Population structure and growth of a non-native invasive clonal plant on coastal dunes in Southern Brazil

1 Universidade Federal de Santa Catarina, Programa de Pós Graduação em Ecologia. Campus Universitário, s/n, Córrego Grande, 88040-900, Florianópolis, SC, Brazil. 2 Universidade Federal de Santa Catarina, Programa de Pós Graduação em Biologia de Fungos, Algas e Plantas. Campus Universitário Trindade, Caixa Postal 476, 88040900, Florianópolis, SC, Brazil. 3 Universidade Federal de Santa Catarina, Departamento de Ecologia e Zoologia, Centro de Ciências Biológicas. Campus Universitário, s/n, Córrego Grande, 88040-900, Florianópolis, SC, Brazil. Cristiana Barbosa1 crisilha@gmail.com


Introduction
Many plant communities are becoming seriously dominated by non-native invasive plant species as a consequence of human-mediated transport and introduction (Rejmánek et al., 2013).In coastal dune systems, different microhabitats may be distinctly susceptible to invasion by non-native species as a consequence of resource availability and stress level (Carboni et al., 2010;Santoro et al., 2012).In coastal ecosystems, a strong environmental gradient along the sea-inland is mainly due to salinity, wind action, temperature, drought and sand instability (Kolb et al., 2002;Lortie and Cushman, 2007;Santoro et al., 2012).Kolb et al. (2002) showed that coastal grasslands which occur in milder conditions (with higher levels of nitrogen and soil moisture) are more susceptible to invasion by nonnative plants than those subjected to harsher conditions (lower levels of nitrogen and soil moisture).On the other hand, Lortie and Cushman (2007) suggested that more severe conditions favored invasion in a coastal dune ecosystem.In a later study, Carboni et al. (2011) related the spread of non-native alien species to intermediate stress conditions on dunes along the coast of Italy.Although evidence can be controversial, most empirical studies show facilitative effects by resident species in severe environments, in the sense that they can buffer neighboring plants from abiotic stressors (He et al., 2013).
The introductions of non-native plants in coastal dune ecosystems result from direct human interventions, mainly for dune stabilization or ornamental purposes (Carboni et al., 2010;Carboni et al., 2011).Introductions may also develop from accidental propagule arrivals from plants cultivated nearby (Kim, 2005).Propagules may also arrive on coastal dunes from the ocean especially with high tides (Aptekar and Rejmánek, 2000).Non-native invasive species impacts on costal dune ecosystems include native species displacement, decreasing soil stabilization functions (Kim, 2005), competition for resources (Santoro et al., 2012), and native species richness or diversity decline (Badano and Pugnaire, 2004;Carboni et al., 2010).
Furcraea foetida (L.) Haw.(Asparagaceae) is an invasive species native to Central and northern South America (García-Mendonza, 2001).In Brazil, populations of F. foetida are currently distributed along the eastern part of the country, invading mainly coastal ecosystems of the Atlantic Forest biome, such as coastal dunes and rocky shores.Records of the species in Brazil date back to the early 19 th century, when it was cultivated for fiber (García-Mendoza, 2001).Furcraea foetida is invasive in several countries, such as South Africa and overseas departments of France (e.g.Reunión Island) (Baret et al., 2006;Crouch and Smith, 2011).Its invasiveness is highly associated with reproduction by thousands of bulbils per plant (Crouch and Smith, 2011).It displaces and competes with surrounding vege-tation, and is controlled for conservation purposes in New Zealand, Hawaii, South Africa, Florida and Brazil (Motooka et al., 2002;Wilcox, 2005;Howell, 2008;Crouch and Smith, 2011;Randall, 2012;Dechoum and Ziller, 2013).
This study aimed to characterize the population structure of F. foetida and habitat conditions, in which it occurs, as well as to assess its potential impacts on plant communities at the Praia Mole, Florianopolis, Brazil.Previous studies showed that plants of Agave desertii and A. macroacantha, species morphologically similar to F. foetida, were found only in sheltered microhabitats in desert environment due to the lower survival rates of seedlings and bulbils in exposed areas where temperature may exceed 70°C in summer (Franco and Nobel, 1988;Arizaga and Ezcurra, 2002).Considering that soil temperature may also exceed 70°C on bare sand during mid-summer in coastal dune ecosystems (Scarano, 2002), we hypothesized that F. foetida is associated with sites with milder habitat conditions generated by greater vegetation cover and height in coastal dune vegetation (hereafter restinga).Based on the impacts caused by introduced plants in coastal dune ecosystems, we also hypothesized that individual plant growth of F. foetida decreases native species richness, diversity, and cover.Taking into account that coastal dune ecosystems are one of the most invaded ecosystems in Brazil (Ziller and Dechoum, 2013), anticipating sites that are most susceptible to invasion by non-native species is key information for designing more effective control and eradication strategies.

