Genotypic diversity and trait variance interact to affect marsh plant performance



  1. Intraspecific diversity can have important effects on population, community and ecosystem processes, yet we have little understanding of the relative importance of genetic- versus trait-based measures of intraspecific diversity.

  2. I conducted a manipulative field experiment of plant (Spartina alterniflora) genotypic diversity and trait diversity to examine their independent and interactive effects on plant performance and community structure. I focused on variation within and among genotypes in plant stem height, a trait that varies substantially across environmental gradients and can be an important predictor of plant competition intensity.

  3. Trait and genotypic diversity interactively affected multiple metrics of plant performance. Both stem density and spatial spread increased with genotypic diversity in the low trait diversity combinations, yet there were negligible to weak negative effects in the high trait diversity treatments. Spartina alterniflora percentage cover also varied with genotypic and trait diversity, but not in a clear linear pattern.

  4. There were no effects of trait or genotypic diversity on associated macrofauna above-ground, yet they interactively affected below-ground measures. Infaunal abundance and sediment oxygen availability mirrored the idiosyncratic response of plant percentage cover.

  5. Despite the interactive effects of genotypic and trait diversity, high trait diversity consistently increased plant performance in genotypic monoculture.

  6. Synthesis. The effects of intraspecific plant trait diversity on a range of plant and community responses in this study reinforce the premise that functional differences underlie ecological effects of genetic diversity and suggest that readily measured trait variance may serve as a valuable predictor of plant performance.


Intraspecific variation due to genetic, environmental and developmental factors can influence ecological processes at the population, community and ecosystem level (Bolnick et al. 2003, 2011; Whitham et al. 2006; Hughes et al. 2008; Dall et al. 2012). For instance, plant intraspecific genetic diversity (most often measured as the number of unique plant genotypes) has been experimentally linked to plant production and associated species abundance and diversity across a wide range of systems (Hughes & Stachowicz 2004, 2011; Johnson & Agrawal 2005; Reusch et al. 2005; Crutsinger et al. 2006; Kotowska, Cahill & Keddie 2010; Drummond & Vellend 2012). However, the mechanisms underlying these effects of genetic diversity, and even the effects themselves, can vary strongly based on genetic composition (Vellend et al. 2010, Hughes & Stachowicz 2011), reinforcing the idea that functional traits are critical to the relationship between diversity and ecological performance (Hooper et al. 2005; Cardinale et al. 2012). In fact, metrics that implicitly (genetic relatedness) or explicitly (functional trait diversity, phenotypic dissimilarity) incorporate information on functional trait variance among genotypes can serve as a better predictor of ecological performance than simply the number of genotypes present (Ellers et al. 2011; Stachowicz et al. 2013), much as phylogenetic relatedness can be more informative for understanding ecosystem function than the number of species (Cadotte, Dinnage & Tilman 2012; Dinnage et al. 2012; Cadotte 2013).

Although effects of genetic diversity necessarily result from corresponding variation in functional traits, we generally lack information on specific genotype–phenotype linkages aside from a few well-studied systems (e.g. Populus hybrid complexes: Schweitzer et al. 2004; Whitham et al. 2006). In addition, it can be challenging to measure a large suite of traits across many genotypes, and the most relevant traits will vary according to the target response variable (Hooper et al. 2005; Stachowicz et al. 2013). Moreover, focusing solely on variation between individuals ignores the potentially significant intraspecific variation occurring within individuals from environmental and/or developmental sources (Bolnick et al. 2003; Callaway, Pennings & Richards 2003; Miner et al. 2005; Albert et al. 2011). For instance, phenotypic plasticity at a single plant trait (e.g. root morphology) can affect competitive ability and species interactions (Callaway, Pennings & Richards 2003). Thus, a trait-based approach focusing on intraspecific trait variance per se, regardless of its source, could serve as an important predictor of population and community-level processes (McGill et al. 2006; Kraft, Valencia & Ackerly 2008; Best, Caulk & Stachowicz 2013).

