Intraspecific aggregation does not increase species richness in dune grasslands

Authors


Correspondence author. E-mail: cfd@dmu.dk

Summary

1. One of the possible mechanisms to maintain a high number of species in a plant community is that intraspecific aggregation due to limited dispersal reduces the importance of interspecific competition.

2. The hypothesis has been examined both theoretically and experimentally in the simplifying case of two species in a uniform abiotic environment, and there is a solid qualitative understanding that intraspecific aggregation reduces the importance of interspecific competition; however, there is considerable uncertainty as to the magnitude of this reduction in plant communities with many species in a heterogeneous environment and its effect on the number of coexisting species.

3. Here, the hypothesis was tested by comparing observed species richness at a site with a compound estimate of the intraspecific aggregation of all the species observed at the site, while taking the expected confounding effects of environmental spatial heterogeneity into account.

4. It is difficult to draw firm conclusions from the non-significant statistical results found in this study, as a lack of significance does not necessarily rule out the presence of an effect. However, based on the observation that species richness was not correlated with the degree of intraspecific aggregation, it was concluded that there was no additional support for the hypothesis that reduced interspecific competition due to intraspecific aggregation is important for maintaining the relatively species-rich flora in North-European dune grassland.

5.Synthesis. It is important to test whether it is possible to generalize the expected effects of intraspecific aggregation from manipulated experiments using only a few plant species to natural plant communities with many species, and the approach taken in this paper is a practically manageable approach.

Introduction

One of the classic questions in plant ecology is the problem of identifying the mechanisms that regulate the number of coexisting species in a plant community (e.g. Silvertown 2004; Stubbs & Wilson 2004), and a promising line of research is to investigate the effect of competition among spatially structured plants. Plants are sessile and dispersed by seeds or vegetative propagules that often end up close to the producing plant, and this striking feature of plant life history will cause plants of the same species to become spatially aggregated (e.g. Watt 1947; Harper 1977; Law, Herben & Dieckmann 1997; Seabloom et al. 2005). Furthermore, in most natural plant communities, plant growth is limited by resources, which the plant has to obtain in competition with neighbouring plants, and the processes of plant–plant interaction will, to a certain extent, determine plant community dynamics including the number of coexisting species (e.g. Harper 1977; Weiher, Clarke & Keddy 1998; Damgaard 2004a,b; Damgaard, Riis-Nielsen & Schmidt 2009); in cases where plant species are spatially aggregated, they will tend to compete with conspecific plants rather than heterospecific plants. Based on the above two general plant ecological phenomena, the hypothesis has been put forward that intraspecific aggregation reduces the importance of interspecific competitive interactions and, consequently, leads to an increase in the number of coexisting species compared to a community of randomly distributed plant species (Rees, Grubb & Kelly 1996).

The hypothesis that intraspecific aggregation reduces the importance of interspecific competition has been supported by the few experiments that have been carried out (Schmidt 1981; Stoll & Prati 2001; Lenssen et al. 2005; Monzeglio & Stoll 2005; Mokany, Ash & Roxburgh 2008). For example, Schmidt (1981) showed that the rate of competitive exclusion was reduced when species were spatially aggregated, and Stoll & Prati (2001) observed that stronger competitors performed worse, and weaker competitors performed better, when species were aggregated compared to when they were sown randomly. Moreover, a number of theoretical studies suggest that intraspecific aggregation (or interspecific segregation) in an otherwise homogenous environment generally leads to an increased likelihood of coexistence, i.e. an increase in the parameter space where coexistence is predicted to occur (e.g. Bolker & Pacala 1999; Neuhauser & Pacala 1999; Chesson 2000; Bolker, Pacala & Neuhauser 2003). However, if interspecific competition is weak relative to intraspecific competition, then spatial structure may reduce the likelihood of coexistence due to demographic extinction of the slightly weaker competitor (Neuhauser & Pacala 1999; Bolker, Pacala & Neuhauser 2003).

