Identifying Species from Marginal Habitats for Novel Use on Green Roofs

Sussman Internship Final Report Toby Liss

December 9, 2016

Cover Photo: Sisyrinchium mucronatum blooming on a serpentine barren in Pennsylvania. (Toby Liss)

Introduction

Green roofs are man-made “marginal” plant habitats with droughty conditions, low nutrient levels, and high environmental stress. This has historically limited green roof plant choices, resulting in low-diversity installations of non-native species. However, natural marginal habitats such as pavement barrens, grasslands, and cliffs experience similar conditions yet support remarkably diverse plant assemblages, hinting at unexplored options for green roof plantings. In addition to surviving on green roofs, marginal plant species may improve green roof function.

Elimination of stormwater runoff is a key green roof function that is not maximized by popular plant selections. Evapotranspiration (ET) moves water into the atmosphere and recharges the soil water holding capacity, but Sedum (the most widely used genus) is a succulent that drastically limits its transpiration during the day. With soil composition on roofs largely limited by weight, improved plant selection is the best opportunity to increase green roof stormwater elimination.

Green roofs are a unique experimental system to study diversity, productivity, and plant assemblage dynamics. As novel assemblages of plants become increasingly common, it will be critical to understand the functional response of such assemblages to a changing climate. This research could enable better green roof plant selection as well as answer basic ecological questions about the influence of diversity on ecosystem function.

Serpentine soils, which develop above shallow serpentinite bedrock, are fairly rare in occurrence although they appear across a wide geographic range. Their unique soil chemistry, toxic to many plant species, contributes to the development of rare ecosystems known as serpentine barrens and serpentine grasslands, which have high levels of plant species diversity and unusually high rates of endemism. Serpentine barrens are an example of a naturally occurring marginal system populated by plants that tolerate stress well. Serpentine soils tend to have low organic content, a low Ca:Mg ratio, high concentrations of metals such as iron and nickel (and to a lesser degree, chromium and cobalt), low organic content, low concentrations of nitrogen, phosphorus, and potassium, and low moisture content (1). Serpentine soils tend to have pHs ranging from slightly acidic to just above neutral. This combination of conditions, quite similar those on green roofs, makes serpentine vegetation a likely candidate pool from which to draw potential new selections for increasing biodiversity, and potentially function, on green roofs.

Study Objectives

At the outset of the study period, this project had two objectives:

1. Construct a comparative analysis framework in order to identify plant species with potential for use on SUNY-ESF’s new Academic Research Building (ARB) and green roofs in general based on their traits and tolerances.

2. Conduct a planting trial of species identified by the comparative analysis framework to assess survivability on a green roof.

As the summer progressed, two additional objectives emerged:

3. Conduct a vegetation survey and watering experiment on existing Gateway Building green roof installation to build long-term knowledge of impacts of irrigation treatments.

4. Draft recommendations for the design of the new Academic Research Building green roof and for research green roofs in general, using language appropriate for non-scientific audiences.

Methods

Objective 1: Plant Trait Comparative Analysis

The starting point for the comparative analysis was a list of species characteristic of the serpentine barrens of southeastern Pennsylvania and northern Maryland (2). Plant traits known to be important in green roofs performance were drawn from the scientific literature, and included sun and moisture tolerances, general habitat distribution, and habit.

A literature search was used to assemble all available information about each plant’s characteristics into a database. Ordinal ranks were assigned to each value for categories with limited possible values. For example, for sun tolerance, plants were described as tolerating full shade, part sun, part shade, or full sun. Plants documented to tolerate full sun were given a score of 3, any combination of part sun/part shade with any other value earned a score of 2, and any plant tolerating full shade earned a score of 1; species with no published data on sun tolerance were scored 0. For plant habit, herbaceous annuals and perennials were scored a 2, while woody shrubs and trees were scored 1 and 0, respectively. Categories with unlimited possible values, such as the text descriptions of a plant’s habitat distribution, were scored according to the judgement of the author on a scale of 1 to 5, again with 0 scored for any plant without published descriptions. Additional small scores increases were given for plants that have high conservation value. Value ranges for all scored categories were decided in terms of relative importance to one another, so that no category was given weight disproportionate to its importance. The largest portion of the total score was derived from published accounts of a species natural distribution and the degree to which that description aligned with known conditions on green roofs.

