Data Package Metadata   View Summary

Study System and Experimental Design. Designed urban experiments can simultaneously address key questions in community ecology as well as complex environmental issues. Vacant and abandoned land is a major concern in many US cities since (1) it is rapidly expanding in many cities, (2) it suffers from a variety of management/mis-management regimes, and (3) while “green”, it harbors a plethora of non-native and/or invasive plant species that can serve as sources of invasion into more “natural” habitats. For this study, we took advantage of a local policy program and adopted 24 vacant lots located in an area of west Baltimore City, Maryland, USA. A vacant “lot” defines a land parcel where a residential rowhouse once stood. All adopted lots were 5 m wide and between 20-25 m long, and all were located within 1 km radius of each other. These lots served as our experimental study units -- a plot -- and each was seeded with different experimentally designed mixtures of native plants. The assembly of these communities was then documented for 3 years post-seeding. Due to logistical constraints, plot position was not uniformly distributed throughout the neighborhood (i.e., building demolition is not a random process) but treatments were allocated randomly to plots. Plots were delineated and separated from each other by a 1m buffer if located adjacent to one another. Unfortunately, lot age since abandonment was unknown, but all were previously occupied by a rowhome of similar construction type.

A species pool consisting of 28 plant species was developed for the experimental seeding of vacant lots (Fig 1). Most of the species were native to the Northeastern USA (all to the USA) and all were expected to be tolerant of the environmental conditions in vacant lots; i.e. neutral-basic soil that is dry and compacted with little organic matter. Seeds were sourced from Ernst Conservation Seeds (https://www.ernstseed.com), and local genotypes used when available. Of the 24 vacant lots, 4 were randomly selected to serve as unseeded controls, and the remaining 20 were seeded with compositionally unique mixes of 12 species each.

The seeded lots were assigned randomly to a fully crossed design of high and low functional and phylogenetic diversity, while maintaining constant species richness across treatments, and ensuring functional and phylogenetic diversity remained uncorrelated. We designed the diversity treatments by first simulating 5000 random communities, consisting of 12 species from our species pool. We then selected mixes that represented the upper and lower quartiles of functional and phylogenetic diversity, while ensuring that the presence of a particular species was not statistically confounded with any diversity treatments. We selected three categorical functional traits to measure functional diversity of the seed mixes: start of bloom time (spring, summer, fall), functional group (forb, legume, grass), and life history strategy (annual or perennial). We estimated functional diversity of each community using the functional dispersion (FDis) index (Laliberte and Legendre 2010) with FD package in R ( R Core Team 2014). We also created a phylogenetic tree to determine phylogenetic diversity of each seeded mix. Using the open-source programs Phylomatic and Phylocom to assemble the phylogenetic tree, we estimated branch lengths and calculated phylogenetic metrics (Webb et al. 2008). We derived our tree from the most recent informal supertree available from the Angiosperm Phylogeny Website (mobot.org/MOBOT/research/apweb/), which is resolved to the family level, and then assigned ages, based on fossil records, to nodes (Wikström et al. 2001), and evenly distributed variance throughout the rest of the tree using the ‘bladj’ function in the phylocomr package in R (R Core Team 2014). We measured phylogenetic diversity of each seeded community using the metric Mean Nearest Taxonomic Distance (MNTD, Webb et al. 2002).

Soil Sampling. To understand the underlying soil conditions which are known to be quite variable in urban ecosystems, we collected approximately equal-sized soil samples from the top 5 cm of soil at each sampling plot using a hand trowel. A single sample was taken at a haphazardly chosen location near the center of the lot. We pooled samples and sent the air-dried soil samples to acquire nutrient availability analyses performed by the Cornell Nutrient Analysis Laboratories (Ithaca, NY) using the Mehlich-I extraction method. We report the following characteristics known to be highly variable in urban environments: soil moisture, pH, organic matter content, P, Al, Ca, Fe, K, Mg, Mn, Zn and NO3.

