Proceedings of the Royal Society B: Biological Sciences
You have accessResearch articles

When and how did the terrestrial mid-Permian mass extinction occur? Evidence from the tetrapod record of the Karoo Basin, South Africa

Michael O. Day

Michael O. Day

Evolutionary Studies Institute, School of Geosciences, University of the Witwatersrand, Private Bag 3, Johannesburg 2050, South Africa

[email protected]

Google Scholar

Find this author on PubMed

,
Jahandar Ramezani

Jahandar Ramezani

Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Google Scholar

Find this author on PubMed

,
Samuel A. Bowring

Samuel A. Bowring

Department of Earth, Atmospheric and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Google Scholar

Find this author on PubMed

,
Peter M. Sadler

Peter M. Sadler

Department of Earth Sciences, University of California, Riverside, CA 92521, USA

Google Scholar

Find this author on PubMed

,
Douglas H. Erwin

Douglas H. Erwin

Department of Paleobiology, National Museum of Natural History, Washington, DC 20013-7012, USA

Google Scholar

Find this author on PubMed

,
Fernando Abdala

Fernando Abdala

Evolutionary Studies Institute, School of Geosciences, University of the Witwatersrand, Private Bag 3, Johannesburg 2050, South Africa

Google Scholar

Find this author on PubMed

and
Bruce S. Rubidge

Bruce S. Rubidge

Evolutionary Studies Institute, School of Geosciences, University of the Witwatersrand, Private Bag 3, Johannesburg 2050, South Africa

Google Scholar

Find this author on PubMed

    Abstract

    A mid-Permian (Guadalupian epoch) extinction event at approximately 260 Ma has been mooted for two decades. This is based primarily on invertebrate biostratigraphy of Guadalupian–Lopingian marine carbonate platforms in southern China, which are temporally constrained by correlation to the associated Emeishan Large Igneous Province (LIP). Despite attempts to identify a similar biodiversity crisis in the terrestrial realm, the low resolution of mid-Permian tetrapod biostratigraphy and a lack of robust geochronological constraints have until now hampered both the correlation and quantification of terrestrial extinctions. Here we present an extensive compilation of tetrapod-stratigraphic data analysed by the constrained optimization (CONOP) algorithm that reveals a significant extinction event among tetrapods within the lower Beaufort Group of the Karoo Basin, South Africa, in the latest Capitanian. Our fossil dataset reveals a 74–80% loss of generic richness between the upper Tapinocephalus Assemblage Zone (AZ) and the mid-Pristerognathus AZ that is temporally constrained by a U–Pb zircon date (CA-TIMS method) of 260.259 ± 0.081 Ma from a tuff near the top of the Tapinocephalus AZ. This strengthens the biochronology of the Permian Beaufort Group and supports the existence of a mid-Permian mass extinction event on land near the end of the Guadalupian. Our results permit a temporal association between the extinction of dinocephalian therapsids and the LIP volcanism at Emeishan, as well as the marine end-Guadalupian extinctions.

    1. Introduction

    An ‘end-Guadalupian’ extinction, distinct from that at the end of the Permian, was first recognized in the marine realm in the 1990s [1,2]. Shortly afterwards it was calculated to be one of the most catastrophic extinction events of the Phanerozoic [3] and since then a considerable body of work has attempted to explore it, focusing on carbonate platforms of southern China and Japan [49]. Some recent studies have identified extinction horizons preceding the Guadalupian–Lopingian boundary (GLB) by several conodont zones [7,8] and therefore conclude that the mid-Permian marine extinctions constitute an intra-Capitanian event. This older age appears to be corroborated by new data from high latitude boreal sequences from Spitzbergen [10]. However, extinction horizons have been identified for a number of Guadalupian invertebrate taxa at various biostratigraphic levels between the late Capitanian Jinogondolella prexuanhanensis/xuanhanensis Conodont Zone and the earliest Lopingian Clarkina postbitteri postbitteri Conodont Zone [6,7]. The ages of the marine extinctions therefore remain unclear but mostly appear to occur before the GLB, for which the current age estimate of 259.1 ± 0.5 Ma is derived from links to the termination of volcanism in the Emeishan Large Igneous Province (LIP) [11,12].

