*Last entry is an hour before closing


*Last entry is an hour before closing


Green Roof Research: Cover and diversity

BRIT Green Roof Research Program

BRIT Urban Ecology Program

More about the project

Byerley Best, B. and R.K. Swadek. 2019. Plant cover & diversity on a young prairie-style green roof relative to slope position & planting assemblage. Poster presentation at Botany 2019, annual conference of the Botanical Society of America, 27-31 Jul 2019, Tucson, Arizona, USA. [link to poster]


Green roofs are an important tool in urban areas for ecosystem services such as storm water management, urban wildscape integration, and reduction of urban heat island effect. Green roofs in the southwest US, however, require additional design specifications (e.g., drought resistance, extreme temperature tolerance) that northern US green roofs don’t typically accommodate. A 20,000-square-foot test roof in North Central Texas was modeled after a local natural prairie system, Goodland Limestone Prairie Barrens/Glades, and incorporated harvested native prairie soil into the planting medium. As part of a larger effort to determine the efficacy of a biomimicry approach to this roof’s design, we examined plant cover and species richness relative to two roof slope positions (upslope and downslope) and three pre-determined planting assemblages at eight six-week intervals beginning about 20 months after the roof was first planted. We predicted that downslope areas would have higher coverage but fewer species due to faster establishment of a few dominant species where the soil moisture might be greater. Conversely, upslope areas should have lower vegetative cover due to drier conditions, but greater species richness from the more open community. The three plant assemblages each began with the same number of known species and an unknown but likely homogenous seed bank from the harvested soil; we thus predicted no difference in richness or cover among the groups when averaged over slope position. We found that species richness did not vary relative to slope position, but cover was consistently greater downslope than upslope. Planting assemblage did not appear to impact either richness or cover differentially when averaged over time. We saw an overall trend in all samples where cover generally kept increasing over time while richness peaked in the spring, troughed in the summer, and began to rebound in the fall, mirroring the bimodal precipitation for the region. This pattern is indicative of annual-dominated systems such as the one upon which the roof was modeled.


  • Two years after planting this green roof, species richness does not differ relative to slope position or vegetation type.
  • Total vegetative cover is higher downslope than upslope.
  • Species composition differed more by slope than by vegetation type.


  1. Slopes are a way to create niches on green roofs and therefore encourage higher diversity. Here the number of species (richness) didn’t differ between slopes but the species composition did, so total roof diversity still increased.
  2. The roof design included three diverse planting assemblages but each used the same soil with a homogenous seed bank. Our results could suggest that design effort could be saved by letting the slope do the work.
  3. The biomimicry design was successful. The patterns of both richness and cover over time mimic what we see in the native prairie analog: a system dominated by annuals that flourish with the spring and fall rains and then subside.
  4. It might not matter, actually. The roof was still establishing during this study. A reexamination at a later date should yield different results, as the effect of perennial dominants should be stronger.


Excerpted from “Soil-based green roofs” (Byerley Best et al. 2015, in Green Roofs Ecosystems):

“When the designers of the headquarters building for the Botanical Research Institute of Texas (BRIT) (Fort Worth, Texas, USA) decided to create a green roof, there were few precedents in the region. The idea of mimicking a local prairie habitat was put forth by community shareholders with hopes of the roof becoming a form of “reconciliation ecology” (Rosenzweig 2003), allowing habitat expansion for native species within a built environment. The landscape architects agreed to a biomimicry approach, and collaborating scientists and students started investigating possible template ecosystems.

The biomimicry strategy, also called habitat template strategy (Lundholm 2006), depends upon comparisons between a self-organizing, natural ecosystem and a model designed to simulate desirable aspects of that system. The sequence was to (1) explore and describe the wild system, (2) create a model system, (3) compare performance of the model system with the wild prototype, and (4) progressively refine the model to optimize desired functions. The advantage of this process brought learning about a poorly understood ecosystem in tandem with developing a regionally appropriate design for living roofs.

