Comparison of establishment methods for extensive green roofs in southern Sweden

ARTICLE IN PRESS

1618-8667/$ - se

doi:10.1016/j.uf

�Correspond

E-mail addr

Urban Forestry & Urban Greening 3 (2005) 103–111

www.elsevier.de/ufug

Comparison of establishment methods for extensive green roofs

in southern Sweden

Tobias Emilsson�, Kaj Rolf

Department of Landscape Management and Horticultural Technology, Swedish University of Agricultural Sciences,

Box 66, SE 23053 Alnarp, Sweden

Abstract

The most common technique for establishment of thin extensive green roofs in Sweden has been using prefabricatedvegetation mats. Our study investigated (1) how the establishment of green roofs in Sweden was influenced by theestablishment method (prefabricated vegetation mat, plug-plant, shoot), substrate composition and species mixture,and (2) whether on-site construction was a possible alternative. The establishment of the vegetation, which in all casesconsisted of succulent species, was recorded using the quadrate point intercept method in fixed plots and the successmeasured as frequency cover.

Prefabricated vegetation mats had higher succulent plant cover than on-site constructed roofs. There was nodifference in succulent plant cover between plots established using plug-plants compared to shoots. Shoot-establishedplots had more moss than the other establishment methods. The commercial substrate ‘Roof soil’ had significantlyhigher succulent plant cover than the other substrates, which might be related to a higher nutrient content. The organiccontent of the non-commercial substrates was rapidly decomposed. The standard species mixture produced a highercover than both the mix developed for northern conditions and the mix with an increased proportion of big leavedspecies. The total cover of the plots was mainly dependent on the cover of two species: Sedum album (L.) and Sedum

acre (L.). Few species managed to establish spontaneously but the establishment of woody species highlighted the needfor proper maintenance.r 2004 Elsevier GmbH. All rights reserved.

Keywords: Crassulaceae; On-site construction; Sedum; Vegetated roofs

Introduction

Green roofs are becoming increasingly popular inmany countries. The interest for green roofs has beenrelated to their capacity to reduce stormwater runoffvolumes and peak flows (Bengtsson, 2002), mitigateurban heat island effects (Akbari et al., 2001) and coolbuildings during summer months (Eumorfopoulou andAravantinos, 1998; Onmura et al., 2001). Green roofs

e front matter r 2004 Elsevier GmbH. All rights reserved.

ug.2004.07.001

ing author.

ess: [email protected] (T. Emilsson).

can also be designed to improve urban biodiversity(Mann, 1998; Brenneisen, 2003).

Installation of green roofs requires larger investmentsthan conventional roofs (Wong et al., 2003). Systemswith thick substrate layers and large plants are especiallyexpensive since they generally require reinforcement andreconstruction of the building unless it was designed forthe extra load from the start. Thin extensive systems cangenerally be built without making any adjustments tobuildings and this reduces the cost of the systemand increases the number of possible roofs that canbe vegetated. Even though the initial cost is high,

ARTICLE IN PRESST. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111104

calculations have shown that the life cycle cost ofextensive green roofs can be lower than the cost ofconventional roofs (Wong et al., 2003). This is due to theextended life expectancy of the roofing membrane andto the reduced energy consumption related to summercooling, which might be an important factor even incountries like Sweden where cooling during summer isbecoming increasingly important in office buildings(Nilson et al., 1997; Wong et al., 2003). Still, the highinvestment can be seen as a barrier to a widespread useof green roofs and much would be gained if extensivegreen roof systems could be installed at a lower cost.

A fundamental part of the success of an extensivegreen roof installation is connected to the establishmentand development of the plant material. Failure of thevegetation during the initial phase means that new plantmaterial has to be brought to the site at an additionalcost, and there is also a risk for erosion of the substrateif it has a lower cover during an extended establishmentperiod (Wolfgang, 2002). The goal of establishment is ahigh cover of the desired vegetation but also survival ofthe established plant species. The guidelines developedby the German organisation ForschungsgesellschaftLandschaftsentwicklung Landschaftsbau E.V (FLL),focus on high cover and state that a green roof shouldhave a projective cover of at least 60% one year afterestablishment (FLL, 2002).

