Toward a better understanding of the regional causes of local community richness

Authors


* E-mail: spharrison@ucdavis.edu

Abstract

Despite widespread acknowledgement that local ecological communities are profoundly shaped by regional-scale influences, including evolutionary and biogeographic processes, this perspective has yet to be widely incorporated into ecological research. Drawing on recent research, we propose four steps towards making regional influences a stronger part of research on the richness of local communities: (1) identifying the regional-scale causes of variation in species richness in the systems ecologists study; (2) testing for effects of regional richness on local richness, using improved observational and experimental analyses to overcome earlier problems; (3) simultaneously analysing environmental influences on regional and local species richness as well as the influence of regional richness on local richness and (4) considering the potential reciprocal effects of local processes on regional richness. In conclusion, we suggest some ways that similar approaches could be applied to other aspects of community structure beyond species richness.

Introduction

In the past several decades, ecologists have become aware that species assemblages are governed not only by local interactions among coexisting species, but also by large-scale biogeographic, historical and evolutionary processes (Ricklefs & Schluter 1993; Ricklefs 2004, 2007). One particular idea that arose from this large-scale outlook, namely that regional propagule supply is an important reason for variation in the species richness of natural communities, was first given prominence by MacArthur & Wilson’s (1967)Theory of Island Biogeography. This idea gained further impetus from many studies showing that the number of species within small localities increased either linearly or asymptotically with the number of species in the larger surrounding regions (local–regional richness or local and regional richness analyses; e.g. Cornell 1985; Ricklefs 1987; Cornell & Lawton 1992; Shurin & Srivastava 2005). The observation that native and exotic species richness are nearly always positively correlated at moderate-to-large spatial extents has helped reinforce the concept of the openness of local communities to immigration from a regional pool (Stohlgren et al. 2008), while the advent of community phylogenetics has provided a powerful new tool for studying regional and evolutionary influences upon local community patterns (Brooks & McLennan 2002; Webb et al. 2002).

Many ecologists are critical, nonetheless, of studies of regional influences on local community richness (e.g. Hillebrand & Blenckner 2002; Hillebrand 2005; Shurin & Srivastava 2005). One issue is conceptual; some find it hard to accept the idea of a regional pool of species that governs the richness of local communities within the region and that is relatively little affected by local ecological processes. Other concerns are more technical; for example, local and regional richness analyses have been criticized for statistical artefacts and problems of interpretation. Even among those who enthusiastically accept the general idea of regional influences on communities, there is still a widespread tendency to focus on the local scale where powerful tools such as manipulative experiments are available, whereas it has been less clear how to make progress on understanding regional influences.

Here we draw upon recent research to try to provide a series of suggestions for improving our understanding of regional influences on local communities. The proposed steps include: first, advancing our understanding of the regional-scale causes of variation in species richness; second, using improved techniques to test the effects of regional richness on local richness; third, integrating external environmental influences on local and regional richness into analyses of the local and regional richness relationship and, fourth, considering the reciprocal effects of local processes upon regional richness. In conclusion, we consider how to extend similar regionally oriented approaches to studying other aspects of community structure beyond species richness.

Understanding large-scale causes of variation in species richness

Under a strictly ecological worldview, in which local interactions and local abiotic conditions are of prime importance, the properties of regions – such as their species richness, or the relationship of productivity to richness across multiple regions – can seemingly only be built up from the properties of the localities they contain. How is it possible, then, to define a ‘regional species pool’ or a ‘regional influence’ on local richness or other community patterns? One answer is that a region has some attributes that are largely independent of its localities, such as its total area, internal environmental heterogeneity and biogeographic history (as, for example, an old region can contain young localities). It is possible for such attributes to affect regional species richness, and, through dispersal, for regional richness to influence local richness. In one early example, Cornell (1985) found that the geographic range sizes of oaks predicted the regional (whole oak species) richness of their gall wasp faunas, and that regional richness in turn predicted local (per-tree) richness of gall wasp communities on different oak species. The effect of regional area upon local richness has recently been termed the ‘echo pattern’ of large-scale species–area relationships (Rosenzweig & Ziv 1999). More recently, Partel (2002) showed that the relationship of soil pH to local plant richness was positive in large geographic regions within which soil pH was generally alkaline, but negative in regions within which pH was generally acidic, and attributed this difference to the larger pool of species adapted to the prevailing pH in each region.