Site preparation and Planting Methods. Sites were prepared and seeded between March and April 2014. Each plot was mowed to approximately 5 cm, prior to application of a foliar herbicide (glyphosate) to remove all existing vegetation. Once the vegetation was dry (approximately 14 days), the soil surface was lightly raked to remove debris and expose the soil surface for seeding. Total seed mass was weighed to the higher end of recommended seeding rates for each plot area. Seeds were mixed on site with approximately 2 l of coarse sand, then broadcast seeded by hand in a cross pattern until the area was evenly covered. Following seeding, no additional management was conducted, apart from mowing to a height of approximately 15 cm, twice a summer in the first season, and once a summer the following seasons. Split-rail fences were installed around the edge of each block of experimental lots (Fig 2), and signs were installed to clarify the intent of the project to neighbors and city officials.

Experimental Design

We designated our treatment groups as:

Data Collection. Plant communities were sampled during peak growing season (July) of each sample year (2014, 2015, 2016), when the vegetation was at a minimum height of 10-cm, which allowed for accurate identification of species. We visually estimated percentage cover of each species in stratified, randomly placed 1-m2 quadrats distributed along two transects running perpendicular to the street and placed randomly along the front of the lot but at least 1 m away from the edge of the lot or any heavily-trafficked walkway. This resulted in a total of eight sampling quadrats per lot.

We identified 130 different species. Seventeen species could not be identified beyond family level, and several cool season grasses could only be identified to genus because they were not in flower at the time of the survey. Because the majority of unknown species were seedlings covering less than 1% of the sampling plot, we excluded them from analyses to avoid complications with interpreting the data. Species identifications were based primarily Del Tredici (2010).

Data analysis. We took several approaches to address the question, “does overcoming dispersal limitation by seeding native species in urban environments and increasing the functional or phylogenetic diversity of potential colonists increase native plant species diversity and abundance in urban vacant land?” First, we ran a one-way analysis of variance on proportional cover of native species and native species richness in the resulting plant community in each of the vacant lot treatments. We ran this analysis both in aggregate across all years, as well as separately by each study year. The groupings we compared were plots that were (i) planted, (ii) cleared but left unplanted, and (iii) unmanaged “control” lots, for which management was identical to ambient vacant lot management conditions. We separated the analysis of phylogenetic diversity of the seed pool (High, Low) from functional diversity (High, Low). As such, we present the results individually for each year (2014, 2015, 2016) as unmanaged plots were not able to be measured in 2014. We do not statistically compare unmanaged plots to the other categories; instead, we visually represented the unmanaged native diversity as a horizontal line for the average level of native diversity (for 2014 and 2015), and with dashed lines around it depicting the standard error. Given that there was no interaction between year × treatment for any of the models, we only present results averaged across years in the following sections, but do follow with analysis of differences following a Tukey’s HSD comparison procedure.

We relied on database trait measures (USDA PLANTS Database, plants.usda.gov; LEDA Traitbase, (Kleyer et al. 2008), the Royal Botanic Gardens Kew Seed Information Database, data.kew.org/sid/) to further address which traits confer native plant species establishment success. These included functional group, bloom time, life history strategy, seed weight, specific leaf area, expected height at maturity, rate of vegetative spread, and root morphology. This effort focused on establishment traits only, which differed from the broader set of traits used to organize the functional diversity seed mixture treatments. We then calculated the percent establishment of each species by determining the number of sites each species was seeded in divided by the number of sites where each species was scored as present at any point in study (which could be larger due to spread to unseeded sites). We also ran a redundancy analysis (RDA) model to determine which components of the matrix of traits predicted establishment success of the seeded species. RDA is a method of constrained ordination which is capable of partitioning variation across multiple explanatory factors. Because the functional trait model was not significant, we estimated establishment success of each species by calculated phylogenetic signal (Blomburg’s K), based on the phylogenetic tree created in Phylocom. This allowed us to determine whether closely related species were more likely to show similar rates in establishment success. This analysis was carried out in the phylocomr package in R. All statistical analyses were carried out in R version 4.0.2 (R Core Team 2020)

Del Tredici, P. 2010. Wild urban plants of the northeast: a field guide. Comstock Publishing, Ithaca, NY.

Laliberte, E., and Legendre, P. 2010. A distance-based framework for measuring functional diversity from multiple traits. Ecology 91: 299-305.

R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Webb, C.O., Ackerly, D.D., McPeek, M.A., and Donoghue, M.J. 2002. Phylogenies and community ecology. Annual Review of Ecology and Systematics 33:475-505.