Green roofs can either be established on-site or bybringing prefabricated vegetation to the roof. InGermany, where most of the development of technologyrelated to production and establishment of green roofshas taken place, green roofs are most often constructedon-site. On-site construction is generally achieved bypumping or lifting the substrate onto the roof and thendistributing shoots, seeds or plug plants. In Sweden,green roofs are mainly applied as prefabricated vegeta-tion mats, which is generally one of the most expensiveways to vegetate buildings but also a method that has alow risk of failure and that ensures instant high plantcover (Krupka, 1992; Schade, 2002; Dunnett andKingsbury, 2004). Vegetation mats are composed ofplants grown in a substrate that is fixed onto a carriermaterial, e.g. geotextile, plastic net or coconut net. Thevegetation mats are lifted to the roof as fully establishedvegetation during construction of the system. Vegetationmats are currently used in southern Sweden andDenmark but there are no comparative studies reportedwhere the less expensive technique of on-site establish-ment has been tested in this climate. The climate is lessextreme in southern Sweden compared to Germany butthe winters are slightly colder. This might influence thesurvival-rate of the newly established succulent species,since decreased substrate depths increases freezinginjuries of succulent plants (Boivin et al., 2001). Littleis known about how establishment of the on-siteconstructed systems compares to the cover of a

prefabricated vegetation and what type of cover andplant composition the consumer can expect whendeciding to use one or the other system.

Substrates are generally the same regardless ofestablishment method. The main component in sub-strates is inorganic material with a high water-holdingcapacity and low density such as pumice, lava, orexpanded clay (Roth-Kleyer, 2001). Recycled materialsuch as crushed roof tiles has also been used as acomponent in roof substrate, even though the density ishigher than in pumice or expanded clay (Roth-Kleyer,2001). The use of recycled material can be a way toreduce the need for transport and to find use for alocally available material that is otherwise worthless.The composition of commercial substrates is sur-rounded with secrecy and patents, while at the sametime the basic idea of substrate composition is readilyavailable in guidelines developed in Germany during thepast 15–20 years (FLL, 2002). Our study is comparingtwo substrates containing crushed roof tiles with acommercial substrate. One of the substrates wasdesigned strictly according to the German guidelinesand the other was designed with a slightly increasedorganic content since this would increase the water andnutrient holding capacity of the substrate and possiblyimprove establishment. Higher water and nutrientholding capacity of the substrate might on the otherhand increase the possibility for establishment of weedsthat would influence the aesthetic characters of the roofnegatively.

The plants used on thin green roofs are succulentsbelonging to the Crassulaceae family. The plants areable to withstand sustained periods without waterthrough both biochemical and morphological adapta-tions. The thin substrate dries out rapidly but thesucculent morphology of the plants enables them tostore large amounts of water and thereby cope withdrought situations. Two of the plants commonly used,Sedum album and Sedum acre, are known to expresscrassulacean acid metabolism (CAM) during droughtperiods (Sayed, 2001). Again, little information isavailable on the survival and establishment rates ofdifferent succulents in the Swedish climate. Most speciescombinations have been developed in Germany and fewsystematic studies on plant performance have beenperformed in Sweden. The species mixes that were testedin our study have been designed for both rapid coverand aesthetics, and all have a high proportion of groundcovering Sedum species, S. album and S. acre, in the mix.Our aim was to test how a standard mix commonly usedin Sweden and Germany compared to a mix composedfor more northern conditions and a more aestheticallypleasing mix with a higher proportion of big-leaved andmore flowering species.

The overall aims of our study were: (1) to describe theeffects of establishment method, substrate composition

ARTICLE IN PRESST. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111 105

and species mixture on vegetation performance andspontaneous establishment by moss and vascularspecies; and (2) to test whether on-site constructioncould be used for establishment of green roofs insouthern Sweden.

The success of the establishment was measured asplant frequency cover and species composition. Otherinteresting results such as aesthetic value of the systemsand the exact cost of installation are briefly discussedbut not analysed in depth.

Table 1. Composition of substrates A and B expressed as

percentage by weight

Composition %-by weight Substrate A Substrate B

Clay 5 5

Broken limestone 8–12mm 5 5

Crushed roof tiles 8–12mm 50 43

Sand 37 37

Organic material (Peat) 3 10

Methods

The study was started in 2000 at the Augustenborgbotanical roof garden situated in the city of Malmo,southern Sweden (551340N, 13110E). The green roofsused in the study were thin ‘extensive green roofs’divided into 105 vegetated roof sections measuring1.15 m� 6.5 m. The roofs were constructed on oldbuildings currently used by Malmo city council andthey were located in an area with a combination ofindustrial and residential buildings. The roofs wereorientated in a north-westerly direction and had aninclination of 41. The green roofs had a maximumwater-saturated weight of around 55 kg/m2, making anyextra adjustments of the roof construction unnecessary.

