Neutral Ecological Theory Reveals Isolation and Rapid Speciation in a Biodiversity Hot Spot

See allHide authors and affiliations

Science  09 Sep 2005:
Vol. 309, Issue 5741, pp. 1722-1725
DOI: 10.1126/science.1115576


South Africa's Mediterranean-climate fynbos shrubland is a hot spot of species diversity, but its diversity patterns contrast strongly with other high-diversity areas, including the Amazon rain forest. With its extremely high levels of endemism and species turnover, fynbos is made up of dissimilar local communities that are species-rich but relatively poor in rare species. Using neutral ecological theory, we show that the relative species-abundance distributions in fynbos can be explained by migration rates that are two orders of magnitude lower than they are in tropical rain forests. Speciation rates, which are indexed by the “biodiversity parameter” Θ, are estimated to be higher than they are in any previously examined plant system.

Regions with many endemic resident species are typically grouped together as biodiversity “hot spots,” but this label obscures fundamental differences among high-diversity systems in species-diversity patterns and in the processes that generate them. The fynbos shrubland of South Africa's Cape Floristic Region (CFR) presents a particularly strong contrast with other diverse systems such as the Amazon rain forest. Amazon rain forest tree communities are dominated by species with large ranges, most of which are locally rare (1, 2), whereas most fynbos shrub species are narrow endemics, and many can be locally abundant (1, 3). Neutral ecological theory (4) enables the quantification of diversity in these contrasting systems in terms of the mechanisms that generate and partition it: rates of speciation and rates of migration.

The CFR contains over 8,000 plant species, of which a majority (69%) are endemic to the region (2). Mediterranean-climate regions are among the most diverse systems outside the tropics, and among these, the CFR has the most species in the smallest area (5). Molecular phylogenetic studies have estimated that speciation rates in some lineages in the CFR are exceptionally high (1, 3). Moreover, CFR plants are thought to have very low migration rates (6). High levels of endemism among CFR species are associated with low observed seed-dispersal distances and short-distance dispersal modes (ant and passive dispersal) (7). Even among Mediterranean-climate regions, the CFR has very high rates of species turnover along habitat and geographic gradients, i.e., high β and γ diversities, respectively (8), which is a spatial pattern that suggests low migration rates (5). Amazon rain forest, in contrast, occupies a much larger area, and can contain as many as 300 tree species in a sample of 600 individuals (extremely high local, or α, diversity) (9). Relatively few Amazon tree species appear to be locally endemic, because even rare species have generally broad distributions, producing comparatively lower β and γ diversities (10, 11). Studies of tropical rain forest trees have shown them to be generally well dispersed, with long mean-dispersal distances and frequent long-distance dispersal events (12, 13).

Hubbell's neutral theory of biodiversity and biogeography (“neutral theory”) provides mechanistic insight into the processes driving these two types of high-diversity systems (4). Neutral theory extends previous ecological work on quantifying and comparing diversity patterns by grounding its descriptions in interpretable ecological processes; its two core parameters represent the processes of speciation and migration (14). The “biodiversity parameter,” Θ, represents the process of speciation within a “metacommunity,” which is a very large collection of similar organisms found across a biogeographic region (4). This parameter is twice the product of the size of the metacommunity Jm (a very large number) and the speciation rate υ, which is the number of new species arising per birth (a very small number). Thus, Θ is proportional to the expected number of new species per generation across the metacommunity and can be used to quantify the diversity-generating capacity of a system. In terrestrial plant systems, Θ has been shown to range from ∼1 in boreal forest to ∼200 in tropical rain forests (4). Within the metacommunity are “local communities,” which are groups of organisms at a particular locality that are linked to the meta-community by migration. The migration parameter m is the proportion of individuals per generation in each local community that come from elsewhere in the metacommunity (4). Given values for m, Θ, and the size of the local community (J), neutral theory produces an expected relative species abundance distribution, called the zero-sum multinomial (ZSM) (15, 16), which can be fitted to sample inventory data.

Neutral dynamics can generate two extreme types of species-rich systems, depending on the amount of migration. A high-migration system (m < 0.1) has species-rich local communities with high evenness in relative abundances. A low-migration system (m < 0.01) is a spatially structured collection of dissimilar local communities that are less species-rich because of local extinction of rare species through stochastic drift and because of correspondingly higher dominance by a few species in each local community. For the same value of Θ, a high-migration system will have higher local diversity, whereas a low-migration system will have proportionally higher species turnover among localities.

