse seeds freely and the seeds germinateuickly. In this paper, self-sowing is defined as seed dispersion fromlanted species. For example Allium spp. which was planted ini-ially self-sow well and this is one of the most dominant species onhe extensive green roof installed in 1985 in Berlin (Köhler and Poll,010). Self-sowing planted species can be valuable or problem-tic, depending on the situation (Hitchmough, 2004). In meadowlanting, self-sowing may be recommended to be planted to fillap, however, if the other plants are displaced, it could present

problem. Plant establishment from vigorous self-sowing seedsan be minimised by using species that do not produce viable seedn the region and cutting before the seed is physiologically capa-le of germination (Hitchmough, 1995). Several studies have been

<7m2

Fig. 1. Overview of green roof.

onducted to identify the plant species that are self-sown inritain on the ground (Clement and Foster, 1994; Hitchmough andoudstra, 1999). However, it is necessary to study self-sowing

lants on green roofs because the microclimate on a green roofends different from that on the ground (e.g. drought, extremeemperature, high light intensity and high wind speed).

This study investigated weeds, including self-sowing plantedpecies, in a semi-extensive green roof, case study in Rother-am, UK. Our first goal was to understand the dynamic changes

n the weeds and self-sowing plants on the green roof over aear. We aimed to suggest appropriate maintenance season andrequency, to reduce maintenance costs and to estimate mainte-ance costs more accurately and to enable more careful selectionf self-sowing plants. Our second goal was to identify which fac-ors (plant species diversity, plant density, mulch, substrate depth)n the planting design may affect weed colonisation. This pro-ides useful information because the planting design tends tonfluence weeds colonisation significantly after the installation ofreen roofs. If planting design influences weed composition, thenppropriate planting design could reduce maintenance costs. Plantpecies diversity, plant density and mulching were studied to testhis hypothesis.

. Methods

This experiment was conducted between February andovember 2006 on the fourth storey of a commercial building;oorgate Crofts Business Centre in Rotherham, North England (Lat-

tude: 53.433◦, Longitude: −1.356◦). An area of 415 m2 of accessiblereen roof was constructed on the roof (770 m2) in the summerf 2005. There was additional building on the third floor, and theuilding was surrounded by the green roof. A picture and floor planf the green roof are shown in Figs. 1 and 2, respectively.

The green roof consisted of a vapour control barrier (Hi-en Universal vapour barrier), 9 cm of insulation (Alumasc BGTolyurethane insulation), a waterproofing membrane (Derbigum),

root barrier (Preventol B2), a geotextile made of polypropyleneith fleece backing for green roof drainage (SSM45), a drainage

ayer (Floradrain FD 40), a filter sheet (SF) and a substrate (Zincoeather and lavender substrate: ≤15% of granules that were

0.063 mm in diameter, salt content ≤2.5%, total porosity 64%, pH.8, dry weight 940 kg/m3, saturated weight 1360 kg/m3, maxi-um water capacity 42%, air content at maximum water capacity

2%, water permeability ≥0.064 cm/s). The substrate analysis was

158 A. Nagase et al. / Ecological Engineering 58 (2013) 156– 164

cPeIImuusgAupiw25

irtvgHBLrPQh‘StuV

p

FO

aT

utt

s1f(n(wtrpm5FitRt1

Jsdhdcoopsai

2

Fig. 2. Floor plan and cross section of green roof.

arried out according to FLL guideline (Alumasc Exterior Buildingroducts, 2006; FLL, 2008). The green roof was divided into a semi-xtensive area, an alpine planting area and a low-edge area (Fig. 2).n the semi-extensive area, 20 cm of the substrate was installed.n the low edge area and the alpine planting area, 2.5 cm of gravel

ulch over 10 cm of the substrate was installed. In these areas,nderneath the substrate, 7.5 cm of crushed brick (Zincolit) wassed to make up the levels. These green roofs were different sub-trate depth but keep the same surface height. The cross section ofreen roof is shown in Fig. 2. All the materials were obtained fromlumasc (Northamptonshire, UK). Approximately half of the plantsed was native UK species and half were non-native species. Thelant list was made for three areas: semi-extensive, alpine plant-

ng and low-edge. The used plants were follows. Forbs and grassesere planted in summer 2005 and bulbs were planted in autumn

005. 43 species of forbs, 5 species of grasses and 6 species of bulbs,4 species in total were planted. The used plants were follows.