In this study, I tested the independent and interactive effects of Spartina alterniflora genotypic diversity and phenotypic trait variance on population and community processes in a 1-year field experiment. Spartina alterniflora is a clonal saltmarsh foundation species that forms the basis of an ecologically and economically valuable intertidal ecosystem along the Gulf and Atlantic coasts of the United States (Pennings & Bertness 2001; Barbier et al. 2011). Because of the high physical stress of the marsh environment, vascular plant species diversity in this ecosystem is typically quite low, which is predicted to increase the relative importance of plant intraspecific diversity (Reusch & Hughes 2006). Spartina alterniflora exhibits extensive genetic (Richards et al. 2004; Travis, Proffitt & Ritland 2004; Edwards, Travis & Proffitt 2005; Hughes & Lotterhos 2014) and phenotypic diversity within sites (Richards, Pennings & Donovan 2005; Richards et al. 2010), and variation in a range of ecologically important traits (e.g. growth rate, morphology and competitive ability) has a genetic basis (Seliskar et al. 2002; Proffitt, Travis & Edwards 2003; Proffitt et al. 2005). However, the relative importance of environment-based versus genetic-based trait variation in this foundation species is unknown. Because saltmarshes are the primary target of costly restoration efforts in coastal habitats in terms of area restored (Grabowski et al. 2012), understanding the sources and impacts of marsh plant trait variation on growth and spatial spread has substantial economic as well as ecological implications.

I focused on trait variation in plant stem height, which varies in saltmarshes both developmentally (e.g. juvenile versus adult stems) and across environmental gradients (e.g. tall form versus short form; Valiela, Teal & Deuser 1978; Gallagher et al. 1988). Plant height can be a strong predictor of plant competition intensity (Cahill et al. 2008), and it is important for plant flooding tolerance in wetland plants (Jung et al. 2010). In addition, previous experiments in saltmarshes have demonstrated that stem height influences both plant–plant (Emery, Ewanchuk & Bertness 2001) and plant–consumer interactions (Hughes 2012; Zerebecki & Hughes 2013). I utilized variance in stem height within and across known plant genotypes to examine the population and community-level effects of genotypic diversity, as well as both inter- and intra-individual trait diversity. I conducted this experiment along the lower edge of an existing marsh to approximate conditions in colonizing marshes.

Materials and methods

Identification and Propagation of Spartina Alterniflora Genotypes

This study was conducted in St. Joseph Bay (SJB), Florida, a shallow protected coastal embayment along the northeastern Gulf of Mexico that is dominated along the shoreline by S. alterniflora. I collected single stems of S. alterniflora separated by at least 10 m with attached roots and rhizome from 16 natural marshes in SJB in summer 2009. I confirmed that the original plants were genetically distinct using eight DNA microsatellite loci designed specifically for this species (Blum et al. 2004; Sloop et al. 2005; see Hughes & Lotterhos 2014 for more details on the genotyping process). The stems were individually planted in a 50:50 mix of commercial potting soil (Miracle-Gro Organic Choice garden soil) and commercial pine mulch in 10.6 L (15.2 cm diameter) pots and propagated in a common glasshouse environment at the Florida State University Coastal and Marine Laboratory (FSUCML; located within 60 miles of St. Joseph Bay). Plants were transferred in November 2010 into wider 6.4-L pots (28.5 cm diameter) with a 50:50 mix of the same potting soil and sand to allow for greater propagation of new stems. Using commercially available soil provided consistency across pots and also prevented destructive sampling of marsh sediments in the field. Pots were then split into multiple 6.4-L pots monthly if the stem density was greater than 10 stems per pot. All plants were irrigated with freshwater daily, immersed in flow-through seawater from the Gulf of Mexico weekly and fertilized monthly with Miracle Gro All Purpose liquid fertilizer (NPK at 24:8:16 with trace elements B, Mo, Cu, Fe, Zn and Mn at a concentration of 2.81 g L−1). During the reproductive period (July to November), pots were visually checked daily and all S. alterniflora flowering stems were removed to maintain the distinct genotypic lines.

