HORTSCIENCE 48(7):929–937. 2013.

Performance of Zoysia matrella ‘Zeon’in Shallow Green Roof Substratesunder Moisture Deficit ConditionsNikolaos Ntoulas1, Panayiotis A. Nektarios, and Efthimia NydriotiLaboratory of Floriculture and Landscape Architecture, Department of CropScience, Agricultural University of Athens, 75, Iera Odos, Athens, 118 55, Greece

Additional index words. compost, digital image analysis, green turf cover, Manilagrass, NDVI,relative water content

Abstract. The use of turfgrasses might provide an additional solution for establishinggreen roofs in urban environments. The aim of the present study was to determineManilagrass [Zoysia matrella (L.) Merr. ‘Zeon’] drought tolerance when grown undergreen roof conditions and under two different irrigation regimes. Treatments included: 1)two extensive green roof substrates {locally produced substrate [3 sandy loam soil:8pumice:4 perlite:4 compost:1 zeolite (by volume)] and a commercially available substratebased on crushed tiles}; 2) two substrate depths (7.5 cm or 15 cm); and 3) two irrigationregimes (3 mm or 6 mm of irrigation every 3 days). Substrate characteristics (particle sizedistribution, saturated and dry bulk density, total porosity, water potential curves, in situsubstrate moisture, pH, electrical conductivity, and nutrient analysis), turfgrass growth,and physiological status [green turf cover (GTC), normalized difference vegetation index(NDVI), and leaf relative water content (RWC)] were determined. During moisture deficitperiods, GTC, NDVI, and RWC were most affected by substrate depth; moderatelyaffected by irrigation regime; and, to a lesser extent, by substrate type. Turfgrass growthand physiological status were best during moisture deficit conditions in the deeper profile(15 cm) using the higher amount of irrigation (6 mm) and the locally mixed substrate.

In recent years, green roof constructionhas been steadily increasing worldwide asa result of the numerous environmental andaesthetic advantages that they provide withinthe often monotonous and unfriendly landscapeof contemporary cities (Claus and Rousseau,2012). City centers are mostly occupied byolder buildings with minimal tolerance ofadditional weight loads. In such cases, exten-sive green roofs with shallow substrate depth,reduced weight, and minimal maintenancerequirements may be a viable option. How-ever, in the absence of governmental incentives,a financial investment is frequently substanti-ated from the residents only if green roofsbecome an accessible and functional space.

Turfgrass growth has seldom been evalu-ated on extensive green roofs mainly as a resultof their water demands, which are higher incomparison with succulents or other xerophyticplants, which conform to the minimal mainte-nance and natural resources input specified

by green roof construction guidelines (FLL-Forschungsgesellschaft LandschaftsentwicklungLandschatsbau. e.v., 2008). Bousselot et al.(2011) evaluated Bouteloua gracilis andBuchloe dactyloides growth in containers withdecreasing moisture content. MacIvor andLundholm (2011) compared survival andcover of different grass species (Festucarubra, Deschampsia flexuosa, Danthoniaspicata) on rooftop modules. In other cases,the studied grasses were mostly ornamentalsrather than turfgrass-covered surfaces such asFestuca ovina glauca (Dunnett and Nolan,2004), Schizachyrium scoparium (Durhmanet al., 2006), F. ovina, Trisetum flavescens,Anthoxanthum odoratum (Nagase and Dunnett,2010), F. scoparia (Dunnett et al., 2008), andKoeleria macrantha (Monterusso et al., 2005;Nagase and Dunnett, 2010).

An increasing number of researchers haveindicated that irrigation might be a necessityfor establishing extensive green roofs in semi-arid regions regardless of the used plants’drought tolerance (Kotsiris et al., 2012b;Ntoulas et al., 2012). Irrigated green roofsare expected to contribute significantly moreas compared with non-irrigated ones to themitigation of the urban heat-island effectthrough the cooling action of plant transpiration(Spronken-Smith and Oke, 1998; Takebayashiand Moriyama, 2009). Crassulacean acid me-tabolism (CAM) or xerophytic plants growingunder severe water stress are not expected tocontribute as much to the cooling of theambient urban temperature during summeras a result of their minimal transpiration. Ithas been proposed that irrigation could be

applied during the first couple years afterestablishment or during severe drought con-ditions to properly establish and sustain thegrowth of green roof plants (Getter and Rowe,2006; Latocha and Batorska, 2007; Williamset al., 2010; Wolf and Lundholm, 2008).Considering these irrigation inputs, it wouldbe of interest to investigate the capacity ofturfgrass to grow under the limited depth ofextensive green roof substrates and to deter-mine their growth response during the Medi-terranean summer drought period.

Turfgrasses have the unique ability to serveall of the three requirements of plants that areused in the urban environment, namely aes-thetics, function, and recreation (Beard andGreen, 1994). Thus, the introduction of turf-grasses on irrigated extensive green roofswould create a usable, functional, and enjoy-able urban surface. In an effort to introduceturfgrass species for irrigated extensive greenroofs in the semiarid Mediterranean region,several criteria have to be met concerning speciesselection, substrate type, and substrate depth.

The effects of extensive green roof sub-strate type and depth on sustainability of bothsucculent and native plant species in differentclimatic regions have been extensively re-searched (Boivin et al., 2001; Dunnett et al.,2008; Durhman et al., 2007; Nektarios et al.,2011a). Substrates have been formulatedfrom materials that are lightweight, maintainadequate moisture for plant growth, facilitatethe removal of excess water, provide nutrients,and possess an appropriate pH and salinity forplant growth (FLL-ForschungsgesellschaftLandschaftsentwicklung Landschatsbau. e.v.,2008; Getter and Rowe, 2006). Inorganicmaterials such as crushed tiles, calcined clay,zeolites, sand, heat-expanded shale (Beattieand Berghage, 2004; Thuring et al., 2010),pumice, and lava (Kotsiris et al., 2012b;Nektarios et al., 2011a; Rowe et al., 2006)have been customarily used on green roofs.Organic substances such as peat and compostshave also been used but at smaller participa-tion percentages to prevent substrate subsidiz-ing as a result of decomposition (Williamset al., 2010).

As substrate depth decreases, plant growthand survival are reduced and fewer plantspecies are able to adapt and survive (Boivinet al., 2001; Dunnett and Nolan, 2004; Dunnettet al., 2008; Durhman et al., 2007; Getter andRowe, 2008; VanWoert et al., 2005). Thesubstrate depths that have been tested forextensive green roofs have varied from 2.5 cmto 15 cm. The difficulty of growing plants inextremely shallow substrates is mainly attrib-uted to the ease of substrate moisture depletionand to the increased substrate temperature ex-tremes (Boivin et al., 2001). However, it hasbeen reported that irrigation has the capacityto compensate for a reduced substrate depthin extensive green roof types. VanWoert et al.(2005) found that plant sustainability and growthwere improved when irrigation was applied andthat irrigation could counteract the difficul-ties imposed by the reduced substrate depth.

