Disentangling the Drivers of β Diversity Along Latitudinal and Elevational Gradients

See allHide authors and affiliations

Science  23 Sep 2011:
Vol. 333, Issue 6050, pp. 1755-1758
DOI: 10.1126/science.1208584


Understanding spatial variation in biodiversity along environmental gradients is a central theme in ecology. Differences in species compositional turnover among sites (β diversity) occurring along gradients are often used to infer variation in the processes structuring communities. Here, we show that sampling alone predicts changes in β diversity caused simply by changes in the sizes of species pools. For example, forest inventories sampled along latitudinal and elevational gradients show the well-documented pattern that β diversity is higher in the tropics and at low elevations. However, after correcting for variation in pooled species richness (γ diversity), these differences in β diversity disappear. Therefore, there is no need to invoke differences in the mechanisms of community assembly in temperate versus tropical systems to explain these global-scale patterns of β diversity.

Some of the most striking and frequently documented patterns in ecology are that species richness in local communities generally declines with increasing latitude and elevation, such that the diversity of many clades peaks in lowland, tropical areas (1, 2). The mechanisms underlying these gradients are often difficult to distinguish because multiple processes operating at multiple scales may govern geographic variation in diversity (3). For example, declines in diversity with elevation and latitude could result from deterministic community assembly processes at local scales (4, 5). Alternatively, spatial variation in local diversity could depend on processes that operate at larger scales (e.g., speciation, extinction, and biogeographic dispersal), which trickle down to affect diversity in the embedded localities (6, 7). One way to disentangle such multiscale effects is to examine patterns of diversity across scales, with a particular focus on β diversity (a measure of compositional differences among samples), which links local (α) to larger-scale (γ) diversity (811). Differences in β diversity along biogeographic gradients have been interpreted as reflecting differences in the ecological processes acting along these gradients, including variation in the range size (11) and dispersal ability (9) of species and in the strength of local processes, such as habitat filtering (8).

As an example of how β diversity decreases with latitude and elevation, and the corresponding changes in α and γ diversity, we use two data sets of woody plants. The first is from 197 locations along a latitudinal gradient spanning more than 100° (12, 13), and the second is a similar set of eight locations spanning a 2250-m elevational gradient in Carchi, Ecuador (14, 15). We define α diversity as the species richness of a single 0.01-ha subplot, γ diversity as the total richness of the 10 subplots (totaling 0.1 ha) at a location, and β diversity as the heterogeneity in species composition (16) among the 10 subplots of 0.01 ha each established at each location, measured as the multiplicative β partition (β=1α¯/γ) (17, 18). This spatial scale is smaller than has been used in many other studies of β diversity, but it is appropriate to capture responses to fine-grained environmental heterogeneity (19), as well as the local neighborhood interactions that are known to strongly influence community assembly in temperate (20) and tropical (21) forests, although it does not capture coarser-grained environmental effects.

In these data sets, sampled woody plant diversity at both smaller (α diversity) and larger (γ diversity) spatial scales declines with increasing latitude (Fig. 1A) (12, 22) and elevation (Fig. 1B) (14). Because γ diversity declines more rapidly along both gradients than does α diversity, β diversity therefore declines with increasing latitude (Fig. 1C) and elevation (Fig. 1D). Thus, these data sets, although collected at small spatial scales, show the same patterns typically seen in larger-scale analyses (8).

Fig. 1

Latitudinal and elevational trends in mean α and γ diversity for woody plants (A and B) drive a significant correlation between latitude and β diversity (C) and elevation and β diversity (D). β diversity is measured as the β partition (Embedded Image).

Although a common explanation for these declines in β diversity would help explain latitudinal and elevational diversity gradients, caution is needed before ascribing any possible ecological mechanisms to these declines in β diversity. It is widely recognized that β diversity is a simple function of α and γ diversity regardless of how it is calculated (e.g., multiplicative, β=γ/α¯; additive, β=γα¯; or β partition, β=1α¯/γ), and, therefore, is not independent of variation in either α or γ diversity (16, 23, 24). Even supposed “true” measures of β diversity (25) can vary simply because of changes in γ diversity (26). Because γ diversity varies along both latitudinal and elevational gradients, its influence on α and β diversity must be accounted for before any ecological explanations are offered.

