Dispersal and recruitment limitation contribute differently to community assembly

Abstract

INTRODUCTION

A fundamental challenge in ecology is to understand how so many species can coexist in highly diverse communities. Traditionally, answers to this question have been sought within the niche theory, but more recent theories have focused on the role of neutral processes, ignoring the differences between species. According to the niche theory, different species can coexist within a community only if they are sufficiently ecologically distinct (Hardin 1960). During the past century, numerous studies have endeavored to search for a precise niche differentiation among coexisting species (Berendse 1983; Hutchinson 1957; Lamont and Bergl 1991; Rocap et al. 2003; Schneider et al. 2004), although we have not yet been able to delineate a range of important patterns in community ecology, such as species abundance distributions (SADs) and species–area relationships (SARs).

The recently developed neutral theory of biodiversity has attracted much attention because of its remarkable success in predicting many community patterns like SADs and SARs. But the basic assumption that all species are equivalent in their competitive ability or fitness has found little empirical support. Current neutral models have assumed that all individuals in a community show equal fitness regardless of species, which is obviously at odds with empirical observations. It is self-evident that some difference, though small, in competitive ability among species is inevitable, and it has been demonstrated that a slight difference in species per capita birth or death rates can result in a much shorter species coexistence time and quite different SADs compared with the predictions of the neutral theory (He et al. 2012; Yu et al. 1998; Zhang and Lin 1997; Zhou and Zhang 2008). In this sense, the neutral theory is fragile.

To defend the robustness of the assumption of ecological equivalence, Hubbell (2001) proposed that dispersal and recruitment limitation, which might be common in especially species-rich communities, can considerably delay competitive exclusion caused by competitive asymmetry. As a consequence, even slow speciation and/or immigration can compensate for species loss caused by drift and competitive exclusion. This proposition stems largely from the theoretical study of Hurtt and Pacala (1995) who examined the consequences of recruitment limitation with both analytical and simulation methods. Hurtt and Pacala demonstrated that competitively inferior species can coexist with competitively superior ones via ‘winning-by-forfeit’ in the case of recruitment limitation. They concluded that recruitment limitation can efficiently lessen the effect of competitive asymmetries among species and prevent even extreme local determinism from strongly regulating the relative abundances of species at the community level. However, it is important to note that Hurtt and Pacala assumed species to be niche-differentiated in their model and it remains unclear whether the effects of dispersal and recruitment limitation in the absence of niche differentiation are as conspicuous as they found. In this context, it has been found that removing niche differentiation from the model of Hurtt and Pacala can lead to very different conclusions: the effects of recruitment limitation were rather limited and insufficient to countervail the relatively rapid competitive exclusion caused by competitive asymmetries among species (Tang and Zhou 2011).

Besides the confounding effects of niche differentiation and dispersal and recruitment limitation, the possibility that dispersal limitation and recruitment limitation can have different influences on community structure has also been neglected in the past. Moreover, the limited studies that theoretically examine the effects of dispersal and recruitment limitation as mentioned above (Hurtt and Pacala 1995; Tang and Zhou 2011) were spatially implicit. In reality, both dispersal and recruitment are restricted within a short distance. The processes of dispersal and recruitment are only relative to those individuals located near the focal plant. Hence, the effects of dispersal and recruitment limitation can only be studied using a spatially explicit model and realistic dispersal modes.

In this paper, we investigated the effects of dispersal and recruitment limitation on delaying competitive exclusion in a homogeneous habitat, and most significantly, without any niche differentiation. Through spatially explicit simulations, we examined the species diversity (measured by Simpson’s diversity index), SAD and SAR under the conditions of slight to moderate competitive asymmetries among species together with dispersal limitation and recruitment limitation. Our simulation results revealed that the effects of even slight competitive asymmetry can only be offset by extremely strong dispersal and recruitment limitation as found before when limited dispersal and recruitment are not modeled in a spatially explicit manner (Tang and Zhou 2011). For the first time, we demonstrated that dispersal limitation is more important to the general shape of both SADs and SARs, whereas recruitment limitation tends to have a bigger impact on offsetting competitive asymmetries and producing neutral patterns.