The study system
The study was carried out between 2009 and 2010 at Praia Mole (27°36' -27°36' S and 48°26' -48°25' W), along the eastern coast of Florianopolis.The regional climate is humid mesothermal (Cfa) according to Köppen-Geiger.The study area consisted of a 48 m wide transect on the front dune, 760 m along the ocean line.The vegetation is dominated by herbs but sub shrubs such as Sophora tomentosa L. and Scaevola plumieri (L.) Vahl may occur (Falkenberg, 1999), usually being irregularly distributed or aggregated in groups.The vegetation is not higher than one meter, and sometimes it can be very sparse or even absent in some sites (Falkenberg, 1999).
Furcraea foetida forms a leaf rosette up to 4 m in diameter and more than 2 m in height.It produces a floral woody peduncle up to 10 m in height in the rosette center.This monocarpic plant can live 5-20 years depending on growing conditions (Francis, 2004).It produces pseudovivipary vegetative propagules (bulbils or plantlets) in large numbers (Hueck, 1953;Elmqvist and Cox, 1996), a rare clonal mechanism that consists in producing vegeta-tive propagules on sexual organs, such as in inflorescences.Adjacent to the study area, three reproductive plants were observed 40, 50, and 130 m away from the front dune.

Data collection
Two transects were set up 10 and 30 m away from the beginning of the vegetation on the front dune.The F. foetida population was surveyed in 180, 16 m 2 plots (4 m x 4 m), 90 per transect.Distance between plots was 2 m, while plots were 2 m away from transect lines, alternatively set on the right and left of the lines.We registered the number of F. foetida plants in each plot as well as rosette diameters and height.We classified the plants in five stages of development based on rosette diameter: bulbils (rosette not formed), small (> 100 cm) and large juveniles (≥ 100 cm and < 250 cm), potentially reproductive adults (≥ 250 cm), and reproductive adults (bearing floral stalks).
A subplot of 1.4 x 1.4 m was defined in the center of each 16 m 2 plot for vegetation structure data.In each subplot the cover percentage of each species, bare soil percentage, and vegetation height were recorded.Cover percentage was visually estimated for each species according to the methodology by Assumpção and Nascimento (2000), as follows: class 1: >0 to 5%; class 2: >5 to 15%; class 3: >15 to 25%; class 4: >25 to 50%; class 5: >50 to 75%; and class 6: >75 to 100%.Mean value of classes was used for statistical analyses.Vegetation height was estimated from the mean of five measurements, 4 taken at plot corners and 1 in the center of each 140 x 140 cm plot.
We sampled 16 young, isolated F. foetida plants and set up 1.96 m 2 plots (1.4 m x 1.4 m) around each plant, centralized within the larger plot.Control plots (without F. foetida) were randomly set at 0.5 m of each F. foetida plant, in order to ensure sampling in a similar microhabitat.We estimated F. foetida and all the other species cover percentage in each plot, and measured F. foetida height, diameter, and number of rosette leaves in October 2009 and October 2010.