An advantage of neutral theory over alternative descriptors of biodiversity patterns is that its parameters represent easily interpreted ecological processes. However, little work has been done to produce parameter estimates for a range of systems and to assess what insight they provide about ecological dynamics. Because most previous applications of neutral theory to plant communities have used tropical rain forest data sets (12, 13, 15, 17), the application of the theory to real-world patterns has been confined to only a few geographic regions and a small portion of parameter space. Here we contrast the Amazon rain forest system with the CFR, which may be the clearest example of a species-rich low-migration system.

The CFR data come from three sites spanning the region—the Cederberg mountains in the northwest, Cape Hangklip in the southwest, and the Zuurberg mountains at the eastern edge. Each of the sites in our data set is an identifiable subregion of the CFR, which consists of a range of hills or mountains and adjacent lowlands (18). For each site, we obtained existing floristic survey data, including hundreds of 5-m by 10-m plots, in which all vascular plants had been identified to species (Table 1) (1922). From these data, we obtained the abundances of all woody plant species for each site (subshrubs, shrubs, and trees) (23). For the comparison to tropical rain forests, we used tree-count data from networks of large plots in the Manu National Park in Peru and Yasuní National Park in Ecuador (10, 11). Estimates for Barro Colorado Island (BCI) in Panama, using tree count data from the 50-ha plot there, were also included as a well-studied baseline (13).

Table 1.

Data sets.

Site Plots Species Individuals
Cederberg (View inline) 191 247 11,765
Cape Hangklip (View inline) 251 247 24,003
Zuurberg (View inline) 69 114 8,916
Manu National Park, Peru (View inline, View inline) 15 648 8,055
Yasuní National Park, Ecuador (View inline, View inline) 14 546 7,613
BCI, Panama (View inline) 1 225 21,457

Observed patterns of diversity and abundance in the CFR are very different from those documented for tropical rain forests (2, 24). Figure 1, which displays species relative abundances at a CFR site (the Cederberg) and an Amazon forest site (Manu National Park), shows that the CFR site has higher dominance by common species, with fewer rare species. Each CFR site contains fewer species than do typical tropical rain forest localities containing equivalent numbers of individuals, but local endemism in the CFR is extremely high, resulting in very high species turnover among sites (18, 25). The mean adjusted Jaccard's Index [a measure of taxonomic similarity between sites that adjusts for sampling effects (26)] for the three CFR sites studied here is 0.024, compared with 0.47 for the two Amazon sites, despite the shorter distances among the CFR sites (∼150 to 400 km versus ∼1500 km).

Fig. 1.

Rank-abundance plots for the Cederberg (CFR) and Manu (Amazon) sites. The triangles (Cederberg, red lines) and circles (Manu, blue lines) show abundance of the ranked species for each site, normalized by total abundance and log transformed. Solid lines show rank abundances predicted by neutral theory for each site, using maximum likelihood parameter estimates. Dashed lines indicate bootstrapped 95% confidence intervals around the predicted rank abundances, quantifying sampling error expected when drawing from a neutral community with known parameters.

We used Bayesian and maximum-likelihood methods to fit the ZSM to species-abundance data from each site (23). Parameter estimates were obtained by using a likelihood derived from the closed-form expression for the ZSM given in Volkov et al. (15), which provides the predicted distribution of abundances for a sample of species from a neutral community. To obtain Bayesian parameter estimates, uninformative priors were placed on the parameters (23).

Bayesian estimates of m for the CFR ranged from 0.0018 to 0.045, which is about two orders of magnitude lower than the estimates for the rain forest sites (0.54 for Yasuní, 0.55 for Manu, and 0.1 for BCI) (Fig. 2 and Table 2). The maximum likelihood estimates for the parameters were consistent with the Bayesian results (Table 2). None of the credible intervals of the CFR estimates of m overlapped those of the Amazon estimates. Within the CFR, the lowest migration estimate is for Cape Hangklip, which is one of the most diverse areas of the CFR (2, 25). Zuurberg, at the eastern edge of the CFR, had much higher estimates of m, approaching those of BCI in Panama.

Fig. 2.

Bayesian estimates of the ZSM parameters m (upper two panels) and Θ (lower two panels). Posterior densities for tropical rain forest sites are shown in gray and CFR sites are shown in black. (A) The low-m sites: Cape Hangklip (solid black line) and Cederberg (dashed black). (B) The higher-m sites: Zuurberg (dotted black), Barro Colorado Island (dashed gray), Yasuní (dotted gray), and Manu (solid gray). (C) Zuurberg (dotted black line), BCI (dashed gray line), Yasuní (dotted gray), and Manu (solid gray). (D) Cederberg (dashed black) and Cape Hangklip (solid black).