A plant’s total score was a sum of the category scores; the resulting list was edited based on practical considerations. Tree and large shrub species which would not be appropriate for a shallow green roof were removed from consideration, as were species that have already been shown in the literature to perform well on green roofs. The top-scoring candidates from the list were cross-checked with available species lists from nurseries in the Northeast and Midwest. A group of 14 species were selected from the top of the scored list that were available for purchase as mature plants (landscape plugs or small containers).

Objective 2: Planting Trial

Wooden frames were constructed to create trial beds with 3, 6, and 10 inches of green roof media. The frames were lined with 1-inch insulation board and landscape filter fabric on all sides. To allow drainage, ½-inch holes were drilled at 12 inches on center across the bottom insulation board of the frame.

The frames were filled with a standard green roof media blend and located on a south-facing roof terrace of a five-story building on the campus of the State University of New York College of

Environmental Science and Forestry (SUNY-ESF). Sun conditions were monitored on the rooftop and the frames were placed so that all frames received full sun throughout the entire day. Three replications of each frame depth were built and arranged into a completely randomized design. The frames were placed as far as possible from the building wall in order to avoid any wind buffering effects of the building.

Five individuals of each of the 14 plant species were installed in the green roof media in September of 2016. The plants were installed in randomized locations, 10 inches on center in a staggered pattern. Excess growth media from the plant production source was gently shaken off of the plants prior to installation. All plantings were watered regularly to maintain moisture during establishment.

Objective 3: Gateway Building Irrigation Experiment

The Gateway Building green roof was designed using species from two native New York marginal habitat types, the alvar grasslands and Great Lakes sand dunes. The plantings were designed with the goal of demonstrating that high biological interest, resilient green roof plantings can be comprised entirely of native species, if species are selected based on an understanding of the conditions tolerated by the plants in their native habitats.

The summer of 2016 was unusually hot, dry, and windy for the region, such that as of the end of June, Syracuse had approximately 3.4 fewer inches of rain than average for that point in the growing season. The Gateway Building green roof was designed as an unirrigated green roof, but many of the plants were dead or dying back by the end of June 2016 (Fig. 1). In order to strike a balance between commitment to the original design intent of the roof and the practical need to protect the investment and aesthetics of the installation, a watering experiment was devised for the roof.

Figure 1. Plant growth on the Gateway roof during June 2016. (Toby Liss)

The alvar grassland (interior beds) and Great Lakes dune (roof perimeter) plantings were each divided into plots of roughly equal area and conditions to receive a randomized control or irrigation treatment (Fig. 1). An initial vegetation survey was conducted at the start of treatments. Three randomly generated points within each plot were sampled using a 0.25 m2 gridded frame centered on the random

coordinates. No samples were taken from within 1 m of any edge of the plots. All living stems of non-grass species were identified and counted within the frame. Percent cover of grass species was estimated using the frame grid for reference. Unoccupied gaps on the soil surface were also quantified using this method. Areas of the dune planting that were outside the safety fence were not surveyed due to access restrictions; however, those areas were counted as part of the plot area, and were given the same treatment as the rest of the plot (i.e., it was possible to water over the fence but not to survey plants). Two triangular areas on the interior portion of the roof were excluded because skylight structures cast a significantly higher amount of shade than is received by the rest of the roof. The large alvar island bed was assigned the watering treatment not at random but by choice, to avoid large dieback in a highly visible area of the roof.

The watering treatment consisted of supplementing the weekly rainfall (measured using data from the weather station located at Syracuse’s Hancock International Airport, just north of SUNY-ESF’s campus) to maintain a seasonal average of around 0.75 inches of rain per week. Application amounts were determined by assessing the flow rate of water from the hose bib on the roof using a known-volume bucket and a stopwatch and then calculating the amount of time required to achieve a 1-inch equivalent of water based on the area of the treatment plot.

Figure 2. Schematic of Gateway roof with planted beds divided into roughly equal-area plots; W = watering treatment, C = control treatment, E = excluded areas. Schematic courtesy of Andropogon Associates, Ltd.

