Niche conservatism and ecological change during the Late Devonian mass extinction

Abstract

Studies of the fossil record can inform our understanding of not only the causes of mass extinctions, but also their effects on biodiversity, ecology and evolution. Here, we examine regional-scale ecological changes resulting from a Late Devonian mass extinction event using brachiopod fossil assemblages from the Appalachian Basin. About half of the species went extinct, but were largely replaced by new immigrant taxa. Both before and after the extinction, the primary gradient in faunal composition was correlated with onshore–offshore position, with a second gradient attributed to frequency of disturbance. Survivors of the extinction displayed a strong degree of niche conservatism along these gradients. Despite these indicators of ecological stability, the pre- and post-extinction faunas were quite distinct at the order level, with atrypids and strophomenids largely replaced by productids, whose spiny shells may have provided greater resistance to disturbance and/or predation. Thus, extinction survivors persisted in similar ecological niches despite environmental perturbations and considerable change in the taxonomic and ecological composition of the regional species pool.

1. Introduction

As anthropogenic environmental changes threaten an increasing number of species, ancient mass extinction events may help illuminate both the causes of mass extinctions and their effects on biological systems [1]. One potential effect of mass extinction and associated environmental change is niche evolution—species evolving to occupy different ecological niches in response to changes in biotic interactions and/or environmental parameters (e.g. [2–5] and references therein). Niche evolution might represent one strategy for surviving the stresses imposed by environmental perturbation and ecological change during an extinction event, just as some modern species are evolving in response to human-induced mortality (e.g. [6–8]). Several palaeontological studies have tested for niche evolution versus niche conservatism in the marine fossil record, considering times of slow, background environmental change as well as regional extinction and immigration events (e.g. [3,5]). These studies suggest that niche conservatism is a common pattern, although niche evolution can be associated with physical or ecological changes, or it can occur slowly over time [9].

The degree to which mass extinctions drive niche evolution is an open question. It is clear that the largest mass extinctions, like the end-Permian, can cause large ecological changes as ecosystems re-evolve complexity after suffering huge losses in diversity [10–12]. Here, we test for niche evolution and ecological change during the Late Devonian (Frasnian–Famennian) mass extinction, traditionally counted as one of the ‘Big 5’ mass extinctions of the Phanerozoic. [13]. Although its extinction magnitude was lower than the events at the ends of the Permian and Cretaceous periods, the Late Devonian event triggered extensive ecological changes in the marine realm [14–17], but it remains to be tested whether it destabilized the niche structure of regional biotas. Although modern extinctions have yet to rise to the level of ancient mass extinctions, the biotic responses to these events may provide helpful context for thinking about future species losses.

During the Late Devonian extinction, an estimated 35% of marine genera went extinct [18], including stromatoporoids, rugose and tabulate corals, ammonoids, placoderms and brachiopods [19–21]. Some authors have argued that diversity losses were further exaggerated by low origination rates [1,22]. The extinction consisted of two pulses, the Lower Kellwasser event and Upper Kellwasser event, which coincided with global cooling, ocean anoxia and positive carbon isotope excursion [23–29].

We focus on the brachiopod fauna of the Appalachian Basin. Brachiopods were common benthic marine suspension feeders at the time, and their calcitic shell had high preservation potential. The two pulses of extinction are marked by dark shales and positive carbon isotope excursions [28,29]. We focus specifically on the first extinction pulse (the Lower Kellwasser event), during which about half of brachiopod species perished [29]. Based on extensive bulk sampling, we examine (i) whether environmental gradients structure the fauna in the same ways before and after the extinction event, (ii) whether surviving species exhibit niche conservatism or niche evolution, and (iii) changes in the higher taxonomic structure of the fauna, as a measure of general ecological change. Ultimately, we aim to understand whether a major environmental and ecological disruption permanently altered the ways in which environmental and ecological conditions controlled species distributions and abundance.

(a) . Geological setting

Upper Devonian strata outcrop along a shallow-to-deep environmental transect in northern Pennsylvania and western New York (figure 1a), with terrestrial deposits to the southeast transitioning to shallow marine sands and then offshore shales to the west [30,32–34]. The first extinction pulse is marked by the Pipe Creek Formation, a regionally extensive, poorly fossiliferous, organic-rich shale to silty-shale that was deposited in a dysoxic to anoxic outer-shelf marine environment.

Figure 1.

