Technical Comments

Comment on “Global Correlations in Tropical Tree Species Richness and Abundance Reject Neutrality”

See allHide authors and affiliations

Science  29 Jun 2012:
Vol. 336, Issue 6089, pp. 1639
DOI: 10.1126/science.1220980


Ricklefs and Renner (Reports, 27 January 2012, p. 464) suggested that strong correlations in the diversity of shared families between isolated tree assemblages reject neutrality. Simulations of a neutral model indicate, however, that isolated assemblages under various configurations of random speciation and extinction do sustain strong correlations in the diversity of shared families. Thus, reported correlations support rather than reject neutral theory.

Assessing the roles of neutral (i.e., random speciation-extinction-immigration) and ecological (e.g., predation, competition, and diseases) processes in macroecological patterns is a topic of considerable recent attention (1, 2). In a new analysis with tropical trees, Ricklefs and Renner (3) showed that isolated assemblages display strong correlations in the diversity of their shared families. They rationalize that neutral processes (i.e., random extinction-speciation-immigration) should yield uncorrelated diversity patterns between independently evolving floras; so, under this null hypothesis, the existence of strong correlations reject neutrality and support the role of homogenizing ecological forces over large scales. Here, I simulated assemblages under different scenarios of random extinction, speciation, and immigration, and show that although neutral processes do indeed cause a differentiation in the constituent species, isolated assemblages can sustain high correlations in the richness of shared families. This suggests that strong correlations in the richness of shared families are expected under neutral theory and, thus, that reported correlations support rather than reject neutrality.

I used the same data of tree diversity analyzed by Ricklefs and Renner (3) [i.e., Southeast Asia (n = 3 sites), the Neotropics (n = 3 sites), and Africa (n = 1 site)] and carried out the following procedure for the neutral model. First, data from all assemblages were merged to create a global pool of species. For each of the seven assemblages, I generated an assemblage with the same number of species but with species randomly selected from the global pool (this should resemble a broad range of initial similarities that the original assemblages might have had before their separation or isolation). In each randomly generated assemblage, I applied an equal probability of extinction to all species, and those species that became extinct were replaced by random speciation of local species or by random immigration from the global pool of species. (Although the assemblages are thought to be long isolated, I modeled the effects of immigration, and in the scenario of complete isolation, immigration was set to zero.) Newly formed species were given a particular probability of also being new genera. After these events of extinction, speciation, and immigration were applied independently to each assemblage, I calculated the correlation in the richness of families shared between all possible pairs of assemblages as calculated in Ricklefs and Renner (3); additionally, I also calculated species and genus similarity between all pair of assemblages using the Sorensen index. This procedure was repeated independently on each subsequent assemblage for 100 time-steps (this was sufficient to reach stable patterns), at which point an iteration of the neutral model ended; I carried out 1000 iterations of the neutral model to obtain average and confidence intervals for neutral expectations.

Simulations under a broad range of values for the parameters of the neutral model (Fig. 1) revealed that neutral assemblages [within (Fig. 1, F to J) and between (Fig. 1, A to E)] can sustain strong correlations in the richness of shared families even under scenarios of complete isolation and after assemblages lose all of their species and genera in common; such a degree of assemblage distinction is rarely observed among actual assemblages, whose genera similarities range from 7% to 72% and species similarities from 0% to 23%. (Such similarities are for all 21 possible comparisons, which include within- and between-region comparisons.) The relationship between genera similarity and family richness correlation in neutral assemblages (lighter gray area in Fig. 2A) revealed that given actual genera similarities, 18 out of 21 possible comparisons (i.e., 86%) have family richness correlations expected under neutral predictions (open circles in Fig. 2B). It is important to note that these neutral correlations are conservative because they were calculated assuming complete isolation of assemblages; adding any level of immigration, as may occur within regional assemblages or between regions early on, should increase the magnitude of the correlations. (Compare the effect of immigration in Fig. 1C versus no immigration in Fig. 1, A, B, D, and E). Interestingly, the only three departures (red filled circles in Fig. 2B) were all below neutral expectations, suggesting that local ecological and environmental factors, when having an effect on assemblage structure, tend to accelerate differentiation between assemblages rather than homogenizing them. Overall, these results suggest that strong correlations in the diversity of shared families between isolated assemblages are expected under neutral theory, and thus the observed correlations support rather than reject neutrality.

Fig. 1

(A to J)A case example of the similarity between two assemblages under different neutral scenarios. The values used for the parameters of the neutral model are indicated above the plots. Colored areas indicate the 95% confidence intervals for the correlation in the richness of shared families (yellow; black lines indicate the mean), genera similarity (green), and species similarity (blue). Upper plots indicate the comparisons between both assemblages and the lower plots the comparisons within one of the assemblages over time (i.e., the comparison of the assemblage at time zero versus the same assemblage at time i). In each of the upper plots, the circle indicates the correlation in the richness of shared families given the genus similarity of the actual two assemblages compared.

Fig. 2

Comparison of neutral versus actual correlations in the richness of families shared between isolated assemblages. For each possible pair of assemblages analyzed, I ran a neutral model using their actual number of species and conservative values for the parameters of the neutral model: Immigration = 0; as demonstrated by Ricklefs and Renner (3), this is a fair assumption. Extinction = 0.1; as noted in Fig. 1, A and B, different values of extinction do not affect the general shape of the patterns. Genus generation = 0.5; I set this value based on the fact that among the analyzed species, 50% belong to only one genus. (A) From the resulting 1000 simulations, I plotted the average genus similarity against the 95% confidence intervals of the correlation between shared families richness (lighter gray area). On that same plot, I plotted the actual genus similarity and family correlation of the two assemblages compared (open black circle). (B) The slide of confidence intervals at each genera similarity for all possible comparisons.

References and Notes

  1. Acknowledgments: I thank the principal investigators of the Center for Tropical Forest Science for facilitating the data from the forest plots used in this analysis and H. Partika and Q. Chen for facilitating computing services. The manuscript benefited from the insightful and constructive comments of R. Ricklefs, K. Gaston, and S. Renner.
View Abstract

Navigate This Article