Report

The Shifting Balance of Diversity Among Major Marine Animal Groups

See allHide authors and affiliations

Science  03 Sep 2010:
Vol. 329, Issue 5996, pp. 1191-1194
DOI: 10.1126/science.1189910

Abstract

The fossil record demonstrates that each major taxonomic group has a consistent net rate of diversification and a limit to its species richness. It has been thought that long-term changes in the dominance of major taxonomic groups can be predicted from these characteristics. However, new analyses show that diversity limits may rise or fall in response to adaptive radiations or extinctions. These changes are idiosyncratic and occur at different times in each taxa. For example, the end-Permian mass extinction permanently reduced the diversity of important, previously dominant groups such as brachiopods and crinoids. The current global crisis may therefore permanently alter the biosphere’s taxonomic composition by changing the rules of evolution.

Although most higher taxa are affected by major marine radiations and extinctions, different groups have peaked in diversity at different times (1). For example, Paleozoic ocean floors were dominated by trilobites, brachiopods, and crinoids, whereas Cenozoic communities were dominated by scleractinian corals and molluscs. Long-term shifts in composition may be explained in two fundamental ways. First, they could result from persistent differences among groups in their styles of diversification (2, 3). On the basis of this theory, the decline in background extinction rates through the Phanerozoic (4) has been attributed to the loss of groups with high intrinsic turnover rates (5). Second, shifts could reflect isolated adaptive radiations or differential responses to mass extinctions. Numerical models of Phanerozoic marine diversification have often combined both perspectives (3, 6, 7).

If diversity trends are determined by group-specific characteristics and these characteristics are invariant, then mass extinctions merely delay the rise of groups that are fated to prosper (3). Furthermore, given these assumptions, it is possible to predict which groups will recover most quickly from the mass extinction that is now occurring. Successful groups would be expected to exhibit logistic dynamics [a decline of growth rates as diversity rises (3, 8, 9)]. Each group’s intrinsic rate of increase and equilibrium diversity level must be quantified in order to test this prediction.

Robust estimates of these dynamical parameters cannot be obtained simply by fitting a line to a fossil diversity curve [a kinetic analysis sensu (3)] because apparent plateaus can be generated by other processes such as random walks. Modeling net diversification rates (a dynamic analysis) removes this ambiguity. These rates are less likely to be misleading because they generally show weak trends and little autocorrelation. They can be quantified easily by taking log ratios of consecutive values (9).

Here, I address these questions using Phanerozoic-scale diversity counts for the 14 most commonly occurring marine animal higher taxa, including six major groups (Fig. 1). The counts were generated from taxonomic occurrences housed in the Paleobiology Database (10). Other recent studies have used the same resource (9, 1113). Data for other groups are not sufficient to yield complete and precise time series.

Fig. 1

Diversity trajectories of six marine groups having standardized diversity levels with peaks of at least 100 genera or medians of at least 20 genera. (A) Data for the most important groups in each evolutionary fauna: Trilobita (dotted line), articulate Brachiopoda (gray line), and Gastropoda (black line). (B) Data for other important groups: Cephalopoda (dotted line), Anthozoa (gray line), and Bivalvia (black line).

The counts were generated by sampling standardization—randomly drawing a comparable amount of data in each time interval (9, 12). However, the method used in this paper differs from simple standardization because it ensures fair representation of uncommon taxa. It works by tracking coverage of the occurrence-frequency distribution, meaning the sum of frequencies of taxa that have been seen at least once in a given subsample. Sampling continues until coverage (as opposed to sheer sample size) reaches a specified target (14). Subsampling and frequency coverage are long-established concepts in the ecological literature, but they have not been bound together methodologically. The algorithm derived here can be used to estimate the relative number of classes in any set of sampling universes regardless of whether the data are biological (14).

Most individual groups were found to diversify without any change in underlying dynamics for periods of at least 200 million years. More specifically, almost every group’s curve exhibits some kind of logistic growth. This finding is based on a nonparametric maximum likelihood analysis comparing observed and predicted changes in diversity (14). Random walk and exponential models can be excluded for eight groups, and multiphase logistic growth models were favored (Akaike weight > 0.75) for Linguliformea, Bryozoa, and Echinoidea. Similarly, previous studies suggested that subclades of particular marine groups such as bryozoans (15), bivalves (16), and gastropods (17) experienced density-dependence. Although the curves for Anthozoa, Trilobita, Gastropoda, and Bivalvia could be fit to an exponential model, support for other models is as strong or stronger in all of these cases.