(a) Map of field area in western New York and northern Pennsylvania. The blue line denotes the Pipe Creek Formation (first extinction pulse) and the red line marks the base of the Dunkirk Formation (slightly above the second extinction pulse). Measured sections at Dansville, NY (DAN), Big Creek, NY (BCP), Cameron, NY (CAM) and Tioga, PA (TGB) are marked by yellow circles. Modified from [30] and [31]. (b) Generalized stratigraphic column through the Frasnian–Famennian transition in shallow marine environments in New York and north-central Pennsylvania. The two extinction intervals are highlighted in red. Modified from [30]. (c) Field photo of the second extinction pulse at the Mills Mills locality, Wiscoy Creek, New York. (d) Field photo of the Pipe Creek Formation (first extinction pulse) and adjacent formations exposed at the Tioga B locality (TGB), Tioga, Pennsylvania.

Underlying the Pipe Creek (preceding the extinction pulse) is the Wiscoy Formation, composed of fossiliferous sandstones and mudstones, which fines upwards towards the Pipe Creek [33]. This formation is divided into the lower informal sandy member and the upper muddy member (figure 1b–d) [30]. The sandy member consists of fine-grained sandstones with muddy interbeds. Hummocks and abundant swales with mud-drapes indicate significant storm reworking and suggest a middle to lower shoreface environment above or near storm wave base. The overlying muddy member is characterized by mudstones with interbeds of silt to very fine sand and has been interpreted as representing an inner shelf environment [30,33]. Each member is characterized by a shallow-to-deep gradient from east to west, as indicated by sedimentological characteristics [30,33].

Immediately above the Pipe Creek Formation (following the extinction pulse) is the Canaseraga Formation. Two members have been defined [30,33]. For this study, we focus on the interval between the Hammond Member and the second extinction pulse (Upper Kellwasser event), which is often highly fossiliferous and spans a period of relative faunal stability between the extinction pulses. We separate this interval into informal muddy and sandy members. Together, these two members represent the same general range in lithology and environment as the two members of the Wiscoy Formation [33], although there are some differences; most notably, swaley bedding is less common in the sandy member of the Canaseraga than in the sandy member of the Wiscoy, indicating less storm reworking [30,31]. As in the members of the Wiscoy Formation, each of these members displays a shallow-to-deep environmental gradient from east to west. The sandy member of the Canaseraga is capped by the Point Gratiot Bed, a thin, laterally extensive dark shale bed marking the second pulse of extinction (the Upper Kellwasser event) (figure 1b,e,f) [30,35].

2. Methods

(a) . Data collection

Bulk fossil assemblages were sampled from the Wiscoy and Canaseraga formations from 34 localities between Tioga, Pennsylvania, and Dansville, New York (figure 1a), representing shallow marine environments above and below storm wave base [33,34,36]. Longer stratigraphic sections were measured and sampled at Tioga, PA (TGB), Cameron, NY (CAM), Big Creek, NY (BCP) and Dansville, NY (DAN) (figure 1a). Rhynchonelliform brachiopods were selected for this study as they are abundant taxa in these strata, representing 100% of the identifiable fauna in many samples. Epibionts and endobionts were abundant in some samples, but their distribution has been analysed as part of a separate project [37]. Gastropods, bivalves, corals, bryozoans and crinoid columnals were occasionally found, but in quantities too small to be meaningful in this analysis. Chonetid brachiopods were excluded; there were probably only two species, which were small and often too poorly preserved to identify [29]. Approximately 14 000 brachiopods (approx. 5800 Wiscoy and approx. 7100 Canaseraga) belonging to 39 species were identified and counted based on multiple sources illustrating the Devonian brachiopod fauna ([33] and references cited therein). All samples included in the analysis had at least 30 identifiable individuals; in some cases, small samples were combined with adjacent beds of similar facies to meet the sample size threshold. A total of 183 samples were included, 87 from the Wiscoy [29] and 96 from the Canaseraga. Many additional samples and fossils were examined but not counted to help verify stratigraphic ranges.

(b) . Ordination

We analysed the Wiscoy and Canaseraga datasets to examine faunal gradients and species' environmental preferences immediately before and after the first pulse of extinction; the two formations were analysed separately owing to the differences in species composition. To control for variability in sample size due to sedimentological and outcrop characteristics, brachiopod abundances were converted to proportions prior to analysis. Differences between samples were characterized using Bray–Curtis dissimilarity. Samples were ordinated using non-metric multi-dimensional scaling (nMDS) to quantify variation in species composition between samples [3,38–40] using the vegan package through R Statistical Software version 2.5-6 [41]. nMDS uses species’ rank order abundances to place samples in multi-dimensional space such that samples of similar composition plot closer together and samples with dissimilar composition plot further apart, displaying gradients in faunal composition. These often correspond to environmental gradients that control species abundances. nMDS was performed in three dimensions with a maximum of 999 iterations and a maximum of 100 random starts. Analyses with a stress value less than 0.15 were seen as a good fit for the data.