METHODS

In all of the simulations, the suitable habitat was assumed to consist of J = 128 × 128 cells (but different community sizes did not change the basic results reported below). The landscape was further assumed to be a torus, with cells on the right edge and their neighbors on the left edge, and with cells on the bottom edge and their neighbors on the top edge. Each cell can contain at most one adult at a time. We subsumed the probabilities of offspring survivorship and offspring establishment into a single, per capita, competitive factor of w (competition coefficient), which is normally distributed as N (1, σ²) among species. We let σ = 0 (neutral case), 0.005, 0.01 and 0.05 in the simulations, respectively.

Dispersal and the per capita fecundity were assumed to be the same across species. Each individual produced F seeds in each generation (F = 1, 10, 100, respectively, in our simulations). Seeds produced by each individual were dispersed according to a fat-tailed dispersal kernel, in which seeds landed between distance r and distance r + dr with the probability of K(r)2πrdr, where:

The parameter L determines the ‘width’ of the dispersal kernel and η describes the ‘fatness’ of the kernel (Chave and Leigh 2002; Rosindell and Cornell 2009). In our simulations, we fixed the value of parameter η = −6, but let L be 1, 8 and 64, respectively. The medium dispersal distances under these parameter values were 0.65, 5.08 and 41.22, respectively.

The death–birth cycle of each generation was modeled as follows. Death was assumed to be random and equally probable for all individuals of all species. At the beginning of each generation, each individual might die with a probability of 5%. All extant individuals and the individuals that died in the current generation produced F seeds, and the seeds then were dispersed according to the dispersal kernel. With the probability of v = 0.001 (different speciation rates were found to yield qualitatively rather similar results, thus not shown here), the empty site was occupied by a new species; otherwise, it would be colonized by one of the seeds arriving at that site. Denote the number of seeds of species i arriving at a given vacant site as f_i and the probability that species i occupied the vacant site was $f_i{w}_i/{\displaystyle \sum _j{f}_j{w}_j}$⁠, where w_i measures the competitive ability of a species, defined as the relative per capita success rate of species i when in competition with other species for an empty site.

In the simulation, we first examined how species diversity decayed with different levels of competitive asymmetry, dispersal limitation and recruitment limitation. At the beginning of each simulation, we constructed a neutral community following the method of Hubbell (2001), and then randomly distributed each individual in the community onto the landscape. We iterated the birth–death cycle for 10 000 generations without speciation and recorded Simpson’s diversity index every 100 generations. Simulations were repeated 100 times for each parameter set and then the medium values of Simpson’s diversity index were computed.

In the second model, we investigated SAD patterns, SARs and the relationship between species competitive ability and species’ rank in abundance under different levels of competitive asymmetry, dispersal limitation and recruitment limitation. For each parameter set, we followed the algorithm described in Hubbell (2001) to form a community. Simulations of birth–death cycles were continued as described above for 30 000 generations, which was long enough for the community to reach a stochastic equilibrium. The reported SAD, SAR and the relationship between species competitive ability and species’ rank in abundance were obtained by averaging over 20 replicates for each parameter set.

RESULTS

As shown in Fig. 1, the decay of Simpson’s diversity with time depended largely on the level of competitive asymmetry among species. In the absence of speciation (v = 0), Simpson’s diversity decreased very slowly as expected in the neutral case of σ = 0, but very quickly when species differed in their competitive ability (σ = 0.005, 0.01 and 0.05) with little dispersal limitation (L = 64) and recruitment limitation (F = 100) (Fig. 1). Mild dispersal limitation (L = 8) and moderate recruitment limitation (F = 10) almost had no effect on delaying competitive exclusion incurred by competitive asymmetry, resulting in similar decreases in the Simpson’s diversity index as those obtained under little dispersal and recruitment limitation. Both severe dispersal and recruitment limitation could delay competitive exclusion and retard the decay of Simpson’s diversity index to some degree. But the effects of dispersal and recruitment limitation on biodiversity maintenance were quite different. In the case of a slight competitive asymmetry (i.e. σ = 0.005), extreme recruitment limitation (F = 1) could efficiently compensate for the loss of biodiversity due to competitive exclusion without dispersal limitation (i.e. L = 64), while even extremely strong dispersal limitation (L = 1) could not work by itself without recruitment limitation.

Figure 1:

the effects of competitive asymmetry, dispersal and recruitment limitation on Simpson’s diversity index. The number of seeds produced by each individual is 1, 10 and 100 from the first to the third row of the panels, respectively.