Objective 4: Guidelines Document for Research Green Roofs

As a part of the planning process for ARB, a document was requested that would communicate to professional designers some of the important design considerations for a green roof whose intended purpose is scientific research. This document was composed based on intern knowledge of the scientific process and a review of the green roof scientific literature.

Results

Objective 1: Plant Trait Comparative Analysis

See Appendix A for the list of plant species generated by comparative trait analysis, including the species ultimately selected for the planting trial.

Objective 2: Planting Trial

W

W W W

W C C

C

C

C C W

W

C

E

E

Plant were not installed until late August 2016. As a result, insufficient time had elapsed at the end of the summer to yet draw meaningful conclusions for this growing season about the survival status of the species tested. However, data will be collected after the last frost in spring of 2017 in order to determine whether any of the species tested are viable for use on green roofs. This data will provide important information about ability to overwinter in the northeastern US climate.

Objective 3: Gateway Building Irrigation Experiment

Distinct differences were observed between treatment and control plots within the first few weeks of beginning the irrigation treatment (Fig. 3). Follow-up vegetation surveys are planned for the spring and fall of 2017 using the same methods. The data will be analyzed to assess changes in density of each species and to make descriptive generalizations about the effect of the irrigation treatment on the vegetative characteristics of the roof plantings.

Figure 3. Stark contrast between irrigated (foreground, far background) and control (center) plots on the Gateway Roof in of 2016. (Toby Liss)

Objective 4: Guidelines Document for Research Green Roofs

See Appendix B for document “Considerations for the Design of a Green Roof Intended for Long-Term Research.” The contents of this document were presented at the Andropogon offices in Philadelphia, PA to the staff of landscape designers, architects, and researchers.

Conclusions

Continuing monitoring of the planting trial species is planned for 2017. Overwinter survival data will be collected in the spring and the plants will be observed over the 2017 growing season. Recommendations will be made for plant selections on the ARB based on the performance of the trial plants. Seeds have been purchased for additional test species from the serpentine barrens; propagation is planned for late winter/early spring of 2017 for additional trial species. Future experiments in this project are planned to assess evapotranspiration rates of individual species and assemblages of species for use on green roofs.

Acknowledgements

I would like to acknowledge the support of several people and organizations who directly contributed to the success of this project:

The Edna Bailey Sussman Foundation for funding this project, including additional support above and beyond the typical award.

Stancill’s, Inc. for their generous donation of green roof media, without which this project would not have been possible.

Lauren Mandel, PLA, ASLA at Andropogon Associates, for sponsoring me as an intern and guiding me in the development of this project.

Tim Toland at ESF, for his expertise and assistance with designing and constructing the planting trial frames, and his assistance with managing the Gateway irrigation experiment.

Donald Leopold, PhD, my major professor at SUNY-ESF, for assistance with developing this project and the generating the opportunity for me to work with Andropogon and be a part of the ARB planning process.

Marin Braco, Darren Damone, and Emily McCoy of Andropogon Associates, each of whom also contributed their ideas and expertise to this work.

Additional employees at Andropogon for their valuable comments and enthusiasm during my presentation.