The nMDS ordinations were used to ask two questions: (i) do the same environmental parameters control faunal gradients before and after the extinction?, and (ii) do survivor species exhibit niche conservatism, maintaining the same preferred environment and environmental tolerance through the extinction and recovery? To quantify preferred environments, we calculated average nMDS scores for each species along both nMDS axis 1 and nMDS axis 2 and compared the scores for surviving species using linear regression. Schizophoria species were grouped together for the analysis as they do appear in both formations but can sometimes be difficult to distinguish owing to preservation. Species with low overall abundance in one or both formations were excluded from the comparison. To quantify niche breadth, we calculated weighted standard deviations of nMDS scores for each surviving species. To make the Wiscoy and Canaseraga analyses comparable, the Wiscoy nMDS scores were first translated onto the Canaseraga nMDS scale using the equation obtained from the regression of Wiscoy nMDS surviving species scores on Canaseraga nMDS surviving species scores. However, comparisons of niche breadths are challenging, and the results should be treated with considerable caution [3]. We also performed a combined analysis to detect systematic changes in niche breadth (that is, whether there was an overall increase or decrease in niche breadth across all species). For this analysis, we centred each species' distribution on zero and calculated a weighted standard deviation on the combined dataset. Niche breadth before and after the extinction was compared using Levene's test for equality of variances, which is less dependent on normal distribution of the data [42]. We also compared the peak abundance values of species before and after the extinction, with peak abundance construed as the maximum proportional abundance of a species (results were the same when peak abundance was calculated as the mean of the highest three values).

3. Results

(a) . Pre-extinction ecological structure

The first nMDS axis is correlated with the onshore–offshore environmental gradient, with samples from the muddier, more offshore muddy member generally plotting more negatively on nMDS axis 1 than samples from the more onshore sandy member (figure 2a). Similarly, within each member, samples from the most distal, northwestern localities (BCP and DAN) plot to the left of samples from the more proximal, southeastern localities (TGB and CAM) [29]. Strophomenids and atrypids were present across the depth gradient, along with members of other orders (figure 2c).

Figure 2.

nMDS plots of samples from the pre-extinction Wiscoy Formation and the post-extinction Canaseraga Formation. In all plots, the x-axis is interpreted as representing onshore–offshore position and the y-axis is interpreted as representing frequency of disturbance. (a) nMDS scores for the Wiscoy Formation samples. Points are coloured by locality and general facies (sandy member versus muddy member). (b) nMDS scores for the Canaseraga Formation samples. (c) nMDS species plot for the Wiscoy Formation. Species that went extinct in the first extinction pulse are marked by open circles and species that survived the first extinction pulse by filled circles. Point colour indicates taxonomic order. (d) nMDS species plot for the Canaseraga Formation. Species that survived the LKW are denoted by filled squares and new species are denoted by open squares. Point colour indicates order.

On the nMDS axis 2, there is a ‘spur’ of data points aligned downward along the axis (figure 2a). Position on this axis reflects the relative abundance of Ambocoelia gregaria, with the most negative samples containing 100% A. gregaria (figure 3a; electronic supplementary material, figure S1). Ambocoelia gregaria is most abundant in muddy laminae interbedded with sandstones bearing hummocks and swales, interpreted as storm deposits [30,33]. The trace fossil Skolithos can be common in these hummocky sands, suggesting opportunistic colonization following storms [43]. Given the tendency of A. gregaria to occur in monospecific assemblages in this setting, we interpret it as an opportunistic taxon associated with environmental disturbance [29,30,43,44].

Figure 3.

Comparison of each surviving species' preferred environment before and after the extinction based on nMDS scores. Normal distribution curves graphically represent the preferred environment (nMDS1 or nMDS2 species score) and the environmental distribution (standard deviation weighted by proportional abundance) of each surviving species before and after the extinction. These are constructed with equal area under the curve, with amplitude not seen as meaningful for this analysis. (a) Comparison of nMDS1 scores (onshore–offshore position). The positive correlation (r² = 0.57) indicates that surviving species generally maintain their relative positions along the onshore–offshore gradient. (b) Comparison of nMDS2 scores (substrate disturbance/stability). The positive correlation (r² = 0.72) indicates that surviving species maintained their general relative position on the substrate disturbance gradient.