    In the terrestrial realm, at least one extinction event among plant species has been identified in the Capitanian Upper Shihhotse Formation of Northern China, suggesting the loss of at least 56% of plant species within this stage [13]. Unfortunately, the age estimates for this sequence rely on palaeomagnetism and missing data within the stratigraphic profile has led to high uncertainty in their correlation with the marine timescale. Among tetrapods, diverse mid-Permian assemblages characterized by dinocephalian therapsids are known around the world but primarily from Russia and South Africa [14]. Younger therapsid faunas of late Permian age do not contain dinocephalians and this has led to the proposal of a global ‘dinocephalian extinction event′ [15], which is best recorded in the Beaufort Group of South Africa's Main Karoo Basin, where the abundance of fossil tetrapods has enabled its subdivision into vertebrate assemblage zones [16]. The extinction of dinocephalian therapsids occurred in the uppermost Tapinocephalus Assemblage Zone (AZ), which is most widely exposed in the southwest of the basin, where it is situated stratigraphically close to the boundary of the Abrahamskraal and overlying Teekloof formations (figure 1). In the past, stratigraphic collecting within the lowest Beaufort strata was limited and most fossil specimens collected from there have imprecise provenance data. Consequently, the stratigraphic ranges of most genera from the Tapinocephalus AZ have been attributed only to that biozone rather than to a discrete stratigraphic interval [16]. Although the disappearance of the diverse dinocephalian fauna had been recognized [17,18], such broad assignments and poor correlation have led to the misinterpretation of Permian tetrapod diversity changes and the extrapolation of an artificial extinction ‘horizon′ at the top of the Tapinocephalus AZ [19].

    Figure 1.

    Figure 1. Minimum stratigraphic ranges of tetrapod genera across the mid-late Permian boundary in the Karoo Basin. The shaded triangles on the left indicate uncertainty surrounding the age of biozone boundaries (electronic supplementary material, Methods). Genera from the Tropidostoma AZ presented here are only those believed to occur in the lowest part of that AZ. The age of the marine extinctions is taken from the literature cited in the text. Star, dated horizon presented here; Wuchi., Wuchiapingian. (Online version in colour.)

    Limited chemostratigraphic data have been used to suggest a correlation between the extinction of dinocephalians in South Africa and the end-Guadalupian marine extinctions [19]. However, carbon isotope correlations for the Capitanian may be dubious [15,20] and most recent studies have instead accepted that the dinocephalian extinction is most probably constrained to the Wordian on the basis of allegedly stronger magnetostratigraphic correlations, thus predating the Capitanian extinctions [8,13,15]. This relies on the restriction of Russian dinocephalian faunas to strata below the Illawarra geomagnetic reversal [21], which has been tentatively assigned to the latest Wordian or perhaps even earlier [2224]. Conversely, the presence of normal polarity in the lower Abrahamskraal Formation suggests the dinocephalian extinctions in South Africa post-date the Illawarra reversal [24]. Although it is possible that the last occurrences of dinocephalians in Eastern Europe and South Africa were diachronous, biostratigraphic correlations support the equivalence of the Tapinocephalus AZ and youngest Russian dinocephalian assemblage (Isheevo) [14,15,25], making the existence of a large temporal discrepancy unlikely. The overlying Pristerognathus AZ, however, has not been directly correlated with an Eastern European equivalent. In most cases, the Pristerognathus AZ is either ignored [14] or is placed rather arbitrarily at an equivalent horizon to the Deltavjatia vjatkensis AZ, probably because of their status as the first ‘post-dinocephalian’ assemblages (e.g. [25]). Kurkin [26] noticed that the distribution of dicynodont therapsids suggests the Deltavjatia AZ may have closer affinities to the late Permian South African Tropidostoma AZ (figure 1), which would leave the Pristerognathus AZ in limbo between the Ulemosaurus and Deltavjatia AZs. The correlation of the Pristerognathus AZ with Eastern Europe remains uncertain but Russian authors consider the dinocephalian extinction in Russia to fall close to the end of the Guadalupian [25,26].