In Bolker, Pacala & Neuhauser (2003), the results of different types of spatial competition models are reviewed and the ecological mechanisms that link intraspecific aggregation with an increased likelihood of coexistence are classified: (i) if the competitively dominant species have relatively poor colonization ability, then, most likely, the dominant species will not outcompete the competitively inferior species at all patches (competition–colonization trade-off hypothesis); (ii) fast growing, but competitively inferior species can persist by dispersing their offspring nearby in order to exploit local resources (spatial successional niche hypothesis); and (iii) competitively inferior species may be locally dominant due to a local founding effect, and the process of competitive exclusion may be irrelevant on an ecological timescale if the speed of competitive exclusion is sufficiently reduced (phalanx strategy). However, an important shortcoming of the existing theoretical and experimental studies on the importance of interspecific competition in aggregated plant communities is that they have been limited to investigating the interaction between only two species. Most plant communities consist of numerous interacting plant species, and it is not trivial to generalize the obtained results on the importance of competition from a few species to many species.

Furthermore, in the above-cited studies on the effect of competition among spatially structured plants, it has been assumed that only the density of different plant species varies in space, and that the environment otherwise is homogenous. However, this is clearly not a realistic assumption in most natural ecosystems (Whittaker 1972; Fitter, Hodge & Robinson 2000). For example, Fridley, Grime & Askew (2009) reported important substrate variation at a spatial scale of only centimetres in a species-rich limestone grassland, which had a significant effect on plant community structure. Understanding population dynamic features of competing plants with limited dispersal in a multi-dimensional environment with spatial heterogeneity is, indeed, a complex endeavour, and the theoretical models necessarily have to be rather simplistic (e.g. Bolker 2003; Snyder & Chesson 2004); nevertheless, the general results of the theoretical models seem rather robust and in concordance with the understanding of most plant ecologists, i.e. that environmental spatial heterogeneity often, but not always (Wilson 2000), will augment the number of different niches at a certain area and as a result increase the expected number of coexisting species (Whittaker 1972; Silvertown et al. 1999; Stubbs & Wilson 2004).

In summary, there is currently a solid qualitative understanding that intraspecific aggregation reduces the importance of interspecific competition, but there is considerable uncertainty as to the magnitude of this reduction in plant communities with many species coexisting in a heterogeneous environment and its effect on the number of coexisting species.

The aim of this paper is to test the hypothesis that the number of coexisting species is augmented due to limited dispersal and ensuing associated intraspecific aggregation and reduced importance of interspecific competition. The hypothesis is tested by comparing observed intraspecific aggregation with species richness in coastal dune vegetation. However, an alternative hypothesis that will lead to a positive relationship between intraspecific aggregation and species richness is niche separation; i.e. through filtering processes, environmental spatial heterogeneity will tend to aggregate subgroups of species at specific environmental conditions from the species in the species pool (Keddy 1992). It will, therefore, be necessary to account for the expected confounding effects of environmental spatial heterogeneity.

Coastal dunes are dynamic ecosystems with reoccurrences of breaches and sand drift due to wind, and the combination of dry and moist spots, north- and south-exposed slopes, and succession processes often provides a mosaic of dune habitats (Warming 1909; Damgaard, Nygaard & Nielsen 2008). In order to account for the confounding effect of environmental spatial heterogeneity, the following strategy was chosen: (i) only to analyse small vegetation plots that had been classified as ‘dune grassland’; and (ii) to analyse the residual variation in species richness without the effect of pH, since the level of pH is known to be highly associated with plant species richness in most habitats (Pärtel 2002) including coastal dune ecosystems (Isermann 2005). In coastal dunes, soil pH depends on the origin of the sand and is correlated with the deposition age and nutritional status of the sand material. Relatively newly wind-deposited and nutrient-rich sand is characterized by a relatively high pH, which by the process of leaching becomes increasingly nutrient-poor and acidic (Stützer 1998).

