Stability, incumbency and ecological reorganization after the Permian-Triassic mass extinction

Abstract

The Permian-Triassic mass extinction (PTME) altered macroevolutionary land-scapes by removing incumbent biota. Here, using terrestrial paleocommunities of the Karoo Basin spanning the PTME, we show that a pre-extinction incumbent configuration of biotic interactions made significant ecological re-organizations or macroevolutionary innovations unlikely. The post-PTME ecosystem initially was more likely to be reorganized, but incumbency was re-established by the Middle Triassic. We argue that the stability of the pre-PTME ecosystem, its subsequent loss, and replacement, resulted from the in-fluence of community-level structure and dynamics on species evolution and survival.

One sentence summary Biotic incumbency at the end of the Paleozoic, based on community functional organization, was destroyed by the Permian-Triassic mass extinction, allowing the development of novel community types.

Introduction

The rise and fall of biotic interactions and incumbent, geologically persistent taxa, are key features of long-term macroevolutionary and macroecological patterns in the fossil record (1). Although the replacement of incumbents and interactions have been explained in various ways (1–6), mass extinctions played particularly important roles. During these transformative events, communities underwent significant compositional and ecological reorganizations, or were re-placed completely by ecologically novel systems, giving mass extinctions an effect on history disproportionate to their contributions to the total number of Phanerozoic extinctions (7). Here we show, using a well-documented series of terrestrial paleocommunities spanning the Permian-Triassic mass extinction (PTME), and a numerical model of community stability (8), that incumbency before the PTME was maintained by a pre-existing community-level structure of biotic interactions. When compared to alternative evolutionary trajectories and ecological reorganizations, this structure was more advantageous to long-term species persistence and co-existence. The structure’s loss through successive waves of extinction resulted in an Early Triassic community in which reorganization would have significantly improved persistence and coexistence. By the Middle Triassic, this recovery stage was replaced by a community in which innovation and reorganization would have been disadvantageous, thereby forming the basis for renewed incumbency.

We studied seven paleocommunities from the Karoo Basin, South Africa, ranging from the late Permian (Wuchiapingian) lower Daptocephalus Assemblage Zone, to the Middle Triassic (Anisian) Cynognathus Assemblage subzone B (31, 35, 36) (Fig. 1; table S1) (12). Taxon composition between successive communities was nearly constant during the Permian, whereas turnover was dramatically greater in the Triassic. We modeled each paleocommunity as an ensemble of food webs of size S (number of species), with a structural complexity determined by the hierarchical and functional partitioning of S into G trophic guilds, and E sets of inter-guild interactions. The ensemble was organized as four nested sets of hypothetical communities of increasingly constrained structural complexity (33). The most inclusive set comprised random networks of size S, constrained by having the total number of interactions drawn from a mixed exponential power law distribution (12); the least inclusive consisted of the observed paleocommunity itself. Nested between these end members were: model communities with S partitioned into G guilds with E links, yielding food webs of structural complexity equal to, but compositionally different from the observed paleocommunity; and models containing the same functional structure of the paleocommunity, but with randomized partitioning of S among guilds. We thus examined the effects of imposing structural features on the ensemble that increasingly required random food webs of size S to be more consistent with observed paleocommunity structure.

Table S1.

Guilds and guild richnesses of all paleocommunities. S is total richness.

Table S2.

Coefficients for each model and community, in Equation 6.

Fig. 1.

Model global stabilities and paleontological context. Upper left panel summarizes the stratigraphic position of each paleocommunity, its taxon richness (S) and taxon continuity (C). Gray bars in S represent the extinction phases. C is calculated as the fraction of vertebrate genera that have persisted from the previous community. (A-U) For each paleocommunity, ΔSe between the observed paleocommunity and (left) random, (middle) guild altered, and (right) richness altered manipulations. ΔSe > 0 represented as cooler colors. ΔSe = 0 uncolored. Paleocommunity acronyms: CAZ-Cynognathus Assemblage Zone (AZ), LAZ - Lystrosaurus AZ, Ph3 - Extinction Phase 3, Ph2 - Extinction Phase 2, Ph1 - Extinction Phase 1, L DAZ - lower Daptocephalus AZ, U DAZ - upper Daptocephalus AZ.