Forbs: Allium schoenoprasum, Armeria juniperifolia, Armeria mar-tima ‘Splendens’, Aster amellus, Calamintha nepeta, Campanulaotundifolia, Centaurea scabiosa, Dianthus deltoides, Erodium cilia-um, Erodium manescavii, Euphorbia cyparissias ‘Fens Ruby’, Galiumerum, Geranium cinereum ‘Ballerina’, Geranium endressii ‘War-rave Pink’, Geranium lucidum, Gypsophila repens ‘Dorothy Teacher’,elianthemum nummularium ‘Wisley Primrose’, Kniphofia ‘Borderallet’, Lavandula angustifolia ‘Hidcote’, Leucanthemum × superbum,imonium latifolium, Nepeta × faassenii, Origanum laevigatum ‘Her-enhausen’, Petrorhagia saxifraga, Phlox douglasii, Primula veris,ulsatilla vulgaris, Salvia × sylvestris ‘Blauhügel’, Sedum acre ‘Goldenueen’, Sedum album ‘Coral Carpet’, Sedum ‘Herbstfreude’, Sedumispanicum ‘Silver Carpet’, Sedum kamtschaticum var. floriferum

Weihenstephaner Gold’, Sedum reflexum, Sedum sexangulare,edum spathulifolium var. purpureum, Sedum spurium ‘Green Man-le’, Sedum telephium ‘Matrona’, Sempervivum arachnoideum, Sileneniflora, Sisyrinchium striatum, Stachys byzantina ‘Silver Carpet’ and

erbascum phoeniceum.

Grasses: Festuca amethystina, Festuca glauca, Helicotrichon sem-ervirens, Melica ciliate and Stipa tenuissima.

hs

ig. 3. Mean monthly temperature and rainfall change in 2006 in Sheffield (Metffice, 2007).

Bulbs: Allium caeruleum, Allium karataviense, Crocus tommasini-nus ‘Whitewell Purple’ Muscari armeniacum, Tulipa tarda andulipa praestans ‘Fusilier’.

Two plant species, H. sempervirens and K. ‘Border Ballet’ weresed as accents in the planting design. Two > 100 m flower heightall species were planted in 12 spots (Kniphofia) and 4 spots (Helic-otrichon), respectively (Fig. 2).

The planting density was approximately 18 plants/m2 in theemi-extensive area, 22 plants/m2 in the alpine plant area and8 plants/m2 in the low edge area. The plants were obtainedrom Chapel Cottage Plants (Cambridgeshire, UK), Van DogeweerdLincolnshire, UK), Barbara Austin Perennials (Wiltshire, UK), Ged-ey Bulbs (Lincolnshire, UK) and Mike Handyside WildflowersCheshire, UK). The size of pot was 9 cm. A drip irrigation systemas installed on the establishment of the green roof in 2005 and

his was used once a week in June and July 2006 because of lowainfall and high temperatures. Except for the above mentionederiod, no irrigation was applied during this study. Maximum andinimum temperatures and precipitation at Sheffield (latitude:

3.383◦; longitude: −1.483◦) for each month in 2006 are shown inig. 3. Weather information for Sheffield is shown because weathernformation for Rotherham was not available. Sheffield is locatedo 15 km away from Rotherham (South West) and same altitude tootherham. The climate was mild; the mean maximum tempera-ure was 25.6 ◦C in July and the mean minimum temperature was.4 ◦C in February.

The quadrats (50 × 50 cm) were set up on the green roof inanuary 2006. There were four combinations of plant species diver-ity and planting density (1) low species diversity and low plantingensity; (2) low species diversity and high planting density; (3)igh species diversity and low planting density and (4) high speciesiversity and high planting density. These four combinations werehosen in areas with mulch (substrate depth of 10 cm) and with-ut mulch (substrate depth of 20 cm). There were four replicationsf each direction (NE, NW, SE and SW) so that 32 quadrats werelaced. When the positions of the quadrats were being determined,everal positions were tried and the positions that fulfilled thebove criteria were chosen. The positions of the quadrats are shownn Fig. 2.

.1. Plant species diversity

The number of plant species was between five and six inigh diversity quadrats and between two and four in low den-ity quadrats when the quadrats were set up. Overall, the quadrats

A. Nagase et al. / Ecological Engin

Fb

wpbqnpFsia

2

hsugm3ptte

Fr

2

wsbwwJ

2

emsdo

3

wrESsseptsolib

ao

ig. 4. Change in the mean number of plant species per quadrat over time. Errorars represent standard errors.

ere set up to capture the most of variety of species which werelanted on the roof at the time of installation. However, in springulbs and some species had not appeared at the time of settinguadrats, emerged in some low density quadrats, increasing theumber of species in these quadrats. Changes in the number oflant species (planted species) per quadrat over time are shown inig. 4. In high diversity quadrats, the number of plant species wastable; however, in low density quadrats, the number of speciesncreased after March because some species such as bulb speciesppeared in spring.

.2. Planting density

The percentage of coverage and height (from the bottom to theighest leaf apex) of each planted plant in each quadrat was mea-ured in each month. The percentage of coverage was measuredsing the 50 × 50 cm quadrat, which was divided into a 5 × 5 cmrid to give a total of 100 squares. High coverage was defined asore than 50% plant cover, and low coverage defined as less than

0% plant cover in January 2006.Changes in the mean coverage of

lanted species per quadrat over time are shown in Fig. 5. Some-imes, plant coverage was more than 100% because the figure washe sum of plant coverage of each species and plants overlappedach other as they grew.

ig. 5. Change in the mean plant coverage (planted species) over time. Error barsepresent standard errors.

tnbso

wbitaa

pchldatoidw

eering 58 (2013) 156– 164 159

.3. Measurement

The number of weeds and self-sowing plants in each quadratere counted monthly between Feb and Nov 2006. If the plant

pecies were too small to identify, they were left until they coulde identified. All weeds and self-sowing plants in each quadratere not removed during the study. The remainder of the roof waseeded six times in this period (late February and on 19 May, 11

uly, 5 August, 11 September, and 12 October in 2006).