In preparation for transplanting to the field site in SJB, genotypes were separated into clusters of interconnected stems with associated roots and rhizomes and transferred from 6.4 L to 4.2 L (11.4 cm diameter) pots in fall 2011. These 4.2-L pots then served as the transplant units for our field experiment. One week prior to the start of the experiment, I counted stem density and measured stem height in each transplant unit; these height data were then used to assign genotypes to trait diversity treatments (see below). During the final week in the glasshouse, all experimental plants were submerged to just above the surface of the pot for 6 h daily in increasing amounts of seawater to facilitate the transition to field conditions.

Field Manipulation of Genotypic and Trait Diversity

I used the S. alterniflora genotypes propagated in the glasshouse to conduct a factorial field experiment of marsh plant genotypic diversity (one, three or six genotypes per plot) and trait diversity (high or low variation in stem height). I quantified non-destructive measures of plant productivity (stem density, stem height, plant percentage cover, plant spatial spread) over time to assess the effects of genetic and trait diversity while avoiding the introduction of disturbance that could interact with the diversity treatments (c.f., Hughes & Stachowicz 2011). Stem density and height are positively correlated with above-ground biomass in this system (Hughes 2012), and they serve as reliable proxies for above-ground biomass production. I also measured treatment effects on (i) the abundance of associated macrofauna (invertebrates and fishes) above-ground, (ii) the abundance of associated macrofauna (invertebrates) below-ground and (iii) sediment characteristics (accumulation rates, organic content, oxygen availability).

The genotypic diversity treatments tested in our experiment fall within the range found in natural S. alterniflora populations in St. Joseph Bay of 1–9 genotypes m−1 diameter plot (mean = 2.92; Hughes & Lotterhos 2014). I used a total of 12 unique genotypes in this experiment; no genotype was used in more than 3 of 12 3-genotype combinations or in more than 6 of 12 6-genotype combinations. Pairwise estimates of relatedness (r; Frasier 2008) of the experimental genotypes ranged from −0.55 to 0.47, with a mean of −0.10. Thus, these genotypes likely represent seedling-derived genets in an outbreeding population. Using stem height data collected in the glasshouse, I calculated the average stem height and coefficient of variation (CV) in height for each genotype. Because I wanted to include developmental sources of trait variation, these data include all stems present in the pots. Thus, variation in stem height may also be representative of variation in allocation to juvenile and adult stems. To assign individual genotypes to the high and low trait diversity genotypic monocultures, I ranked all genotypes by their CV in height; the six genotypes with the lowest CVs were assigned to the low trait diversity treatment, and the six genotypes with the highest CVs were assigned to the high trait diversity treatment. For the 3- and 6-genotype polycultures, I ranked each genotype by average stem height and then systematically created unique genotype combinations that either had low variation in height across genotypes (i.e. all short genotypes or all tall genotypes; N = 6 per level of genotypic diversity) or high variation in height across genotypes (i.e. a mix of short and tall genotypes; N = 6 per level of genotypic diversity). Low and high variance genotypes were distributed evenly across the genetic polyculture treatments. Initial trait diversity among genotypes in genetic polyculture was higher on average than initial trait diversity within genotypes in genetic monoculture, yet average stem height did not vary across treatment (Fig. 1).

Figure 1.

Trait diversity across treatments at the beginning of the experiment. (a) Variation in stem height was significantly different between high trait diversity and low trait diversity treatments at all levels of genotypic diversity. (b) There was no significant difference in average stem height across trait diversity or genotypic diversity treatments. Error bars represent ±1SE.

All experimental plots began with six transplant units, regardless of whether they were a monoculture or polyculture. Limitations on the number of shoots available per genotype necessitated these low initial densities. Initial stem density in the experimental plots did not vary across genotypic or trait diversity treatments. I planted two replicates of each genotype in monoculture and two replicates of each unique 3- or 6-genotype combination, for a total of 72 experimental plots (N = 12 per genotypic and trait diversity combination).