The green roofs were composed of three layers, theuppermost being a vegetation layer consisting of a 4 cmthick growing substrate and the vegetation. The secondlayer was a filter layer in the form of a geotextile, whichprevented small particles from being washed from thesubstrate layer into the drainage layer or out from thesystem. The third layer was a drainage layer in the formof recycled foam material, sold under the product nameAquaTop.

All plots were given a fertiliser addition of 15 g/m2 atthe time of establishment and in spring the followingyear. The fertiliser was a 50:50 mixture of a long-termfertiliser (Multicote 8M extra N:18-P:6-K:12) and aconventional fertiliser (ProMagna N:11-P:5-K:18).

The cost of the vegetation layer in these experimentalplots was approximately 32 h/m2 for prefabricatedvegetation, 23 h/m2 for plug-plant establishment and14 h/m2 for shoot establishment.

Establishment methods

Three establishment methods, pre-made vegetationmats, plug plants and shoots, were used in the study.The vegetation mats were made of a geotextile and a soilsubstrate that was reinforced by a plastic net. Thesubstrate in the vegetation mats was the same as thesubstrate used in the on-site establishment. The vegeta-tion mats were established using the same methodology

as for the shoot-established surfaces but after establish-ment they were kept in a sandpit and watered regularly.The vegetation mats were established during spring 2000and this meant that in reality they were 4 months olderthan the on-site established surfaces at the start of theexperiment and at the start of the experiment they hadalmost full cover. The plots established with plug plants,each having a soil-root volume of 65 cm3, wereestablished with a density of 25 plants/m2 of theindividual species mixture. Shoots were established bydistributing 150–200 g shoots/m2 of the different speciesmixtures. The same type of thin plastic net as in thevegetation mats was used to reinforce the substrate inshoot establishment.

Substrates

The first substrate was bought from the Swedish greenroof company VegTech and is referred to in this study as‘Roof soil’. The composition of the substrate is notknown exactly but it is a commercially availablesubstrate that is composed of a natural soil mixtureimproved by the addition of lava rock, expanded clay,organic material and clay. The two other substrates,substrates A and B, were mixed by us according to thelist of content in Table 1. The main difference betweenthe two substrates A and B was in their organic mattercontent and in the amount of crushed tiles (Table 1).

Plant mixes

All three types of plant mixtures used in this studyinvolved different combinations of the succulent species.The first mixture was a standard mix commonly used forgreen roof establishment in Sweden. The second mix wasdesigned for more northern conditions and had a largeproportion of S. acre. The last mix had a higherproportion of big-leaved species and species that haveintense flowering (Table 2).

Experimental design

Establishment method, substrate composition andspecies composition were varied in a factorial fashionbut due to time constraints prefabricated vegetation

ARTICLE IN PRESS

Table 2. Composition (%) of the three species mixtures standard, northern and big-leaved mixed used in the establishment

Species Standard mix Northern mix Big-leaved mix

Sedum acre (L.) 40 70 30

Sedum album (L.) 40 10 30

Sedum rupestre (L.) 5

Sedum sexangulare (L.) 5 10

Hylotelephium ewersii (Ledebour) 5 5

Phedimus floriferus ‘Weihenstephaner Gold’ (Praeger) 10

Phedimus hybridus (L.) 5

Phedimus kamtschaticus (Fischer & C.A. Meyer) 5

Phedimus spurius (M. von Bieberstein) 10 10 10

Table 3. Experimental setup studying the effect of establishment method, substrate and species composition

Species mix Roof soil Substrate A Substrate B

Standard mix Pre-fabricated mats __ __

__ __

__ __

Plug-plants Plug-plants Plug-plantsShoots Shoots Shoots

Northern mix Pre-fabricated matsPlug-plants Plug-plants Plug-plantsShoots Shoots Shoots

Big-leaved mix Pre-fabricated matsPlug-plants Plug-plants Plug-plantsShoots Shoots Shoots

The first part of the experiment involves the column at the left and the second part includes all grey shaded areas.

T. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111106

mats were only constructed on the commerciallyavailable substrate Roof soil (Table 3). The unbalanceddesign meant that the analysis had to be divided in twoparts. In the first part of the experiment only Roof soilwas involved. This part of the experiment investigatedthe effect of the establishment method (plugs, shoots orpre-made mats) and species mix (standard, northern, orbig-leaved) on plant cover. In the second part of theexperiment, three types of substrates (Roof soil,substrate A and substrate B), three species mixes(standard, northern, or big-leaved), and two types ofestablishment methods (plugs and shoots) were involvedand the effects of these parameters on cover weremeasured. All treatments were randomly assigned toplots.

Vegetation survey

The survey of the vegetation was made in autumn2001, one year after the roof vegetation had beenestablished. The vegetation of the plots was analysedusing a quadrate point-intercept method (Greig-Smith,1983). The points were arranged in a 45 cm regular13� 13 grid with each point spaced 37.5 mm apart. The

grid was constructed by two nets arranged at 25 mmdistance from each other. The use of two grids ensuresperpendicular recording of vegetation cover. The cross-hairs had a diameter of 2mm. The grid was suspended10 cm above the substrate layer.

All vegetation that was covered by the projection ofthe cross-hairs was recorded as present. Great care wastaken to record all vegetation layers by carefully movingthe higher layers without disturbing the lower. In mostcases, the vegetation consisted of a single layer. Vascularplant species, moss or lack of vegetation was recorded.The point-intercept measurements were complementedby a survey of plants species in every plot in order toinclude species with low cover.

The vegetation was measured in three fixed quadrantsper experimental plot. The first replicate was randomlylocated in the upper part of the roof, the secondreplicate was randomly located in the middle part andthe third replicate was randomly located in the lowerpart of the roof, resulting in a blocked experimentaldesign.

All succulents were identified and labelled accordingto Eggli (2003). All other vascular plants were identifiedand labelled according to Flora Europaea (Tutin et al.,1968–1980, 1993).

ARTICLE IN PRESS

Fig. 1. Frequency cover of succulent plants one year after

establishment when different species mixtures and establish-

ment methods were used on Roof soil substrate. Symbols show

mean values and bars represent standard error of the mean

value =Vegetation mats =Shoots =Plug plants).

T. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111 107

Soil analysis

Soil was collected from three randomly selected plotsfor each treatment. The three soil samples from eachtreatment were mixed into one bulk sample on which allanalyses were made. All soil analyses were preformed intriplicate. Soil density and total pore space wereanalysed after Proctor hammer compaction (FLL,2002). Organic matter content was estimated as loss onignition (5501, 15 h) and calculated as % dry weight. pHwas measured in 0.01 M CaCl2 solution using a WTWpH electrode SenTix 21. Determination of the CaCO3

was carried out according to the method of Scheibler(Hoffmann, 1991). Phosphorus and potassium contentswere determined by calcium-acetate-lactate extractionand magnesium was determined after CaCl2 extraction.Plant available mineral nitrogen was extracted withCaCl2 and analysed after steam distillation. Detaileddescriptions of chemical soil analyses methods are givenin Hoffmann (1991).

Statistical analysis

The plant frequency cover data collected using thepoint intercept method were transformed using anarcsin(x0.5) transformation in order to get homogeneousvariance (Underwood, 1998). A factorial ANOVA wasused to test for significant differences between thetreatment combinations and where differences existed,means were separated by a Tukey test. The establish-ment of the succulent plants failed on one plot locatedclose to the projection of a higher roof and it wastherefore excluded from the analysis. Throughout theanalysis, the threshold for significance was set atPo 0:05: Data are presented as mean7SE. All statis-tical analyses were performed using the SPSS vs. 10statistical programme.

Results

The establishment was successful on most plots, butthere were substantial differences in the frequency coverbetween the different establishment methods, speciesmixtures and soil substrates.

The first part of the experiment, involving only theRoof soil substrate, showed a significant difference incover between the establishment methods but alsobetween the species mixtures. The prefabricated vegeta-tion mats that had been pre-grown for 4 months beforethe experiment and were fully established at the start ofthe experiment and still had a higher cover than shoot-and plug plant-established plots one year after establish-ment (Fig. 1). The mean succulent plant cover on theprefabricated mats was more than 80% for all species

mixtures. There was no significant difference betweenthe two other establishment techniques on the Roof soil;they both had a mean succulent plant cover ofapproximately 50–60%. The standard species mixture,when used on the Roof soil substrate, had a significantlyhigher cover than the other species mixtures. There wasno difference between the mix selected for more north-ern conditions or the mix with more big-leaved species.