.4. Statistical analysis

One-way analysis of variance (ANOVA) was used to detect theffects of different plant species diversity, planting density andulch (with and without) on the number of weeds. Linear regres-

ion was carried out to identify the relationship between plantingensity and number of weeds. All statistical analysis was carriedut using Minitab Release 14 software.

. Results

During the 9 months of this study, only nine plant weed speciesere found to have colonised in a total of 32 quadrats on the green

oof. These species were Cardamine hirsuta, Cerastium fontanum,pilobium montanum, Geranium molle, Picris echioides, Poa annua,enecio jacobaea, Sonchus oleraceus and Taraxacum officinale. Thesepecies are commonly observed in gardens and the urban land-cape in the UK and no invasive weeds were observed in thisstablishment period. Their life form, habitat, phenology and dis-ersal are summarised in Table 1. All these species are nativeo the UK. Half of the species are annuals, and the remainingpecies are perennials. Although their habitats are a wide rangef places, the similarity is their habitats tend to be dry-openands. Many plant species exist as rosettes until growth conditionsmprove, and they tend to overwinter. Most species are dispersedy wind.

Tracking changes in numbers of individual weeds show over-ll numbers dropped after August but individual species had theirwn peaks (Table 2). C. hirsuta, E. montanum and S. jacobaea werehe most commonly found species on this green roof. The highestumber of C. hirsuta were found in winter and spring, and the num-er dramatically declined in summer. E. montanum and S. jacobaeahowed higher numbers in early summer. In contrast, the numbersf G. molle and P. echioides increased in autumn.

The effect of plant species diversity on the mean frequency ofeeds per quadrat over time is shown in Fig. 6. Overall, larger num-

ers of weeds were observed in the high diversity quadrats thann the low-diversity quadrats. However, the difference was statis-ically significant only in February. The difference between high-nd low-diversity quadrats in the number of weeds became smallfter August when the number of weeds decreased.

The effect of plant density on the mean frequency of weedser quadrat over time is shown in Fig. 7. Similar changes in weedolonisation were observed in high- and low-density quadrats;owever, the number of colonising weeds was higher in the

ow-density quadrats than the high-density quadrats. The effect ofensity was statistically significant in April, May, August, Octobernd November. In addition, there was a significant negative rela-ionship between the number of colonising weeds (mean number

f weeds in low-density quadrats and high density quadratsn each month) and planting density (y = 113–2.90 x, R2 = 51.2%,f = 1,18, F = 18.87, P < 0.01) (Fig. 8). This result suggests that lesseeds colonised in high-density quadrats. Even in low density, it

160 A. Nagase et al. / Ecological Engineering 58 (2013) 156– 164

Table 1Characteristics of weeds which were found on the green roof.

Life form Habitat Phenology Dispersal

Cardamine hirsuta Annual, sometimes biennials Well-drained sandy orcalcareous soils

Pants exist as rosettes of pinnate-compound leavesuntil flowering stems develop at maturity; Flowersfrom January to June

Ballistic

Cerastium fontanum Perennial All types of unshaded dry landhabitats

Winter green. Flowers from April to September.Seed shed from June onwards

Animal, wind

Epilobium montanum Perennial Widely distributed on rocky,disturbed shaded ground

Overwintering as short above- or below-groundstolons. Flowers from June to August, Set seedsfrom June to September

Wind

Geranium molle Winter or more rarely,summer-annual

Mainly recorded fromlimestone outcrops anddistributed in wide habitatswith bare ground such aswasteland and pasture.

Variable, depending on the season of germination.Autumn-germinating plants overwinter asrosettes. Flowers within the period from April toSeptember. Sets seeds from June to October.

Ballistic

Picris echioides Annual or biennial Waysides, hedgebanks, fieldmargins, rough places and cost

It is an autumn- or spring-germinating plant.Spring/summer flowering but can flower at anytime of the year.

Wind

Poa annua Short-lived perennial Occurs in a great variety ofdisturbed situation, but mostcommon on arable land anddisturbed fertile soil

Leaves, flowers and fruits may be found during allseasons. Most typically summer annual but canbehave as winter annual in droughted habitats.

Animal, wind

Senecio jacobaea Perennial Widely dispersed, particularlyin rocky habitat, but restrictedto habitats with at least a littlebare ground.

Seeds germinate mainly in autumn and seedlingsoverwinter in a leafy condition. Some seedsgerminate in spring. Ripe seeds area dispersedfrom August until winter.