In October 2011, I created 72 1 m diameter plots in two rows of 36 plots (separated by at least 1.5 m) along the seaward edge of a natural S. alterniflora marsh in St. Joseph Bay, FL. Above- and below-ground vegetation was first removed by hand, and then, plots were tilled with a gas-powered tiller to ensure that all existing root material was killed or removed. Treatments were assigned to plots in a randomized complete block design, with two plots of each genetic diversity × trait diversity combination per block (Fig. S1). Transplants were planted into the plots over a 2-day period in November 2011, and associated invertebrates above-ground were counted at monthly intervals for 1 year. In June 2012, I also measured the height of three stems per transplant (when present) in each plot; due to relatively low clonal expansion rates during the winter and spring following planting, it was possible to track individual transplants until July 2012. After 11 months (October 2012), I measured environmental conditions (sediment oxygen availability, sediment organic content, sediment salinity, tidal height) in each plot. Sediment oxygen availability was estimated by measuring redox potential at 5 cm depth with a redox probe (Thermo Scientific Sure-Flow Probe) in four of the six experimental blocks; tide height and logistical constraints prevented sampling of the remaining two blocks. Sediment organic content was measured as percentage mass loss-on-ignition from a 5-cm deep and 2-cm wide sediment core taken in each plot divided into surface (0–1 cm) and subsurface (1–5 cm) portions. Sediment porewater salinity was collected from a depth of 5 cm in each plot using a porewater sipper and measured using a refractometer (Atago 2491 Master-S/Mill alpha, Atago USA, Inc., Bellevue, WA, USA). I also measured water height (cm) in each plot over a 10-min period at slack high tide to examine relative differences in tidal height. In addition to the environmental measurements, I quantified stem density, S. alterniflora percentage cover (by counting the number of occupied squares in a gridded quadrat), spatial spread (measured as (i) the longest distance in metres between two S. alterniflora stems in a plot and (ii) plot area) and the height of six stems per plot. I also counted associated invertebrates above-ground and sampled associated fish abundance and diversity by deploying a single cylindrical minnow trap (length = 42 cm; diameter at widest point = 22 cm; diameter of openings at each end = 2.5 cm) in each plot over one night-time high tide. To estimate above- and below-ground plant dry mass as well as infaunal abundance, I destructively sampled the experiment after all other measurements were completed by taking three 10-cm deep sediment cores in each plot. This depth included the majority of the S. alterniflora rooting zone in these plots.

Statistical Analyses

I conducted two analyses to evaluate the consistency of stem height and variance in stem height over the course of the experiment. First, I compared the average stem height of each genotype in the glasshouse to the average stem height of that genotype across all treatments in the field after 8 months (June 2012) using linear regression. Secondly, I compared the CV in height at the plot level from the time of planting to the CV in height at the plot level in both June 2012 and again in October 2012 using linear regression. In all analyses, I included trait diversity and genotypic diversity as fixed factors (including all possible interactions) to see if consistency in height or height variability differed by treatment.

I used a model selection approach to examine the independent and potentially interactive effects of S. alterniflora genotypic diversity and trait diversity (variation in stem height) on population and community parameters. For the experimental data taken in October 2012, I treated trait diversity and genotypic diversity as fixed factors and experimental block as a random factor. Because above-ground invertebrate abundances were low in October 2012, I also analysed the abundance of the three most common species (the snails Melongena corona, Terebra dislocata and Littoraria irrorata) and the combined abundance of all species over the course of the experiment, with trait diversity and genotypic diversity as fixed factors and sampling date and experimental block as random factors. By comparing the difference between the Akaike information criterion corrected for small sample sizes (AICc; Burnham & Anderson 1998) of a particular model and the lowest AIC observed (the AIC difference, or dAIC), I evaluated which of the models best explained the observed data (Bolker 2008). Models with dAIC values separated by <2.0 were considered not significantly different than one another (Richards 2005). I also calculated the Akaike weight (w) as the model likelihood normalized by the sum of all model likelihoods, with values close to 1.0 indicating greater confidence in the model. Akaike weights for the best models are provided in the text. All candidate models and their dAIC scores and Akaike weights are provided in appendices. Analyses were conducted using the lmer function in the lme4 package and the AICctab function in the bblme package, R statistical software, version 2.11.1 (R Foundation for Statistical Computing, Vienna, Austria).

To evaluate whether observed diversity responses differed from additive expectations, I calculated predicted values for a given genotypic diversity and trait diversity combination by summing the average monoculture response of each component genotype and dividing by six (the number of initial transplants).