The first objective of this research was tocompare the growth of Manilagrass [Zoysia

Received for publication 8 Feb. 2013. Accepted forpublication 19 May 2013.We thank Egreen S.A. for donating green roofmaterials, L. Cambanis S.A. for providing thecompost, and LAVA Mining & Quarrying S.A.for donating pumice.This article is a portion of a Ph.D. dissertation to besubmitted by N. NtoulasThe mention of a trade mark, proprietary product,or vendor does not imply endorsement by theauthors nor does it imply approval to the exclusionof other products that may also be suitable.1To whom reprint requests should be addressed;e-mail ntoulas@aua.gr.

HORTSCIENCE VOL. 48(7) JULY 2013 929

matrella (L.) Merr. ‘Zeon’] in two differentsubstrates. The second objective was to inves-tigate the impact of increasing either substratedepth or the amount of irrigation on Manila-grass growth and recovery potential during andafter moisture deficit periods, respectively.

Materials and Methods

Experimental design and setupA 2-year outdoor study was conducted on

a rooftop at the Laboratory of Floriculture andLandscape Architecture, Agricultural Univer-sity of Athens, Athens, Greece (lat. 37�59# N,long. 23�42# E, 35 m a.s.l.) from 19 Aug. 2009until 14 Oct. 2011. The study comprised 40plots of 0.81 m2 each (0.9 m · 0.9-m internaldimensions). Experimental arrangement wasmultifactorial with three factors: two substratetypes, two substrate depths, and two irriga-tion regimes replicated five times.

The experimental plots were constructedon a building roof, which was covered withthe following consecutive layers: 1) vaporbarrier (Eshaelastic; Alfa Alfa Energy SA,Attika, Greece); 2) 20-mm polystyrene ther-mal insulation with aluminium foil at oneside (Tempra; Vaportherm, Alfa Alfa EnergySA); 3) paintable primer to enable appropri-ate contact of waterproofing asphalt panesonto thermal insulation (Extralac; Alfa AlfaEnergy SA); and 4) two layers of waterproofingasphalt panes (Eshagum Antiroot B2; Alfa AlfaEnergy SA) with a weight of 4 kg·m–2 in-cluding a root-repellent substance (Preventol).

On top of the asphalt panes, woodenboards 2 cm thick with variable height wereused to create adjacent plots and preventmechanical and hydrological continuity be-tween the plots and the surrounding area.Within each experimental plot, the asphaltpanes were covered with a protection mat,which was a synthetic cloth made of non-rotting synthetic polyester fibers having 5 mmthickness and a dry weight of 0.47 kg·m–2 thatalso acts as a water reservoir by retaining5 L·m–2 of water according to the manufac-turer claims (SSM 45; Zinco, Egreen, Athens,Greece). A drainage layer made of recycledpolyethylene with 25 mm height and a weightof 1.7 kg·m–2 (FD25; Zinco, Egreen) with water-retaining troughs and openings for ventilationwas placed over the protection cloth. Thedrainage layer had the capacity to store 3 L·m–2

serving as additional water storage. Withineach plot, the drainage layer was covered bya non-woven geotextile (SF; Zinco, Egreen)made of thermally strengthened polypropylene,having 600 mm thickness, a mass of 0.1 kg·m–2,apparent opening size of D90 = 95 mm, andwater flow rate of 0.07 m·s–1. The filter sheetwas used to prevent fine particle migrationfrom the substrate toward the drainage layerto ensure that it would not clog and wouldfunction effectively. The geotextile was sta-pled to the side boards of each experimentalplot and cut below the final substrate surfaceto interrupt the continuation of geotextile to-ward the ambient environment and thus min-imize any potential wick-like capillary watermovement toward the atmosphere.

The experimental plots were randomlyfilled with two substrates. The first one(S15:Pum40:Per20:C20:Z5) comprised sandyloam soil (S), pumice (Pum), perlite (Per),compost (C), and clinoptilolite zeolite (Z) atthe following volumetric proportions: 3sandy loam soil:8 pumice:4 perlite:4 com-post:1 zeolite. The materials were mixeduniformly. The sandy loam soil had 77.0%sand, 7.8% silt, 15.2% clay and 0.703%(w/w) organic matter, a pH of 8.63, and anelectrical conductivity (EC) of 80 mS·cm–1.Pumice had a granulometry of 0.05 to 8 mm(LAVA; Mineral & Quarry A.D., Athens,Greece). Zeolite was a clinoptilolite witha granulometry of 0.8 to 2.5 mm (S & BIndustrial Minerals A.D., Athens, Greece)and perlite (Perloflor; ISOCON A.D.,Athens, Greece) had a particle distributionof 0.25 to 5 mm. The compost comprisedstraw, sawdust, yard waste (clippings andwood chips), and dairy cow, horse, andchicken manure. The second substrate wascommercially available (Zinco, Egreen)based on crushed tiles (the exact compositionand proportions falls under company pro-prietary restrictions).

Half the plots were randomly assigned toeach of the two substrate depths (either 7.5 cmor 15 cm). Plots were then filled to their re-spective depths (7.5 and 15 cm) with thetwo substrate mixtures. After substrates wereplaced into the plots, light compression andleveling were applied.

Substrate measurementsThe substrate granulometry was deter-

mined using a series of sieves of varioussizes (ASTM D 2974-07a, 2007) that vi-brated on a mechanical shaker (VibratorySieve Shaker AS 200 basic; Retsch Ltd.,Haan, Germany). The particle distributionof each material was plotted on a logarithmic

scale to investigate their compliance with thecorresponding FLL-ForschungsgesellschaftLandschaftsentwicklung Landschatsbau. e.v.(2008) specifications.

Water potential curves of the substrateswere determined according to Nektarios et al.(2011b) using a 100-cm column comprised of20 rings, each 5 cm in height. Total porosityand dry and saturated bulk density of eachsubstrate were determined using lysimeters15 cm in height with a 30.5 cm inner diameter(Nektarios et al., 2011a). The pH and EC ofthe substrates were determined using a 1:5(w/w) substrate-to-water extraction.