Materials and methods

Vegetation sampling and classification methods

Ninety-six Danish coastal dune sites were monitored from 2004 to 2008. At each site, 20–60 plots were placed randomly, and in each plot species abundances were measured by the pin-point method in frames of 50 × 50 cm with 16-grid point (Kent & Coker 1992; Damgaard 2009). The plots from 22 sites were revisited every year with GPS accuracy, and the rest of the sites were only visited once. In some of the plots, a soil sample was taken and the pH in the soil was determined (Damgaard, Nygaard & Nielsen 2008).

Based on a detrended correspondence analysis (DCA) (Hill 1979), a supervised classification model was developed on the plots that contained more than two species. The first three DCA axes were used as predictor variables in a quadratic discriminant analysis (Venables & Ripley 1997), where eight a priori determined types of dune vegetation acted as the response variable, and each plot was classified as the most likely vegetation type (for further details see Damgaard, Nygaard & Nielsen 2008). Only the plots that were classified as belonging to the ‘dune grassland’ vegetation type were used in further analyses.

Dune grassland is characterized as more or less closed perennial grassland on calcareous fixed sand with a relatively rich flora of grasses and herbs as well as cryptogams. A total of 391 taxa (only the higher plants were consistently determined to species level) were observed in the classified dune grassland plots, of which Carex arenaria, Festuca rubra, Agrostis capillaris, Galium verum and Hypochoeris radicata were observed most consistently. The Shannon diversity index (Exp(H)) was 74, the Simpson index (1/H) was 32, and the species abundance distribution did not depart significantly from a log-normal distribution (for further details see Damgaard, Nygaard & Nielsen 2008).

Intraspecific aggregation

The intraspecific aggregation at the level of the community was estimated for all observed species at a site by a single parameter using the Pólya–Eggenberger distribution, which is a generalization of the multinomial distribution, which quantifies the departure from an independent and random distribution of hits in pinpoint measurements (Damgaard & Ejrnæs 2009). Consider the data from several pinpoint frames which are randomly placed at a site. The stochastic variable Yij is the number of pins that, when inserted vertically through one of the m grid points in pinpoint frame j into the vegetation, touch individuals of plant species i, and inline image is an estimate of the plant cover of plant species i in frame j. The multivariate probability mass function of the number of hits in frame j of the k species present at the site, inline image, may be described by a multivariate Pólya–Eggenberger distribution (mPE-distribution):

image(eqn1)

where inline image, inline image, and inline image(Johnson, Kotz & Balakrishnan 1997).

Since the sum of the estimates of the plant cover of the k species is different from one and may vary between frames, njis allowed to vary among frames. The vector a is the mean number of hits of the k species in the sampled pinpoint frames at the site. The parameter c measures the degree of intraspecific aggregation for all species in the plant community. If c = 0, then the species are randomly dispersed and the mPE-distribution degenerates to the multinomial distribution.

The mean of the mPE-distribution is independent of c and equal to the mean of the uncorrelated multinomial distribution, inline image. The variance depends on the distribution of nj and the parameter c. More specifically, if c > 0, then the variance of the number of hits will be augmented relative to the multinomial distribution, opposite, if c < 0, then the variance of the number of hits will be smaller than the variance predicted by the multinomial distribution.

In order to illustrate what is meant by the intraspecific aggregation of all the species in a plant community and to demonstrate the type of data that is needed for quantifying the intraspecific aggregation parameter c, a series of simple abundance matrices of four species and three plots is shown in Table 1 together with the corresponding maximum-likelihood estimate of parameter c. These simple examples also provide a notion of the scale of parameter c.

Table 1.   Examples of the calculation of intraspecific aggregation using parameter c (eqn 1) in an example with four species measured in three plots. The numbers represent an abundance measure, e.g. the number of grid points where the species is hit in a pinpoint frame. Only the bold numbers are varied among the four cases
SpeciesABCDABCDABCDABCD
Plot 18611510112161102011
Plot 212441124411244112441
Plot 315032150321503215032
 inline imageinline imageinline imageinline image

The log-likelihood function of the multivariate probability mass function was maximized for each site and the maximum-likelihood estimate of c was used to represent the degree of intraspecific aggregation at the site. A high value of inline image indicates a high degree of intraspecific aggregation at the site (Table 1).