Average stem height of each S. alterniflora genotype in the glasshouse was a strong predictor of average height after 8 months of growth in the field experiment (F1,10 = 26.21, P < 0.001; R2 = 0.72; height in field = 0.90 × height in glasshouse + 12.39; Fig. 2). In contrast, the predictive power of initial height variation for height variation observed over the course of the experiment differed by trait diversity treatment after 8 months (F1,64 = 4.86, P = 0.03; Fig. 2): the low trait diversity plots showed a positive relationship between initial and final trait variation (CV at 8 months = 0.17*Initial CV + 0.49; R2 = 0.15), yet there was no relationship in the high trait diversity plots (CV at 8 months = 0.02*initial CV + 0.53; R2 = 0.01; Fig. 2). This difference in the predictability of trait diversity treatments did not persist at 11 months (F1,62 = 1.97, P = 0.16). In addition, the relationship between initial and final trait diversity did not differ depending on the number of genotypes present at either time period (P > 0.35). In all subsequent analyses, I used the initial trait diversity assignments.

Figure 2.

Stem height as a metric of trait variability. (a) Initial stem height was a strong predictor of stem height after 8 months for the 12 genotypes used in the experiment, regardless of genotypic diversity treatment. (b) Initial plot-level variation in stem height was correlated with plot-level variation in stem height after 8 months for the low trait diversity treatment, but not the high trait diversity treatment.

Spartina alterniflora genotypic diversity and trait diversity interactively affected multiple measures of plant production, including S. alterniflora stem density, spatial spread and percentage cover (Fig. 3 and Table S1). In general, the positive effects of high trait diversity on plant productivity were most obvious in genotypic monoculture. The model including an interactive effect of trait and genotypic diversity had the highest weight, particularly for spatial spread (w = 1.0). For stem density and percentage cover, the interactive model had only moderately higher likelihood (density w = 0.50; percentage cover w = 0.32) than models including an additive effect of trait diversity alone (density w = 0.20; percentage cover w = 0.31) or additive effects of genotypic and trait diversity (density w = 0.22; percentage cover w = 0.23; Table S1). Both stem density (y = 0.72x + 41.24, R2 = 0.75; Fig. 3a) and spatial spread (y = 0.09x + 0.93, R2 = 0.85; Fig. 3b) increased with genotypic diversity in the low trait diversity combinations, yet there was no effect of genotypic diversity on stem density (y = −0.10x + 46.20, R2 = 0.01) and a weak negative effect on spatial spread (y = −0.06x + 1.54, R2 = 0.97) in the high trait diversity treatments. Percentage cover of S. alterniflora varied across genotypic and trait diversity treatments, but not in a clear linear pattern (Fig. 3c). Observed stem density and spatial spread in polyculture were consistent with additive expectations based on the performance of the component genotypes in monoculture (Fig. S2).

Figure 3.

Experimental effects of Spartina alterniflora genotypic and trait diversity on plant (a) stem density, (b) spatial spread and (c) percentage cover. The effects of trait diversity on plant production were strongest in genotypic monoculture, particularly for spatial spread. Genotypic diversity increased density and spatial spread in the absence of trait diversity. Error bars represent ±1SE.

The null model, excluding effects of genotypic and trait diversity, provided the greatest explanatory power for per stem above-ground biomass at the end of the experiment (w = 0.88; Table S1). Below-ground, the null model (w = 0.58) and a model including an effect of trait diversity (w = 0.38) had the greatest explanatory power (Table S1). There was also no difference in S. alterniflora reproductive stem production by genotypic or trait diversity (null model w = 0.75). Trait diversity alone (w = 0.38) and additive effects of trait and genotypic diversity (w = 0.18) influenced S. alterniflora stem height, but these differences were very slight (Fig. S3): average stem height was higher in the high trait diversity plots, and it decreased with increasing genotypic diversity (y = −0.35x + 36.49, R2 = 0.43).