Table 2.

Parameter estimates under neutral theory ZSM distribution. Estimates for neutral theory parameters Θ and m are shown. Both maximum likelihood estimates (MLE) and Bayesian estimates [posterior (Post.) mode] are presented; the 0.025 and 0.975 quantiles for the Bayesian estimates quantify the uncertainty associated with the estimates.

Site Θ m
MLE Post. mode 0.025 quantile 0.975 quantile MLE Post. Mode 0.025 quantile 0.975 quantile
Cederberg 394.6 370 198.9 2350.5 0.0055 0.0053 0.0032 0.0090
Hangklip 640.3 530 328.3 3033.7 0.0020 0.0018 0.0014 0.0029
Zuurberg 30.2 29 15.3 52.5 0.037 0.045 0.015 0.13
Manu 187.4 183 154.9 228.3 0.53 0.55 0.34 0.81
Yasuní 152.8 153 121.9 188.6 0.50 0.50 0.34 0.80
BCI 47.7 45 31.3 67.0 0.093 0.10 0.048 0.44

Estimates of Θ for the CFR were very high, peaking in the southwest (Cape Hangklip) at Θ = 530, which is more than twice the highest value yet estimated for any rain forest (∼200) (4). Because the posterior distribution for Θ in low-migration communities is right-skewed (Fig. 2D), we present the posterior mode rather than the posterior mean as a conservative estimate (Table 2). The credible interval for Θ for Cape Hangklip does not overlap the interval for Manu, the richer of the two Amazon communities, and this value is the highest estimate of Θ for any plant system yet examined. As with m, there is a longitudinal trend in Θ within the CFR: the highest Θ values coincide with the lowest migration rates in the southwest.

The eastern part of the CFR, including the Zuurberg site, has more-even seasonal rainfall patterns than does the southwestern CFR, and the vegetation contains a higher proportion of widely distributed, well-dispersed forest and thicket species (27). The lowest migration rates are thus associated with the highly seasonal winter rainfall conditions in the southwest, as well as with the highest levels of endemism and species turnover (28). The southwestern site (Cape Hangklip) is also geographically central, whereas the eastern site (Zuurberg) is the most peripheral, suggesting that edge effects may be reducing species diversity at the periphery of the region.

Combining the data from the subregions within each region (CFR and Amazon) produces a proxy for samples from the two meta-communities. By using the composite data, we can assess the importance of migration limitation by comparing the fits of models with and without a migration parameter (16). Estimates of Θ for the metacommunity neutral model with no migration limitation were performed by maximizing the likelihood of the metacommunity sampling distribution (16). The assumption of unlimited migration drastically lowers the estimate of Θ for the CFR (from 697 to 88.9), whereas Θ for the Amazon is only moderately reduced (251 to 219). The reduction in the CFR estimate is not surprising, because if migration limitation is ignored, the lower species richness can be accommodated only by lower values of Θ. The assumption of unlimited migration is clearly unsatisfactory for both systems, because a likelihood ratio test strongly favors the model with migration limitation in both cases [for Amazon: χ2 = 24.6, 1 degree of freedom (d.f.), P < 0.001; for CFR: χ2 = 19.4, 1 d.f., P < 0.001].

Estimates of Θ can be converted to estimates of per capita speciation rate if it is possible to estimate the number of individuals in the metacommunity. With a much larger area, the Amazon rain forest metacommunity almost certainly contains at least as many individuals as does fynbos, which covers <50,000 km2. The high estimates for Θ thus imply a substantially higher per capita speciation rate in the CFR. These results are consistent with the prevailing view that the western CFR is an extremely migration-limited system with extraordinarily high speciation rates. Each of the local communities in our data set is an identifiable subregion of the CFR that consists of a range of hills or mountains and adjacent lowlands. Thus, the CFR metacommunity is topographically fragmented, which predicts low migration rates, consistent with our results. With these low migration rates, it is likely that local communities will be sufficiently isolated to allow ecological drift to cause divergence among communities, because few individuals per generation will be exchanged (19). Further, populations of individual species will be genetically isolated, so that genetic drift will tend to cause divergence and the formation of new species. Thus, our results support the view that the fynbos metacommunity consists of topographical islands isolated not by water but by drier lowlands. This pattern contrasts sharply with that found in tropical rain forests, which exhibit high connectivity over long distances. Within the CFR, the association of the highest values of Θ, and thus speciation rates, with the lowest migration rates supports the view that isolation over short spatial scales (1 to 100 km) has played a role in generating, as well as structuring, the high diversity of the CFR.

Supporting Online Material

Materials and Methods

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article