The Model Ecosystems: Fort Worth Prairie Barrens and Glades
Exploration refined project participants’ perception of the plant communities associated with limestone, leading to a fundamental distinction between “barrens” and “glades” (Quarterman 1989; Homoya 1994; Baskin & Baskin 2003; Swadek & Burgess 2012). Barrens are habitats where shallow soil over bedrock restricts plant growth, vegetation usually occurring on soils 5–25 cm (2–10 inches) deep over weathered limestone. Glades by contrast have extensive areas of exposed bedrock, with distinctive plant communities restricted either to soils in deeper crevices or soils less than 5 cm (2 inches) deep. In general, both of these habitats are dominated by short perennial bunchgrasses, prickly pear cactus, and yucca, which are hardy, drought-tolerant species, and annual grasses and forbs, which can complete their life cycle with seasonal rains. Patches of bare soil with cryptogamic crusts are also common.”

Woodland (background), Walnut barrens with Indian paintbrush ( Castilleja indivisa), and a Walnut glade (foreground) in the spring. In the late summer this habitat is typically very dry due to minimal, shallow soils. ( Photo Rebecca Swadek)


  • A total of 34 native species were chosen for the roof. Planting assemblages were designed to simulate native prairie plant communities, centering each assemblage of species around a single dominant species: Opuntia phaeacantha, Yucca pallida, and Bouteloua curtipendula (or Buchloe dactyloides).
  • Our sampling plots were 0.25 m^2 in size. This size was chosen based on species area curves for our native prairie analog. Plot locations were randomized by superimposing a grid over the assemblage sections across the roof, denoting a boundary for upslope vs. downslope, and using a random number generator to assign final sampling areas. Grid squares were 1.5 m^2, creating an allowance for smaller plots to move slightly and account for local trampling effects and impeding roof hardware or structural obstacles.
  • Sampling occurred every 6 weeks from March through November 2012. Six plots per assemblage (9 per slope position) were assessed at each sampling interval.
  • For each plot, we listed every plant species, their relative percent cover, and whether alive of standing dead. We also calculated the total live vegetative cover as well as cover of bare ground and litter.
Roof map with planting assemblages and location of plots
Example of 2012 roof plots of all three assemblages and two slope positions

RESULTS TO DATE (in progress)

  • Out of 38 species included in the planting assemblages, 22 species were observed in or immediately near the plots during the 2012 sampling period. 94 additional volunteer species were observed in or near the plots.

Top 5 dominant families by number of volunteer species

FamilyNo. spp.
Species richness did not differ by slope or by assemblage. 
While percent vegetative cover was consistently greater downslope than up, cover did not vary by assemblage.
Percent dissimilarity of species for three assemblages. Cactus and grass assemblages are least dissimilar (i.e., most similar), with 39% species dissimilarity (39% of their species are unique to only one assemblage; they share 61% of their species). When species lists are compared relative to slope (ignoring assemblage), they are more dissimilar than when compared by assemblage. Downslope and upslope plots only share about half of their species (49% dissimilar, 51% similar). 
  • Volunteers. Plant species found on the roof that were not part of the three planting assemblages have three potential sources: seed bank from the native soil incorporated into the planting medium, potting soil of nursery-grown transplants used in the assemblages, and colonization. Colonization events were likely both pre-installation (while roof trays were tended at ground-level for three months in a temporary nursery) and post-installation (bird- and wind dispersal). Tracking provenance of volunteers will help us refine establishment and management recommendations.
  • What is the trajectory of the plant communities on the roof compared to the native prairie it was designed to mimic? Will the cover of spring annuals decrease over time towards a climax community? We don’t expect so. The Fort Worth Prairie (which includes Goodland/Walnut barrens modeled here), much like the rest of the southwest, experiences variation in annual species richness and cover based on winter and spring rainfall. We hope to reassess cover and diversity in 2022, ten years after the first assessment.
  • Click here to view a live video feed of the roof
BRIT green roof, May 2015
BRIT green roof, Nov 2018
Rebecca Swadek and Brooke Byerley Best on the roof, Aug 2013

Research Team