Turfgrass establishment and maintenanceManilagrass was established by plugging

on 19 Aug. 2009, 1 year before the initiationof moisture deficit study to achieve adequateand uniform growth between plots. Plugshaving dimensions of 5 cm · 5 cm wereprepared from washed sod. At each experi-mental plot, 20 plugs were planted in a grid-like pattern with a spacing of 15 · 15 cmbetween the plugs. Zoysia matrella ‘Zeon’was selected as the turfgrass species becauseit exhibits several desirable characteristicssuch as: 1) increased tolerance to extendedperiods of moisture deficit and shade incomparison with other cultivars (Morris,2001; Qian and Engelke, 1997); 2) lateral(creeping) growth habit with limited ver-tical growth that require minimal mowing(Volterrani et al., 2008); 3) natural preven-tion of weed encroachment as a result ofvigorous root, stolon, and rhizome produc-tion that creates a compact and thick mat (DeLuca et al., 2008a); 4) cold tolerance thatpermits earlier green up in spring and longerretention of fall color (De Luca et al., 2008b;Patton and Reicher, 2007); and 5) tolerance toinsect and disease problems (Anderson et al.,2007; Patton, 2010).

Fig. 1. Monthly precipitation and maximum, minimum and average air temperature for the 2 study years(2010 and 2011). The gray areas correspond to the two moisture deficit periods in which dailymaximum, minimum, and average air temperature are presented.

930 HORTSCIENCE VOL. 48(7) JULY 2013

During the establishment period (19 Aug.2009 until 1 July 2010), plots were irrigatedaccording to seasonal demands. More specif-ically from 19 Aug. to 15 Oct. 2009, plotswere irrigated once a day to promote Man-ilagrass establishment and prevent moisturestress. Irrigation initiated again after dor-mancy termination (20 Mar. 2010) and wasapplied every other day until the initiation ofmoisture deficit treatments.

Plots were left unmowed during 2009 asa result of the very limited upward growth ofManilagrass sward. In 2010 and 2011, plotswere mowed to a height of 60 mm in bi-weekly intervals from June until the initiationof moisture deficit periods. Water-solublegranular fertilizer (Complesal 12-12-17 witha nitrogen content of 6.5% as NH4

+ and5.5% as NO3, 12N–5.2P–14.1K–1.2Mg–8.0S;Agrevo Hellas S.A., Athens, Greece) wasapplied at a rate of 25 g·m–2 during Manila-grass establishment on 21 May, 25 June, and18 July 2010.

Water inputs during moisture deficitperiods and recovery

Moisture deficit treatments were repeatedin time (years 2010 and 2011). In 2010,moisture deficit period initiated on 10 Aug.and ended on 7 Sept., whereas in 2011, itinitiated on 10 Aug. and ended on 20 Sept.Termination of each moisture deficit periodcoincided with the first fall rainfall for bothstudy years.

Moisture deficit periods. Before the initi-ation of each stress period, all plots wereirrigated to saturation to produce uniform wetconditions. From then on, irrigation was ap-plied every 3 d at a rate of 3 mm for the lowirrigation regime and 6 mm for the highirrigation regime. Based on 10-year data ofthe Laboratory of General and AgriculturalMeteorology of the Agricultural University ofAthens, the average actual evapotranspirationfor the months of July and August is 5 mm·d–1

(Kotsiris et al., 2012a). Thus, the appliedirrigation regimes of 6 mm and 3 mm appliedin 3-d intervals corresponded to 40% and 20%of actual evapotranspiration, respectively. Plotswere hand-irrigated with a hose equippedwith a Wohnungswasserzähler Typ ETR flowmeter having a delivery accuracy of 0.1 L(Rossweiner Armaturen & Messgeräte GmbH& Co. oHG, Rosswein, Germany).

Recovery periods. In fall of both studyyears, plots were left without irrigation toinvestigate their recovery based on naturalrainfall events. After winter, irrigation wasapplied daily in Spring 2011 until the initia-tion of the moisture deficit period to promoteManilagrass growth and to reach uniformturfgrass cover among treatments.

MeasurementsVolumetric substrate moisture content

was determined every 3 d before irrigationwith the use of a WET-2 sensor that measureda cylindrical sampling area of 65 mm in depth ·45 mm in width (Delta-T Devices, Cambridge,U.K.) connected to a HH2 handheld moisturemeter (Delta-T Devices) calibrated using the

substrate-specific methodology of Kargas et al.(2013).

Percentage of GTC was determined usingdigital images of each plot taken on eachirrigation date with a Canon IXUS 100 ISdigital camera mounted on the top of a sealedbox (0.7 m · 0.7 m · 1 m in height), whichwas constructed to secure consistent lightingconditions throughout the image-acquisitionprocedure. The box was equipped with fourfluorescent lamps of 1200 lumens each,which were mounted on the upper, inner sideof the box. Camera settings were maintainedconstant during the study comprising an

aperture of F3.2, white balance set to fluo-rescent, ISO sensitivity set to 200, and focallength of 33 mm. The images were in JPEGformat with a size of 1600 · 1200 pixels.Each image was cropped using Adobe Photo-Shop CS3 software (Adobe Systems Inc., CA)to remove any unwanted image portion. GreenTurf Cover percentage was determined usingdigital image analysis SigmaScan Pro Version5.0 software (SigmaScan Pro; Systat SoftwareInc., Chicago, IL) (Bousselot et al., 2010;Karcher and Richardson, 2005; Olmsteadet al., 2004; Richardson et al., 2001). Greenleaves were selectively identified in the

Fig. 2. (A) Particle size distribution of the two substrates (S15:Pum40:Per20:C20:Z5, where S = sandy loamsoil; Pum = pumice; Per = perlite; C = compost; and Z = zeolite in volumetric proportions that areindicated by their subscripts and commercial based on crushed tiles). The gray area represents thespecified particle size range by FLL-Forschungsgesellschaft Landschaftsentwicklung Landschatsbau.e.v. (2008) guidelines. (B) Water potential curve for each substrate. Values are the mean of threereplications (± SE).

HORTSCIENCE VOL. 48(7) JULY 2013 931

images by setting the hue range from 47 to107 and saturation from 0 to 100. In eachimage, GTC percentage was determined bydividing the number of green pixels with thetotal pixel count of each image.

Spectral reflectance was determined beforeirrigation between 1200 and 1400 HR usingtwo double-channel SKR 1800 sensors (SkyeInstruments Ltd., Powys, U.K.). Sensors hadcenter wavelengths of 650 nm (red) and 800 nm[near infrared (NIR)] and a bandwidth of �10 nm. The sensors were connected to a displaymeter (SKL 925 SpectroSense2+.GPS; SkyeInstruments Ltd.) and mounted on a telescopichand-held pole (SKL 910; Skye InstrumentsLtd.) with the downward-facing sensor 1.30 mabove the turf surface, giving a sampling areaof 0.25 m–2. Red and NIR reflectances wereused to calculate NDVI values.