The recent rapid growth in phylogeny-based techniques, combined with the increasing interest of ecologists in evolutionary influences, has offered hope for a more explicit understanding of the causes of regional-scale species richness. We briefly review some major evolutionary and historical explanations for variation in regional richness, including rates of diversification, niche conservatism and differential dispersal (for a more detailed overview, see Harrison & Cornell 2007).

To test whether regions differ in species richness because of different rates of diversification, i.e. speciation minus extinction, the simplest approach is a sister clade comparison, in which one asks (for example) whether clades at lower latitudes are consistently larger than their sister taxa at higher latitudes. A more complex approach is to analyse lineage-through-time plots to compare rates of diversification among clades. Using these and related phylogeny-based methods, some recent studies have found evidence that higher diversification rates can explain the greater species richness of regions at lower latitudes, as well as differences among other types of regions and among taxa within the same region (e.g. Mittelbach et al. 2007; Ricklefs 2007; Wiens 2007).

Alternatively, some regions may have higher species richness due to niche conservatism (Wiens & Donoghue 2004), i.e. intrinsic and extrinsic limitations on the ability to adapt and disperse, which tend to confine some taxa to the geographic or climatic zones in which they originated. As applied to the latitudinal richness gradient or the productivity–richness relationship, for example, the niche conservatism hypothesis suggests that there are more species in warm and mesic regions because the majority of modern higher taxa originated in warm and mesic environments, and many of these taxa cannot adapt fully to cooler or drier conditions (Wiens & Donoghue 2004; Harrison & Grace 2007; Hawkins et al. 2007; Wiens 2007).

A third class of explanations for variation in regional richness is the differential dispersal of taxa among geographic regions. Dispersal among regions takes place at time scales intermediate between those of speciation and extinction on one hand and competition and predation on the other, and thus may form a critical link between the evolutionary and ecological determinants of community structure (Ricklefs 2004). While it is a difficult class of processes to study, Roy & Goldberg (2007) have devised a model-fitting approach for testing the role of differential dispersal in explaining regional patterns in the richness and age distributions of taxa, and Moore & Donoghue (2007) have recently developed a rigorous method for identifying key dispersal events in lineages.

A few studies have attempted to use evolutionary methods to explain patterns in local as well as regional richness. For example, McPeek and colleagues (e.g. McPeek & Brown 2000) concluded that there were more species of Enallagma damselflies in lakes with than without fish because of the particular history of diversification and post-glacial colonization of this group in the eastern USA, which has produced a regional species pool with fewer dragonfly-adapted than fish-adapted Enallagma. Dragonflies are top predators in fishless lakes, and only some lineages within Enallagma have evolved the behaviours necessary to resist dragonfly predation; Other phylogenetic analyses of local assembly patterns within clades are reviewed by Brooks & McLennan (2002).

Narrow taxonomic scope is a major limitation on these approaches to understanding regional species richness. In general, it has not yet been possible to reconcile the rigor of phylogenetic evidence with the taxonomic breadth of complete ecological communities; in other words, all potentially interacting species as opposed to only the members of a single clade. Clearly, ecologists will continue to need a broad array of approaches to understanding regional influences, including correlative studies and descriptive biogeography as well as phylogenetic analyses (e.g. Qian et al. 2007).