The structurally complex models represent hypothetical alternative evolutionary pathways, ranging from minimally divergent (species richnesses within guilds vary from observed because of differential rates of origination/extinction), to more divergent histories where ecological innovation, origination and extinction could result in alternative community types. If all histories were possible, incumbency would emerge if a particular type of community consistently supported a greater number of stably coexisting species, and thus greater species persistence relative to other communities. We estimated levels of coexistence, Se, associated with real and hypothetical communities using a tractable model of interspecific interactions (8)(12).

Population sizes, X, are functions of intrinsic population growth rate, r, both positive and negative interspecific trophic interactions (β), and normalized carrying capacity. X^* is population size modified by interspecific interactions as c comprise scaling factors ranging between zero and one, that modify the average interaction strengths of the community. Populations become extinct because larger values of r increase the variability of population trajectories (Fig. S1, S2), and more negative interaction strengths decrease species feasibilities, but the system always settles dynamically to a stable state, yielding Se. The system thus undergoes a May-Wigner transition (8, 14), with Se declining nonlinearly with increasing S, number of interspecific interactions, or (12) (Fig. S3).

Fig. 1.

Single species dynamics in the model outlined in Equation 3. A - population trajectories when r equals 1.5 (red), 2.2 (blue) and 2.6 (purple). B - bifurcation diagram.

Fig. 2.

Population trajectories and phase plots (attractors) of taxa from the lower Daptocephalus Assemblage Zone (lDAZ), illustrating three types of dynamics. A - stable oscillation or limit cycle of a freshwater bivalve. B - quasi-periodic oscillation of a freshwater fish. C - chaotic dynamics of an omnivorous insect. Color bar illustrates time step (t). Upper plots show a subset of 10,000 time steps. Lines projected at base of attractor plots show the trajectories for the subsets.

Fig. 3.

S declines as the model is iterated, asymptotically yielding Se. Se declines with increasing r and average negative interaction strength, as determined by the scaling factor c1. Both plots are of the lDAZ community, simulated at three different values of r and c1, with 10 food web simulations at each value.A - r equals 2 (black), 2.5 (blue) and 3 (red). c1=0.1. B - c1 equals 0.1 (black), 0.5 (blue), 1.0 (red). r=2.

We explored the dependence of Se on r and for each random, alternative or observed model of a paleocommunity by simulating Equation 1 at values of r ranging from 2-4. Average interaction strengths were drawn randomly from a uniform distribution ranging from 0-1 and were scaled by factors ranging from 0.1-1 (c in Equation 2). Thirty species level food webs were simulated for each model at each parameter set (r, −c₁β, c₂β), for a total of 27,000 food webs per model per community. Differences between observed and hypothetical models (ΔSe) were tested with parameter dependent t-tests (12), and visualized as regions of parameter space where the global stability of the models differ significantly (Fig.1A-U).

where is a regression estimate of Se (Fig. S4). One community model typically did not outperform another at every parameter set (although see Fig. 1D, 1F, 1J and 1M). Specific values of intrinsic population growth rates and biotic interactions determine the circumstances under which a model may be expected to support a significantly greater number of species compared to another. By contrasting observed paleocommunity structures to random communities and alternative model communities, several important conclusions may be drawn regarding the basis for a pre-PTME pattern of incumbency, the loss of incumbency, and the ecological transformation wrought by the PTME.

Fig. 4.

Visualization of Se and ΔSe throughout the r,c1,c2 parameter space, for the lDAZ community. Color intensity shows the value of Se in A and B. A - observed lDAZ community. B - random food web with S and the total number of predator-prey interactions equal to lDAZ’s S (S=138). C - ΔSe of the observed model minus the random model. Values of ΔSe=0 are not colored. Note that ΔSe ≥ 0 everywhere, except for a small region where r ≥ 3.5 and 0.2 ≤c1≤0.3.