Wind

Sonchus oleraceus Winter or summer annual Frequently recorded fromdemolition sites andwidespread on disturbedplaces.

Autumn-germinating plants overwinter asrosettes, spring-germinating plants overwinter asachenes. Flowers and fruits from May onwards inautumn-germinating plants and June onwards inspring-germinating plants

Wind

Taraxacum officinale Perennial A common constituent of allbut except aquatic habitat.

A small rosette of leaves overwinters. Flowers fromMay to October but mainly from April to June.

Wind

A

wp

awnrwJ

wfirvda

4

eqwgnemTi

TT

dapted from: Klingman et al. (1982) and Grime et al. (1988).

as possible to reduce the number of weeds significantly after thelant cover reached over 50%.

The effect of the combination of mulch (with and without)nd substrate depth (10 and 20 cm) on the mean frequency ofeeds per quadrat over time is shown in Fig. 9. The combi-ation of gravel mulch and shallow substrate was effective ineducing weed invasion. Differences were greatest in monthshen high weed number was observed (February, March and

une).During the experiment, 29 species which showed self-seeding

ere found in 32 quadrats. The total number of self-seeded plantsound in 32 quadrats over time is shown in Table 3. This resultndicates that the seeds of many plant species used on this green

oof are not dormant and they germinate easily. In particular,ery high numbers of seedlings of A. schoenoprasum, C. rotun-ifolia, Festuca spp. and P. saxifraga occurred after flowering inutumn.

oatw

able 2otal number of individual weeds over time (total of 32 quadrats).

February March April May Jun

Cardamine hirsuta 478 442 448 306 224Cerastium fontanum 0 18 6 10 9Epilobium montanum 40 42 59 68 148Geranium molle 0 0 0 0 0Picris echioides 1 3 1 2 41Poa annua 1 5 2 4 3Senecio jacobaea 118 120 137 144 123Sonchus oleraceus 2 9 8 4 3Taraxacum officinale 1 0 0 0 0

Total 641 639 661 538 551

Most seeds set from May to June.

. Discussion

A limited number of weeds (nine species), was found on thextensive green roof over a period of nine months. Except for theuadrat areas, weeding was performed six times a year and someeed species were removed before they set seeds. This result sug-

ests that six times weeding per year was efficient to keep lowumber of weed colonisation during the establishment for semi-xtensive green roofs in North England. Six times weeding per yearay be relatively high maintenance for semi-extensive green roofs.

he maintenance requirement depends on how much tidiness ismportant for the green roof.

The low water-holding capacity of green roof means the period

f drought stress for weeds is more severe and more frequent thant ground level (i.e. the window of opportunity for seed germina-ion is shorter, and many plants will die before setting seeds). Mosteeds in the quadrats did not flower; therefore, they germinated

e July August September October November

159 35 23 16 5 3 1 0 11 0

132 20 30 34 44 0 0 12 11 10

39 33 46 66 54 3 5 1 3 1

67 35 101 46 69 3 1 0 10 1

0 0 1 0 2

406 130 214 197 186

A. Nagase et al. / Ecological Engineering 58 (2013) 156– 164 161

Table 3Total number of self-sowing over time (total number of 32 quadrats).

February March April May June July August September October November

Allium schoenoprasum 0 0 21 36 18 22 569 10,300 10,400 3052Aster amellus 3 0 0 0 22 25 24 15 28 32Calamintha nepeta 0 0 3 0 0 4 113 87 65 103Campanula rotundifolia 1 0 0 10 9 9 19 34 240 2492Dianthus deltoids 9 24 35 80 107 109 93 55 49 56Erodium manescavii 25 29 55 80 76 106 78 119 149 173Euphorbia cyparissias ‘Fens Ruby’ 0 0 15 55 50 50 51 60 39 53Festuca spp. 397 840 622 1210 445 460 398 628 3019 3365Galium verum 0 0 0 7 2 2 6 8 0 1Geranium endressii ‘Wargrave Pink’ 0 0 0 0 0 0 0 1 0 1Gypsophila repens ‘Dorothy Teacher’ 0 0 0 0 11 3 0 0 0 0Kniphofia ‘Border Ballet’ 0 0 0 0 0 0 1 4 101 167Leucanthemum × superbum 3 0 0 0 17 24 18 41 53 50Melica ciliate 0 0 0 0 0 0 0 0 1 540Nepeta × faassenii 0 0 0 0 0 0 0 0 1 0Origanum laevigatum ‘Herrenhausen’ 0 1 0 0 0 1 10 11 87 202Petrorhagia saxifrage 0 12 46 47 20 30 35 15,809 11,082 6407Primula veris 0 2 1 0 0 0 0 0 0 0Pulsatilla vulgaris 0 0 0 0 1 1 0 0 0 0Sedum acre ‘Golden Queen’ 10 4 21 28 22 21 39 51 25 24Sedum album ‘Coral Carpet’ 6 11 14 18 11 11 11 15 11 7Sedum hispanicum ‘Silver Carpet’ 3 4 11 27 25 16 15 14 13 14Sedum kamtschaticum var. floriferum