Improving analyses of the local and regional richness relationship

The majority of attempts to understand regional contributions to local richness in natural communities have used regressions of local on regional richness, or local and regional richness analyses, although a minority have used manipulative experiments. Positive local and regional richness relationships have been found widely (Cornell & Lawton 1992; Shurin & Srivastava 2005), including in hyperdiverse communities (Karlson et al. 2004; Witman et al. 2004) and across a broad range of spatial scales (Hillebrand & Blenckner 2002; Shurin & Srivastava 2005; Freestone & Harrison 2006; Cornell et al. 2008). It was originally argued that positive linear local and regional richness relationships implied regional control over local richness, whereas asymptotic relationships implied that local richness was constrained by local ecological interactions (Ricklefs 1987). However, local and regional richness analyses have been much criticized, in part because of sampling and statistical artefacts, and in part because of theoretical problems with their interpretation. We consider each of the most critical problems and discuss possible solutions and limitations, and then discuss the advantages and disadvantages of experimental approaches.

One statistical issue is pseudosaturation, or the tendency for curvilinear local and regional richness relationships to arise if local richness is underestimated or the regional pool is overestimated. Underestimation of local richness results from disproportionate under-sampling of rare species in species-rich regions (Caley & Schluter 1997). Overestimates of the regional pool occur because local richness is usually measured in a single habitat, whereas regional richness is often taken from taxonomic compendia that combine all habitats within a biogeographic region. The supposed pool thus includes species that do not occur in the habitat of interest (Cornell & Lawton 1992; Srivastava 1999). Curvilinearity may also occur if the locality is defined at the excessively small scale of a few individuals, where the numbers of individuals set an upper limit on richness regardless of increases in regional richness (Loreau 2000). At the other extreme, linear relationships at very large local scales may occur because large localities subsume multiple habitats, substantial beta diversity and therefore a large proportion of regional diversity, producing autocorrelation at the two scales (Huston 1999; Loreau 2000; Hillebrand & Blenckner 2002).

Underestimation can be corrected by the use of indices such as Fisher’s alpha or the Chao estimator (Gotelli & Colwell 2001), which use the frequency of rare species to estimate true local richness. Overestimates of regional richness can be avoided if accurate natural history information makes it possible to include in the species pool only those species that are able to occupy the habitat of interest (e.g. Zobel et al. 1998). Autocorrelation can be avoided when truly independent datasets are available to measure local and regional richness (Srivastava 1999), and/or by choosing scales for sampling local richness that are small enough to minimize internal environmental heterogeneity. Random placement null models are another useful way to deal with the autocorrelation and underestimation problems (Belmaker et al. 2008). One way to avoid arbitrary definitions of local and regional scales is to examine the consistency of the relationship across different spatial scales (Cornell & Karlson 1997; Hillebrand & Blenckner 2002; Shurin & Srivastava 2005). Such an approach increases the power to detect curvilinear patterns, which are often missed when the local scale is too large relative to the regional scale (Shurin & Srivastava 2005).

These solutions to sampling bias were employed in a field study of coral species assemblages along a regional biodiversity gradient in the west-central Pacific Ocean (Karlson et al. 2004; Cornell et al. 2008). Data were collected on species-relative abundances, and separate regional richness estimates for flat, crest and slope habitats were used. Five regions were sampled (large, widely spaced island groups) along the gradient, at three local scales, comprising 10 m transects 100–2 m apart, sites 103–4 m apart on 15 islands and islands 104–6 m apart within each region. Slopes of the regressions of log-LSR on log-RSR showed a weak tendency to decrease with decreasing scale using the raw data, suggesting a slight but non-significant tendency towards curvilinearity at the smallest (single transect) spatial scale (Fig. 1). However, when the Chao-1 estimator was substituted for raw richness, the trend disappeared, suggesting that it may have been caused by undersampling; all slopes were linear, suggesting local communities in this case are open to enrichment from the regional species pool.

Figure 1.

 Log-local vs. log-regional species richness of corals at three spatial scales along a regional biodiversity gradient in the west-central Pacific Ocean that includes the world’s most diverse coral assemblages. The local scales comprised transects 100–2 m apart, sites 103–4 m apart and islands 104–6 m apart. Local richness was averaged over all samples at each scale, and separate calculations of local and regional richness were made for each habitat. All habitats fall on the same regression line, and none of the regression slopes differs significantly from 1, indicating linearity of the relationship at all scales (Cornell et al. 2008).

