Diversity partitioning during the Cambrian radiation

Significance

Many mechanisms of the Cambrian Explosion have been proposed, but rigorous quantitative analyses of biodiversity dynamics are scarce, although they may shed light on important factors. Using a comprehensive database and sampling standardization, we dissect global diversity patterns. The trajectories of within-community, between-community, and global diversity during the main phase of the Cambrian radiation revealed a low-competition model, which was probably governed by niche contraction and the increase of predation at local scales. At continental scales, the increase of beta diversity was controlled by the high rate of community turnover among adjacent continents. This finding supports the general importance of plate tectonics in large-scale diversifications.

Abstract

The fossil record offers unique insights into the environmental and geographic partitioning of biodiversity during global diversifications. We explored biodiversity patterns during the Cambrian radiation, the most dramatic radiation in Earth history. We assessed how the overall increase in global diversity was partitioned between within-community (alpha) and between-community (beta) components and how beta diversity was partitioned among environments and geographic regions. Changes in gamma diversity in the Cambrian were chiefly driven by changes in beta diversity. The combined trajectories of alpha and beta diversity during the initial diversification suggest low competition and high predation within communities. Beta diversity has similar trajectories both among environments and geographic regions, but turnover between adjacent paleocontinents was probably the main driver of diversification. Our study elucidates that global biodiversity during the Cambrian radiation was driven by niche contraction at local scales and vicariance at continental scales. The latter supports previous arguments for the importance of plate tectonics in the Cambrian radiation, namely the breakup of Pannotia.

Whittaker (1) decomposed regional (gamma) diversity into local (alpha) and turnover (beta) components. Although this concept is widely used and the principal levels of biodiversity are well explored in time and space, there are few analyses on the relationship of alpha and beta diversity in deep time during major evolutionary radiations. The potential of Whittaker’s concept to provide insights into evolutionary processes has been demonstrated in a few paleobiologic studies (e.g., refs. 2 and 3). The question as to how the Cambrian radiation was partitioned within and between communities is old (4), and not resolved. The Cambrian Explosion has been characterized as the rapid increase of both biodiversity and morphological disparity (5–7). The causes of this unique evolutionary radiation have been suggested to be abiotic (8), ecologic (9, 10), and genetic (5, 11) factors and their complex interplay (12, 13). Here, we derive potential triggers of the Cambrian radiation from the way diversity was partitioned geographically and environmentally during the main diversification phase.

We use sampling-standardized analyses of fossil occurrence data from the Paleobiology Database to derive accurate time series of alpha, beta, and gamma diversity from the Ediacaran into the earliest Ordovician. Gamma diversity is understood as global genus richness, whereas beta diversity is used to characterize turnover between individual assemblages as well as depositional environments or geographic areas (Materials and Methods). We test whether the Cambrian increase in gamma diversity was preferentially governed by beta or alpha diversity. We also separated environmental and geographic beta to identify the main driver of beta diversity. Environmental beta measures faunal dissimilarity between different habitats, whereas geographic beta (also termed geodisparity; ref. 14) measures faunal dissimilarity between different geographic regions. The relationship of alpha and beta diversity during the main diversification interval in the early Cambrian is assessed as a function of global diversity, which can elucidate underlying processes (15).

Results

Raw gamma diversity exhibits a strong increase in the first three Cambrian stages (informally referred to as early Cambrian in this work) (Fig. 1A). Gamma diversity dropped in Stage 4 and declined further through the rest of the Cambrian. The pattern is robust to sampling standardization (Fig. 1B) and insensitive to including or excluding the archaeocyath sponges, which are potentially oversplit (16). Alpha and beta diversity increased from the Fortunian to Stage 3, and fluctuated erratically through the following stages (Fig. 2). Our estimate of alpha (and indirectly beta) diversity is based on the number of genera in published fossil collections and, thus, may be affected by monographic biases. However, the same basic pattern is seen in diversity estimates of paleocommunities with abundance data (Fig. S1).

Fig. 1.