Appendix A

Score Species Name Common Name Habit

18 Minuartia michauxii rock sandwort HA/HP

17 Phlox subulata moss phlox, moss-pink, creeping phlox HP

16.5 Panicum philadelphicum Philadelphia panic-grass HA

16.5 Sporobolus neglectus small rush grass, small dropseed HA

16 Asclepias verticillata whorled milkweed HP

15.5 Asclepias viridiflora green milkweed, green comet milkweed HP

15.5 Sporobolus vaginiflorus poverty dropseed, sheathed rush grass HA

15.5 Symphyotrichum ericoides white heath aster HP

15.5 Trichostema brachiatum false pennyroyal HA

15 Lilium philadelphicum wood lily HP

14.5 Eragrostis spectabilis purple lovegrass, tumblegrass HP

14.5 Lespedeza capitata round-headed bush-clover, round-headed lespedeza HP

14.5 Lobelia spicata spiked lobelia, palespike lobelia HP

14.5 Phemeranthus teretifolius round-leaf fameflower, quill fameflower HP

14 Dichanthelium acuminatum tapered rosette grass, Lindheimer panic-grass HP

14 Dichanthelium dichotomum cypress panic-grass, annulus panic-grass HP

14 Lespedeza virginica slender bush-clover, slender lespedeza HP

14 Phlox pilosa downy phlox, prairie phlox, fragrant phlox HP

14 Schizachyrium scoparium little bluestem HP

14 Viola sagittata var. sagittata arrow-leaf violet HP

13.5 Aristida oligantha prairie three-awn HA

13.5 Asclepias purpurascens purple milkweed HP

13.5 Carex glaucodea blue sedge HP

13.5 Paspalum setaceum slender beadgrass, Muhlenberg's hairy beadgrass HP

13.5 Quercus ilicifolia scrub oak, bear oak SD

13.5 Sorghastrum nutans Indian-grass HP

13 Antennaria plantaginifolia plantain-leaf pussytoe, woman’s-tobacco HP

13 Bouteloua curtipendula side-oats grama, tall grama HP

13 Carex retroflexa reflexed sedge HP

13 Cyperus lupulinus Great Plains flatsedge, sand sedge HP

13 Lechea minor thyme-leaf pinweed HP

12.5 Aletris farinosa white colicroot HP

12.5 Cerastium velutinum var. villosissimum serpentine chickweed, long-haired barrens chickweed HP

12.5 Lonicera dioica wild honeysuckle VP

12.5 Penstemon hirsutus hairy beardtongue HP

12.5 Pinus rigida pitch pine TE

12.5 Saxifraga virginiensis early saxifrage HP

12.5 Sporobolus heterolepis prairie dropseed HP

12.5 Vernonia glauca Appalachian ironweed, tawny ironweed, broadleaf ironweed HP

12 Andropogon gyrans Elliott's beardgrass HP

12 Chamaecrista fasciculata partridge-pea, prairie senna HA

12 Comandra umbellata bastard toadflax HP

12 Juncus secundus lopsided rush HP

12 Oenothera fruticosa ssp. glauca* sundrops, narrow-leaf evening-primrose HP

12 Quercus prinoides dwarf chestnut oak, dwarf chinkapin oak SD

12 Spiranthes vernalis spring ladies'-tresses, grass-leaved ladies tresses HP

12 Stylosanthes biflora pencil-flower HP

11.5 Castilleja coccinea Indian paintbrush HA

11.5 Cerastium velutinum var. velutinum barrens chickweed, field chickweed HP

11.5 Danthonia spicata poverty-grass, poverty oatgrass HP

11.5 Desmodium marilandicum Maryland tick-clover, smooth small-leaf tick-trefoil HP

11.5 Houstonia caerulea bluets, Quaker-ladies, azure bluet HP

11.5 Luzula multiflora field woodrush, common woodrush HP

11.5 Quercus marilandica blackjack oak TD

11.5 Sabatia angularis common marsh-pink, rose-pink HA

11.5 Solidago nemoralis gray goldenrod HP

11.5 Tridens flavus purpletop HP

11 Agrostis hyemalis fly-away grass, ticklegrass, winter bentgrass, southern hairgrass HP

11 Fragaria virginiana wild strawberry HP

11 Pycnanthemum tenuifolium narrow-leaf mountain-mint, slender mountain-mint HP

11 Sisyrinchium mucronatum needletip blue-eyed-grass HP

10.5 Cirsium horridulum yellow thistle, horrible thistle HA

10.5 Cystopteris fragilis fragile fern HP

10.5 Desmodium paniculatum panicled tick-trefoil HP

10.5 Galium boreale northern bedstraw HP

10.5 Heliopsis helianthoides ox-eye, smooth ox-eye HP

10.5 Juniperus virginiana eastern red-cedar TE

10.5 Liatris spicata blazing-star HP

10.5 Muhlenbergia mexicana Mexican muhly, satingrass HP

10.5 Pinus virginiana Virginia pine TE

10.5 Polygala verticillata whorled milkwort HA

10.5 Quercus stellata post oak TD

10.5 Scleria triglomerata whip-grass, whip nut-rush HP

10.5 Scutellaria serrata showy skullcap HP

10.5 Setaria parviflora perennial foxtail, marsh bristle-grass HP

10.5 Shepherdia canadensis buffalo-berry, buffaloberry SD

10.5 Smallanthus uvedalius bear's foot, leaf cup HP

10.5 Solidago bicolor silver-rod, white goldenrod HP

10.5 Symphoricarpos albus common snowberry SD

10 Arabis lyrata lyre-leaf rockcress, lyrate rockcress HB/HP

10 Ceanothus americanus New Jersey tea SD