We conclude that it is possible to overcome the statistical obstacles to testing for significant effects of regional on local richness. We suggest that positive local and regional richness relationships provide some evidence for the openness of local communities to regional enrichment, although the degree of linearity or nonlinearity by itself may not be enough to determine the strength of local interactions. Observational local and regional richness analyses should ideally be complemented by experiments, as we discuss below (e.g. Fox et al. 2000; Shurin et al. 2000). Theoretical questions have also been raised about the interpretation of local and regional richness relationships from a metacommunity perspective, in which local and regional richness may have reciprocal influences; we will consider these in Local influences on regional richness: the metacommunity perspective section.

Manipulative experiments addressing the effects of regional processes on local richness have generally involved altering the size of the regional species pool, such as by adding seeds from nearby sites into experimental plots and determining the effect on the richness and composition of the local community (Turnbull et al. 2000). In a now-classic experiment, Tilman (1997) added seeds of 0–54 native species to plots containing varying numbers of resident species. The effects of the additions on the local community were dramatic. The number of species that became established continued to increase with the number of species added, indicating strong effects of the regional pool on local richness. Many of the newly established species have persisted in the plots for > 10 years (D. Tilman, pers. comm.). Nevertheless, richer plots were more resistant to establishment, suggesting that local processes also played a role in limiting the richness of the community. A subsequent study demonstrated that species in the same functional group as the residents were less likely to become established, probably through competitive inhibition (Fargione et al. 2003).

Even stronger resistance to establishment was observed in an experimental manipulation of the regional pool of zooplankton species occupying fishless ponds in northern Michigan (Shurin et al. 2000). When an average of 12.9 species were experimentally added to the ponds, some species were able to colonize successfully, but over 91% went extinct and there were no observable effects of the additions on diversity in the ponds. As in the Tilman (1997) experiment, successful establishment correlated negatively with the diversity of resident species, suggesting that species interactions played an important role in preventing establishment, despite the strong linear relationships between local and regional richness in the zooplankton system. However, when resident species biomass was experimentally reduced, successful establishment by the added species increased dramatically (Shurin et al. 2000).

In an aquatic microcosm experiment, Fox et al. 2000 found a saturating relationship between regional (initial) and local (final) species richness. They attributed this relationship to the fact that the richest regional pools, but not the less-rich pools, apparently included some species that were highly vulnerable to exclusion via local species interactions. This result highlights the importance of ensuring that regional species richness is not confounded with regional species composition.

Regional species pools of moss-inhabiting microarthropods were altered by placing moss fragments from different sites around a defaunated local moss patch and observing the process of colonization for 16 months (Starzomski et al. 2008). Regional richness had no effects on local richness, a pattern that the authors attribute to pseudosaturation. However, the composition of the regional pool did affect the richness and composition of the local community, as did assembly time and seasonality. Other experimental studies confirm the effects of the composition of the regional pool (Fox et al. 2000; Fukami 2004a,b).

The experiments described above generally support the joint importance of local processes and regional pools in limiting local richness, and show that empirically, strong interactions and openness to regional enrichment are not mutually exclusive. One resolution to this apparent paradox is provided by Tilman’s (2004) stochastic niche theory, which proposes that propagules can establish successfully in a saturated community as long as their arrival coincides with an appropriately timed disturbance. Another possibility is that local richness may be maintained not only by competitive relations but by a continuous influx of propagules from other parts of a spatially heterogeneous region, as in mass-effects metacommunity models (Leibold et al. 2004). Some experimental evidence from herbaceous plant communities supports this possibility (Mouquet et al. 2004).