The cover of moss showed a somewhat differentpattern. The only variable that had a significantinfluence on the moss cover was the establishmentmethod. Plots established with shoots had significantlyhigher moss cover than plots established with plugplants or plots established with vegetation mats. Therewas also significantly less moss on the vegetation matscompared to the plug plant-established plots. Thecomposition of the species mixture used had nosignificant impact on the moss cover on Roof soilsubstrate (Fig. 2).

The total cover of succulent plants and moss wasaffected by both establishment method and speciesmixture (Figs. 1 and 2). The prefabricated vegetationmats again had the highest cover, followed by shoot-established plots and finally plug plants. The standardmix had a significantly higher cover than the big-leavedspecies mix. There was also a block effect for the totalcover, which increased towards the edge of the roof.

The second analysis, excluding the prefabricatedvegetation mats, showed a significant effect for bothsoil substrate and species mixture on frequency cover ofsucculent plants. However, there was no difference in

ARTICLE IN PRESST. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111108

succulent plant cover between plug plants and shoot-established plots. There was also a significant interactionbetween establishment method and soil substrate, whichshowed a negative relationship of shoot establishmenton substrate A. In general, the Roof soil had asignificantly higher cover than the other substrates andthe standard mixture had the significantly highestfrequency cover of the species mixtures tested (Fig. 3).The mean succulent plant cover was below 40% whenthe northern mix or the big leaved mix were used forshoot establishment on substrate A or plug planting onsubstrate B (Fig. 3).

Fig. 2. Frequency cover of moss one year after establishment

when different species mixtures and establishment methods

were used on Roof soil substrate. The figure shows mean

values and standard error. ( =Vegetation mats =Shoots

=Plug plants).

Fig. 3. Impact of establishment method, species mixture and substra

standard error. ( =Roof soil, =Substrate A, =Substrate B).

The moss cover in the second analysis showed adependency on species mixture and establishmentmethod, but not on soil substrate. The use of thestandard species mixture resulted in less moss than theother two species mixtures. The use of shoots generallyresulted in more moss. A significant interaction showeda negative impact on moss cover when Roof soil wasused in combination with plug-plants (Fig. 4). There wasalso a block effect for the moss cover, which increasedtowards the edge of the roof.

The total cover of succulent plants and moss showedsignificant positive effects of establishment with Roofsoil, when looking at all plots established with plugplants or shoots (Figs. 3 and 4). There were alsosignificant interactions between establishment methodand soil substrate, which meant that the combination ofthe shoot establishment on substrate A had a reducedcover. There was also a significant block effect as thetotal cover increased towards the bottom of the plots.

The total cover of the plots was basically made up ofS. acre and S. album. The other species had never morethan 7% of the total mean succulent cover. S. album wasespecially favoured by the establishment with thestandard mix on vegetation mats where it constitutedclose to 90% of the total succulent cover.

Few plants managed to establish spontaneously onthe thin extensive green roofs used in the study. Most ofthe plants were common ruderals but there were alsosome uncommon calcicole species such as Saxifraga

tridactylites and Saxifraga granulata. A total of 13different species were found to have established sponta-neously on the roof. The establishment was occasionaland showed no apparent pattern. The following plantswere found on the roofs: Cerastium semidecandrum (L.),Senecio vulgaris (L.), Cerastium pumilum (Curtis),Arabidopsis thaliana (L.), Poa alpina (L.), Epilobium

spp. (L.), Acer campestre (L.), Taraxacum sect. (Weber),

te on succulent plant cover. The figure shows mean values and

ARTICLE IN PRESS

Fig. 4. Impact of establishment method, species mixture and substrate on moss cover. The figure shows mean values and standard

error ( =Roof soil, =Substrate A, =Substrate B).