10 Heuchera americana alumroot HP

10 Linum intercursum sandplain wild flax HP

10 Polygonum tenue slender knotweed, pleat-leaf knotweed HA

10 Potentilla canadensis dwarf cinquefoil HP

10 Rosa carolina pasture rose, Carolina rose SD

9 Symphyotrichum depauperatum** depauperate aster, serpentine aster HA/HP

n/a Calamintha arkansana*** Arkansas calamint HP

bold Indicates species included in the planting trial

* ssp. fruticosa substituted because glauca was not commercially available

** Species below score cutoff included because of expert recommendation and observed habitat distribution

*** Species not from serpentine barrens, included in trial because of expert recommendation

Species excluded because already growing successfully on ESF campus

Species excluded because of size impractical for shallow substrate depth

Appendix B

Considerations for the Design of a Green Roof Intended for Long-Term Research

Replication:

In any scientific experiment, it is crucial that the experimental conditions be replicated so that statistical inferences can be drawn from the differences (if any) found between the different “treatment” conditions (e.g. high water, medium water, low water). For example, you could grow two plots of plants, each with a different treatment condition, and measure the average runoff from each plot. However, without replicating the treatments (having at least two plots at each depth you are testing), you cannot perform statistical determinations of the significance of the differences in runoff volume between those two plots. Differences between them could be due to chance occurrences that are difficult to detect, such as one plot being located in a slightly warmer location, slight variations in media composition, or a subtle pattern in the prevailing winds. Without replication, there is no way to draw systematic conclusions about the true effects of the treatments you tried. The bare minimum number of replications in any scientific study is two.

A well-designed research green roof will allow the replication of a variety of treatment conditions that may be of interest. As you can see, as additional treatment variables are incorporated into a study, the number of individual plots needed tends to multiply. For example, if you wanted to test two different irrigation conditions at two depths, you would have four unique treatments to test. With the minimum number of replications being two, you would need eight individual plots in order to accomplish this experiment.

In general, a green roof designed for research should have distinct sections where varied treatment conditions can be created in replicated sets, each able to be measured independently for the results of interest, such as water runoff, vegetation growth, etc. In experimental studies, the assignment of treatments to plot locations is randomized. As such, any features that one plot has (such as the availability of irrigation) should be the same in all plots, so that treatments can be randomly assigned to a set of plots that are as equal as possible in their base conditions prior to the application of any treatments. It is advisable to complete a sun/shade study and consider possible rain shadows prior to deciding the arrangement of the plots in order minimize light, temperature, and moisture variability between the plots.

Watersheds:

Given the importance of replication in scientific studies and the interest in green roofs as a storm water management tool, the quantities of water entering and leaving the green roof are of high interest to researchers. A research green roof should have distinct watersheds so that all runoff from water striking a particular treatment plot on the roof should exit the roof from a unique and known outflow point.

Given the need for replication, a minimum of eight unique watersheds is recommended. Figure 1 shows an example of how this might be accomplished. A major watershed division is created using a ridgeline along the center of the roof, with each side subdivided into four treatment plots. Plots in the same watershed could be further subdivided if the experimenter were not interested in measuring runoff.

Figure 1. Watershed design for a research green roof, shown without substrate. (Toby Liss)

For plots in the same plane, it would be necessary to have a vertical physical barrier beneath the substrate surface to separate the water flow. V-shaped diverters channel the water flow to an exit point where instrumentation can be installed to measure flow volume and rate.