Integrating environmental influences on regional and local species richness

While some influences on species richness appear inherently regional in scope, such as regional area, heterogeneity, age and evolutionary history, and other influences appear obviously local in their action, such as competition and other biotic interactions, other aspects of the environment present more of a conceptual and methodological problem. Gradients in temperature, precipitation, productivity and latitude, for example, have all been shown to correlate with species richness at both local and regional scales, although the form and strength of these relationships may often differ across scales (e.g. Mittelbach et al. 2001, Chase & Leibold 2003; Hillebrand 2004). When regional and local richness are correlated with one another, but both are also influenced by abiotic environmental gradients, what can we conclude about regional and other influences on local richness? Are the environmental gradients in question best interpreted as acting locally, through ecological interactions, or regionally, through evolutionary or biogeographic mechanisms? Could an observed local and regional richness relationship be an artefact of environmental conditions that are shared at the local and regional scales, rather than reflecting cause and effect?

One possible approach to such problems is to include measures of the environment in statistical models of the local and regional richness relationship, using structural equation modelling (SEM; Grace 2006). In such a model (Fig. 2), some environmental factors may affect local richness directly, others may act directly on regional richness and affect local richness only indirectly, and still others may have direct (and possibly different) effects on richness at both scales. This approach has the potential to shed clearer light on the scales at which the environment affects richness, and thereby to aid in the search for plausible mechanisms.

Figure 2.

 Conceptual framework for a multiscale structural equation model of the correlates of species richness (Harrison et al. 2006).

In a study of plants on serpentine soils in California (Harrison et al. 2006), the conceptual model assumed that regional (80–5000 km2) species richness positively affected local (1000 m2) species richness. Climate (e.g. rainfall, temperature, productivity) and spatial structure (area and isolation of serpentine patches) were hypothesized to affect both regional and local richness directly, although potentially in different ways; these variables were measured at both regional scale and local scales. Historical variables (e.g. age of exposure of serpentine within a region) were expected to act directly only on regional richness, while conditions in the study plots (e.g. soil fertility, animal disturbance) were expected to act directly only on local richness. SEM was used to evaluate the significance and strength of the direct and indirect paths in this model. The strongest predictor in the model was the remotely sensed measure of productivity known as NDVI (normalized difference vegetation index), which, when measured at the regional scale, explained 50% of the variation in regional richness. Whether measured regionally or locally, however, productivity exerted only indirect and weak effects on local richness. Regional richness significantly predicted both total and residual variation in local richness (Fig. 3).

Figure 3.

 Results of a multiscale structural equation model of the correlates of species richness (Harrison et al. 2006). Width of arrows indicates the relative strengths of pathways. A ‘+’ indicates positive effect and a ‘+/−’ indicates a unimodal effect.

Productivity is among the most widely studied correlates of local species richness (e.g. Grace 1999; Mittelbach et al. 2001), although cross-study evidence suggests that its relationship to richness is stronger and more consistently positive at large regional scales (Hawkins et al. 2003; Currie et al. 2004), as is the latitudinal diversity gradient (Hillebrand 2004). The study described above is the first to show that productivity acts directly on regional richness and affects local richness only indirectly. This may cast some doubt upon ecological mechanisms for the productivity–richness relationship, such as that greater productivity enhances the ability of competing species to coexist locally, and instead suggests the relationship may reflect history or evolution. Consistent with this, later analyses showed that the positive productivity–richness pattern is driven by the dominance of the regional species pool by species with evolutionary affinities to high-productivity conditions (Harrison & Grace 2007). Similarly, Hawkins et al. (2007) found the regional productivity–richness relationship was stronger in bird lineages of tropical than temperate derivation, and Partel et al. (2007) found the classic unimodal local productivity–richness relationship prevailed only in the temperate zone, where fewer species are adapted to high-productivity conditions than in the tropics. These studies suggest that productivity may be a regional-scale influence on richness, and that evolutionary niche conservatism may be a strongly contributing mechanism.