Leaf RWC of Manilagrass was measuredas described by Turner (1981). A 1.5-g leafsample was collected from fully expandedleaves in each plot every 3 d, before irrigationbetween 1200 and 1400 HR. Leaves wereimmersed in deionized water for 24 h, andthe turgid weight of each sample was deter-mined. Subsequently, the leaves were oven-dried for 48 h at 75 �C, and their dry weightwas determined. The RWC was calculatedbased on the following formula:

RWC ¼ ðFW � DW ÞðTW � DW Þ · 100

where FW = fresh weight, DW = dry weight,and TW = turgid weight of the leaf tissue. RWCmeasurements ended whenever Manilagrassturned dormant and leaves’ green color was lost.

Meteorological data. Throughout the du-ration of the research, air temperature andprecipitation were monitored by the weatherstation of the Laboratory of General andAgricultural Meteorology of the AgriculturalUniversity of Athens, Athens, Greece, lo-cated 15 m away from the experimental site(Fig. 1).

Statistical analyses. Data were subjectedto statistical analysis using JMP Version 8statistical software (SAS Institute Inc., Cary,NC) following multivariate analysis of vari-ance (MANOVA). As a result of statisticaltreatment differences between years, data werepresented separately for each year. However,similar trends were observed in the data asdiscussed subsequently. The main researchinterest was focused on treatment effects forturf growth and quality factors (GTC, NDVI),RWC, and substrate moisture at each separatesampling date. MANOVA was performed tocompare treatment effects within each sam-pling date as a three-way factorial design(substrate type · substrate depth · irriga-tion regime). Means were separated usingFisher’s protected least significant differenceat a 0.05 P level (P < 0.05).

Results and Discussion

Substrate characteristics. Particle size dis-tribution differed between the two substrates.Only the substrate formulated by locally avail-able materials was within the particle size

distribution guidelines suggested by FLL-Forschungsgesellschaft LandschaftsentwicklungLandschatsbau. e.v. (2008) for extensive greenroof construction. The crushed tiles substrateslightly deviated from the guidelines becauseit was coarser for all particles sizes exceeding2 mm in diameter (Fig. 2A).

The substrate that was formulated bylocally available materials exhibited higherwater retention capacity for all tensions (Fig. 2B).Its increased water capacity was attributed tothe finer texture and to the creation of greaterporosity within the substrate that favored theretention of water at low tensions.

Saturated and dry bulk density was less inthe substrate formulated by locally availablematerials (8.2% at saturation and 19.3% whendry) compared with the crushed tiles sub-strate as a result of the lightweight nature ofits constituents. By contrast, total porosity ofsubstrate formulated by locally availablematerials was higher by 6.5% compared withthe crushed tiles substrate (Table 1).

Apart from the physical differences, sub-strates significantly differed concerning theirchemical characteristics as well. The substratewith crushed tiles had an organic content thatwas 2.8 times greater compared with thesubstrate composed of locally available mate-rials (Table 1). The increased organic matterof the crushed tiles substrate resulted in highertotal nitrogen, potassium, magnesium, andsodium compared with locally formulated sub-strate. Calcium was also significantly higherin the crushed tiles substrate presumably asa result of the excess of calcium contained inthe crushed tiles. In contrast, phosphorus wasthe only element that was found in similaramounts in both substrates. The pH andEC of both substrates were similar andwithin the optimum levels for green roofs(FLL-Forschungsgesellschaft Landschaftsent-wicklung Landschatsbau. e.v., 2008; Table 1).

Moisture deficit periods. When all datawere pooled over each study year, it wasobvious that substrate depth was the mostsignificant factor with deeper substrates pro-viding larger values for GTC, substrate mois-ture content, NDVI, and RWC comparedwith shallow ones (Table 2). Similar resultshave been reported by several researchers forvarious plants species growing in different

green roof substrate depths. In most cases,deeper substrates favored plant growth andsurvival and increased the number of plantspecies capable of providing sustainablegrowth on green roof systems. Nektarioset al. (2011a) found higher growth index ofthe native Dianthus fruticosus ssp. fruticosus,growing in 15-cm rather than 7.5-cm depth,whereas Dunnett et al. (2008) reported thatproductivity, species richness, and diversityof 15 herbaceous perennial grass and herbspecies were higher when a substrate depth of20 cm was used compared with a depth of10 cm. Similarly for 12 Sedum species Getterand Rowe (2008) found increased growth andcoverage at 10 cm and 7 cm compared with4-cm substrate depth. Benvenuti and Bacci(2010) reported a significantly greater vegeta-tion cover and growth of 20 species originatedfrom dry habitats produced when substratedepth increased from 10 cm to 15 cm.

Irrigation regime was also significant butto a lesser degree compared with substratedepth. The higher irrigation regime of 6 mmincreased GTC only in 2011, whereas NDVIand RWC increased in both study years. Incontrast, substrate type was indifferent forGTC, NDVI, and RWC Manilagrass values,although moisture content of locally formu-lated substrate exhibited increased moistureretention capacity (Table 2).

In situ substrate moisture content. Duringthe moisture deficit periods (10 Aug. until 7Sept. 2010 and 10 Aug. until 20 Sept. 2011),substrate moisture quickly declined duringthe first 10 d for all treatments (Fig. 3). Dif-ferences between moisture content were evi-dent for all treatments but the most pronouncedones occurred between the two substratedepths. The substrate formulated by locallyavailable materials retained higher moisturecompared with the crushed tiles substrate inboth study years except from the five lastsampling dates of 2010 (Fig. 3). These resultswere expected as a result of the observeddifferences in water potential curves betweenthe two substrates (Fig. 2B), where locallyformulated substrate retained higher moisturecontent compared with the crushed tiles one.

Substrate depth was the most influentialfactor because it provided the largest differ-ences for all measurements compared with

Table 1. Nutrient analysis, chemical and physical properties of substrate S15:Pum40:Per20:C20:Z5 (S =sandy loam soil; Pum = pumice; Per = perlite; C = compost; Z = zeolite in volumetric proportions thatare indicated by their subscripts) and commercial substrate based on crushed tiles.z

S15:Pum40:Per20:C20:Z5 Crushed tiles Method of analysis

Organic matter (%) 2.78 ± 0.18 7.71 ± 0.33 Loss on ignitionTotal N (%) 0.180 ± 0.02 0.46 ± 0.04 Kjeldahl methodExtractable P (mg·kg–1) 107.8 ± 3.8 113.7 ± 3.0 ColorimetricallyExchangeable K (mg·kg–1) 1,919 ± 28.2 1,303 ± 24.5 Atomic absorption

spectrophotometry(GBC 932A/A)

Exchangeable Mg (mg·kg–1) 415 ± 5.6 788 ± 7.3Exchangeable Ca (mg·kg–1) 1,510 ± 12.3 3,928 ± 28.2Exchangeable Na (mg·kg–1) 487 ± 3.2 633 ± 5.1pH 7.86 ± 0.006 7.55 ± 0.006 1:5 (w/w) extractionElectrical conductivity (mS·cm–1) 908.60 ± 5.49 877.3 ± 25.12 1:5 (w/w) extractionSaturated bulk density (kg·L–1) 1.320 ± 0.02 1.438 ± 0.03 LysimeterDry bulk density (kg·L–1) 0.824 ± 0.006 1.021 ± 0.004 LysimeterTotal porosity (%) 45.9 ± 0.7 43.1 ± 1.1 LysimeterzValues are the mean of three replications (± SE).N = nitrogen; P = phosphorus; K = potassium; Mg = magnesium; Ca = calcium; Na = sodium.