Global genus-level diversity of marine animals from the Ediacaran to the earliest Ordovician. (A) Raw counts of the number of genera (sampled-in-bin). Note log scale of y axis. (B) Sampling-standardized genus-level diversity (sampled-in-bin) based on shareholder quorum subsampling with 70% frequency coverage per stage. The brown line refers to all marine genera, and the blue line excludes archaeocyaths. Ma, million years ago. Ava, Avalon assemblage; Dru, Drumian; For, Fortunian; Guz, Guzhangian; Jia, Jiangshanian; Nam, Nama assemblage; Pai, Paibian; St2, Stage 2; St3, Stage 3; St4, Stage 4; St5, Stage 5; St10, Stage 10; Tre, Tremadocian (Ordovician) Whs, White Sea assemblage.

Fig. 2.

Alpha diversity and beta diversity from the Ediacaran to the earliest Ordovician based on unweighted by-list subsampling of 45 collections per stage. Error bars are SDs of 100 subsampling trials.

Beta diversity can be biased by the variation of geographic clustering among sites over time. Therefore, we test the correlation between the median paleogeographic distance of collections and beta diversity, as well as between the median distance of grid centroids and beta diversity. There is no significant correlation in both cases (median distance between collections/beta: ρ = 0.1, P = 0.727; median distance between grid centroids/beta: ρ = −0.24, P = 0.418), implying that geographic clustering does not cause a significant bias.

Mass extinctions may affect diversity at multiple levels. Several episodes of profound turnover and mass extinction have been noted at the Ediacaran-Cambrian boundary and throughout the Cambrian (17, 18). Our sampling-standardized analysis confirms high extinction rates throughout the Cambrian (Fig. 3). There is no significant correlation between extinction rates and either alpha or beta diversity (extinction rate/alpha: ρ = −0.35, P = 0.266; extinction rate/beta: ρ = 0.287, P = 0.366), indicating that the observed extinctions are unlikely to introduce a substantial bias on alpha-beta-gamma diversity patterns.

Fig. 3.

Sampling-standardized extinction and origination rates through the Ediacaran and Cambrian. Rates are per-capita rates of Foote (70) but not standardized by stage durations (71). Error bars are SDs of 100 subsampling trials.

We find a strong correlation between global genus richness and beta diversity (ρ = 0.93, P < 0.001), whereas there is no significant correlation between global genus richness and alpha diversity (ρ = 0.42, P = 0.137). The same relationship is evident with a moving-window approach of five successive stages (Fig. S2). Here the beta-gamma link is significant over the interval from Stage 2 to the Drumian. The strong correlation between beta diversity and gamma diversity could be biased by the multiplicative approach we are using to derive beta diversity from both alpha and gamma diversity, which are independently assessed (Materials and Methods). Although this method probably has a better biological underpinning than the additive approach (15), a correlation between gamma and beta diversity is still evident when using an additive method (ρ = 0.84, P < 0.001). Therefore, gamma diversity in the Cambrian was largely governed by differentiation among communities or assemblages rather than by genus packing within assemblages.

An alpha-beta-gamma plot during the time of unconstrained diversification, that is, in the first three Cambrian stages, reveals that alpha diversity initially increased faster than beta diversity, but subsequently, beta diversity increased more rapidly (Fig. 4). This pattern is predicted for a low-competition scenario (15). The pattern of beta diversity is similar for environmental and geographic beta, which both increased during the first three stages and then stabilized (Fig. S3). This finding suggests that taxonomic gradients among habitats and geographic regions were equally important. To disentangle geographic turnover, we assess geodisparity in three spatial intervals of paleogeographic distance among 5 × 5° paleogeographic grids. The result shows (Fig. 5) that turnover in assemblage composition increased gradually with paleogeographic distance in the intervals of less than 2,000 km and greater than 4,000 km, likely reflecting normal Cambrian distance-decay patterns (19). However, turnover in the 2,000- to 4,000-km interval increased dramatically with geographic distance (Table S1), and this spatial interval is the distance typically measured between adjacent paleo-continents in the early Cambrian (Fig. S4). Plate-tectonic configuration in the early Cambrian is poorly constrained (20), but the Neoproterozoic formation of Gondwana is well established (21, 22). Beta diversity within Gondwana shows more muted fluctuations than global beta diversity. Subtracting the global beta from Gondwana beta reveals a major increase in Stage 3 (Fig. S5). Because this increase is concurrent with the main increase of gamma diversity, geodisparity appears to be the main driver of the Cambrian diversity peak.