‘Weihenstehaner Gold’0 0 0 0 1 1 0 0 0 0

Sedum sexangulare 0 4 12 23 3 3 4 3 7 7Sempervivum arachnoideum 0 0 2 2 3 3 1 1 0 0Silene uniflora 0 1 0 1 10 5 1 0 0 5Stachys byzantina ‘Silver Carpet’ 0 0 2 2 2 5 5 32 41 46Stipa tenuissima 117 356 145 216 57 72 80 70 80 527

0

16

gwosis

glFWh

Fc

mcotcsThn

Verbascum phoeniceum 0 0 0

Total number of plants 10 12 15

radually or their seeds may have continued to enter the roof. Thereas one exception; many flowers and seed heads of C. hirsuta were

bserved. C. hirsuta may spread its seeds by itself on the roof. Somepecies would have been transported initially by wind, and oncen situ can produce large seed crops which may remain viable foreveral season (Archibold and Wagner, 2007).

It is important to note that this study is particularly relevant toreen roofs at establishing stage. Weed species on this green roof isikely to change over time; some species may disappear over a year.

or example, C. hirsuta is a typical pioneer plant (Roxburgh andilson, 2003); they tend to colonise and spread in the beginning,

owever, they usually disappear in later. Therefore, this species

ig. 6. Effects of diversity on the number of weeds per quadrat over time. Values areompared within the same month. *P < 0.05. Error bars represent standard errors.

yTa

Fc

1 1 1 2 0 1

22 24 22 22 20 22

ay not exist for a long time. Dunnett et al. (2007) studied plantolonisation on extensive green roofs in Sheffield after six yearsf instalment and showed 35 colonising species were identified inotal. S. jacobaea, E. montanum, S. oleraceus and T. officinale wereommon plant species which could be found in their study and thistudy. These species may stay be present over a number of years.he number of weeds on this green roof may increase over time,owever, the number of plant species varied from year to year witho apparent significant tendency according to roof are over a 20-

ear period on two extensive green roofs in Berlin (Köhler, 2006).his study was conducted in a newly established green roof over

period of 9 months, further research is necessary to understand

ig. 7. Effects of density on the number of weeds per quadrat over time. Values areompared within the same month. *P < 0.05. Error bars represent standard errors.

162 A. Nagase et al. / Ecological Engin

F(c

tt

arcdfdcfpydifhfb

A

Fos

tEtfrabntgnTtr

tblDohsibrawtwtiWemsta

r

ig. 8. The relationship between number of colonising weeds and planting densityy = 113–2.90x, R2 = 51.2%, df = 1, 18, F = 18.87, P < 0.01). Number of weeds and plantoverage are in low-density quadrats and high-density quadrats in each month.

he dynamic changes that occur in weed populations in the longererm.

This green roof studied in the present study was installed foresthetic reasons and, except for the quadrat areas, weeds wereemoved regularly. However, notably, some weeds have a highonservation value, and selective weeding is recommended if bio-iversity is important; the capacity of green roofs to act as habitatsor colonising species confers substantial benefits for urban bio-iversity (Francis and Lorimer, 2011). Weeds may contribute toreate diverse plant structure and microhabitats. All the weedsound in this study were UK native plant species. The use of nativelants on green roofs has attracted considerable attention in recentears (Butler et al., 2012). Native plants are frequently used in bio-iverse green roofs to create habitats and provide resources for

nvertebrates (Brenneisen, 2006; Kadas, 2006). Some of the weeds,ound in this study, are host plants for butterflies. For example, C.irsuta is a host for Pieris napi, G. molle for Aricia agestis and P. annuaor Thymelicus acteon. In addition, P. echioides and T. officinale can

e an important nectar source (Dennis, 2010).

A large number of weeds was identified until July; however afterugust, the number decreased significantly. This result is related to

ig. 9. Effects of substrate depth and mulch on the number of weeds per quadratver time. Values are compared within the same month. *P < 0.05. Error bars repre-ent standard errors.

wsmppnpatcio

iossfecArrget

eering 58 (2013) 156– 164

he plant phenology of the three dominant weed species; C. hirsuta,. montanum and S. jacobaea. Their germination rate decreased, andhey began to die after August. This result suggests that weedingrom spring to early summer is particularly important for this greenoof because weeds must always be removed before they flowernd set seeds (Weiler and Scholz-Barth, 2009). After August, it maye possible to reduce the frequency of maintenance. Notably, theumber of S. jacobaea plants increased again in September becausehis species germinates in autumn as well as spring. If an autumn-erminating plant species such as S. jacobaea is the most domi-ant species on a roof, weed management in autumn is essential.herefore, weed management should be planned after studyinghe phenology of the dominant weeds e.g. seed germination andeproductive biology.