Hypotheses based upon the existence of a larger regional species pool that influences local-scale richness have previously been proposed to explain the relationships between diversity and productivity (Grace 1999; Huston 1999), diversity and pH (Partel 2002), diversity and invasibility (Wardle 2001), diversity and stability (Loreau et al. 2001), and other ecological phenomena. We suggest that it may be possible to test such hypotheses using statistical models similar to the one described above, in which the influence of regional richness on local richness is a core structural assumption. Such models could also address concerns that have been raised about circularity in local and regional richness analyses; in a model with multiple environmental predictors, SEM could be used to test whether regional richness causes (can uniquely explain variation in) local richness, whether local richness causes (can uniquely explain variation in) regional richness, or whether the local and regional richness relationship disappears once the shared responses of local and regional richness to environmental influences are taken into account.

So far, there have been few studies of the kind suggested in this section, although Grace et al. (2007) used a similar approach to investigate whether local species richness drives local community biomass or vice versa. The major limitation is the requirement for data on regional and local environments as well as regional and local richness. Also, while SEM is the only statistical technique that allows any variable – in this case, regional richness – to both act on some variables and be acted on by others, it has been criticized for its inability to test for complex nonlinear relationships, and Wootton (1994) has argued persuasively that the predictions of SEM models should be tested by experiments.

Local influences on regional richness: the metacommunity perspective

In large part due to the influential work of Ricklefs and colleagues (Ricklefs 1987, 2004, 2007; Ricklefs & Schluter 1993), ecologists have tended to equate embracing the regional scale with acknowledging the importance of evolution and biogeography in ecology. Yet it is inescapably true that regions are made up of localities, and that local ecological processes and dispersal among localities must reciprocally contribute to regional patterns. Moreover, just as ecologists have tended to focus on very small spatial scales, evolutionists interested in questions of species richness have tended to focus on single clades at very broad, often continental or global geographic scales. Perhaps the greatest challenge in the study of regional influences on local communities is the need for more analyses at the scale of regions much smaller than continents yet much larger than collections of field plots – scales of perhaps 101–106 km2. This has been suggested as roughly the scale at which the species pools are assembled that are among the most significant drivers of local community composition (Ricklefs 2004). Within such regions, species richness is likely to be strongly affected by meso-scale influences such as dispersal, local extinction and spatial heterogeneity, which form a link between the very large-scale processes of speciation and (global) extinction on the one hand, and the local-scale processes of competition and predation on the other.

Metacommunity theory, in which regions are portrayed as collections of local communities linked by dispersal, provide an important framework for considering the potential reciprocal interactions between local and regional richness and other aspects of community structure. One of the fundamental questions asked by metacommunity models, for example, is under what conditions locally incompatible competitors can coexist at the regional scale. In a synthesis, Leibold et al. (2004) identified four types of metacommunity models according to the roles played by dispersal and environmental heterogeneity: species-sorting, mass-effect, neutral and patch dynamics models. These models generally predict that the regional coexistence of strong competitors is possible at low-to-intermediate dispersal rates, whereas when dispersal is sufficiently high, local competitive exclusion may propagate to the regional scale and reduce regional richness (e.g. Caswell & Cohen 1993; Mouquet & Loreau 2003). The species-sorting and mass-effects metacommunity models provide an additional reason why regional richness may be relatively resistant to local processes; these models emphasize regional environmental heterogeneity and niche differences, which expand the possibilities for locally incompatible species to coexist stably at the regional scale (e.g. Mouquet & Loreau 2003).

When competition or other interactions are strong, rates of dispersal among localities are high and spatial refuges are absent, metacommunity theory predicts that local dynamics can feed back to reduce regional (metacommunity) richness. Empirical studies have demonstrated the interacting effects of local community dynamics, among-community dispersal and regional species composition in aquatic microcosms and mesocosms; when connectivity among patches is manipulated, the results generally support the predictions that local and regional richness are maximized at low-to-intermediate dispersal (e.g. Holyoak & Lawler 1996; Forbes & Chase 2002; Cadotte 2006). In systems with unmanipulated spatial structure, such as zooplankton in natural ponds or inquilines in pitcher plants, it has usually been easier to test the local-scale than regional-scale predictions of metacommunity models (e.g. Shurin & Allen 2001; Cottenie et al. 2003; Miller & Kneitel 2005). However, Shurin & Allen (2001) showed that patterns of predator and prey zooplankton richness across multiple natural ponds met the predictions of a system-specific metacommunity model.