These findings refute the idea that diversity trends, including successive replacements of groups, are a function of exponential or random growth interrupted by mass extinctions (6, 7, 18). Such arguments only ever seemed plausible because biases in older data sets, including the greater quantity and quality of data in the Cenozoic, created the appearance of a steep post-Paleozoic increase (12). These biases also obscured large, rapid shifts in diversity such as the Cambrian explosion and the mid-Jurassic radiation.

Sepkoski (3) categorized marine animals into so-called Cambrian, Paleozoic, and Modern evolutionary faunas on the basis of shared curve shapes. This categorization has served as a benchmark for macroevolutionary research at the Phanerozoic scale [for example, (19)]. Sepkoski (2, 3, 20) also modeled the rate and timing of taxonomic replacements by assuming that each fauna has diversified logistically and that the underlying logistic parameters have held constant through the Phanerozoic. If so, then recoveries from any mass extinction should be predictable, and extinctions should not alter the fate of the major groups (3).

Curve-shape similarities are easily recognized by applying multivariate ordination methods (20), and a factor analysis of the 14 diversity curves recovers the three original categories (Fig. 2A). Furthermore, the summed diversity trajectory of each evolutionary fauna is proportionately about the same in either Sepkoski’s original family-level compendium (3, 20), his genus-level compendium (21), or the data set analyzed here (Figs. 1 and 3). The general resilience of the proportions is noteworthy because the global diversity curve (Fig. 3) differs from all published curves that stem from Sepkoski’s compendia (3, 21, 22). The older curves feature proportionately lower Cambrian and Modern faunal diversity in the Paleozoic. These differences result from methods of sampling and counting taxa (12, 14) and do not reflect important discrepancies in the underlying data (12).

Fig. 2

Multivariate ordinations of diversity estimates for 14 major marine animal higher taxa ranging across 50 time intervals (Fig. 1). Similar results were obtained (14) by analyzing Sepkoski’s family- and genus-level compendia (21). Points are identified by the first two letters of group names. Groups are Trilobita, Linguliformea, Graptolithina, and Conodonta (Cambrian fauna, gray circles); Anthozoa, Ostracoda, nonlinguliform Brachiopoda (articulates, Ar), Cephalopoda, and Crinoidea (Paleozoic fauna, open squares); and Bryozoa, Bivalvia, Gastropoda, Echinoidea, and Chondrichthyes (Modern fauna, solid circles). (A) Factor analysis of raw counts with zero values included, as in (2). (B) Principal coordinates analysis of logged and differenced counts. Similar results are yielded by other methods, such as hierarchical cluster analysis (14).

Fig. 3

Sampling-standardized Phanerozoic diversity curve for the three marine evolutionary faunas. Values are summed curves for constituent groups that were generated independently (14). Unlabeled area represents groups not assigned to a fauna.

Because net diversification rates yield evidence about process instead of pattern, it is necessary to show whether Sepkoski’s scheme also summarizes such rates adequately. The best approach is to perform a principal coordinates analysis on a Euclidean distance matrix derived from the rates. Doing so allows ignoring undefined values for extinct groups while using the same similarity metric that underlies factor analysis (14). This ordination fails to separate the three faunas cleanly: Although some structure is evident, the Paleozoic groups are widely scattered (Fig. 2B).

Sepkoski himself went into great detail about exceptions to his model (3, 20), and later analyses of his data also called the evolutionary fauna scheme into question (23). Furthermore, a three-fauna pattern can be generated by factoring data produced with a simple simulation (6), and little correspondence exists between traditional faunal membership and ecological attributes such as mineralogy, motility, physiological buffering, trophic level, or habitat affinity (13, 24). Thus, the three-fauna pattern seems only to summarize coincidental peaks in diversity trajectories (20).