Spartina alterniflora genotypic and trait diversity did not predict the abundance of the above-ground invertebrate community associated with these plants over the course of the experiment, either collectively or individually (null model w > 0.78; Table S2). Similarly, there was no effect of either measure of plant diversity on the abundance of fish associated with these plots after 1 year of growth (null model w = 0.80; Table S2). However, models including additive (w = 0.22) and interactive (w = 0.35) effects of genotypic and trait diversity, as well as effects of trait diversity alone (w = 0.30) predicted the total abundance of macrofaunal invertebrates (polychaetes, bivalves, gastropods) below-ground after 1 year (Fig. 4a, Table S2). Infaunal abundance differed idiosyncratically with genotypic and trait diversity in a pattern similar to plant percentage cover.

Figure 4.

Spartina alterniflora genotypic diversity and trait diversity interactively affect (a) below-ground invertebrate abundance and (b) sediment oxygen availability. Error bars represent ±1 SE.

Genotypic and trait diversity did not explain variation in sediment temperature, sediment porewater salinity, sediment organic content or tidal height (null model w > 0.75; Table S3). Interestingly, sediment oxygen availability varied interactively with plant genotypic and trait diversity (w = 0.82) in a manner similar to S. alterniflora percentage cover and infaunal abundance (Fig. 4b).


Genotypic diversity in Spartina alterniflora affected several metrics of plant production, but these effects varied according to the degree of variation in a key plant trait. When initial trait variation was high, there was generally no effect, or even a slight negative effect, of genotypic diversity on plant productivity. However, genotypic diversity had positive effects on stem density and spatial spread when initial trait variation was low. These results suggest that when intraspecific variation is low (i.e. genotypes with low trait variance growing in monoculture), the addition of either genotypic diversity or trait diversity has a positive impact on plant performance, yet these positive effects do not continue increasing linearly as additional genotypes are added. These results highlight the need to consider diversity at multiple taxonomic levels, as interactions among levels of diversity may be common (e.g. Hughes, Best & Stachowicz 2010; Crawford & Rudgers 2012, 2013; but see Fridley & Grime 2010).

Spartina alterniflora genotypic diversity and trait variation also interactively affected vascular plant percentage cover, but this relationship was idiosyncratic. Experimental plots had higher plant cover with high initial trait diversity when one or six genotypes were present, but the reverse was true when three genotypes were present. This finding could indicate increased competition with increased genotypic diversity that is only alleviated when trait variance is also high. Regardless of the underlying mechanisms, the effects of diversity on plant cover most likely caused the similar patterns in the abundance of associated macrofauna below-ground, as well as sediment oxygen availability. Plant structure above- and below-ground can provide refuge for infaunal organisms (Pennings & Bertness 2001); thus, infaunal abundance often mirrors plant cover. The response of plant cover and infaunal abundance, in turn, likely played a role in the similar pattern in sediment oxygen availability, as both plant roots and bioturbation by infauna can increase sediment oxygen levels (Bertness 1985, Pennings & Bertness 2001). However, the available data cannot rule out the possibility that variation in sediment oxygen availability was a cause, rather than a consequence, of variation in plant percentage cover (Castillo et al. 2005).

The effects of plant trait variation on a range of plant and community responses in this study reinforce the premise that functional differences underlie ecological effects of genetic diversity (Hughes et al. 2008; Ellers et al. 2011; Stachowicz et al. 2013). Interestingly, these effects occurred even though initial trait variation within a plot was not a strong predictor of trait variation in that same plot eight or eleven months later, particularly in the high trait variation treatments. The effects of trait variation on plant production were most obvious in genotypic monoculture: genotypes that exhibited greater initial variance in stem height had higher stem densities, spatial spread and percentage cover than less variable genotypes over the course of a year. In genotypic polyculture, the effects of initial trait diversity varied depending on both the response variable and the number of genotypes present. This contrast between the trait diversity treatments in genotypic monoculture and polycuture suggests that the source of trait variation may be important for its ecological effects: diversity in monoculture derived from variance within genotypes, whereas trait diversity in polyculture resulted primarily from differences between genotypes (but also included variance within genotypes). The differences between genotypic monoculture and polyculture in this study could also manifest due to the peculiarities of stem height as a trait: high variation in stem height within genotypes may be an indicator of actively growing and/or more productive genotypes (i.e. those that have a mix of shorter juvenile stems and taller adult stems). Additional manipulations that decouple trait and genotypic variation are needed to determine the generality of these results.

