Plant establishment on unirrigated green roof modules in a subtropical climate

Shallow-rooted plants were studied on unirrigated modular green roof trays. Four species had 100% survival, six had varied survival rates and five had no survivors. These outcomes suggest that water conservation practices can be an effective approach for green roofs.


Introduction
The use of green roofs as a self-reliant and low-input technology emerged in Europe, Scandinavia and the UK, where climate conditions are favourable for maintaining green roof vegetation without irrigation. Criteria for selecting vegetation for unirrigated green roof systems are articulated in the 'Guidelines for the Planning, Construction and Maintenance of Green Roofing', commonly referred to as the German FLL Guidelines for green roofs (FLL 2008). The guidelines state, 'green roofs are designed to depend primarily on precipitation for their water supply ' (p. 47), and thus the guidelines define performance expectations for substrate depth, composition and stability, porosity, nutrient-and waterholding capacity as well as plant watering requirements.
In Europe, hundreds of plant species have been identified for use on green roofs (Cantor 2008). As green roof technology continues to emerge beyond the climates of Europe, more information on the performance of a range of plant species on green roofs is needed. For example, in the semi-arid regions of Australia, the lack of data regarding plants for green roofs is a barrier to the growth of the green roof industry (Williams et al. 2010). In central Taiwan, 31 species of green roof plants were grown in planting dishes to test drought tolerance and seven species demonstrated normal growth (Liu et al. 2012). In North American climate zones that are much drier and warmer than Europe (Kottek et al. 2006), little is known about viable plant species on green roofs. In a recent review of North American green roof vegetation research, 40 succulent species and 94 herbaceous species were identified on green roofs across 15 ecoregions . Only a few species of plants have been found to support vegetation on shallow unirrigated green roofs in climates that frequently experience heat stress and drought Sutton et al. 2012).
Extensive-type green roofs are typically shallow (,12 cm) and are dominated by succulents, whereas simple-intensive green roofs are typically deeper (12-35 cm) and can accommodate forbs and grasses (Kö ehler 2003(Kö ehler , 2007Dunnett and Kingsbury 2004;Dunnett and Nagase 2007;Sutton et al. 2012). Succulents are a popular choice for shallow unirrigated green roofs because of their ability to tolerate well-drained soil, drought conditions and shallow substrate depths (Earth Pledge, 2005;Snodgrass and Snodgrass 2006;Cantor 2008). Many succulents found to be successful in Europe have also performed well on green roofs in North America in the Upper Midwest (Durhman et al. 2006;Rowe et al. 2012), the Pacific Northwest (Hauth and Liptan 2003), New England and Nova Scotia (Lundholm et al. 2010;MacIvor and Lundholm 2011;Barker and Lubell 2012). With the growth of the green roof market in the southern USA, it is important to understand which plant species are viable on green roofs in southern climates with or without irrigation.
The purpose of this research was to identify species and their survival on unirrigated green roofs in a subtropical climate that is characterized by hot and dry summers interspersed with large precipitation events. We selected 15 species that were known for their drought tolerance as well as their ability to withstand the occasional winter freeze. In addition, these species were chosen for their ability to maintain their root system in shallow, welldrained, soils. We hypothesized that all of these species would show good survival without supplemental irrigation after an initial irrigated establishment period.

Study site characteristics
The research site was located in College Station, Texas (30837 ′ N, 96820 ′ W), which lies south and east of the geographic centre of Texas in a humid subtropical climate (Larkin and Bomar 1983) at 150 m elevation. College Station typically experiences daytime maximum temperatures .32.0 8C over 100 days a year. College Station's mean annual precipitation is 1011.2 mm, although only 233.9 mm falls between June and the end of August, which coincides with high diurnal temperatures (Table 1).