932 HORTSCIENCE VOL. 48(7) JULY 2013

substrate type and irrigation regime in bothstudy years (Table 2). The deeper green roofprofiles (15 cm) exhibited higher moisturecontent during the first six sampling dates

(Fig. 3). However, this situation was reversedafter the sixth and seventh sampling datesfor 2010 and 2011, respectively, because themoisture content at shallow depth (7.5 cm)

reached higher values compared with 15-cmdepth substrates. This pattern occurred inboth study years and has also been observedin other similar green roof studies (Ntoulaset al., 2012). The higher moisture content inthe shallow substrates was caused by theearlier induction into dormancy of the turf-grass growing in shallow depth comparedwith the deeper substrates as substantiatedby GTC and NDVI measurements (Figs. 4and 5). The earlier induction into dormancyreduced turf sward transpiration in shallowsubstrates and resulted in substrate moisturecontent increase. In contrast, plants growing indeeper substrates were still transpiring andthus depriving the substrate from its moisture.Irrigation regime also affected substratemoisture content because the 6-mm irriga-tion exhibited higher moisture content com-pared with 3 mm irrigation in both studyyears (Fig. 3).

Green turf cover. The GTC was affectedby all factors but to a different extent for eachone of them. Based on the observed magnitudeof the differences among the three factors,GTC was mostly influenced by substrate depthbut not as much by irrigation regime andsubstrate type (Fig. 4; Table 2). Differenceswere more pronounced during the first studyyear (2010) compared with the second one(2011) for all treatments, presumably as a re-sult of lower summer temperatures in 2011compared with 2010 (Fig. 1). Manilagrassgrowth started to decline 6 d after the initiationof moisture deficit and was completely haltedafter 18 d for all treatments. Limited growthand lack of turfgrass species survival havebeen observed in studies where restrictedsubstrate depth was combined with extremedrought conditions. Using digital image anal-ysis, Steinke et al. (2011) compared growthresponse of various bermudagrass and buffa-lograss cultivars under restricted (10 cm) andunrestricted soil depth during consecutive60-d drought. They reported that in 10-cmdepth, no grass survived and the sward wascompletely browned off within the first 20 dof drought imposition, whereas all entriessurvived when planted on unrestricted soildepth and subjected to the same droughtlevel.

GTC exhibited minimal differences be-tween the two substrate types during the firststudy year (Fig. 4). In that case, GTC wasslightly greater in the locally formulated sub-strate compared with the crushed tiles. Thiscould be attributed to the higher water contentof locally formulated substrate compared withthe crushed tiles one according to Figures 2Band 3. Taking into consideration that fertiliza-tion was not applied during moisture deficitperiods, the increased nutrient status of thecrushed tiles substrate (Table 1) was expectedto increase shoot and leaf production. Suchlush growth has been documented to negativelyaffect the drought tolerance of turfgrasses aswell as of other plant species. Ebdon et al.(1999) found increased evapotranspiration asshoot growth increased after nitrogen fertil-ization in Poa pratensis, whereas Nagase andDunnett (2011) reported that plant drought

Table 2. Statistical significance of the measured parameters.z

Treatment

Moisture deficit period

Substratemoisture

Greenturf cover NVDI

Leaf relativewater content

------ % v/v ----- -------- % ------- ------------------- -------- % -------

2010 2011 2010 2011 2010 2011 2010 2011

Substrate (S)S15:Pum40:Per20:C20:Z5

y 12.18 11.72 42.86 33.79 0.6284 0.6316 62.25 66.93Crushed tile 9.59 8.42 41.49 33.39 0.6132 0.6299 59.89 66.40

Significance *x *** NSw NS NS NS NS NS

Depth (D)Shallow (7.5 cm) 10.09 9.07 35.17 28.81 0.5737 0.5856 56.20 62.60Deep (15 cm) 11.68 11.07 49.17 38.37 0.6679 0.6760 65.94 70.64

Significance NS * *** ** *** *** * *

Irrigation (I)3 mm 10.20 10.87 39.08 29.86 0.5980 0.5820 56.94 61.686 mm 11.57 9.28 45.27 37.32 0.6436 0.6795 65.20 71.56

Significance NS NS NS * * *** * *LSD 2.132 1.660 8.489 7.097 0.0409 0.0292 7.90 7.54

InteractionsS · D NS NS NS NS NS NS NS NSS · I NS NS NS NS NS NS NS NSD · I NS NS NS NS NS NS NS NSS · D · I NS NS NS NS NS NS NS NS

zMeans derive from pooled data over each moisture deficit period of either 2010 or 2011.yS = sandy loam soil; Pum = pumice; Per = perlite; C = compost; Z = zeolite in volumetric proportions thatare indicated by their subscripts.x*, **, *** Significant at P < 0.05, 0.01, and 0.001, respectively.w

NS = nonsignificant at P < 0.05.NDVI = normalized difference vegetation index.

Fig. 3. Substrate moisture content (% v/v) as affected by substrate type (S15:Pum40:Per20:C20:Z5, where S =sandy loam soil; Pum = pumice; Per = perlite; C = compost; and Z = zeolite in volumetric proportionsthat are indicated by their subscripts and commercial based on crushed tiles), substrate depth (7.5 or 15cm), and irrigation regime (3 mm or 6 mm) during the moisture deficit periods. Values are the mean offive replications. Bars represents Fisher’ least significance difference (LSD) at P < 0.05.

HORTSCIENCE VOL. 48(7) JULY 2013 933

tolerance was reduced as the organic amend-ment of the substrate increased as a result ofthe promotion of lush foliage growth.

Substrate depth provided the most pro-nounced differences in both study years be-cause the deeper substrates improved GTCfor the whole duration of the study (Fig. 4).GTC improvement in the deeper profiles wasattributed to their increased water contentcompared with the shallow ones (Fig. 3). It wasalso expected that deeper substrates wouldhave produced more roots, thus exploitingwater from deeper layers of the 15-cm profileas compared with 7.5-cm substrate depth and,therefore, better withstood drought stress(Huang et al., 1997; Marcum et al., 1995)and, concomitantly, produced higher GTC(Karcher et al., 2008).