In order to minimize the number of downspouts from the roof, the flow from multiple watersheds can flow into a single downspout, as long as there is an access point where separate flows could be measured before they combine. For example, researches have devised a simple sealed barrier that allows instrumentation to be inserted prior to the combination of flows (3) (Fig. 2).

Figure 2. Watershed flow separator to allow independent measurement of multiple watersheds that flow into a combined downspout. Top left: overhead view of a downspout without flow

separation (measures a single watershed); bottom left: measures two watersheds; right: oblique view of a divided downspout (3).

Depth:

Given the multiplicative nature of experimental treatments (see “Replication”), a uniform depth is the most practical unless a very large roof is being designed. Variable depths create variable conditions that are hard to account for in experiments that aren’t trying to examine the effects of substrate depth, creating challenging “nuisance factors.” With a single-depth roof, temporary frames could potentially be set on the substrate surface and filled with additional substrate to create altered depth conditions for a particular experiment if necessary (Fig. 3), if they are within the weight limits of the roof.

Figure 3. A technique for increasing media depth on a subplot of a research green roof. (Toby Liss)

Instrumentation:

Any instrumentation or access points should be preferentially located at roof level, and made accessible without harnesses or other special safety equipment to maximize their utility to researchers. If feasible, data loggers should be incorporated into the design. These devices allow researchers to record measurements at defined intervals without being present on the roof. Data loggers with wireless or satellite capability store the data and then allow the researcher to wirelessly pull the data points, or automatically upload the data to cloud storage. A summary is presented in Table 1.

Table 1. Summary of desirable green roof instrumentation.

Function Device(s) Measure runoff flow rates and volumes for each watershed

Flow meters installed at downspouts

Measure soil temperature (at surface or below ground)

Temperature probes above or below substrate surface

Measure soil moisture Moisture probes below substrate surface

Measure precipitation, wind speed/direction, air temperature

Weather station centralized on roof

Control quantity and timing of irrigation Irrigation control panel Measure irrigation rates and volumes Flow meter incorporated into irrigation system Apply irrigation at researcher discretion Manual hose bib Record data at designated intervals and upload wirelessly

Roof devices wired into a data logger with wireless or satellite capability

Measurement/ coordinate system:

Vegetation is commonly sampled at regular coordinate intervals, or using coordinates generated randomly. Having a built-in grid with marked measurements along the edges of plots would eliminate the need to measure the same plots repeatedly to locate sample points. Metered edges of the roof or plots should be arranged perpendicularly. Although the metric system is the scientific standard, In the United States it’s probably worthwhile to include English units as well. Markings should be at least every centimeter or 0.5 inches. A system with an anchor point that could slide along the marker (similar to the sliding weights on a balance) would allow researchers to temporarily attach cording and easily find any pair of coordinates on the roof (Fig. 4).

Figure 4. Sliding anchor point system for locating coordinates on a green roof. (Toby Liss)

Irrigation:

Install an irrigation system, even if it will be used infrequently. This will allow the possibility for future experiments to easily apply irrigation treatments. Irrigation must be planned carefully so that the individual plots can be watered separately without crossover. Irrigation should be available in all plots so that future treatments can be assigned randomly and so that long term differences in soil conditions don’t arise between plots. Hose bibs should also be installed to allow for spot watering as needed.

Literature Cited

1. Brooks RR. Serpentine and its vegetation: a multidisciplinary approach. Dioscorides Press; 1987.

2. Latham R. Pink Hill Serpentine Barrens Restoration and Management Plan for the John J. Tyler Arboretum, Media, Pennsylvania. 2008 Dec 1 [cited 2016 Apr 4]; Available from: http://www.tylerarboretum.org/wp-content/uploads/2013/04/Latham-PinkHillReport-2008.pdf

3. Voyde E, Fassman E, Simcock R, Wells J. Quantifying Evapotranspiration Rates for New Zealand Green Roofs. J Hydrol Eng. 2010 Jun;15(6):395–403.

4. Latham, R E. 2015. Characteristic native vascular flora of the Pennsylvania serpentine barrens (version 2015-03-26). Unpublished; 5 pp. (contact: rel@continentalconservation.us)