Species richness

Since the number of plots classified as ‘dune grassland’ varied among the sites and it is expected that species richness (the observed number of taxa) at a site increases with the number of sampled plots (Gray, Ugland & Lambshead 2004), the observed species richness was adjusted to the expected species richness if 40 plots had been sampled.

The species accumulation curve of all plots from sites with 20 or more plots classified as ‘dune grassland’ was modelled byinline image, where x is the number of plots, S is the observed species richness, and a and z are shape parameters (Fig. 1), and the adjusted observed species number was defined by: inline image, where S40 is the expected species richness if 40 plots had been sampled.

Figure 1.

 Species accumulation curve of dune grassland plots. The solid line is the fitted model (inline image) assuming normally distributed residual variance inline image.

Statistical analyses

The estimated spatial aggregation parameter, inline image, the adjusted species richness, S40, and the mean pH were calculated for all sites with 20 or more plots classified as ‘dune grassland’. For the sites that were revisited annually, the average values across years were calculated. The resulting sets of estimates (inline image, S40, pH) from 92 sites were treated as independent data points and analysed in linear regression models assuming normally distributed residual variation. The distribution of the residual variation in the different models was checked by plotting the residuals against the predicted values. All the analyses were done using Mathematica (Wolfram 2007).

Results

The spatial composition of the plant species was more aggregated than a random pattern at all sites (likelihood ratio test of inline image; P < 0.0001 for all sites), i.e. at all sites, the mPE-distribution fitted the plant abundance data significantly better than the multinomial distribution, which is the most appropriate null model of such multivariate discrete data. There was a sizeable variation in the degree of intraspecific aggregation among sites (mean = 8.50; SD = 2.74), but, there was practically no variation among years at the 22 sites that had been sampled repeatedly.

The pH in the soil had a significant positive effect on the adjusted species richness in the dune grassland (Fig. 2; P < 0.0001), whereas it had no significant effect on the degree of intraspecific aggregation (Fig. 3; P = 0.69). When adjusted species richness was regressed on both pH and the degree of intraspecific aggregation, then the effect of the degree of intraspecific aggregation was not significant (Table 2), and, in fact, since pH and the degree of intraspecific aggregation were uncorrelated (Fig. 3), the degree of intraspecific aggregation had no significant effect on the adjusted species richness (Fig. 4; P = 0.77).

Figure 2.

 The adjusted species richness in dune grassland sites plotted against the average pH at the site.

Figure 3.

 The degree of intraspecific aggregation, inline image, plotted against the average pH at the site.

Table 2.   Analysis of variance of the effect of pH and intraspecific aggregation, inline image, on the adjusted species richness
Sourced.f.Sums of squaresMean sum of squaresF statisticP value
pH1112851128546.29< 0.0001
inline image11.3711.3710.00560.94
Error8921697243.78  
Total9132983   
Figure 4.

 The adjusted species richness in dune grassland sites plotted against the degree of intraspecific aggregation, inline image.

Thus, species richness was not correlated with the degree of intraspecific aggregation – neither in the case when species richness was corrected for the effect of pH nor when species richness was uncorrected. Consequently, in this study, I did not find additional support for the hypothesis that reduced interspecific competition due to intraspecific aggregation is important for maintaining the relatively species rich flora in North-European dune grassland.

Discussion

The importance of reduced interspecific competition due to intraspecific aggregation in controlling plant community dynamics has previously been demonstrated both theoretically (e.g. Bolker, Pacala & Neuhauser 2003) and empirically (e.g. Schmidt 1981), but these demonstrations have been done with only two competing species. It is important to test whether it is possible to generalize the result observed for a few plant species to plant communities with many species, and the approach taken in this paper, i.e. to compare observed species richness at a site with a compound estimate of the intraspecific aggregation of all the species observed at the site, is a practically manageable approach.