Modular green roof trays
Plants were grown in 0.61 × 0.61 m rigid plastic modular green roof trays (Fig. 1A) (TectaGreen TM , Tecta America Corp., Skokie, IL, USA). The modules were 11.4 cm deep (4.5 ′′ ) with 8.9 cm (3.5 ′′ ) depth of FLL-compliant growth media (Rooflite w drain, Skyland, Avondale, PA, USA) and a 2.54 cm (1 ′′ ) depth of expanded shale filled inside the drainage retention cups (Fig. 1A). A non-degradable landscape fabric was provided by the green roof vendor and was placed between the two layers of substrate materials to maintain their separate functions. No fertilizer was applied during the investigations. Further details regarding the green roof system used in this study are described in greater detail in Aitkenhead-Peterson et al. (2010).

Plant selection
Our initial list of prospective plants included over 100 species. Several variables were used to narrow down the list, including a plant's reported capacity for drought tolerance or avoidance, cold and heat tolerance, ability to withstand sustained exposure to solar radiation and wind, adaptability to shallow substrates, capacity to reproduce, nativity to the region and plant life-form. Since the green roof substrate we investigated was shallow (8.9 cm) and reliable, and green roof guidelines for Europe (FLL 2008) suggest that forbs and grasses need more than 12.7 cm to thrive on green roofs, we looked to other forms of plants such as succulents and subshrubs. However, one species of grass (Nassella tenuissima) was investigated. The 15 species we investigated (Table 2) are a mix of native and exotic shallowrooted species that exhibit good resistance to drought

Plant installations
The initial investigation ran from 2 April 2009 to 19 October 2010. All plants were removed at the end of the study. Twenty-seven plants were established in three monoculture replicate trays (n ¼ 3) for Delosperma cooperi, Sedum kamtschaticum and Phemeranthus calycinus syn. Talinum calycinum. Nine 5-cm-deep nursery-grown plant plugs were installed and spaced 20.32 cm apart from each other in rows (Fig. 1B). Nine planted module replicates were placed in a completely randomized arrangement along the edge of the study platform (Fig. 1C). The second plant installation study began on 10 March 2010 and ended on 19 October 2010. All plants were left in place at the end of plant measurements. Nine additional green roof modules were installed at the research site identical to and adjacent to those used in the first study. New plant species included: Lampranthus spectabilis (10 cm spacing), Malephora lutea (10 cm spacing) and Sedum mexicanum (10 cm spacing). We began to investigate denser plant spacing to increase shading on the growth media and to help retain soil moisture.
Delosperma cooperi, Sedum moranense and P. calycinus were also planted but were mixed in trays with three of each species for a total of nine plants in three trays (n ¼ 3). Twenty-seven plants of Bulbine frutescens and S. moranense were studied in monoculture plantings (20.32 cm spacing) with nine plants per three tray replicates (n ¼ 3).
The third plant study began on 16 February 2011 with seven additional species including: Graptopetalum paraguayense, Dichondra argentea, Stemodia lanata, Nassella tenuissima, Manfreda maculosa, Myoporum parvifolium and Sedum tetractinum. Several species from the previous investigation were also re-examined: B. frutescens, M. lutea, S. mexicanum, L. spectabilis, S. kamtschaticum and P. calycinus. Plants were completely randomized into nine module replicates by three groups: succulents only, herbaceous species, and a mix of succulents and herbaceous species, with three trays per group (n ¼ 3). Plants were spaced 5-10 cm apart and were not organized in rows, to achieve a vegetative cover of mixed species.

Maintenance
Irrigation was applied only during the first several weeks of establishment, and only when natural rainfall