Metacommunity models have led to an important conceptual critique of local and regional richness analyses, by showing that it is possible for either linear or curvilinear local and regional richness relationships to arise under a variety of assumptions (Shurin & Srivastava 2005). Most significantly, linear local and regional richness relationships may arise even in the presence of strong competition, depending on the relative rates of dispersal (Caswell & Cohen 1993; Mouquet & Loreau 2003; Fukami 2004a), disturbance (Caswell & Cohen 1993; Mouquet et al. 2003; Tilman 2004; Hillebrand 2005) and/or predation (Shurin & Allen 2001). It has also been suggested that curvilinear relationships can arise even without competition (He et al. 2005; Fox & Srivastava 2006). However, by providing the basis for improved null models, metacommunity theory may be useful in refining local and regional richness analyses just as it has been in identifying their weaknesses (e.g. Mouquet & Loreau 2003; Cadotte 2006; Hugueny et al. 2007).

For example, Hugueny et al. (2007) adapted a simple metacommunity model by Hastings (1987) to create a null model for the system of three Daphnia species in rockpools on Baltic Sea islands, where colonization and extinction had been measured in hundreds of rockpools for multiple years (Bengtsson 1989; Pajunen & Pajunen 2003). The observed relationship of local (rockpool) to regional (island) Daphnia species richness was asymptotic. The model by Hugueny et al. (2007) assumed that all Daphnia and all rockpools were identical, Daphnia were either present or absent in rockpools, and competition among the three species would cause a linear increase in local extinction with the number of species per pool. Colonization of pools by each species was modelled in two ways: as a function of the frequency of occupied pools, or as a constant rate, reflecting uncertainty about whether this system shows Levins-like (all populations small and transient) or mainland-island (some populations large and permanent) structure. The model was parameterized and tested with independent datasets. The empirical local and regional richness relationship could only be fit by the model with competition and a Levins-like metacommunity structure (Fig. 4); this model also fits the observed average local species richness per island better than either a no-competition model or a mainland–island model (Hugueny et al. 2007).

Figure 4.

 Observed (points) and predicted (lines) relationships between local and regional species richness for the rock pool Daphnia system. Predictions are based on the Levins-like and mainland–island metacommunity models with competition. For comparison, the predicted relationship without competition (identical in both Levins-like and mainland–island models) is shown. Vertical bars represent the 95% confidence region around the predictions of the Levins-like model (see Hugueny et al. 2007 for details).

Extending metacommunity ideas to large-scale natural systems remains a major empirical challenge. The studies by Shurin & Allen (2001) and Hugueny et al. (2007) illustrate the utility of making predictions about local and regional richness based on detailed metacommunity models, as a basis for improved inferences about underlying processes. However, it must be acknowledged that relevant parameters such as the rates of extinction and colonization of localities will be very difficult to measure in most systems, especially when species have cryptic life-history stages.

Finally, we reiterate that in the absence of the detailed information required to test metacommunity theory, the question of whether regional richness is strongly affected by local processes could be treated as a statistical problem. One possible approach is to test whether local richness predicts residual variation in regional richness or vice versa, as outlined in the preceding section. Another approach has been to analyse the correlation between native and exotic species richness at multiple spatial scales; for example, Stohlgren et al. (2008) used a comprehensive analysis of plant invasions to argue that the addition of new species to communities seldom if ever results in the regional-scale extinction of resident species. Based on metacommunity theory, plausible explanations for his results include weak interactions, spatial heterogeneity and/or low-to-moderate connectivity among communities within regions.

Conclusions: beyond species richness?