Table 4. Chemical and physical characteristics of three extensive green roofs substrates

Substrate A Substrate B Roof soil

Density (dry) g/cm3 1.4770.02 1.4870.01 1.3770.01a

Total pore space % 44.0170.79 43.3470.42 46.6770.35a

Organic content % 1.0270.01a 1.6070.03b 5.2570.04c

pH 7.4970.04 7.4970.02 7.3570.00a

CaCO3 (g/100g dry soil) 1.5470.17 1.6470.15 10.8770.79a

P2O5 (mg/100g dry soil) 3.3870.82 2.8670.53 3.9871.40

K2O (mg/100g dry soil) 5.0070.00 6.3370.33 15.0071.00a

Mg (mg/100g dry soil) 1.0370.12a 2.0070.06ab 2.7770.38b

N mineral (mg/100g dry soil) 1.6270.2 1.5170.16 1.7870.21

N total (mg/100g dry soil) 68.57710.86 67.4374.72 219.27739.29a

Values are mean7SE. Significant differences between values not sharing a common index letter.

T. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111 109

Poa annua (L.), Hieracium pilosella (L.), S. tridactylites

(L.), Erophila verna (L.) and S. granulata (L.).The Roof soil differed from substrates A and B in all

of the variables investigated except available phosphor-ous, magnesium and mineral nitrogen (Table 4).Differences between substrates A and B were onlyfound in respect to the organic content of the substrate.The organic content of substrate A decreased from 3%to 1.02% and that of substrate B from 10% to 1.60%(Table 4).

Discussion

Results from the first year showed that establishmentwith prefabricated vegetation achieved higher coverthan the other methods. The high cover of theprefabricated vegetation mats can be explained by thefact that they were pre-grown under favourable condi-tions for several months prior to establishment and that

they had an almost complete cover at the start of theexperiment. The plants on the on-site established plotsdid not grow to a comparable cover during the first yearand this means that vegetation mats can have anadvantage if high cover is needed during the first year.However, the vegetation mats that were used on ourresearch plots were more than twice as expensive as thevegetation layer of shoot establishment and close to30% more expensive than the vegetation layer of theplug plant establishment.

It was interesting to note that the study did notindicate any difference in the cover of succulent plantsbetween the other two methods. Plug plants aregenerally believed to be less sensitive to environmentalconditions than shoots, since the pre-grown plants havea developed root system and canopy, but there wasnothing in this study that suggested an advantage forthe use of plug plants in a Swedish climate. In this study,the establishment of shoots functioned just as well anddue to its lower price would be the preferable method.There were some problems with birds removing some of

ARTICLE IN PRESST. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111110

the plug plants in their search for food and it is possiblethat this had a negative effect on the possibility of plugplants to create a high cover.

At the same time, the plots established with shootsgenerally had more moss than the other establishmentmethods. Moss is usually no problem on green roofs,since it adds to the total cover and absorbs water. It canalso have an aesthetic advantage since it remains greenand growing even in the late autumn and winter.However, it can become a problem if too large aproportion of the roof is totally covered with moss.Moss has a tendency to attract birds that are looking forfood and nesting material. Birds dig in the soil substratein their search for food and spread moss and plantsparts, which can eventually end up in the rain gutter andcause clogging. Some people might also view roofs withfood-seeking birds and a high moss cover as a lessattractive outcome of an investment in a green roof.Future studies will show whether the moss increases andcomes to dominate the shoot-established system.

The commercial substrate used in this study led to abetter performance of the succulent vegetation, some-thing that was most probably due to the higher totalnitrogen content (Fig. 4). The positive effect of nitrogenfertilisation on Sedum vegetation has been shown inother studies (Fischer and Jauch, 2002). At the sametime, it is important to remember that the nitrogencontent of roof substrates is meant to be low in order tominimise the negative effects on stormwater quality andthat too high nitrogen concentrations might make theplants more susceptible to freezing injury (FLL, 2002).The higher organic content of substrate B compared toA did not have any direct effect since the organicmaterial in both substrates was almost completelydecomposed during the first year. The peat materialthat was used in our substrate was not suited for a greenroof substrate since it did not resist decomposition. Therapid decay of materials, and subsequent decrease innutrient- and water-holding capacity, may have negativeimpact on the ability of plants to grow at a high rate,spread and maintain a desirable aesthetic character. Therapid decay of organic material may also haveconsequences for the quality of stormwater, since itmight result in leakage of nutrients to the stormwatersystem. The decay and transport of organic substancesfrom green roofs can in itself also have effects on thecolour and quality of the water (Marx and Kolb, 2002).It is therefore important to choose a form of organicmaterial that is stable and that will be able to maintaina nutrient-holding capacity over an extended periodof time.