Plant measurements
Monthly plant growth measurements [growth index (GI)] and photographs were taken for D. cooperi, S. kamtschaticum and P. calycinus in 2009, and B. frutescens, D. cooperi, S. moranense, S. kamtschaticum and P. calycinus in 2010. A plant GI was devised as a measurement of the volumetric plant canopy area (cm 3 ) and porosity of each plant's canopy. This method is a modification to the measurement method initially used by Schroll et al. (2009). An idealized sphere was taken of the plant canopy with the longest width by the longest perpendicular width (Schroll et al. 2009); however, we also measured the mean canopy height as well. The GI was calculated by multiplying the height of the plant canopy by the twodimensional area of the plant canopy and the estimated percentage of live growth occupying the area of the canopy. Since plants were intermixed in trays, photographic or quadrant grid methods would not allow measurement of growth for intermixed species as the plants matured. Dead plants were left in place and were not included in the descriptive analyses. Weeds were not pervasive but were removed so that only the studied species were allowed to compete for resources.
For 2011 plant installations, photographs were taken once a month and a plant health rating was calculated at the end of 1 year of growth on 4 April 2012. The visual inspection resulted in plant health ratings based on the following: 1 ¼ severe decline; 2 ¼ some discolouring; 3 ¼ slight distress; 4 ¼ plant is healthy; 5 ¼ healthy and evidence of reproduction. Monthly growth means and standard errors of species cover of modular trays were analysed statistically to determine the growth rates and survival.
Species differences in plant health ratings and GI analyses were analysed using analysis of variance (ANOVA) (Stata 12 software, StataCorp, College Station, TX, USA) using a mean value for each species per replicate tray (usually n
Of the species studied in depth using a parametric survival analysis, P. calycinus was the only species with 100 % survival from Day 0 to Day 600 (Fig. 2). Median survival time for D. cooperi (655 days) was longer than that for S. kamtschaticum (223 days) and B. frutescens (191 days), while S. moranense (158 days) had a shorter median lifespan than all species except B. frutescens ( Fig. 2; Table 4).
Plant growth during the 2009 growing season for D. cooperi outperformed all other species with a maximum GI of 1131 cm 3 in December, but quickly declined after cold air temperatures damaged plants and top growth remained minimal and never fully recovered (Fig. 3) top growth had died back and it was assumed that the plants were dead by the end of the experimental measurements in October. All of the 2010 D. cooperi were left in place and none of them emerged in 2011; however, top growth emerged during the spring of 2012 but no measurements were taken. During July, S. kamtschaticum achieved a maximum GI of 1452 cm 3 , which was a greater volume than for any other species and maintained dominance until August (Fig. 3). The species P. calycinus performed consistently throughout the hottest and driest periods; however, its dormancy cycle begins in early fall, thus its GI of zero from November 2009 to March 2010 and October 2010 was due to dormancy (Fig. 3). During April 2010, the GI for B. frutescens was the highest of all species at 1507 cm 3 , but plant growth began to decline after maximum daytime air temperatures were consistently over 37.0 8C and dry conditions persisted (Fig. 4). The GI for S. moranense modules peaked at 1468 cm 3 in June, but declined quickly thereafter and all the plants were dead after 150 days (Figs 2 and 4).
Four species of the 2011 installations survived without any losses, including G. paraguayense, P. calycinus, M. maculosa and M. lutea (Table 3). Several species suffered some mortality, including B. frutescens, N. tenuissima, L. spectabilis, and S. mexicanum (Table 3). There were several species with no surviving plant replicates, including D. argentea, S. lanata, M. parvifolium, S. tetractinum and S. kamtschaticum (Table 3) 3 Graph showing a comparison of species (initial study) monthly GI (cm 3 ) means with maximum and minimum air temperatures (8 8 8 8 8C) and precipitation events (mm). The arrow points to the climate event when maximum air temperatures did not rise above freezing and the minimum temperature for the day was 27.8 8C.
rating of 3.7 and G. paraguayense had a mean health rating of 3.5 (Table 5).