Data analysis
The population structure of F. foetida was characterized by plant density descriptions in the 180 plots, frequency at each development stage, and spatial distribution patterns using the Morisita dispersion index (Id) and the χ 2 test for significance of getting far away from randomness.We calculated the importance value index (IV) for F. foetida and of sampled plant species.
We conducted a principal component analysis (PCA) using Canoco software, version 4.5 (Ter Braak, 1995) and developed a correlation matrix to assess whether plots with F. foetida were correlated to structural restinga parameters.Variables used in this analysis were bare soil percentage, Σ absolute cover of all species, average vegetation height, richness, and diversity (H' Shannon-Wiener, with Log 10 ).Furcraea foetida was excluded from the estimates of the vegetation parameters.
Plant height and diameter of F. foetida were correlated (Spearman r = 0.75, p < 0.0001), so only diameter, number of leaves and cover were used to check whether these parameters varied between 2009 and 2010.Generalized linear mixed models (GLMM) were fitted assuming the Poisson distribution for number of leaves and the Gamma distribution for diameter and cover.Year was considered a fixed effect while each plant ID was considered a repeated measured replicate, therefore a random effect.
GLMM were also used to check whether the growth of F. foetida plants has led to changes in native plant communities.In these models, the explanatory variables year (2009 and 2010) and condition (plots with or without F. foetida, called treatment and control, respectively), as well as the interactions between year and condition were considered fixed effects.Plots were considered a random effect.Furthermore, the sum of absolute cover of the five species with highest IV (Smilax campestris, Rumohra adiantiformis, Diodella radula, Remirea maritima, and Sophora tomentosa, respectively; Table 1) were compared fitting zero-inflated GLMM assuming the negative binomial distribution for the response variable.Year and condition as well as their interaction were considered fixed effects, while plots were considered a random effect.All GLMM analyses and figures were produced with R software (version 3.2.2) (R Core Team, 2015) using the "glm-mADMB" and the "sciplot" packages.
Axis 1 in the PCA explained 47.2% of the variation with bare soil (0.59), Σ absolute cover of all species (-0.56), and vegetation height (-0.53) as the most important variables (Figure 1).The most important variables on axis 2 were richness (-0.66) and diversity (-0.69), explaining 38.5% of variation (Figure 1).Axis 1 separated most of the plots with F. foetida on the negative side with higher values of vegetation cover and height.Considering leaf number and cover percentage, F. foetida plants significantly increased in size between 2009 and 2010 (Tables 2 and 3).There was a significant increase in species richness from 2009 to 2010, but it did not differ between treatments (Table 4).On the other hand, na-tive species cover decreased between 2009 and 2010, but also did not vary between treatments (Table 4).Species diversity did not vary between years or between treatments (Table 4).
The assessment of cover percentage for the five species with highest IV values shows that D. radula, R. maritima, and S. campestris increased in treatment plots and decreased in control plots between 2009 and 2010 (Figure 2, Table 5).Conversely, R. adiantiformis cover decreased in treatment plots and increased in control plots.Sophora tomentosa cover did not differ between conditions or years (Figure 2, Table 5).