A high irrigation regime (6 mm) providedbetter GTC measurements compared with thelower regime of 3 mm. Differences betweenirrigation regimes followed those observed atsubstrates’ moisture content measurements(Figs. 3 and 4). Similar results were reportedby Fu et al. (2004) for Z. japonica visualquality ratings. The researchers observeda quick decline in visual quality ratings 20 dafter water stress imposition, when irrigation

was applied at 20%, 40%, and 60% of ac-tual evapotranspiration. The 20% irrigationresulted in the most severe visual qualityreduction followed by 40% irrigation. In ourcase, GTC reduction occurred within 10 d afterthe initiation of moisture deficit treatments,but this difference could be attributed eitherto reduced substrate depth compared withFu et al. (2004) or to a different response ofZ. japonica and Z. matrella to water stress asreported by White et al. (2001).

Normalized difference vegetation index.As an overall indicator of Manilagrass con-dition, NDVI further supported the findingsof previous measurements (Fig. 5). Substratetype provided significant differences onlyduring the first moisture deficit period (2010)when locally produced substrate exhibitedincreased NDVI values compared with thecrushed tiles substrate.

In contrast to substrate type, substrate depthprovided the most pronounced differences asin the case of GTC (Table 2). Although NDVIvalues were similar at the initiation of eachmoisture deficit period, the deeper substrate(15 cm) exhibited larger NDVI values com-pared with the shallow depth of 7.5 cm duringboth study years (Fig. 5).

The beneficial effects of deeper profilescould be attributed to several factors. First,deeper substrates contribute to increased watermoisture content as a result of their capacityto contain larger amounts of water (Fig. 3).Second, deeper substrates have been foundto ameliorate temperature extremes (Boivinet al., 2001). This effect is expected to im-prove further in highly porous media such asthose used for green roofing as a result of theinsulating performance of the air entrapped inthe substrate porosity (Kotsiris et al., 2012a).Third, increasing the substrate depth mighthave a direct impact on the inherited droughttolerance mechanism of each species. Huang(1999) pointed out that the drought tolerancemechanism of B. dactyloides and to a lesserextent of Z. japonica is stress avoidance byreducing surface roots and redistributingthe root system into deeper depths to exploitdeeper water reservoirs. However, on greenroof systems, shallow substrate profiles pro-hibit the expression of this drought avoidancemechanism and thus render the plants to in-creased stress. By increasing the substratedepth, the avoidance mechanism could bepartially used and might contribute to theincreased NDVI measurements.

Fig. 4. Green turf cover (%) as affected by substrate type (S15:Pum40:Per20:C20:Z5, where S = sandy loam soil; Pum = pumice; Per = perlite; C = compost; and Z =zeolite in volumetric proportions that are indicated by their subscripts and commercial based on crushed tiles), depth (7.5 cm or 15 cm), and irrigation regime(3 mm or 6 mm). Values are the mean of five replications. Bars represents Fisher’ least significance difference (LSD) at P < 0.05. The gray areas correspond tothe moisture deficit period of each study year (2010 and 2011).

934 HORTSCIENCE VOL. 48(7) JULY 2013

The higher irrigation regime resulted ingreater NDVI values in comparison with thelow irrigation regime for both study years.Differences were not as pronounced as forsubstrate depth during the first study year(2010). However, during the drought periodof the second study year (2011), differencesincreased substantially (Fig. 5). The increaseddifferences between the two irrigation regimeswere contributed to the lower temperatureextremes that occurred in 2011 compared with2010 that permitted better growth of turfgrasssward receiving the high irrigation regime.

Leaf relative water content. RWC is ameasurement that indicates the physiologicalstatus of the plant through the determina-tion of cellular water deficit. A high RWC isindicative of a plant with a higher turgidityand, thus, a better physiological status. TheRWC was similar between the two substratesfor both study years further substantiatingprevious GTC and NDVI observations that de-termined a similar behavior of the two sub-strates (Fig. 6).

In contrast, RWC was greater when plantswere grown in the deeper substrates andunder the 6-mm irrigation regime. Carmo-Silva et al. (2009) reported that Z. japonica

RWC started to decrease only after substratewater content decreased below 10%. Simi-larly, in our study, RWC decreased abruptlywhen substrate moisture content was 10% to15% v/v depending on the treatment (Figs. 3and 6). Fu et al. (2004) determined that waterstress imposition on Z. japonica ‘Meyer’ in-itiated RWC reduction after �20 d. Theyreported that RWC reduction was severe (lessthan 50% RWC) after 55 d when the appliedirrigation was 20% of the actual evapotrans-piration, whereas the 40% irrigation alwaysremained above 60% RWC. In our case, RWCreached 50% for both irrigation regimes (40%and 20% of actual evapotranspiration) afteronly 9 to 12 d from the initiation of moisturedeficit cycle depending on the treatment(Fig. 6). The significantly reduced time forRWC to reach 50% reduction in our study wasattributed to the limited green roof substratedepth compared with the unrestricted one ofFu et al. (2004).

Recovery. After the moisture deficit im-position, Manilagrass recovery potential wasable to be assessed only in 2010 because inyear 2011, natural precipitation was minimalin September and occurred mostly in October.Thus, Manilagrass did not have the necessary

time to recover before its induction to winterdormancy.

In 2010, the moisture deficit period endedin September coinciding with the first fallrainfalls and it was able to recover from alltreatments as indicated by GTC (Fig. 4) andNDVI (Fig. 5). Because irrigation was notprovided, recovery speed depended only onnatural fall rainfall and thus ManilagrassGTC did not reach the pre-deficit levels andfluctuated in accordance with natural pre-cipitation (Figs. 1 and 4). Two weeks after theoccurrence of precipitation events, Manilagrassrecovery reached 20% to 25% GTC, whichcoincides with the percent recovery (22%)reported by Qian and Fry (1997) for Z. japonica2 weeks after the termination of water deficittreatments.

During fall recovery in 2010, differencesobserved during the moisture deficit periodcontinued to exist indicating a carry on effecton Manilagrass GTC and NDVI. The positiveimpact of deeper substrate depth and higherirrigation was evident, whereas differencesbetween substrate types were eliminated.

In the recovery period of 2011, a singlemeasurement was performed because turf-grass growth and quality, as indicated by

Fig. 5. Normalized difference vegetation index as affected by substrate type (S15:Pum40:Per20:C20:Z5, where S = sandy loam soil; Pum = pumice; Per = perlite; C =compost; and Z = zeolite in volumetric proportions that are indicated by their subscripts and commercial based on crushed tiles), substrate depth (7.5 cm or 15cm), and irrigation regime (3 mm or 6 mm). Values are the mean of five replications. Bars represents Fisher’ least significance difference (LSD) at P < 0.05. Thegray areas correspond to the moisture deficit period of each study year (2010 and 2011).