In previous ecological studies, plant species diversity was foundo confer resistance to invasion because more diverse assem-lages utilise the available resources more completely, leaving

ittle resource space for individuals of new species (Levine and’Antonio, 1999). In this study, however, an opposite tendency wasbserved in that the larger number of weeds colonised areas ofigh species diversity, although there was a significant effect ofpecies diversity only in February. Indeed, recent theoretical stud-es have consistently supported the predicted negative relationshipetween plant species diversity and invasiveness, although theesults of empirical studies have been decidedly mixed (Levinend D’Antonio, 1999). Understanding of how invasiveness variesith plant species diversity is complicated by the fact that varia-

ion in plant species diversity is controlled by, and thus covariesith disturbance, resource availability, physical stress, competi-

ors, consumers, etc., the factors that are also known to influencenvasiveness (Rejmánek, 1989, Huston, 1994, Robinson et al., 1995,

iser et al., 1998). Researchers have recently started to applycological methods, largely developed in non-urban locations, toetropolitan regions (Steiner, 2011). Further research is neces-

ary to determine how increasing plant diversity could improvehe short- and long-term functioning of green roofs (Cook-Pattonnd Bauerle, 2012).

A higher planting density (coverage of more than 50%) couldeduce weed invasion significantly. This is supported by Rao (1999)ho suggested that increasing planting density by using a higher

eedling rate and narrower planting spacing is an important weedanagement technique as it enhances the competitiveness of

lanted species by suppressing or smothering weeds. Larger plantopulations create shading, which prevents weed seed germi-ation, emergence and establishment. However, a higher plantopulation is dependent on growth habit, leaf orientation, durationnd other characteristics. Indeed, the results of this study suggestedhat weed invasion is affected not only by a high plant cover per-entage, but also by factors such as plant structure. Further researchs necessary to study how plant characteristics, including morphol-gy, affect weed invasion.

As hypothesised, it was shown that gravel mulch reduced weednvasion. The area with gravel mulch was shallower than that with-ut gravel mulch; this may also have helped to reduce weed inva-ion because a shallow substrate is generally drier than a deep sub-trate. In this green roof, white gravel mulch was particularly usefulor interest in winter, when many plants disappeared. However, theffectiveness of mulches in suppressing weeds may depend on theharacteristics of the mulch, including depth, texture and colour.ccording to Hitchmough (2004), the most effective mulches forestricting colonisation through weed seed rain are synthetic mate-

ials such as polythene woven weedmats, followed by very coarserades of bark, wood chips, and coarse mineral aggregates. It is nec-ssary to consider the effectiveness and cost of mulch as well ashe design of green roofs. However, the abovementioned research

Engin

wctd

ootdNigsaceiost

5

weoipsswpsastcuAmlpawtwtu

A

Rgpf

R

A

A

A

B

B

B

B

C

C

C

D

D

D

E

F

F

F

G

GG

HH

H

H

H

K

K

K

K

K

L

L

L

L

M

M

M

N

N

A. Nagase et al. / Ecological

as carried out on the ground; textile-type mulches may be unac-eptable on roofs where access is difficult or winds could blowhe material away. Further research is necessary to determine howifferent mulches affect weed invasion on green roofs.

Twenty-nine plant species out of the 54 planted species plantedn the green roof in the present study were self-sowing. The seedsf many plant species used on this green roof are not dormant andhey germinate easily.Other planted species may be self-sown butid not germinate because they required winter chilling/dormancy.ot only the number of self-sowing plants but also their survivabil-

ty, establishment and growth rate may affect the performance ofreen roofs. It was observed that self-sown E. manescavii, E. cyparis-ias ‘Fens Ruby’ and Festuca spp. established well and grew rapidlynd these species could be invasive. Since a high planting densityould reduce weed invasion significantly, desirable plants that gen-rate seedling is another strategy to reduce bare surface area, ass selecting plant species that grow quickly. In contrast, seedlingsf some species such as O. laevigatum ‘Herrenhausen’ and Sedumpp. grew slowly, and these species tend to remain small in size;herefore, they are less likely to disturb other planting.

. Conclusion

In this study, it was shown that six times weeding per yearas efficient to keep low number of weed colonisation during the

stablishment in semi-extensive green roofs. Knowing phenologyf expected weeds (especially dominant weeds) allows target-ng maintenance to remove them before they set seeds. Someerennials that are commonly planted for green roofs are self-owing. Vigorous self-sowing seeds can be minimised, however,elf-sowing plants may be useful to fill gap and reduce number ofeeds. It was possible to reduce weed colonisation using an appro-riate planting design such as a higher planting density and shallowubstrate with 2.5 cm depth of mulch. If the results described herere applied to the initial planting design and weed phenology istudied after the installation of green roofs, it should be possibleo develop low maintenance green roofs, thus reduce maintenanceosts. In this study, a practical green roof was used for improvednderstanding of the phenology of weeds and self-sowing plants.lthough the study benefitted from high replication and a com-ercial maintenance regime, the commercial design of the roof

imited the power of the experimental design; the differences inlant diversity and density were small and the gravel was onlypplied to the thin substrate. Further research is necessary to studyeeds under in more controlled planted green roofs to the fac-

ors that affect that weed colonisation on green roofs. This studyas carried out during the establishment period, therefore, fur-

her research on long term green roof maintenance is required tonderstand dynamic change of green roofs over the time.