A far more important problem with Sepkoski’s model is that equilibrium diversity levels have changed once or even twice in most taxonomic groups (table S1). The overall diversity trend (Fig. 3) also suggests offset logistic growth (table S1) (9). These changes in equilibrium points are unpredictable in the sense that they come at very different geological times and are equally likely to involve increases or decreases, as shown by the lack of correlation between early Paleozoic and Cenozoic carrying capacities (Spearman’s ρ = 0.264; P = 0.435) (table S2).

Sepkoski’s model implies that average diversification rates are a good predictor of long-term success. If so, then dominant groups such as bivalves may be expected to recover quickly from the current mass extinction. However, there is no strong correlation of median net diversification rates and Plio-Pleistocene diversity for living groups (ρ = 0.437; P = 0.179) (table S2). The result changes if we add zero diversity values to represent the three extinct groups (ρ = 0.654; P = 0.011), but this observation merely demonstrates that groups with unfavorable diversification rates had already been sifted out (5) by the early Mesozoic.

The lack of a relationship between average rates and eventual success is a side effect of the Permo-Triassic mass extinction, which was the worst in the history of life on Earth (4, 9). The groups that might otherwise have been the most diverse today can be identified by starting with the end-Permian diversity estimates, ignoring changes going into the Early Triassic, and using the post-Paleozoic rates to project forward up to the Recent. In contrast, these adjusted diversity estimates are well correlated with median diversification rates (ρ = 0.664; P = 0.031). Moreover, there is a strong negative correlation between median overall rates and rates for this boundary (ρ = –0.655; P = 0.034). Examples of otherwise successful groups that suffered include anthozoans, bivalves, gastropods, crinoids, and “articulate” brachiopods. The first three eventually recovered, but the last two stagnated at low diversity levels. Effectively, then, the Permo-Triassic mass extinction reset the clock on diversification and overturned the balance of groups.

These factors together explain why individual diversity curves are so idiosyncratic (Fig. 1). For example, the key “Cambrian fauna” clade Trilobita rose immediately to dominance and then declined swiftly, unlike any other group (Fig. 1A). The curve for nonlinguliform brachiopods (articulates) includes two large, distinctive Paleozoic peaks (Fig. 1A) that relate to radiations in particular subclades (14). Anthozoa and Cephalopoda were both important throughout the Paleozoic and Mesozoic (Fig. 1B), but peaks in their curves do not match, and shelled cephalopods were the only large group to suffer a large and irreversible decline at the Cretaceous-Paleogene boundary.

Meanwhile, the two major existing classes—Bivalvia and Gastropoda—achieved their prominence in different ways. The former increased throughout the Phanerozoic (16), although the current data depict a shallow trend (Fig. 1B). Meanwhile, gastropods tracked a modest plateau throughout the Paleozoic and Mesozoic before suddenly rising to a new equilibrium level near the end of the Cretaceous (Fig. 1A). This clear offset reflects the adaptive radiation of the carnivorous clade Neogastropoda and corresponds to the Mesozoic marine revolution, which involved an “arms race” between gastropods and shell-crushing predators (25).

Global diversity should rebound from today’s extinction crisis within roughly the equivalent of a geological period (9). However, it is not possible to predict changes in taxonomic composition (table S1), which are more ecologically and evolutionarily important than total counts of taxa. The most severe extinctions also have unexpected impacts on the relative abundance of closely related groups (26), the shape of species-abundance distributions (11), and patterns of epifaunal and infaunal tiering (27). Thus, it would be unwise to assume that any large number of species can be lost today without forever altering the basic biological character of Earth’s oceans.

Supporting Online Material

www.sciencemag.org/cgi/content/full/329/5996/1191/DC1

Materials and Methods

Figs. S1 to S9

Tables S1 and S2

References

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. I thank J. Sepkoski and D. Raup for asking the questions I seek to answer. I am grateful to M. Foote, S. Holland, G. Hunt, A. Miller, T. Olszewski, and P. Wagner for their suggestions; M. Kosnik and A. Miller for reviews; and M. Foote for verifying that my subsampling algorithms were programmed correctly. Numerous contributors to the Paleobiology Database made this study possible, and I am particularly grateful to M. Clapham, A. Hendy, and W. Kiessling for recent contributions. Research described here was funded by donations from anonymous private individuals having no connection to it. This is Paleobiology Database publication 117.
View Abstract

Navigate This Article