Fig. 4.

Alpha-beta-gamma plot for the first three Cambrian stages, the time of continuous diversification. Note log scale of all axes.

Fig. 5.

Distance-decay curves during the first three Cambrian stages plotted as taxonomic dissimilarity among 5 × 5° paleogeographic grids. Blue circles, Stage 3; green circles, Stage 2; red circles, Fortunian. The black lines denote separate regressions for three distance intervals.

Finally, although we binned all fossil collections into the currently recognized Ediacaran assemblages and Cambrian stages (Materials and Methods), the Cambrian timescale remains in flux (e.g., ref. 23). This problem raises the question whether our results might change with further revisions of the timescale. We therefore tested how the changes in international correlations and calibrations over the last 20 y have affected the basic results. Binning our data to the traditional Siberian subdivision of the Early Cambrian and lumping all Ediacaran assemblages, as well as Middle and Late Cambrian stages (24), shows some differences such as an extended peak in gamma diversity in the Atdabanian and Botomian, but the basic patterns are robust (Fig. S6): Gamma diversity peaked in the late Early Cambrian, alpha diversity exhibits a profound increase until the Atdabanian, and beta diversity lagged behind alpha diversity in the earliest Cambrian and then caught up.

Discussion

Although a surprising diversity of benthic organisms is now known from the Ediacaran (7), diversity was substantially lower at the gamma and alpha level than in the Cambrian (Figs. 1 and 2). The high taxonomic turnover recognized at the Ediacaran-Cambrian transition (Fig. 3) that marks the disappearance of Ediacaran soft-bodied organisms may be tempered by secular variation in global taphonomic regimes (25), but can also be attributed to ecosystem engineering and predation (26). Environmental perturbations were invoked to explain the extinction of the few skeletal organisms (27).

That the main pulse of the Cambrian radiation was in the early Cambrian has long been known (28, 29), but defining the time of peak diversity has been hampered by problems with stratigraphic correlations. Our improved data and updated stratigraphic assignments demonstrate a pronounced origination and diversification pulse in the earliest Cambrian (Fig. 3). This pulse may be exaggerated by the closure of the Ediacaran taphonomic window with widespread soft-body preservation and the opening of a novel taphonomic window by the widespread acquisition of skeletons (30). Cambrian diversity rose to a prominent peak in Stage 3, which is roughly equivalent to the Siberian Atdabanian and most of the Siberian Botomian (31). The subsequent decline in global diversity may have been partly due to the Sinsk extinction event in Stage 4 (32–34), but environmental factors are more plausible to explain the lack of further diversification in the later Cambrian. For example, prolonged tropical warming has been suggested to explain the scarcity of metazoan reefs in the later Cambrian (35) and would also fit the gamma diversity trajectories given that cooling is thought to have facilitated the subsequent Ordovician radiation (36). Alternatively, a late Cambrian rise in oxygen levels has been proposed to explain the onset of the Ordovician radiation (37). By inference, oxygen limitation may have hindered further diversification after the early Cambrian.

The relationship between alpha and beta diversity during the time of unbounded diversification (Fig. 4) suggests that a low-competition model (15) best explains the Cambrian radiation. When competition is low, the addition of new species to a community happens either by exploitation of previously unused resources or by packing more species in marginal ecospace (38). This process will initially not shrink niches of preexisting taxa such that alpha diversity will increase steeply, whereas beta diversity will initially increase only moderately. As diversification continues, competition will increase, niches will contract, and consequently beta diversity will increase profoundly, whereas alpha diversity levels off. Niche contraction facilitates niche partitioning and, thus, changes of taxonomic composition among habitats. Predation is a potential factor for keeping the system in a low-competition mode (39). Predation is first recognized in the latest Ediacaran (40, 41) and well-documented in the early Cambrian (42, 43). Apex predators are first common in Cambrian Stage 3 (44, 45). Therefore, escalation is a plausible trigger of the Cambrian radiation in terms of both biomineralization (46, 47) and community ecology (48). Predation universally defeats competition in benthic marine systems (49), suggesting that the Cambrian radiation was not exceptional in this aspect.