    Recent U–Pb CA-TIMS geochronology from the South African Beaufort Group has provided precise age constraints for several volcanic ash beds intercalated with Permian biozones within the Main Karoo Basin, from which the top of the Tapinocephalus AZ was determined to be older than 261.241 ± 0.088 Ma [27]. This was based on a date considered to be within the lower part of the overlying Pristerognathus AZ; however, the dated horizons are located in the southeast of the Karoo Basin, where few fossil specimens have been recorded from these biozones and thus biostratigraphic resolution is correspondingly poor. The precise age of the dinocephalian extinction event therefore remains poorly constrained.

    2. Material and methods

    (a) Genus stratigraphic ranges

    The stratigraphic ranges of 54 tetrapod genera, represented by 1953 specimens from the Abrahamskraal Formation and overlying Poortjie Member of the Teekloof Formation, were recalculated from the stratigraphic positions of all known genus occurrences within the Karoo Basin. Generic occurrences were taken initially from the Beaufort Fossil Vertebrate GIS database, maintained by the Evolutionary Studies Institute (ESI) in Johannesburg, which records locality data for 13 596 fossil specimens curated in South Africa that are identified to genus level. The GIS data were imported into .kmz format and thereafter into Google Earth from where the surface geology could be seen. To this were added 37 specimens from foreign or other collections and 103 specimens collected during ongoing fieldwork by the ESI (see the electronic supplementary material, Methods). All 1953 fossil specimens used in this study are listed in the electronic supplementary material, table S1.

    The Karoo Basin (approx. 600 000 km2) was split into 10 sectors to account for lateral variation in stratigraphy (figure 2), within which fossil specimens occurring in the Abrahamskraal Formation and the Poortjie Member of the Teekloof Formation were given stratigraphic positions using 50 m stratigraphic bins for the Abrahamskraal Formation and Lower, Middle and Upper divisions for the Poortjie Member (see the electronic supplementary material, Methods and table S1). This enabled the range of each genus to be calculated within each sector, which was subsequently used in the constrained optimization (CONOP) analysis (see the electronic supplementary material, table S2). A composite section showing tetrapod-stratigraphic ranges throughout the basin was compiled using correlation of the Abrahamskraal–Teekloof Formation contact.

    Figure 2.

    Figure 2. Map of the lowest Beaufort Group in the South African Karoo Basin. Star, locality of tuff sample (K1202-B1) on the farm Puntkraal, Sutherland district, Northern Cape Province. Alternating long and short dashed lines, boundaries of the sectors used when calculating the stratigraphic position of fossil specimens. The Philippolis sector is isolated as no fossil material has been found in the lowest Beaufort strata between the eastern limit of the Victoria West sector and the vicinity of Philippolis. (Online version in colour.)

    (b) Constrained optimization (CONOP)

    Owing to the uneven recovery of fossils across the Karoo Basin, the 10 sectors differ in detail concerning the sequence of genus appearances and disappearances. The best record combines information from all sectors (see the electronic supplementary material, table S2). When sections are combined by lithological correlation, using presence or absence values within stratigraphic bins, genus range-ends are grouped at equivalent intervals. Such grouping tends to inflate generic richness, so to mitigate this we also sought fully resolved composite sequences of first and last appearances using CONOP. The resulting composite sequences achieve best-fit event ordering in relation to all the locally preserved sequences and do not rely upon lithostratigraphic boundaries for correlation between sectors. Best-fit was found by minimizing the net adjustments required to make the sequence in every sector match a single sequence. Adjustments were limited to range extension and subject to the constraint that any viable sequence must reproduce all observed coexistences of genera. Sequences were optimized, from a random starting sequence, using the simulated annealing heuristic as implemented by the CONOP algorithms [28]. After an optimal sequence was reached, the spacing between events was scaled using the mean thickness between adjacent events in the local sectors.

    The optimization does not use the local positions of lithostratigraphic boundaries. To provide a link between this arbitrary scale and the lithostratigraphy, lithostratigraphic boundaries were entered as separate events in each section. Thus, they are not forced to be correlative, but find their position in the composite according to the local taxon ranges. The algorithm was constrained to maintain only the known superpositional order of the formations. This was achieved by adding a stratigraphic ‘pseudo-section’ that included only the lithostratigraphic boundaries—one event level for each boundary. The optimized composite sequences then emerged with a suite of local events for each boundary that did not overlap with events for other boundaries and could be attached as conservative estimates to the arbitrary scale (figure 3).