Discussion
Our findings indicate that there are several species that performed well in south-central Texas with minimal watering during establishment and no watering thereafter even though area climate conditions were warmer and drier than long-term means, especially during 2011 when College Station was under exceptional drought conditions from 5 April 2011 to 28 March 2012 (Nielsen-Gammon 2011). On several green roofs in South Florida (tropical climate) establishment of plants on shallow green roofs (14 cm deep) without irrigation was tested and it was recommended that a minimum depth of 15 cm was needed to support plant growth (Livingston et al. 2004). Our plant establishment findings are the first report of species establishing on very shallow (,12 cm deep) green roofs in a humid subtropical climate with minimal irrigation during establishment and termination of irrigation thereafter. Our results demonstrate that it is possible to find plant species that can survive and even thrive on very shallow unirrigated green roofs in warm subtropical climates.
The top-performing species for survival included P. calycinus, G. paraguayense, M. lutea and M. maculosa. Only one species, P. calycinus, was found re-seeding onto other nearby green roof trays. This can be a desirable trait for green roofs, especially if the green roof is dominated by plant species that spread only by rhizomes or shoots, as long as the species is not aggressive or invasive in the landscape. Several species had decent survival rates and appeared to suffer from cold temperatures during their establishment year. Malephora lutea exhibited complete mortality during the winter of 2010, when an unusually long period (.24 h) of below freezing temperatures occurred (Fig. 4), but no mortality for those installed in 2011. All of the L. spectabilis died during 2010, but there was 44 % survival during the winter of 2011. The persistence of those two species through April 2012 demonstrates that they can establish when winter conditions are not abnormally cold (Table 1).
Delosperma cooperi showed a surprising capacity to survive even though top growth had ceased to exist. Plants suffered total canopy dieback in 2010, and failed to produce top growth during the following year, but re-emerged after consistent rainfalls returned to the region in late 2011 and early 2012. Our findings reinforce those by Rowe et al. (2012), where they also found changes in plant species success over multi-year periods. Dichondra argentea, S. lanata, M. parvifolium, S. tetractinum and S. kamtschaticum failed to establish in 2011, which was probably due to record drought and heat conditions during the spring and summer of 2011. Several S. kamtschaticum had successfully established during the 2009 experiment; however, irrigation was applied during 2009 in early summer. In 2011, plants were installed 2 months earlier than in 2009 and irrigation was stopped earlier. The establishment period with irrigation was apparently not long enough to help S. kamtschaticum survive the extreme dry 2011 summer without supplemental irrigation beyond May. College Station did not experience unusually high levels of night-time humidity throughout the 2011 growing season compared with the 2009/2010 growing seasons, and therefore the high mortality of S. kamtschaticum is probably not due to high night-time humidity. We suggest that S. kamtschaticum, D. argentea, S. lanata and S. tetractinum are still worthy of further study during more normal climate conditions or with deeper substrates, or perhaps with consistent irrigation throughout the entire first growing season. These species are drought and heat tolerant, but they have demonstrated difficulty establishing with limited irrigation under extreme drought and high-temperature conditions.
All of the species with 100 % survival were succulents. One gramminoid species was investigated (N. tenuissima) and it had a 22 % survival rate (Table 3). In terms of plant form, all of the top performers were erect plants except M. lutea. It is possible that the vertical stature favours survival in high-light environments by minimizing interception of solar radiation and thus reducing heat load and potential transpiration rates.
Since representatives of both native and non-native species had high survival, adaptability of plant species to the conditions of the substrate and microclimate (high light, wind exposure, local precipitation patterns) is perhaps a more important predictor of success than nativity of the species to the region (Durhman et al. 2006(Durhman et al. , 2007. From an ecological perspective, however, it would be better to make use of native plants on green roofs where possible, to provide habitat for native resident and migrating wildlife and insects. Testing plants for green roofs that also provide for local or migrating wildlife is plausible and could help conserve biodiversity with green roof vegetation (Kowarik 2011).  Table 5 Results of one-way ANOVA comparison (Bonferroni) of mean health ratings (measured on 4 April 2012) testing the significance between species.

Conclusions and forward look
The difference between species health rating means is shown, and those that are statistically significant (P , 0.05) are highlighted in bold. Plant species health rating key: 0 ¼ N. tenuissima, S. kamtschaticum and S. mexicanum-had varied performance. Five species-D. argentea, S. lanata, M. parvifolium, S. moranense and S. tetractinum-had no survivors. It is possible that the species with varied performance may perform better if provided with irrigation or deeper substrates. The outcomes of this study demonstrate that there may be several plant species for use on shallow unirrigated green roofs in humid subtropical climates, showing that green roofs are a viable alternative roofing type in spite of the more challenging climatic conditions.

Sources of funding
Funding for this study was provided by Texas A&M University, College of Architecture Research and Interdisciplinary Council Grant Program (CRIC); materials were donated by TectaAmerica TM , Rooflite TM , Emory Knoll Nursery and Joss Growers Nursery.

Contributions by the authors
All the authors contributed to a similar extent overall.