cknowledgements

We express our appreciation to Mortgage Crofts Centre inotherham for providing research field, to career-support pro-ramme for woman scientists in Chiba University for founding toroof our English, to Dr. Algirdas Paskevicius in Chiba Universityor helping drawing figures.

eferences

lumasc Exterior Building Products, 2006. Green roof substrates. www.

alumasc-exteriors.co.uk (accessed 01.06.06).

llaby, M. (Ed.), 2006. Oxford Dictionary of Plant Sciences. Oxford University Press,New York.

rchibold, O.W., Wagner, L., 2007. Volunteer vascular plant establishment on roofsat the University of Saskatchewan. Landscape Urban Plan. 79, 20–28.

N

eering 58 (2013) 156– 164 163

renneisen, S., 2006. Space for urban wildlife: designing green roofs as habitats inSwitzerland. Urban Habitats 4, 27–36.

erndtsson, J.C., 2010. Green roof performance towards management of runoff waterquantity and quality: a review. Ecol. Eng. 36, 351–360.

utler, C., Orians, C.M., 2011. Sedum cools soil and can improve neighbouring plantperformance on a green roof. Ecol. Eng. 37 (11), 1976–1803.

utler, C., Butler, E., Orians, C.M., 2012. Native plant enthusiasm reaches newheights: Perceptions, evidence, and the future of green roofs. Urban For. UrbanGreen 11, 1–10.

arter, T., Fowler, L., 2008. Establishing green roof infrastructure through environ-mental policy instruments. Environ. Manage. 42 (1), 151–164.

lement, E.J., Foster, M.C., 1994. Alien Plants of the British Isles: A Provisional Cat-alogue of Vascular Plants (Excluding Grasses). Botanical Society of the BritishIsles, London.

ook-Patton, S.C., Bauerle, T.L., 2012. Potential benefits of plant diversity on vege-tated roof: a literature review. J. Environ. Manage. 106, 85–92.

unnett, N., Kingsbury, N., 2008. Planting Green Roofs and Living Walls, 2nd ed.Timber Press, Portland, OR.

unnett, N., Nagase, A., Hallam, A., 2007. The dynamics of planted and colonis-ing species on a green roof over six growing seasons 2001–2006:influence ofsubstrate depth. Urban Ecosyst. 11 (4), 385–398.

ennis, R.L.H., 2010. A Resource-Based Habitat View for Conservation: Butterfliesin the British Landscape. Wiley-Blackwell, Oxford.

milsson, T., 2008. Vegetation development on extensive vegetated green roofs:influence of substrate composition, establishment method and species mix. Ecol.Eng. 33 (3–4), 265–277.

orschungsgesellschaft Landschaftsbau Landschaftentwicklung e.V. (FLL), 2008.Guidelines for the Planning, Construction and Maintenance of GreenRoofing—Green Roofing Guideline. FLL, Bonn.

rancis, R., Lorimer, J.A., 2011. Urban reconciliation ecology: the potential of livingroofs and walls. J. Environ. Manage. 92 (6), 1429–1437.

unk, J.L., Cleland, E.E., Suding, K.N., Zavaleta, E.S., 2008. Restoration throughreassembly: plant traits and invasion resistance. Trends Ecol. Evol. 23 (12),695–703.

etter, K.L., Rowe, D.B., 2006. The role of green roofs in sustainable development?Hort Sci. 41 (5), 1276–1285.

rant, G., 2006. Extensive green roofs in London. Urban Habitats 4 (1), 51–65.rime, J.P., Hodgson, J.G., Hunt, R., 1988. Comparative plant ecology. HarperCollins

Publishers Ltd, London.itchmough, J., 1994. Natural neighbours. Landscape Des. April, 16–22.itchmough, J., 1995. Perennial pleasures in the urban sward. Horticulturist 4 (2),

1–9.itchmough, J., 2004. Herbaceous perennials. In: Hitchmough, J., Fieldhouse, K.

(Eds.), Plant User Handbook. Blackwell Science, Oxford, pp. 325–342.itchmough, J., Woudstra, J., 1999. The ecology of exotic herbaceous perennials

grown in managed, native grassy vegetation in urban landscapes. LandscapeUrban Plan. 45 (2–3), 107–121.

uston, M.A., 1994. Biological Diversity: The Coexistence of Species on ChangingLandscapes. Cambridge University Press, Cambridge.

adas, G., 2006. Rare invertebrates colonizing green roofs in London. Urban Habitats4, 66–86.

lingman, G.C., Ashton, F.M., Noordhoff, L.J., 1982. Weed science: principles andpractices, second edition. John Wiley & Sons, New York.

osareo, L., Ries, R., 2007. Comparative environmental life cycle assessment ofgreenroofs. Build. Environ. 42 (7), 2606–2613.