    Figure 3.

    Figure 3. Generic richness and generic last-appearance events. (a) Raw generic richness data for composite stratigraphic section. H, Hoedemaker Member; K, Karelskraal Member; P, Poortjie Member; T, Tropidostoma AZ. (b) Uncertainty band (shaded) for best-fit composite sequences of last-appearance events per five composite sequence units (solid lines). Dashed lines, percentile outcomes for 300 random reordering events using the same genus durations. (c) Uncertainty band for total generic richness within the upper Abrahamskraal Formation and lower Teekloof Formation. The position of lithostratigraphic boundaries can be constrained to uncertainty bands (hatched boxes on x-axis). (Online version in colour.)

    Because the CONOP analysis requires knowledge of first and last appearances, genera whose minimum stratigraphic ranges in any one sector were poorly constrained were excluded from the analysis. Genera with uncertain taxonomic validity were also excluded (see the electronic supplementary material, Methods and table S2).

    (c) U–Pb geochronology

    Zircons were separated from a white tuff layer (sample K1202-B1) 3.5 m above the basal sandstone bed of the Poortjie Member of the Teekloof Formation in the Sutherland district of the Northern Cape Province (electronic supplementary material, figure S1). Six single zircons were selected based on grain morphology using a binocular microscope and analysed by the U–Pb isotope dilution thermal-ionization mass spectrometry (ID-TIMS) technique following the detailed procedures described in Ramezani et al. [29] and used in Rubidge et al. [27]. All zircons were pre-treated by a chemical abrasion (CA-TIMS) method modified after Mattinson [30] to mitigate the effects of radiation-induced Pb loss, and spiked with the EARTHTIME ET535 mixed 205Pb–233U–235U tracer prior to dissolution and analysis. Uranium and lead isotopic data reduction, date calculation and propagation of uncertainties were carried out using computer applications and algorithms of Bowring et al. [31] and McLean et al. [32]. Complete U–Pb data appear in the electronic supplementary material, figure S1 and table S3.

    A sample date is calculated based on the weighted mean 206Pb/238U date of a statistically coherent cluster of the youngest zircon analyses consisting of three or more data points, which is interpreted as the age of pyroclastic eruption and a maximum estimate of the depositional age of the tuff. Uncertainties are reported at 95% confidence level and follow the notation ±X/Y/Z Ma, where X is the internal (analytical) uncertainty in the absence of all external errors, Y incorporates the U–Pb tracer calibration error and Z includes the latter as well as the decay constant errors of Jaffey et al. [33]. Complete uncertainties (Z) are necessary for comparison between age data from different isotopic chronometers (e.g. U–Pb versus 40Ar/39Ar).

    3. Results

    (a) Taxonomic turnover

    We present a comprehensive late Guadalupian tetrapod biostratigraphy based on a faunal analysis of 1915 specimens, belonging to 52 tetrapod genera (excluding gorgonopsians and temnospondyls) recovered from the lower Beaufort Group throughout the main Karoo Basin. Extensive fieldwork and map-based labwork allowed fossil specimens to be assigned to one or more 50 m stratigraphic bins within the upper Abrahamskraal Formation and lower Teekloof Formation of the Karoo Basin. Raw richness counts within stratigraphic bins suggest a decline from 35 genera to 10 from the Moordenaars Member to the mid-Poortjie Member, representing approximately 71% loss of generic richness (figure 3a).

    Because preserved ranges underestimate true ranges [34], the severity of extinction can be independently and more finely resolved using CONOP. Several equally good solutions all reveal a significant clustering of last appearances, which supports the existence of coordinated extinctions and indicates that genus richness was lost as a result of an increased extinction rate, not merely a drop in originations (figure 3b). In the majority of best-fit sequences recovered by CONOP there is a peak in last appearances within the Karelskraal Member of the uppermost Abrahamskraal Formation (figure 3b). Instantaneous generic richness may have peaked between 29 and 35, before falling to between 7 and 8, thus suggesting a statistically significant 74–80% loss of generic richness between the Karelskraal Member and the middle Poortjie Member (figure 3c). These data suggest that immediately preceding the peak in last-appearance events per-taxon extinction counts were approximately 0.15, rising more than threefold to approximately 0.5 at the peak before declining to approximately 0.2 after generic richness reached its minimum.