(b) . Post-extinction ecological structure

As in the Wiscoy samples, the first nMDS axis for the Canaseraga samples corresponded with onshore–offshore position (figure 2b). Samples from the most distal (northwest) location, BCP, plot to the left, whereas those from the most proximal (southeast) site, TGB, plot to the right. For a given location, samples from the muddy member plot to the left of those from the sandy member. Thirteen brachiopod species present in the Wiscoy disappeared during the first extinction interval and 12 new species appeared in the Canaseraga [29,30]. All strophomenids and atrypids present in the Wiscoy went extinct, with only one new atrypid appearing in the basin after the extinction. Productids dominated the new species (figure 2d).

As in the pre-extinction plots, position on the nMDS axis 2 reflects the relative abundance of A. gregaria, which increases in proportional abundance along a ‘spur’ of data points projecting downward (figures 2b and 3b). Although this species was never as abundant as it was in the Wiscoy, it was most abundant near the transition from muddier to sandier sediments at the CAM section.

(c) . Preferred environment

nMDS axis 1 scores for species present in both the Wiscoy and the Canaseraga samples were correlated, with an r² of 0.57 (figure 3a). This indicates that, generally, these species were similarly positioned along an onshore–offshore transect before and after the extinction. nMDS axis 2 scores for species present in both the pre- and post-extinction strata were also strongly correlated, with an r² of 0.72 (figure 3b), indicating that survivor species maintained similar relative levels of tolerance to disturbance before and after the extinction. However, the correlation in axis 2 scores is strongly driven by low values in A. gregaria (electronic supplementary material, figure S1).

(d) . Niche breadth and peak abundance

Overall, niche breadth did not change significantly as a result of the first extinction pulse; when all surviving species are combined, the variance on nMDS axis 1 was not significantly different in the Wiscoy and Canaseraga datasets according to Levene's test (F = 0.18, p = 0.67). However, the standard deviations of individual survivor species were not strongly correlated across the extinction pulse, although, as noted, such comparisons are difficult in fossil datasets [3]. Notably, the surviving species that are highly abundant in both the Wiscoy and Canaseraga demonstrate little change in niche breadth through the extinction (e.g. Floweria chemungensis; electronic supplementary material).

Peak abundance of survivor species in the Wiscoy and Canaseraga were uncorrelated, with an r² of 0.25 (maximum proportional abundances of species) and an r² of 0.26 (mean of the highest three proportional abundances) (electronic supplementary material).

4. Discussion

Mass extinction can reshape biological communities in many ways. Using the Appalachian Basin as a model, we examine how the niche structure, ecology and overall faunal gradient structure respond to environmental perturbation and changes in the regional species pool.

(a) . Ecological stability and niche conservatism

In the Appalachian Basin, members of the orders Atrypida and Strophomenida became extinct preferentially (figure 2c) [29]. New species that appeared after the first extinction pulse largely belong to the orders Productida and, to a lesser extent, Rhynchonellida (figure 2d). The shift to the dominance of productids, an order characterized by spines used for stability and/or defence, suggests a shift to a new ecologically distinct state across the shelf [45–47] that affected marine ecosystems worldwide [48], although it occurred unusually abruptly in the Appalachian Basin [29]. This new state persisted for the remainder of the Palaeozoic Era [48].

Despite these large-scale changes in faunal composition, nMDS analyses of pre- and post-extinction faunas show that brachiopod faunas in the Appalachian Basin were structured similarly before and after the extinction event. In both time slices, the primary gradient in faunal composition (nMDS axis 1) is interpreted as onshore–offshore position, as indicated by facies and geographical position along the habitat transect (figures 1a and 2). Likewise, nMDS axis 2 correlates with the abundance of A. gregaria in both analyses; as noted above, A. gregaria is interpreted as a rapid colonizer that blooms in frequently disturbed environments. Both before and after the extinction, these Ambocoelia-dominated assemblages only occur at intermediate depths, imparting a wedge shape to the plotted data. Thus, the same environmental gradients (onshore–offshore position and disturbance) structure the fauna before and after the extinction in the same manner.