öhler, M., 2006. Long-term vegetation research on two extensive green roofs inBerlin. Urban Habitats 4, 3–26.

öhler, M., Poll, P.H., 2010. Long-term performance of selected old Berlin green-roofs in comparison to younger extensive green roofs in Berlin. Ecol. Eng. 36,722–729.

evine, J.M., D’Antonio, C.M., 1999. Elton revisited: a review of evidence linkingdiversity and invisibility. Oikos 87 (1), 15–26.

evine, J.M., 2000. Species diversity and biological invasions: Relating local processto community pattern. New Series 288, 852–854.

uckett, K., 2009. Green Roof Construction and Maintenance. McGraw-Hill, NewYork.

undholm, J., Maclvor, J.S., MacDougall, Z., Ranalli, M., 2010. Plant species and func-tional group combinations affect green roof ecosystem functions. PLoS One 5(3), e9677, http://dx.doi.org/10.1371/journal.pone.0009677.

acIvor, J.S., Lundholm, J., 2011. Performance evaluation of native plants suited toextensive green roof conditions in a maritime climate. Ecol. Eng. 37, 407–417.

et Office, 2007. Historic station data in Sheffield UK, Available fromhttp://www.metoffice.gov.uk/climate/uk/stationdata/sheffielddata.txt

olineux, C.J., Fentiman, C.H., Gange, A.L., 2009. Characterising alternative recycledwaste materials for use as green roof growing media in the UK. Ecol. Eng. 35,1507–1513.

aeem, S., Knops, J.M.H., Tilman, D., Howe, K.M., Kennedy, T., Gale, S., 2000. Plantdiversity increases resistance to invasion in the absence of covarying extrinsicfactors. Oikos 91 (1), 97–108.

agase, A., Dunnett, N., 2010. Drought tolerance in different vegetation types for

extensive green roofs: effects of watering and diversity. Landscape Urban Plan.97, 318–327.

agase, A., Dunnett, N., 2011. The relationship between percentage of organic matterin substrate and plant growth in extensive green roofs. Landscape Urban Plan.103, 230–236.

1 Engin

N

R

R

R

R

R

S

S

S

T

W

W

64 A. Nagase et al. / Ecological

agase, A., Dunnett, N., 2012. Amount of water runoff from different vegetationtypes on extensive green roofs: effect of plant species, diversity and plant struc-ture. Landscape Urban Plan. 104, 356–363.

ao, V.S., 1999. Principles of Weed Science, 2nd ed. Science Publisher, New Hamp-shire.

ejmánek, M., 1989. Invasibility of plant communities. In: Drake, J.A., Mooney, H.A.,Di Castri, F., Groves, R.H., Kruger, F.J., Rejemánek, M., Williamson, M. (Eds.), Bio-logical Invasions: A Global Perspective. , 2nd ed. John Wiley, Chichester, pp.369–398.

obinson, G.R., Quinn, J.F., Stanton, M.L., 1995. Invasibility of experimen-tal habitat island in California winter annual grassland. Ecology 76 (3),786–794.

owe, D.B., Getter, K.L., Durhman, A.K., 2012. Effect of green roof media depth on

Crassulacean plant succession over seven years. Landscape Urban Plan. 104(3–4), 310–319.

oxburgh, S.H., Wilson, J.B., 2003. Stability and coexistence in a lawn community:experimental assessment of the stability of the actual community. Oikos 88 (2),409–423.

Z

Z

eering 58 (2013) 156– 164

ailor, D.J., Hagos, M., 2011. An updated and expanded set of thermal property datafor greenroof growing media. Energy Build. 43 (9), 2298–2303.

nodgrass, E.C., McIntyre, L., 2010. The Green Roof Manual: A Professional Guide toDesign, Installation, and Maintenance. Timber Press, Portland, OR.

teiner, F., 2011. Landscape ecological urbanism: origins and trajectories. LandscapeUrban Plan. 100, 333–337.

onietto, R., Fant, J., Ascher, J., Ellis, K., Larkin, D., 2011. A comparison of bee com-munities of Chicago green roofs, parks and prairies. Landscape Urban Plan. 103(1), 102–108.

eiler, S., Scholz-Barth, K., 2009. Green Roof Systems: A Guide to the Planning,Design and Construction of Building Over Structure. John Wiley and Sons, NJ.

iser, S.K., Allen, R.B., Clinton, P.W., Platt, K.H., 1998. Community structure andforest invasion by an exotic herb over 23 years. Ecology 79 (6), 2071–2081.

imdahl, R.L., 2004. Weed-crop Competition, A Review, 2nd ed. Blackwell Science,Oxford.

hang, Z., Shen, L., Tam, V.W.Y., Lee, W.W.Y., 2012. Barriers to implement exten-sive greenroof systems: A Hong Kong study. Renew. Sustain. Energy Rev. 16,314–319.