We show that the increase in beta diversity was the principal driver of increasing gamma diversity during the Cambrian radiation. The pivotal role of beta diversity in structuring global diversity patterns has also been suggested for early Cambrian reef communities (50). Beta diversity strongly depends on both the sizes of sampling area and sampling units (51). In our case, local community differentiation was driven largely by niche contraction. However, at regional or continental scales, beta diversity or community differentiation between regions does not necessarily increase when niches contract, because environmental gradients can be strongly interrupted by dispersal limitation and topographic isolation. Therefore, there must be other factors that drive beta diversity at continental scales in the Cambrian. The steep distance decay in the 2,000- to 4,000-km distance interval (connecting adjacent continents or distant parts of Gondwana; Fig. S4) suggests that the continents became biogeographically distinct in the first two Cambrian stages, and that deep oceans between continental shelf areas caused effective migration barriers, which likely enhanced allopatric speciation (52).

In summary, the drivers of beta diversity varied with spatial scale: (i) at local scales, an increase of beta diversity was due to niche contraction, which, in turn, may have been fueled by predation; and (ii) at continental scales, beta diversity was governed by the strong increase of provincialism among paleo-continents. This increase was accompanied by profound continental reconfigurations often referred to as the breakup of the supercontinent Pannotia (20, 53). Pannotia assembled in the interval from 650 to 550 Ma (54). Its breakup was characterized by the opening of the Iapetus and Ægir oceans (55), resulting in the separation of Laurentia, Baltica, Siberia, and Gondwana, but also orogenies leading to the amalgamation of distinct cratonic blocks within Gondwana. The disassembly of Pannotia (particularly the separation between Laurentia and Gondwana) is suggested to have started close to the onset of the Cambrian radiation (Fortunian; ∼541 Ma) and ended before Cambrian Stage 3 (∼521 Ma) (20). This rapid disassembly of Pannotia corresponds well with the increase of geodisparity (Fig. 5).

Although the breakup of Pannotia is sometimes discussed as a potential trigger of the Cambrian radiation (54), the concept is not widely used. However, our results support the contention that the disassembly of Pannotia increased geodisparity and, thereby, beta and gamma diversity in the Cambrian. Disassembly of supercontinents is rare and always seems to have profound evolutionary consequences. Examples are (i) the breakup of the supercontinent Rodinia some 750 Myr ago has been linked with global glaciations, which are often held responsible for the emergence of metazoans (56); (ii) further continental dispersal in the Ordovician is thought to have facilitated the Ordovician diversification (57); and (iii), the link between the breakup of Pangea in the Early Jurassic and the Mesozoic-Cenozoic radiation has long been suggested (58) and is supported by sampling-standardized diversity curves (59).

Alternatively, the capacity of biological dispersal could have changed substantially in the early Cambrian. Our knowledge of dispersal in Ediacaran biota is limited but at least some Ediacaran taxa were able to cross oceanic basins and achieve a cosmopolitan distribution (26, 60). Cambrian animals probably dispersed with nonplanktotrophic larval stages (61), which would limit their geographic distribution relative to animals with planktotrophic larvae (62). Therefore, the increase of geodisparity we observe during the early Cambrian could be due to the widespread appearance of animals with larval stages combined with continental disassembly. In other words, biological innovation—the evolution of larval stages (61)—in tandem with supercontinent breakup best explains the increase of geographic beta diversity. Our study provides evidence for niche partitioning, plate tectonics, and key innovations as strong and persistent evolutionary forces, and the Cambrian radiation is no exception, although it was more substantial in quality than later radiations. Future work should explore the links between the ecological dynamics described here and the astonishing increase of morphological disparity in the early Cambrian.