To account for effects of variation in γ diversity, we first explored the relation between γ diversity and β diversity in the absence of any process other than random sampling. We found that expected β diversity increased with γ diversity, which can be shown algebraically for the multiplicative β partition (Fig. 2A) (27), or with a simple simulation model using a wide variety of other traditional β diversity metrics (fig. S2). This expected relation between β diversity and γ diversity holds regardless of the specific scales used to measure α and γ diversity (e.g., Fig. 2A).

Fig. 2

The relation between β and γ diversity (A) expected algebraically, based on the mean probability that a species occurs in a subplot when sampled from a larger species pool where abundances follow a lognormal distribution. Curves represent β-diversity values, measured as the β partitionEmbedded Image, for 10 subplots each composed of n individuals, as indicated. Similar relation are observed in empirical data from woody plants along a latitudinal (B) and elevational (C) gradient. See supporting online material for simulations showing similar relation for other common measures of β diversity and for samples generated from uniform abundance distributions.

Furthermore, in the woody plant data sets presented here, the correlation between γ diversity and observed β diversity along either the latitudinal (Fig. 2B) or elevational (Fig. 2C) gradient was consistent with the pattern expected, solely on the basis of random sampling of individuals from the species pool. Because of this consistency, it is not yet parsimonious to infer that ecological mechanisms (e.g., niche-based processes or habitat associations) drive the observed differences in community structure along these biogeographic gradients. Instead, a null modeling approach is first needed to determine if β diversity deviates from the expectations of a random (stochastic) assembly process and whether the magnitude of the deviation varies along latitudinal and elevational gradients.

Using the woody plant data sets, we compared observed patterns of β diversity to patterns generated by a null model. The null model randomly shuffles individuals among subplots while preserving γ diversity, the relative abundance of species at the location, and the number of individuals per subplot (28). This explicitly corrects for γ dependency (fig. S4) and provides expected values of β diversity for each site based solely on random sampling from the species pool.

It was surprising that the null model analysis revealed that β diversity is generally greater than expected at nearly all locations along both latitudinal and elevational gradients (Fig. 3). This suggests that species tend to be more aggregated within local subplots than expected by chance (29). Aggregation across the range of species pools, climates, and forest types in our study could be explained by habitat filtering (30), dispersal limitation (31), and/or priority effects (32). However, the magnitude of the deviation did not vary systematically along latitudinal or elevational gradients (Fig. 3, C and D). In other words, after correcting for differences in species pool size, β diversity was the same both at tropical and temperate sites and at high- and low-elevation sites. This means that the net outcome of local community assembly processes is consistent (in terms of their effect on β diversity) across these gradients (33) at the scale of our study.

Fig. 3

Patterns in observed (red) and expected (black) β diversity of woody plants along a latitudinal (A) and an elevational (B) gradient and patterns in the β deviation, a standard effect size of β-diversity deviations from a null model that corrects for γ dependence, with either latitude (C) or elevation (D).

Taken together, our results indicate that variation in β diversity across broad biogeographic gradients is more likely to be driven by γ diversity than by differences in the mechanisms of community assembly (e.g., niche versus neutral) (32, 34); range size and dispersal; or density-dependent interactions (21, 35). Therefore, there may be no need to invoke different local assembly processes when trying to explain latitudinal or elevational differences in β diversity. Instead, a more plausible explanation is that variation in biogeographic or regional processes sets the size of the species pool (3), and the combined influence of local processes acts in a consistent way across large-scale diversity gradients (33) to produce the patterns of species turnover that are ubiquitous in the natural world.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We are grateful to A. H. Gentry, the Missouri Botanical Garden, and numerous additional collectors who contributed to the latitudinal data set. The data sets are available in the original publications or electronically from SALVIAS ( This work was conducted as part of the Gradients of β-diversity Working Group supported by the National Center for Ecological Analysis and Synthesis (NCEAS), a center funded by NSF (grant EF-0553768); the University of California, Santa Barbara; and the state of California. N.J.B.K. was supported by the National Science and Engineering Research Council of Canada CREATE Training Program in Biodiversity Research. L.S.C. was supported by an NCEAS postdoctoral fellowship. N.J.S. was supported by U.S. Department of Energy Program for Ecosystem Research DE-FG02-08ER64510. J.C.S. was supported by an NSF Postdoctoral Fellowship in Bioinformatics (DBI-0906005).
View Abstract

Navigate This Article