First, the decline of Se with increasing population growth rates (r) or negative interaction strengths (−β) depends critically on the higher level, functional organization of the community. Compartmentalization into guilds slows the decline relative to random, uncompartmentalized food webs of equal S, regardless of whether compartments are observed or hypothetical. The parameter range under which compartmentalized (guild-structured) communities would support significantly more species than random communities (ΔSe > 0) depends on S and the structure of the community. During the PTME, when S was lowest, compartmentalization would have permitted greater species coexistence under all circumstances of r and β, relative to random communities (Fig. 1J and 1M). Although this is generally true of pre-and post-PTME paleo-communities, there are cases, when r and −β are both high, or −β has been scaled by a factor of approximately 0.2, that ΔSe ≤ 0 (Fig. 1G and 1P). Those cases represent conditions under which paleocommunity structure would have been more ecologically and evolutionarily labile; structure could be altered by changes to species properties or species composition, without a negative impact on stability.

Second, during the late Permian, observed paleocommunities would have been more globally stable than alternatively structured communities (ΔSe > 0), along a gradient of decreasing r and −β (Fig. 1K,1N,1Q,1T). Observed paleocommunities prior to the PTME could be less stable at greater values of r and −β (ΔSe < 0) (Fig.1Q,1T), whereas during the extinction, ΔSe > 0 (Fig. 1K,1N). These contrast with models that preserve guild structure, but randomize the partitioning of S among guilds (Fig. 1L, 1O, 1R, 1U). Those hypothetical communities would possess the same interacting lineages as the paleocommunities, but would be distinguished by having different diversification and extinction rates. In those cases, the hypothetical models are generally more stable than the paleocommunities themselves within much of the parameter range. Thus, the pre-existing Permian community structure was maintained by its pattern of biotically interacting lineages, but not specifically the numbers of species involved in those interactions.

Third, no patterns of incumbency could have been established in the Early Triassic. The Early Triassic third phase of the PTME comprised survivors from the Permian Karoo Basin ecosystem, and likely immigrants from neighboring regions (15). Structural re-organizations of this assemblage could have generated more stable and persistent alternative communities (Fig. 1E), at parameter values encompassing more realistic ecologies when compared to pre-PTME communities (Fig. 1Q, 1T). Moreover, ΔSe > 0 throughout the parameter space when compared to alternative partitionings of S (Fig. 1F), showing that although the community comprised species capable of surviving the PTME, options for subsequent within-guild variation of taxon richness were very narrow. Changes via origination and extinction within existing lineages would have yielded less stable communities, producing a macroevolutionary dead-end. The succeeding Lystrosaurus Assemblage Zone (LAZ) has been noted for its unusually rich amphibian fauna (30,38), numerical dominance of the herbivorous therapsid Lystrosaurus (38), accelerated ontogenetic development of some tetrapods (18, 19), and unusual ecological dynamics (21, 32). Our results show that both variation of guild richnesses (Fig. 1I), and reorganizations of guild structure (Fig. 1H) would have been overwhelmingly likely to result in more persistent species and stable communities than those observed. The LAZ community itself would have persisted unchanged, with lineages therefore apparently incumbent, only if negative species interactions were exceedingly weak (Fig. 1H). This contrasts with the later, Middle Triassic Cynognathus Assemblage Zone (CAZ), wherein any changes to guild richnesses or paleocommunity structure were unlikely to yield greater persistence or stability (Fig. 1B, 1C). The CAZ community signaled a return to incumbency in the Karoo ecosystem.

Paleocommunity compositional and structural constancy prior to the PTME resulted from greater global stability compared to communities that could have arisen via macroevolutionary variation. This incumbency was generated by the persistence of interacting lineages, even during the PTME (Fig. 1), and the communities could have accommodated changing taxon diversities while maintaining ecological functions. Successful ecological reorganization was unlikely. The extinction and replacement of most lineages led to an Early Triassic community (LAZ) that would have been more ecologically and evolutionarily variable, and labile. We posit that patterns of ecological reorganization and evolutionary innovation in the wake of mass extinctions are therefore accounted for, at least in part, by the destruction of highly stable, preexisting community structures, and the subsequent time required for the development of new systems of stably coexisting, biotically interacting lineages. The rapidity with which taxon richness increased in the wake of the PTME further suggests that recovery does not require extended intervals of time for the evolution of new biotic interactions. Instead, time is required for the evolution of persistent assemblages of interactors and interactions.