Furthermore, there is strong evidence for niche conservatism in the survivors of this extinction pulse, as these species maintained their relative placement on both depth and substrate disturbance gradients (figure 2). Species position on these gradients was correlated with statistical significance despite the limited number of survivor species available for comparison (a low number of survivor species is, of course, an issue inherent to the study of mass extinctions). Likewise, there is no evidence of a systematic shift towards wider or narrower environmental tolerances after the extinction (electronic supplementary material, data file) [5], although, again, such comparisons should be viewed with caution [3]. The same environmental gradients seem to have controlled species' distributions, and survivors did not significantly shift their positions on these gradients [3]. Overall, these data indicate that the positions of ecological niches of survivors were relatively stable through the extinction. However, the peak abundances of species were uncorrelated between the two formations, suggesting that species may have become more or less dominant in these niches.

Finally, species richness was also quite stable at the regional level, with similar numbers of species occurring in the Wiscoy samples and in the Canaseraga samples (electronic supplementary material). In other words, the victims of the first pulse of extinction were replaced by an approximately equal number of new species. These new species are presumed to be, at least in part, immigrants from other basins, as they are not closely related to taxa we found in the Wiscoy. However, some may have been present in the Wiscoy, but unsampled. Intriguingly, the second pulse of extinction had few victims and was not followed by a pulse of immigration [29,30].

(b) . Niche conservatism and environmental change

Similar methods have been used to study fossil marine invertebrates, including many brachiopods, from the Upper Ordovician of the Cincinnati Arch Region during a faunal turnover known as the Richmondian Invasion [5,22,40,49,50]. These studies detail changes in benthic faunal gradient structure, especially in light of the introduction of new benthic taxa, as the region experienced gradual changes in climate [5,40]. As in our study, faunal gradients were structured by a shallow–deep depth gradient and secondarily by substrate stability. However, unlike in our study, considerable niche evolution occurred during the Richmondian Invasion [5,40,50]. A variety of factors could be responsible for the contrasting faunal responses between the Ordovician and Devonian examples, but the most obvious is the nature and pace of the environmental changes associated with each event. The first pulse of extinction represents a relatively rapid drop in global temperatures and expanded marine dysoxia, followed by a return to similar conditions post-extinction [25,27,29]. Environmental change may have been too rapid, extreme or non-directional for species to adapt via natural selection. This fluctuation in environmental conditions contrasts with the slower, longer-term climate changes in the Ordovician example, where directional selective pressures could have driven niche evolution [5,40].

Another possible response to mass extinction is reduced competition and/or predation, and as a result, the expansion of species’ niche breadths [5,51]. Although it is difficult to rigorously compare niche breadths of individual species from these data [3], we detected no overall change in niche breath from the pre-extinction samples to the post-extinction samples. Possibly, niche breadth was controlled primarily by environmental factors (i.e. depth and substrate disturbance), such that changes in community composition had little effect. Alternatively, the lack of change in niche breath may reflect the fact that brachiopod species richness recovered quickly after the extinction, and our post-extinction samples post-date this recovery. It is possible that ecological gradient structure and niche breadths were transiently disrupted during the extinction and immediately afterward (e.g. during the deposition of the Pipe Creek Formation and lowest beds of the Canaseraga Formation), but we lack sufficient samples from this interval to test this possibility (i.e. we have not located outcrops that preserve the proper environments owing to facies migration during the Pipe Creek transgression).

In sum, these examples suggest that the effects of major Earth-system changes on the environmental distributions of species are context-specific, depending on the nature, magnitude and duration of environmental change.

5. Conclusion

The Appalachian Basin brachiopod fauna experienced significant turnover during the first pulse of the Late Devonian extinction event, which eliminated about half of the species pool. Owing to a combination of selective extinction and immigration, ecological dominance shifted to clades that were distinct in morphology and likely ecology. This shift reflects global trends in brachiopod faunal composition. However, common brachiopod species that survived the mass extinction were largely unaffected by these taxonomic and ecological changes. Their spatial distribution was still controlled by environmental gradients in depth and disturbance, and they demonstrated niche conservatism along these gradients, tracking their preferred habitats through changes in temperature, sea level and community composition. Thus, ecological reorganization was driven primarily by turnover in species composition due to extinction and immigration, rather than by secondary evolutionary reactions by the remnants of the pre-extinction fauna. However, survivor species did change in peak abundance across the event, suggesting that some were more successful in the aftermath of the extinction, while others were less successful.

Data accessibility

The datasets and code supporting this article have been uploaded as electronic supplementary material [52].

Authors' contributions

S.K.B.: conceptualization, data curation, formal analysis, writing—original draft, writing—review and editing; J.Q.P.: data curation, formal analysis, methodology; J.A.B.: data curation, methodology; A.M.F.: investigation; A.M.B.: conceptualization, data curation, funding acquisition, methodology, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the National Science Foundation (grant nos EAR-0922186 and EAR-1738121).