HORTSCIENCE VOL. 48(7) JULY 2013 935

GTC and NDVI measurements, remainedstagnant as a result of the limited Septemberrainfall events (only 3.4 mm of precipitation,on 20 Sept. 2011; Fig. 1). In 2011, rainfallsoccurred mostly in October and thus, Man-ilagrass entered its winter dormancy beforeits recovery from water deficit treatments.

After the end of winter dormancy in Spring2011, Manilagrass was irrigated to reacha similar level of GTC that was necessary toinitiate the second moisture deficit period. Inthe spring period after dormancy, only deepersubstrates improved Manilagrass GTC (Fig. 4)and NDVI (Fig. 5) compared with shallow

profiles. In contrast, substrate types and irri-gation regime were similar except from a sin-gle date when high irrigation and substrateformulated by locally available materials hadgreater GTC (Fig. 4).

Conclusions

Substrate depth was the most significantfactor that improved growth and droughtresistance of Z. matrella ‘Zeon’, further con-firming that substrate depth is among themost important factors on extensive greenroof systems. The deeper substrate resulted in

improved drought tolerance of Manilagrasscompared with the shallow one. However,because load is a crucial issue on green roofconstruction, if substrate depth must be re-duced to 7.5 cm, then irrigation should exceed6 mm every 3 d to have adequate growth andsuccessfully overcome summer moisture def-icit periods.

Literature Cited

Anderson, W.F., M.E. Snook, and A.W. Johnson.2007. Flavonoids of zoysiagrass (Zoysia spp.)cultivars varying in fall armyworm (Spodop-tera frugiperda) resistance. J. Agr. Food Chem.55:1853–1861.

ASTM D 2974-07a. 2007. Standard test methodsfor moisture, ash, and organic matter of peatand other organic soils. American Society forTesting and Materials, West Conshohocken, PA.

Beard, J.B. and R.L. Green. 1994. The role ofturfgrasses in environmental protection and theirbenefits to humans. J. Environ. Qual. 23:452–460.

Beattie, D.J. and R.D. Berghage. 2004. Green roofmedia characteristics: The basic, p. 411–416.Proc. of 2nd North American Green Roof Con-ference: Greening Rooftops for SustainableCommunities Conference, Portland, OR, 2–4 June2004. The Cardinal Group, Toronto, Ontario,Canada.

Benvenuti, S. and D. Bacci. 2010. Initial agro-nomic performances of Mediterranean xero-phytes in simulated dry green roofs. UrbanEcosyst. 13:349–363.

Boivin, M., M. Lamy, A. Gosselin, and B. Dansereau.2001. Effect of artificial substrate depth onfreezing injury of six herbaceous perennialsgrown in a green roof system. HortTechnology11:409–412.

Bousselot, J.M., J.E. Klett, and R.D. Koski. 2010.Extensive green roof species evaluations usingdigital image analysis. HortScience 45:1288–1292.

Bousselot, J.M., J.E. Klett, and R.D. Koski. 2011.Moisture content of extensive green roof sub-strate and growth response of fifteen temperateplant species during dry down. HortScience46:518–522.

Carmo-Silva, A.E., A. Francisco, S.J. Powers, A.J.Keys, L. Ascensão, M.A.J. Parry, and M.C.Arrabacxa. 2009. Grasses of different C4 sub-types reveal leaf traits related to droughttolerance in their natural habitats: Changes instructure, water potential and amino acid con-tent. Amer. J. Bot. 96:1222–1235.

Claus, K. and S. Rousseau. 2012. Public versusprivate incentives to invest in green roofs: Acost benefit analysis for Flanders. Urban For.Urban Green. 11:417–425.

De Luca, A., M. Volterrani, M. Gaetani, N. Grossi,P. Croce, M. Mocioni, F. Lulli, and S. Magni.2008a. Warm-season turfgrass adaptation innorthern Italy. Proc. of the 1st ETS Conf.,19–20 May 2008, Pisa, Italy. p. 75–76.

De Luca, A., M. Volterrani, M. Gaetani, N. Grossi,P. Croce, M. Mocioni, and F. Lulli. 2008b.Warm season turfgrass adaptation in Europenorth of the 45� parallel, p. 496–501. In: Crews,D. and R. Lutz (eds.). Science and golf V. Energyin Motion, Inc., Mesa, AZ.

Dunnett, N., A. Nagase, and A. Hallam. 2008. Thedynamics of planted and colonising species ona green roof over six growing seasons 2001–2006: Influence of substrate depth. UrbanEcosyst. 11:373–384.

Dunnett, N. and A. Nolan. 2004. The effect ofsubstrate depth and supplementary watering onthe growth of nine herbaceous perennials in a

Fig. 6. Leaf relative water content (%), as affected by substrate type (S15:Pum40:Per20:C20:Z5, where S = sandyloam soil; Pum = pumice; Per = perlite; C = compost; and Z = zeolite in volumetric proportions that areindicated by their subscripts and commercial based on crushed tiles), substrate depth (7.5 cm or 15 cm), andirrigation regime (3 mm or 6 mm) during the moisture deficit periods. Values are the mean of fivereplications. Bars represents Fisher’ least significance difference (LSD) at P < 0.05.

936 HORTSCIENCE VOL. 48(7) JULY 2013

semi-extensive green roof. Acta Hort. 643:305–309.

Durhman, A.K., D.B. Rowe, and C.L. Rugh. 2006.Effect of watering regimen on chlorophyllfluorescence and growth of selected green roofplant taxa. HortScience 41:1623–1628.

Durhman, A.K., D.B. Rowe, and C.L. Rugh. 2007.Effect of substrate depth on initial growth,coverage, and survival of 25 succulent greenroof plant taxa. HortScience 42:588–595.

Ebdon, J.S., A.M. Petrovic, and R.A. White. 1999.Interaction of nitrogen, phosphorus, and po-tassium on evapotranspiration rate and growthof kentucky bluegrass. Crop Sci. 39:209218.

FLL-Forschungsgesellschaft LandschaftsentwicklungLandschaftsbau e.V. 2008. Richtlinien f€ur diePlanung, Ausf€uhrung und Pflege von Dach-begr€unungen. Richtlinien f€ur Dachbegr€unun-gen (Guideline for the planning, execution andupkeep of green-roof sites). Bonn, Germany.

Fu, J., J. Fry, and B. Huang. 2004. Minimum waterrequirements of four turfgrasses in the transi-tion zone. HortScience 39:1740–1744.

Getter, K.L. and D.B. Rowe. 2006. The role ofextensive green roofs in sustainable development.HortScience 41:1276–1285.

Getter, K.L. and D.B. Rowe. 2008. Media depthinfluences Sedum green roof establishment.Urban Ecosyst. 11:361–372.