The alternative species mixtures developed to increasethe succulent cover during establishment did not workas well as the standard mix. The advantage found for thestandard species mixture can mainly be attributed tothe higher S. album content. The vegetatively spreading

S. album was hardy and functioned well as a groundcover on this thin soil. In most cases there were nodifferences between the northern mixture and themixture with a higher content of big leaved species.The big-leaved species had little influence on the totalcover of the plots but they had an important impact onthe aesthetic function of the roof. However, thesecomplementary species were found on most plots. Itwas also found that S. acre might acquire a degradedlook after flowering, especially on the more nutrient-richRoof soil, where dense stands of grey dead flower stalksproduced an unattractive appearance. It is interestingto note that the prefabricated vegetation mats gaveS. album a competitive advantage. Establishment byplug plants and shoots more closely followed the initialspecies composition.

The absolute cover of the surfaces established on-sitewas close to the threshold value of 60% cover defined byFLL (2002), but it was only the standard species mix onthe Roof soil that had a cover that was higher than 60%.Some of the on-site established plots that had beenestablished with the northern or the big-leaved mix didnot achieve a high succulent plant cover, e.g. the plotsestablished with the northern and big-leaved species mixusing plug plants on substrate B and the shootestablishment on substrate A fell below 40% cover,which can hardly be seen as a successful establishment.

Some of the spontaneous established species wereprobably brought to the research plot with the shootmixtures, plug plants or vegetation mats, since no suchbiotopes are known to be present in the surroundings. Itis important to note that the occasional establishment ofspecies such as A. campestre highlights the need forproper maintenance. Woody species can cause seriousproblems with root penetration of the waterproofmembrane on the roof, even if these plants occur atlow frequency and with low cover. Annual maintenanceis needed even on 4 cm thin roofs. Furthermore, the lowcapability for spontaneous establishment questions theability of thin extensive green roofs with engineeredsubstrates to function as stepping stones, exchangebiotopes and refuges for rare plant species that havedifficulties in surviving in the urban environment. Thegreen roofs might have a higher biological value thanconventional roofs, but it seems unlikely that the thinroof type used in this study can compare with, forexample, an urban park or grassed land.

Conclusion

Our research shows that on-site establishment is afunctional alternative to vegetation mats when establish-ing green roofs in Sweden. On-site construction is stillrare in Sweden and some of the variables, in particular

ARTICLE IN PRESST. Emilsson, K. Rolf / Urban Forestry & Urban Greening 3 (2005) 103–111 111

the development of new substrates, have to be improvedin order to have rapid plant cover. The varying resultsfrom the three different species mixtures also showsthat there is important work to be done regardingthe development of new mixtures that ensure high coverand aesthetically pleasing vegetation during the firstyear. Vegetation mats have an advantage in exposedsites or when high initial cover is needed for otherreasons, but the differences in cost between theprefabricated vegetation mats and establishment byshoots or plug plants can motivate the use of on-siteestablishment on more roofs. The benefits of greenroofs to the urban environment are not fully realiseduntil there is a substantial amount of green roofs inthe urban landscape and developing cheaper establish-ment methods would be one step towards buildingmore green roofs. The economic conditions have tobe investigated in more detail before on-site establish-ment can be applied in full scale but it seems eco-nomically feasible given the low cost achieved for ourplots.

The establishment is only the first part of the life of agreen roof and an important future study would be thecomparison between establishment success and long-term success and especially the long-term maintenancerequirements of the installation since these factors arerelated to the life-cycle cost of the green roof.

Acknowledgements

This work was funded by FORMAS. The authorswould like to thank the two anonymous reviewers andJan Erik Mattson for valuable comments on manu-scripts, and Jan-Eric Englund for help with thestatistical analysis.

References

Akbari, H., Pomerantz, M., Taha, H., 2001. Cool surfaces and

shade trees to reduce energy use and improve air quality in

urban areas. Solar Energy 70, 295–310.

Bengtsson, L., 2002. Avrinning fr(an grona tak (Runoff from

Green Roofs). Vatten 58, 245–250 (In Swedish, with

English summary).

Boivin, M.-A., Lamy, M.-P., Gosselin, A., Dansereau, B.,

2001. Effect of artificial substrate depth on freezing injury

of six herbaceous perennials grown in a green roof system.