    (b) Age of the TapinocephalusPristerognathus AZ boundary

    Here we report a new U–Pb zircon date (CA-TIMS method) of 260.259 ± 0.081/0.14/0.31 Ma based on the weighted mean 206Pb/238U date of four overlapping zircon analyses from a tuff bed situated 3.5 m above the basal sandstone of the Poortije Member of the Teekloof Formation in the western Karoo Basin (Northern Cape Province). Biostratigraphically, this is well constrained by local fossil occurrences and basin-wide fossil datasets to the uppermost Tapinocephalus AZ, close to the boundary with the overlying Pristerognathus AZ and thus provides a precise maximum age constraint on the TapinocephalusPristerognathus AZ boundary (figures 1 and 4). This supersedes the previously reported U–Pb date of more than 261.241 ± 0.088 Ma (or older) for the same AZ boundary in the southeastern Karoo Basin [27] (see the electronic supplementary material, Discussion).

    Figure 4.

    Figure 4. Recalibrated ages for Permian Beaufort Group biozones. Dated horizons are shown in lithostratigraphic position. Hashed boxes, uncertainty surrounding stratigraphic position of assemblage zone boundaries relative to dated horizons. New date in bold. All units are scaled to time. Asterisk, date obtained from [27].

    4. Discussion

    Of the higher taxa, the greatest levels of extinction occur in the clade Dinocephalia, which went extinct with its 16 genera present in the pre-extinction Moordenaars Member. Secondly, all three stem-group genera of Tapinocephalus AZ pareiasaurian parareptiles became extinct, although more advanced, crown-group pareiasaurs had reappeared in the Karoo Basin by the Late Permian Tropidostoma AZ [35,36]. The disappearance of both genera of varanopids in this interval heralded the final extinction of pelycosaur-grade synapsids. Among dicynodonts and therocephalians, no supra-generic losses occurred within the extinction interval and the decrease in generic richness was only moderate. The end-Tapinocephalus AZ extinction was therefore characterized by a decrease in generic richness across all major taxa, although perhaps nearly half of this resulted from the elimination of the species-rich dinocephalian clade. Conversely, the extinction of the less diverse basal pareiasaurs and varanopid synapsids had less impact of overall generic diversity. The fact that the dinocephalians and basal pareiasaurs were the largest animals in the Tapinocephalus AZ alludes to selectivity against ecological traits associated with large body size.

    Combined with a U–Pb CA-TIMS maximum age constraint of 255.2 ± 0.16 Ma for the top of the Cistecephalus AZ [27], our new date of 260.259 ± 0.081 Ma from the uppermost Tapinocephalus AZ suggests the combined duration of the Pristerognathus, Tropidostoma and Cistecephalus AZs would have been approximately 5 Myr (figure 4). This contrasts with an estimated duration of 3 Myr for the overlying Dicynodon AZ alone [27]. The duration of the Tapinocephalus AZ is more difficult to estimate as although U–Pb zircon ages (in situ secondary ion mass spectrometry) ranging from 264 to 268 Ma have been obtained from the lower Abrahamskraal Formation [24], these have uncertainties on the order of 2–6 Myr and are not biostratigraphically well constrained. The biochronology of the Permian Beaufort Group would therefore benefit most from more precise age constraints on the Eodicynodon AZ and the very lowest Beaufort Group strata.