Materials and Methods

Data.

Fossil occurrence data of Ediacaran to Tremadocian age were entered into the Paleobiology Database (PaleobioDB; paleobiodb.org) and downloaded alongside all other occurrences on May 23, 2014. This dataset comprises 7,117 collections and 39,737 taxonomic occurrences. We assigned each collection to one of 14 stages based on the three widely recognized assemblages in the Ediacaran (63), the 10 stage subdivision of the Cambrian period according to the 2012 International Commission on Stratigraphy time scale (31), and the Tremadocian of the Ordovician (SI Materials and Methods, Fig. S7, and Table S2).

Plate tectonic configurations and paleopositions of fossil occurrences are based on Scotese’s paleomap software (64). Paleopositions were automatically provided upon download of the occurrences from the PaleobioDB. The concept of Pannotia is implemented in Scotese’s reconstructions, and although newer reconstructions differ in some aspects such as the degree of continental dispersal in the earliest Cambrian, they agree in that the Early Cambrian was a time of continental disassembly (SI Materials and Methods).

Diversity Estimation.

To account for sampling heterogeneity (65), diversity has been estimated with sampling standardization. Sampling standardization can be achieved with random draws of the same number of fossil collections, taxonomic occurrences, or a fixed sum of frequency of genus abundance in each time interval (59, 66). We generally assessed diversity by counting taxa actually sampled in an interval (sampled-in-bin, SIB) and, thus, did not interpolate taxon occurrences between their first and last appearance in the fossil record. This counting method has the advantage that edge effects at the beginning and end of time series are avoided (59). We used shareholder quorum subsampling (66) with a quorum of 70% to estimate gamma diversity. Alpha and beta diversity as well as extinction and origination rates were assessed by drawing 45 taxonomic collections at random (by-list unweighted method; UW) (67) and averaging results across 100 subsampling trials.

Alpha diversity was estimated as the number of genera found in a collection, where a collection can vary from a small sample taken from a single bed to a whole formation at a regional scale. To test whether this approach biases our estimate of alpha, we have also measured alpha diversity in the subset of collections for which specimen abundances are provided. We used the Shannon–Wiener Index (68) for those collections that reported at least 80 specimens.

We use two commonly used measures of beta diversity: (i): $\beta =\gamma /\overline{\alpha }$, where $\gamma$ is total diversity at global scales, and $\overline{\alpha }$ is the mean genus richness within assemblages. This metric is a global measure of beta diversity based on presence-absence data (1, 38); and (ii), a pairwise dissimilarity measure based on the Bray–Curtis index (69), which assesses the pairwise dissimilarity among sampling units with respect to environmental or geographic variables.

Environmental beta was measured by pooling collections in the same basic environmental setting and comparing the faunas between environments that differed by one, two, or three environmental categories. Environmental beta is thus estimated by the mean dissimilarity in the context of habitat disparity. Environmental categories were tropical/nontropical (separated at 30° absolute paleolatitude), carbonate/siliciclastic (distinguished by dominant lithology, marls were ignored), and shallow-water/deep-water (separated by storm wave base).

Geographic beta was estimated in the context of paleogeographic distance. To create a paleogeographic distance matrix, we applied the geodisparity method introduced by Miller et al. (14). Occurrence data for a given stratigraphic interval were pooled in 5 × 5° paleogeographic grids. A dissimilarity matrix of these pooled data was computed, and the data were then parsed into 2,000-km intervals of great circle distances between grid centers. We focused on three geographic scales: within a continent (0–2,000 km), between adjacent continents or within a larger continent (2,000-4,000 km), and between remote continents (greater than 4,000 km). The turnover, or the rate of change in community composition, is represented by the slope of the relationship between dissimilarity and geographic distance. This distance–decay relationship was estimated by linear regression (Table S1).

Statistical Tests.

All statistical tests are nonparametric applying Spearman’s rho (ρ). Time-series data were tested for autocorrelations before performing correlation tests. Because there are no significant autocorrelations in any of the relevant variables (alpha, beta, gamma, extinction rate), correlations are based on nondifferenced data.