HortTechnology 11, 409–412.

Brenneisen, S., 2003. Okologisches Ausgleichspotenzial von

Extensiven Dachbegrunungen—Bedeutung des Ersatz-

Okotpos fur den Arten- und Naturschutz und die

Stadtentwicklungsplanung. Dissertation, Universitat Basel,

Basel (in German).

Dunnett, N., Kingsbury, N., 2004. Planting Green Roofs and

Living Walls. Timber Press, Oregon, USA pp. 254.

Eggli, U., 2003. Illustrated handbook of succulent plants.

Crassulaceae, Springer, New York, pp. 458.

Eumorfopoulou, E., Aravantinos, D., 1998. The contribution

of a planted roof to the thermal protection of buildings in

Greece. Energy and Buildings 27, 29–36.

Fischer, P., Jauch, M., 2002. Dungung von extensiven

Dachbegrunungen. Dach+Grun 11, 22–28 (In German).

FLL, 2002. Richtlinie fur die Planung, Ausfuhrung und

Pflege von Dachbegrunungen. Forschungsgesellschaft

Landschaftsentwicklung Landschaftsbau E.V., Bonn

(in German).

Greig-Smith, P., 1983. Quantitative Plant Ecology, 3rd ed.

Blackwell, Oxford.

Hoffmann, G., 1991. Die Untersuchung von Boden–Metho-

denbuch 1. VDLUFA-Verlag, Darmstadt (in German).

Krupka, B.W., 1992. Dachbegrunung: Pflanzen- und Vegeta-

tionsanwendung an Bauwerken. Ulmer, Stuttgart, pp. 508

(in German).

Mann, G., 1998. Vorkommen und Bedeutung von Bodentieren

(Makrofauna) auf begrunten Dachern in Abhangigkeit von

der Vegetationsform. Dissertation, Universitat Tubingen,

Tubingen (in German).

Marx, I., Kolb, W., 2002. Mineralische Zusatze fur Eins-

chrichtsubstrate in Dachbegrunung. Dach+Grun 11 (1),

23–28 (in German).

Nilson, A., Uppstrom, R., Hjalmarsson, C., 1997. Energy

efficiency in office buildings: lessons from Swedish projects,

Swedish Council for Building Research & NUTEK,

Stockholm.

Onmura, S., Matsumoto, M., Hokoi, S., 2001. Study on

evaporative cooling effect of roof lawn gardens. Energy and

Buildings 33, 653–666.

Roth-Kleyer, S., 2001. Vegetationstechnische Eigenschaften

mineralischer Substratkomponenten zur Herstellung von

Vegetationstrag- und Dranschichten fur bodenferne Begru-

nungen. Dach+Grun 10, 4–11 (in German).

Sayed, OH., 2001. Crassulacean acid metabolism 1975–2000, a

check list. Photosynthetica 39, 339–352.

Schade, C., 2002. Eigenschaften und Anwendung von Vege-

tationsmatten in der extensiven Dachbegrunung. Disserta-

tion. Institut fur Grunplanung und Gartenarchitektur,

Universitat Hannover (in German).

Tutin, T.G., Burges, N.A., Chater, A.O., Edmondson, J.R.,

Heywood, V.H., Moore, D.M., Valentine, D.H., Walters,

S.M., Webb, D.A. (Eds.), 1993. Flora Europaea, 2 Edition.

Vol. 1. Cambridge U.P., Cambridge.

Tutin, T.G., Heywood, V.H., Burges, N.A., Moore, D.M.,

Valentine, D.H., Walters, S.M., Webb, D.A. (Eds.),

1968–1980, Flora Europaea, 1 Edition. Vol. 2–5. Cam-

bridge U.P., Cambridge.

Underwood, A.J., 1998. Experiments in Ecology: Their

Logical Design and Interpretation using Analysis of

Variance. University Press, Cambridge.

Wolfgang, E., 2002. Mangel bei Begrunungen. In: Dachab-

dichtung—Dachbegrunung. Teil 1, Fehler: Ursachen,

Auswirkungen und Vermeidung Fraunhofer-IRB-Verl.,

Stuttgart, pp. 186 (In German).

Wong, N.H., Tay, S.F., Wong, R., Ong, C.L., Sia, A., 2003.

Life cycle cost analysis of rooftop gardens in Singapore.

Building and Environment 38, 499–509.