Phenotypic plasticity has been linked to increased competitive ability and fitness in a range of taxa (Callaway, Pennings & Richards 2003; Miner et al. 2005; Stomp et al. 2008). For example, plants with greater plasticity in stem elongation had higher fitness than those with less plasticity (Schmitt, McCormac & Smith 1995; van Kleunen & Fischer 2001), consistent with the positive effects of within-genotype stem height variance shown here. Stem height is correlated with plant nutrient content in S. alterniflora (A.R. Hughes, unpubl. data), and it may be correlated with other traits (e.g. root morphology, resource use) that influence plant competition and performance (Callaway, Pennings & Richards 2003; Ashton et al. 2010). However, the exact mechanism linking variance in stem height to increased plant performance in this experiment remains uncertain. Also unclear is the source of variation within genotypes. Novel phenotypes can arise from DNA methylation, RNAi or chromatin remodelling, even among identical genotypes (Jablonka & Raz 2009; Herrera, Pozo & Bazaga 2012). Such epigenetic variation can contribute to functional diversity (Cubas, Vincent & Coen 1999; Johannes et al. 2009; Bossdorf et al. 2010; Zhang et al. 2013) and could provide an explanation of some of the within-genotype variability observed here.

Spartina alterniflora trait and genotypic diversity did not influence associated community abundance or diversity above-ground in this study. The experimental plots were located at the leading edge of the marsh and had relatively low stem densities, and the abundance of common associated invertebrates was very low compared with the adjacent established marsh. Thus, it is unclear whether S. alterniflora diversity has no effect, or whether the low numbers of associated species simply made it difficult to detect any effects. In contrast to invertebrates, fishes were abundant at high tide, yet their numbers were highly variable spatially. Their increased mobility may reduce the likelihood that they will respond to plant diversity at the spatial scale (~1 m2) of my experimental plots. Given the strong evidence from previous work for effects of plant genetic identity and diversity on associated animals (e.g. Hughes & Stachowicz 2004; Johnson & Agrawal 2005; Crutsinger et al. 2006; Crawford & Rudgers 2013; Hughes, Moore, & Piehler 2014), additional tests of the effects of S. alterniflora diversity on the above-ground community are warranted.

The experimental plots were located at the lower edge of the existing marsh and, thus, are representative of naturally colonizing marshes (Proffitt, Travis & Edwards 2003; Civille et al. 2005), as well as the initial stages of restoration efforts when planting density is low due to logistical and financial constraints (Broome, Seneca & Woodhouse 1986). The fact that both trait and genotypic diversity were important for stem density and spatial spread in this experiment is consistent with prior findings that S. alterniflora intraspecific diversity can be important for density and spatial spread during the colonization stage (Wang et al. 2012). Abiotic stress from inundation and salinity is high in sparsely vegetated areas (Bertness 1992; Pennings & Bertness 2001), and these strong abiotic forces could have enhanced (c.f., Reusch et al. 2005) or overwhelmed (c.f., Johnson et al. 2008) the effects of genotypic diversity. The relative importance of genotypic and trait diversity in colonizing versus established marshes remains unclear. However, the consistency of plant performance in the treatments that started with high trait diversity, regardless of genotypic diversity, suggests that variation at a single, easily measured and ecologically relevant trait may serve as a useful proxy for more complicated and expensive measures of diversity in restoration efforts.


Jim and Lillian Hughes kindly provided access to the field site. Robyn Zerebecki helped extensively in the glasshouse and the field. R. Coker, A. Dillon, E. Field, A. Moore, M. Murdock, E. Pettis and T. Rogers also helped in the laboratory and field. D. Levitan graciously provided access to his laboratory. C. Hays, C. Richards, J. Stachowicz, R. Zerebecki and 2 anonymous referees provided constructive comments on the manuscript. This is contribution number 310 from the Northeastern University Marine Science Center. The author has no conflict of interest to declare. This study was supported by National Science Foundation Grant DEB-0928279 to A.R. Hughes.

Data accessibility

R scripts uploaded as Supporting Information (Appendix S1). Data deposited in the Dryad Repository (Hughes 2014).