Open in new tabDownload slide

The SADs of the metacommunities at equilibrium changed from log-series distribution to lognormal-like as dispersal limitation was enhanced (Fig. 2). This extends the prediction of the spatially implicit neutral model, in which a metacommunity always follows a log-series SAD (Hubbell 2001; Ostling 2011; Volkov et al. 2003, 2007; Zhang et al. 2012). Contrary to dispersal limitation, the general shape of SADs remained almost the same under different levels of recruitment limitation (Fig. 2).

Figure 2:

the effects of competitive asymmetry, dispersal and recruitment limitation on SAD patterns. The number of seed produced by each individual is F = 1, 10 and 100 from the first to the third row of the panels, respectively. Species are binned into Log₂ classes (1, 2–3, 4–7, 8–15, …).

Open in new tabDownload slide

As illustrated by Zhou and Zhang (2008), species richness considerably decreased and SADs departed greatly from those predicted by the neutral model when species differed in their competitive abilities (Fig. 2). The most significant discrepancy in SADs between neutral and non-neutral scenarios occurred when there was little dispersal and recruitment limitation. SADs remained almost the same as dispersal and recruitment limitation were enhanced to moderate levels (L = 8 and F = 10). Extremely strong dispersal limitation and recruitment limitation could each delay competitive exclusion caused by competitive asymmetry, but in different ways. For instance, extreme recruitment limitation, even without dispersal limitation, could still efficiently delay competitive exclusion when competitive asymmetry was very slight (σ = 0.005) and resulted in similar SADs compared with that produced by the neutral model. By contrast, extremely strong dispersal limitation (L = 1) alone could not produce such an effect without recruitment limitation (Fig. 2).

When competitive asymmetries were introduced into the neutral model, competitively superior species would definitely dominate the community, while other species behaved essentially neutrally (Fig. 3). This was the case even when the interaction of a slight competitive asymmetry (σ = 0.005) and severe dispersal and recruitment limitation produced neutral-like SADs. In such cases, species rank in abundance still correlated positively with species competitive ability for common species (Fig. 3).

Figure 3:

the relationship between species competition coefficient (w) and species rank in abundance under different levels of competitive asymmetry, dispersal and recruitment limitation. Parameter values are the same as in Fig. 2.

Open in new tabDownload slide

Fig. 4 illustrates SARs under different levels of competitive asymmetry, dispersal limitation and recruitment limitation. Dispersal limitation had an important effect on the general shape of SARs, especially in neutral and nearly neutral cases. SARs changed from a power law to a logarithmic law as the dispersal distance increased, consistent with previous findings (Chave et al. 2002). In contrast, recruitment limitation had little effect on the shape of the species–area curves.

Figure 4:

the effects of competitive asymmetry, dispersal and recruitment limitation on the SARs. Parameter values are the same as in Fig. 2.

Open in new tabDownload slide

In general, the number of species found in a given area always decreased with increased competitive asymmetry, especially in the absence of dispersal and recruitment limitation (Fig. 4). Moderate dispersal limitation (L = 8 in our simulations) and moderate recruitment limitation (F = 10) would cause the SARs to depart significantly from the neutral one. Again, the separate effects of extreme dispersal and recruitment limitation on SARs in the case of a slight competitive asymmetry (σ = 0.005) were different. The SAR under a very slight competitive asymmetry (σ = 0.005) resembled that found in the neutral case even if only recruitment was extremely limited (i.e. F = 1 in our simulations), but extremely strong dispersal limitation alone could not maintain a neutral-like SAR (Fig. 4).

DISCUSSION

In this paper, we investigated the ability of dispersal and recruitment limitation in delaying competitive exclusion by means of spatially explicit simulations. As expected, we found that dispersal and recruitment limitation could indeed delay competitive exclusion to some extent, but their effects were quite moderate unless extremely strong dispersal and recruitment limitation were assumed. Most importantly, we clarified that the effects of dispersal limitation are different from that of recruitment limitation on community structure. Dispersal limitation largely affects the shape of SADs and SARs, while recruitment limitation is more effective in delaying competitive exclusion caused by competitive asymmetry.