Discussion
Our first hypothesis was corroborated considering that F. foetida was positively associated with milder environmental conditions, more specifically with greater native plant cover and vegetation height.Conversely, regarding our second hypothesis, although F. foetida plants showed   significant increase in cover and number of leaves over the one-year period of the study no difference was found for native species richness, diversity, and cover between invaded and uninvaded plots.Other authors have showed that species richness remained stable in invaded and uninvaded plots by different invasive plant species although some of them only displaced the dominant native species and had no effects on non-dominant species (Sax, 2002;Mason and French, 2008;Powell et al., 2013;Dong et al., 2015).Hejda et al. (2009) found that decreases in species richness in invaded plots by different invasive plant species significantly interacted with species-specific differences in cover between the invading and native dominant species.In our study, although we cannot be completely certain that changes in cover of dominant species were due to growth of F. foetida plants, it is likely that these different species-specific effects of F. foetida on dominant native species probably result in changes of local plant communities.
Considering that the assessment of invasion impacts on native species is strongly influenced by space scales (Vilà et al., 2011;Pyšek et al., 2012;Powell et al., 2011;Powell et al., 2013;Ricciardi et al., 2013), and change dramatically over time (Flory and Clay, 2010;Dostál et al., 2013;Powell et al., 2013), long-term studies carried on across the invaded range of F. foetida could shed light on which are the main invasion impacts and which biotic and environmental factors are related to these impacts.In addition, experimental studies in which invaded plots could be compared to those from which F. foetida has been removed offers a straightforward method to demonstrate that ecological differences between these plots are linked to the effects of alien species (Kumschick et al., 2015).However, in order to certify that the outcomes of these experiments are due to the invasive species removal, the confounding effects of disturbance must be minimized (Kumschick et al., 2015).
Despite being associated with higher and denser vegetation, F. foetida plants had on average higher cover than native dominants and were taller than native plants.Species that can grow vigorously, reaching higher cover/biomass (large and succulent leaves), and are taller than members of invaded resident communities, including dominant species, tend to cause the strongest impacts on these communities (Brabec and Pyšek, 2000;Hedja et al., 2009).
The F. foetida population studied in the present work consisted predominantly of juvenile not reproductive plants and bulbils in establishment.Some biological traits of F. foetida plants such as vegetative reproduction, and very large, succulent leaves may be associated with establishment success.These traits were related to invasiveness across five Mediterranean islands for increasing colonization capacity in disturbed habitats under water depletion (Lloret et al., 2005).The current presence of large juveniles, potentially reproductive adults, and adult plants near the study area may result in massive reproduction (García-Mendoza, 2001).If this occurs the invasion will increase significantly, in time potentially generating more impact on local plant communities.
Considering prevention as the cheapest and most effective approach in invasive species management we recommend the removal of F. foetida before the studied population becomes able to expand (Ziller and Dechoum, 2013).Management priorities must consider population structure, first controlling reproductive plants in order to refrain bulbil production followed by large juveniles, and then younger plants.Dense restinga sites must also be considered a management priority due to establishment success in plots with higher vegetation cover.Public awareness campaigns on non-native invasive species and their impacts on natural ecosystems are important to hinder the species ornamental use in coastal areas.Finally, legal regulations are essential to guide the use of the species and give control and eventual eradication a viable prospect.To sum up, we consider that F. foetida may be an imminent threat to the studied plant communities, considering that mid to long-term changes that may be caused by the current species specific effects of F. foetida on the plant community, summed to a likely population growth due to the potential reproduction of large juvenile plants.Furthermore, considering that the ocean is a potential dispersion vector of the species propagules, its arrival in sites susceptible to its establishment may cause impacts on coastal ecosystems in a regional scale.That said, we conclude that immediate control actions, public awareness campaigns and legal regulations are essential to minimize F. foetida impacts on local and regional biodiversity.

Figure 2 .
Figure 2.Cover (%; mean + SE) of three of the five native species with highest importance value (IV) in the sampled plots at the Praia Mole, Florianopolis, southern Brazil.The three species represented are those whose cover significantly changed between 2009 and 2010.White bars = Control (plots without F. foetida); black bars = Treatment (plots with F. foetida).

Table 1 .
Population structure and growth of a non-native invasive clonal plant on coastal dunes in Southern Brazil Floristic list and importance value (IV) of species sampled in restinga vegetation at Praia Mole, Florianopolis, southern Brazil.Ac: absolute cover (%), Af: absolute frequency (%), Rc: relative cover (%), Rf: relative frequency (%), and IV: importance value of species = Rc + Rf. *non-native species.

Table 1 .
Continuation.Population structure and growth of a non-native invasive clonal plant on coastal dunes in Southern Brazil

Table 2 .
Number of leaves, diameter, and cover (mean ± SE) of F. foetida plants (n= 16) sampled at Praia Mole, Florianopolis, Brazil.*denotes significant differences between 2009 and 2010 in each of the parameters.Different letters indicate significant differences between years for each of the parameters.

Table 3 .
Generalized linear mixed model results for Furcraea foetida (N=16) leaves, diameter, and plant cover measured in 2009 and 2010 in restinga vegetation at Praia Mole, Florianopolis, southern Brazil.Fixed effect: year, random effect: plant ID.Significant p values in bold.

Table 4 .
Generalized linear mixed model results for all species richness, diversity and Σ of absolute cover in plots with and without Furcraea foetida in restinga vegetation at Praia Mole, Florianopolis, southern Brazil.Fixed effects: year and condition (with and without F. foetida); random effect: plots.Significant p values in bold.

Table 5 .
Generalized linear mixed model results for the five species with highest importance value of species (IV) in plots (N = 180): Σ absolute cover with and without Furcraea foetida in restinga vegetation at Praia Mole, Florianopolis, southern Brazil.Significant p values in bold; IV: Rc + Rf.