In light of the well-recognized need for ecologists to adopt a larger spatial and temporal perspective, we have tried to articulate some concrete ways in which progress can be made in understanding the influence of processes at the regional scale. We have tried to identify areas of confusion and limitations of methodology or data, as well as to report new and emerging insights. We have restricted our discussion to species richness, as it is perhaps the most accessible aspect of community structure, but we think our suggested approaches can be extended to other aspects as well. For example, consider the question of why the top–down influences of predators are greater in some communities than others. Paralleling the approach we have taken, one could ask, first, what is the regional biogeographic history of the predator lineages in question; second, does their species composition vary significantly among discrete regions, suggesting species pool effects; third, how do environmental influences relate to the species composition and impacts of predators at both regional and local scales and, fourth, are local community dynamics plausible explanations for the regional presence or absence of particular predators? In keeping with a very large-scale view, Losos (1996) concluded that the strong differences in predator assemblages between North America, East Africa and Southeast Asia resulted from the intercontinental biogeographic history of the Carnivora. But explanations based upon the environment, local dynamics or dispersal among localities are possible in other cases, particularly at smaller scales.

Regional influences on some other aspects of species composition have been suggested in several of the examples we have described. For example, the changing relative abundances of species of different functional or biogeographic groups within local communities, across gradients of latitude, productivity or soil pH, may be the result of variation in the composition of regional species pools (e.g. Grace 1999; Harrison & Grace 2007; Partel et al. 2007). Metacommunity dynamics have been shown to influence the regional presence or absence of dominant predators or competitors, with substantial effects on the structure as well as the richness of local communities. For example, a strong competitor led to increased community similarity in well-connected microcosm metacommunities (e.g. Forbes & Chase 2002), and the local abundances of competing pitcher plant inquilines were affected by manipulating levels of dispersal (Miller & Kneitel 2005).

One of the greatest recent advances in understanding large-scale influences on local community structure has come from the new field of community phylogenetics, which compares the degree of taxonomic relatedness among members of a local community to their expected relatedness based on a regional species pool (e.g. Webb et al. 2002). Although the focus is on phylogenetic species composition, there are parallels to studies of local and regional richness: first, because both regional and local community data are required, and second, because historical vs. ecological explanations are being contrasted. The usual expectations are that coexisting species will be less related than expected based on the regional pool (i.e. overdispersed) if competition predominates, while they will be more related than expected based on the regional pool (i.e. underdispersed) if regional species pools constitute a dominant limitation on local community membership. Also similarly to analyses of richness, the scale at which regions and localities are defined may influence the results (Cavender-Bares et al. 2006; Swenson et al. 2006). As the regional scale increases, local communities tend to switch from over- to underdispersion, and it has been suggested that this switch can be used to indicate the scale at which regional influences become dominant controls over local community membership (Swenson et al. 2006). However, the interpretation of scale-dependence in community phylogenetic structure is still in its early stages (Cavender-Bares et al. 2006).

Hardy & Senterre (2007) have recently proposed an analytical framework in which the phylogenetic structure of a community can be partitioned into within-locality (alpha) and among-locality (beta) components, exactly analogous to the traditional multiscale partitioning of species richness. This new approach offers a potentially powerful way to ask whether there are parallel patterns in these two aspects of the local and regional geographic structure of communities. At this stage, community phylogenetic analyses dealing with entire communities (as opposed to particular clades) are limited by their requirement for complete phylogenies, and the kinds of sampling biases that may arise in such analyses are just beginning to be explored (Hardy & Senterre 2007).

No single study has yet combined more than one of the steps we suggest here. Data limitations remain perhaps the most formidable obstacles. Nonetheless, we conclude with the optimistic suggestion that the ever-increasing availability of large environmental, biogeographical and phylogenetic databases, combined with the new analytic tools ranging from SEM and geostatistics to community phylogenetics, offers the possibility that creative ecologists and evolutionists will fulfil the hope expressed in classic works such as Geographical Ecology by MacArthur (1972) and Species Diversity in Ecological Communities edited by Ricklefs & Schluter (1993): a fuller understanding of the large-scale regional processes that shape local ecological communities.

Acknowledgements

We thank T. Fukami, J. Chase, M. Holyoak and two anonymous referees for insights that greatly improved an earlier version of the manuscript.

Ancillary

Advertisement