Although the importance of dispersal and recruitment limitation on species distribution and species coexistence has been widely recognized (Freestone and Inouye 2006; Hubbell 2001; Hurtt and Pacala 1995; Verheyen and Hermy 2001; Wehncke et al. 2003), few studies have addressed the extent to which they become important. In fact, dispersal and recruitment limitation often operate in concert, and their effects are difficult to distinguish. Furthermore, they may work together with other processes to affect community dynamics. For instance, both niche differentiation and recruitment limitation contribute to species coexistence in the theoretical study of Hurtt and Pacala (1995). Although Hurtt and Pacala concluded that an arbitrarily large number of species can coexist for a very long period in the presence of both niche differentiation and recruitment limitation, they were unable to evaluate the relative importance of these two processes in their model. Following Tang and Zhou (2011), but in a more realistic, spatially explicit manner, this study provides a better understanding of community assembly of competitively asymmetric species under dispersal and recruitment limitation. We showed that neutral species distribution patterns do not necessarily imply neutral mechanisms because competitive exclusion and severe dispersal and recruitment limitation can interact to produce a SAD that is exactly predicted by the neutral model (left-upper panel in Fig. 2). However, in such communities, species abundance positively correlated with their competitive ability despite an emerged neutral pattern.

In real communities, although dispersal limitation and recruitment limitation were widely observed, it is not easy to distinguish the effects of dispersal and recruitment limitation from other processes (Ehrlen and Eriksson 2000; Pinto and MacDougall 2010; Yu et al. 2004). For instance, Ehrlen and Eriksson (2000) manipulated patterns of dispersal limitation and patch occupancy in seven temperate forest herbs by experimental sowing. They found limited recruitment due to a lack of seeds in six out of the seven species investigated. Seed size was positively related with the successful establishment of seeds but negatively correlated with patch occupancy, suggesting that competitive ability and dispersal limitation interact to determine the patch occupancy in these herbs. A recent research demonstrated that dispersal limitation and environmental structure interacted to restrict the occupation of an optimal habitat of a savanna plant Viola praemorsa (Pinto and MacDougall 2010). But unlike previous studies that solely reported the presence of dispersal and/or recruitment limitation, Pinto and MacDougall used a multi-scale combination of statistical-, experimental- and performance-based demographic approaches to isolate the contributions of dispersal distance and environment-based limitations to the distribution of this plant at different scales. They found weak habitat matching in spite of significant niche-based environmental responses of V. praemorsa. Dispersal limitation restricted the occupation of the optimal habitat at the coarse scale, whereas a lack of autocorrelation of environmental variables prevented this plant from aggregation at the fine scale.

Besides, dispersal and recruitment limitation themselves are frequently confounded in previous studies, either conceptually or functionally. In some field studies, recruitment limitation was referred to as a combined consequence of low and uncertain seed supply, difficulties in seedling establishment and restricted dispersal (Eriksson and Ehrlen 1992; Tilman 1997). In this paper as in many other studies (Clark and Ji 1995; Hurtt and Pacala 1995; Tang and Zhou 2011), recruitment limitation stands only for the shortage of seed production and dispersal limitation is defined as limited dispersal. We first demonstrated the differential effects of dispersal and recruitment limitation on macroecological patterns such as SADs and SARs, which provided useful insights into community assembly.

All in all, we conclude that moderate dispersal and recruitment limitation may be insufficient to overwhelm the effects of fitness differences among species and lead to neutral community dynamics. This might be taken as a compelling argument for bringing niche differentiation into neutral framework (Gravel et al. 2006; Leibold and McPeek 2006). The conception of ‘emergent group’ may be an efficient way to fulfill this purpose (Herault 2007). According to Herault, an emergent group is a set of species that have a similar functional niche owing to ecological convergence. Species within an emergent group obey the zero-sum random drift, whereas species from different emergent groups coexist through some kind of trade-offs. Thus, species present in the same community are either different or adequately similar (Scheffer and van Nes 2006). It has been suggested that such an integrative model may produce similar SADs as the purely neutral model (Chisholm and Pacala 2010), possibly providing an explanation for the common observation of neutral patterns in real communities.

FUNDING

National Natural Science Foundation of China (30970543, 31030014), Program for New Century Excellent Talents in University (2006-0907) and the Fundamental Research Funds for the Central Universities.

Conflict of interest statement: None declared.

References