Huang, B. 1999. Water relations and root activitiesof Buchloe dactyloides and Zoysia japonica inresponse to localized soil drying. Plant Soil208:179–186.

Huang, B., R.R. Duncan, and R.N. Carrow. 1997.Drought-resistance mechanisms of seven warm-season turfgrasses under surface soil drying: II.Root aspects. Crop Sci. 37:1863–1869.

Karcher, D.E. and M.D. Richardson. 2005. Batchanalysis of digital images to evaluate turfgrasscharacteristics. Crop Sci. 45:1536–1539.

Karcher, D.E., M.D. Richardson, K. Hignight, andD. Rush. 2008. Drought tolerance of tall fescuepopulations selected for high root/shoot ratiosand summer survival. Crop Sci. 48:771–777.

Kargas, G., N. Ntoulas, and P.A. Nektarios. 2013.Moisture content measurements of green roofsubstrates using two dielectric sensors. Hort-Technology 23:177–186.

Kotsiris, G., A. Androutsopoulos, E. Polychroni,and P.A. Nektarios. 2012a. Dynamic U-valueestimation and energy simulation for greenroofs. Energy Build. 45:240–249.

Kotsiris, G., P.A. Nektarios, and A.T. Paraskevopoulou.2012b. Lavandula angustifolia growth andphysiology is affected by substrate type and

depth when grown under Mediterranean semi-intensive green roof conditions. HortScience47:311–317.

Latocha, P. and A. Batorska. 2007. The influenceof irrigation system on growth rate and frostresistance of chosen ground cover plants on exten-sive green roofs. Ann. Warsaw Univ. of Life Sc.–SGGW, Hort. and Landscape Architecture 28:131–137.

MacIvor, J.S. and J. Lundholm. 2011. Performanceevaluation of native plants suited to extensivegreen roof conditions in a maritime climate.Ecol. Eng. 37:407–417.

Marcum, K.B., M.C. Engelke, S.J. Morton, andR.H. White. 1995. Rooting characteristics andassociated drought resistance of zoysiagrasses.Agron. J. 87:534–538.

Monterusso, M.A., D.B. Rowe, and C.L. Rugh.2005. Establishment and persistence of Sedumspp. and native taxa for green roof applications.HortScience 40:391–396.

Morris, K. 2001. National zoysiagrass test 1996.National Turfgrass Evaluation Program. FinalReport 1997–2000, NTEP No. 01-15.

Nagase, A. and N. Dunnett. 2010. Drought toler-ance in different vegetation types for extensivegreen roofs: Effects of watering and diversity.Landsc. Urban Plan. 97:318–327.

Nagase, A. and N. Dunnett. 2011. The relationshipbetween percentage of organic matter in sub-strate and plant growth in extensive greenroofs. Landsc. Urban Plan. 103:230–236.

Nektarios, P.A., I. Amountzias, I. Kokkinou, andN. Ntoulas. 2011a. Green roof substrate typeand depth affect the growth of the native speciesDianthus fruticosus under reduced irrigationregimens. HortScience 46:1208–1216.

Nektarios, P.A., S. Kastritsis, N. Ntoulas, and P.Tsiotsiopoulou. 2011b. Substrate amendmenteffects on potted plant production and dryweight partition of Lantana camara. Hort-Science 46:864–869.

Ntoulas, N., P.A. Nektarios, K. Spaneas, and N.Kadoglou. 2012. Semi-extensive green roofsubstrate type and depth effects on Zoysiamatrella ‘Zeon’ growth and drought toleranceunder different irrigation regimens. Acta Agri-culturae Scandinavica, Section B—Plant. SoilSci. 62:165–173.

Olmstead, M.A., R. Wample, S. Greene, and J.Tarara. 2004. Nondestructive measurements ofvegetative cover using digital image analysis.HortScience 39:55–59.

Patton, A.J. 2010. Selecting zoysiagrasses: Pesttolerance and other criteria. Golf Course Mgt.78:80–85.

Patton, A.J. and Z.J. Reicher. 2007. Zoysiagrassspecies and genotypes differ in their winterinjury and freeze tolerance. Crop Sci. 47:1619–1627.

Qian, Y.L. and J.D. Fry. 1997. Water relations anddrought tolerance of four turfgrasses. J. Amer.Soc. Hort. Sci. 122:129–133.

Qian, Y.L. and M.C. Engelke. 1997. Evaluation ofzoysiagrass genotypes for shade tolerance.Texas Turfgrass Res. J. TURF-97:1–11.

Richardson, M.D., D.E. Karcher, and L.C. Purcell.2001. Quantifying turfgrass cover using digitalimage analysis. Crop Sci. 41:1884–1888.

Rowe, D.B., M.A. Monterusso, and C.L. Rugh.2006. Assessment of heat-expanded slate andfertility requirements in green roof substrates.HortTechnology 16:471–477.

Spronken-Smith, R.A. and T.R. Oke. 1998. Thethermal regime of urban parks in two cities withdifferent summer climates. Intl. J. RemoteSens. 19:2085–2104.

Steinke, K., D. Chalmers, J. Thomas, and R. White.2011. Bermudagrass and buffalograss droughtresponse and recovery at two soil depths. CropSci. 51:1215–1223.

Takebayashi, H. and M. Moriyama. 2009. Study onthe urban heat island mitigation effect achievedby converting to grass-covered parking. Sol.Energy 83:1211–1223.

Thuring, C.E., R.D. Berghage, and D.J. Beattie.2010. Green roof plant responses to differentsubstrate types and depths under various droughtconditions. HortTechnology 20:395–401.

Turner, N.C. 1981. Techniques and experimentalapproaches for the measurement of plant waterstatus. Plant Soil 58:339–366.

VanWoert, N.D., D.B. Rowe, J.A. Andresen, C.L.Rugh, and L. Xiao. 2005. Watering regime andgreen roof substrate design affect Sedum plantgrowth. HortScience 40:659–664.

Volterrani, M., N. Grossi, F. Lulli, and M. Gaetani.2008. Establishment of warm-season turfgrassspecies by transplant of single potted plants.Acta Hort. 783:77–84.

White, R.H., M.C. Engelke, S.J. Anderson, B.A.Ruemmele, K.B. Marcum, and G.R. Taylor.2001. Zoysiagrass water relations. Crop Sci.41:133–138.

Williams, N.S.G., J.P. Rayner, and K.J. Raynor.2010. Green roofs for a wide brown land:Opportunities and barriers for rooftop green-ing in Australia. Urban For. Urban Green.9:245–251.

Wolf, D. and J.T. Lundholm. 2008. Water uptake ingreen roof microcosms: Effects of plant speciesand water availability. Ecol. Eng. 33:179–186.

HORTSCIENCE VOL. 48(7) JULY 2013 937