A possible positive relationship between intraspecific aggregation and species richness may be caused by the reduced importance of interspecific competition as well as niche separation along environmental gradients, and it was assumed to be critical that it was possible to account for the expected confounding effects of environmental spatial heterogeneity. It was, consequently, decided to (i) only to analyse small vegetation plots that had been classified as ‘dune grassland’ and (ii) analyse the residual variation in species richness without the effect of pH, which is known is to be highly associated with plant species richness. However, in the case of the investigated dune grasslands, the result that species richness was not correlated with the degree of intraspecific aggregation was robust for soils with different pH.

It is difficult to draw firm conclusions from the non-significant statistical results found in this study, as a lack of significance does not necessarily rule out the presence of an effect. However, based on the observation that the residual variation in species richness was not correlated with the degree of intraspecific aggregation, it was concluded that there was no evidence that intraspecific aggregation in itself, by reducing the importance of interspecific competition, had a positive effect on the number of coexisting plant species in dune grasslands.

The measured level of intraspecific aggregation depends on the distance between the grid points in the pinpoint frames. When the distance between the grid points is reduced, then it is more likely that the same individual plant is hit in two neighbouring grid points. In this study, the distance between the grid points was 12.5 cm, and this distance is not sufficient to rule out that the same individual plant is hit more than once. The reported intraspecific spatial aggregation is, consequently, not a measure of the aggregation of individual plants, and the comparison with a random pattern is, therefore, not realistic (Damgaard & Ejrnæs 2009). Furthermore, a high estimate of the parameter c thus, in principle, may be an indication of a site with relatively large individuals rather than a high intraspecific spatial aggregation. This is probably not the case for dune grasslands, where the majority of species either are rather small or are composed of many relatively small ramets. In fact, the concept of individual plants in dune grasslands is questionable, since they often are dominated by tussocks of perennial grass species with vegetative propagation.

Generally, one has to be careful when inferring from an observed pattern to an underlying causal process, since different ecological processes can result in the same pattern. It could be argued that plant species diversity in the relative labile coastal dune ecosystems mainly is controlled by disturbance–colonization processes and that observed species richness is a poor proxy for the expected number of coexisting species at ecological equilibrium. However, dune grasslands are only found on fixed dunes, which are relatively stable compared to the white dunes closer to the beach, and have a more or less closed plant cover. Furthermore, the observed positive relationship between soil pH and species richness suggests that it is the younger, relatively nutrient-rich dunes with a high pH, that have the highest species richness, which, again, indicates that species richness is not controlled by the rate of colonization. On the contrary, there is some support for the hypothesis that density-dependent effects are important for regulating plant communities in fixed dunes; in a demographic study of four winter annuals, Rees, Grubb & Kelly (1996) observed significant intraspecific density-dependent effects.

The use of observed species richness as a proxy for the expected number of coexisting species at ecological equilibrium is an important assumption of the present analyses and, apart from the above-mentioned criticism that the dune plant community may not have had enough time to reach a stable species richness, species richness is underestimated, since (i) only a fraction of the site was sampled and the observed species richness still showed an increasing response to the sampling effort (Fig. 1) and (ii) the diversity of lichens and mosses was not determined to species level. However, there were no a priori reasons why the compound measure of intraspecific aggregation consistently should be biased at either high or low levels of species richness.

The presented analysis corroborates the hypothesis that species richness of higher plants in Danish dune grassland is controlled by the origin and leaching process of the sand material that forms the dunes, such that calcareous and young nutrient-rich dunes with a high soil pH were associated with a high species richness of higher plants (Warming 1909; Damgaard, Nygaard & Nielsen 2008). The decrease in species richness of higher plants continues in the later stages of the North-European coastal dune ecosystem, where leached dune grassland may turn into grey dunes and, eventually, may end up as dune heaths, which are dominated by a few species (Warming 1909; Damgaard, Nygaard & Nielsen 2008). However, if the diversity of lichen and mosses had been included in the study, it is possible that another picture would have emerged, i.e. species richness may have decreased with pH in the dune grassland (Isermann 2005).

Ancillary