    Our U–Pb date of 260.259 ± 0.081 Ma for the uppermost Tapinocephalus AZ constrains the end-Tapinocephalus AZ extinctions to the Late Capitanian [37]. This age is predicated upon the age of the GLB being isochronous with or shortly post-dating the uppermost felsic ignimbrites of the central Emeishan LIP, for which an age of 259.1 ± 0.5 Ma has recently been retrieved [12]. Combined with additional U–Pb CA-TIMS ages from intrusive rocks of the Emeishan LIP, this age indicates rapid emplacement between ca 260 Ma and 259.1 ± 0.5 Ma [11,12] and suggests that the end-Tapinocephalus AZ extinctions may have coincided with the initial stages of volcanism within the Emeishan LIP, as has been done for the extinctions of at least some marine organisms [7,8].

    The presumed close temporal relationship between the marine and plant extinctions and the Emeishan LIP has prompted much debate surrounding a link between the two (e.g. [7,11,13,3840]). The new age of approximately 260 Ma for the top of the Tapinocephalus AZ in South Africa is consistent with a further temporal link between Emeishan volcanism and tetrapod extinctions. If this was the case then it would circumstantially implicate Emeishan volcanism as a causal factor. The existence of a concurrent tetrapod extinction event in the non-marine Karoo Basin, situated thousands of kilometres from the Emeishan LIP, would also suggest an extinction mechanism with a global influence, such as atmospheric disruption through the release of volcanic gases and large-scale contact metamorphism of biogenic rocks [11,39,41,42]. New geochemical data from the GLB at geographically disparate localities is not consistent with global warming resulting from the large-scale release of methane or volcanic CO2 [20] and stable or cooling conditions may in fact have prevailed [43]. Acidification, short-term cooling and volcanically induced darkness, possibly linked to the onset of Emeishan volcanism, appear to be the most probable causes of extinction for marine animals at the end of the Guadalupian [39] and this could have affected terrestrial ecosystems through reduced photosynthesis. While it is tempting to conclude the existence of concurrent marine and terrestrial extinction events driven by LIP volcanism, proving concurrence is extremely difficult as the duration and timing of the Emeishan LIP and the age of the GLB are at present insufficiently constrained. The mechanisms driving elevated extinction rates among mid-Permian tetrapods are also poorly understood and there is as yet no robust geochemical evidence to support an increased volcanic influence in the Karoo.

    5. Conclusion

    Although not as catastrophic as that at the end of the Permian, the mid-Permian extinction on land was a significant event in which the tetrapod fauna experienced a 74–80% loss of generic richness. This was mostly because of the extinction of all dinocephalian therapsids and, although many basal genera disappeared and all major tetrapod taxa suffered some loss of generic richness, overall higher taxon extinctions were not severe. The ecological gap left by the loss of the diverse large-bodied elements of the fauna and the widespread reduction in generic richness probably became governing factors in the appearance of large gorgonopsians and dicynodonts and of small therocephalians in the late Permian Tropidostoma and Cistecephalus AZs [35].

    The date (approx. 260 Ma) for the base of the Poortjie Member of the Abrahamskraal Formation provides a high-precision maximum age constraint on the TapinocephalusPristerognathus AZ boundary in the Main Karoo Basin and indicates that the mid-Permian extinction event occurred close to the end of the Capitanian. This is later than the most recent estimates have suggested. Temporal constraints on Emeishan volcanism and the marine as well as the floral extinctions must be tightened before the existence of a global intra-Capitanian extinction can be properly assessed, but the new late Capitanian age for the terrestrial extinction event does support the possibility that all phenomena occurred within the same estimated time frame.

    Data accessibility

    The datasets supporting this article have been uploaded as part of the electronic supplementary material.

    Authors' contributions

    B.R. conceived the initial idea for the study and provided data on the stratigraphic position of fossil material in existing collections. M.D., B.R. and F.A. conducted the fieldwork. M.D. compiled fossil data. S.B. and J.R. are responsible for the U–Pb CA-TIMS analyses. P.S. conducted the CONOP analyses. All authors wrote the final manuscript and aided in the interpretation of the results.

    Competing interests

    We declare we have no competing interests.

    Funding

    This work was made possible by financial support to M.D., B.R. and F.A. from the Palaeontological Scientific Trust (PAST) and its Scatterlings of Africa programmes, as well as the National Research Foundation (NRF). The support of the DST/NRF Centre of Excellence in Palaeosciences (CoE in Palaeosciences) towards this research is hereby acknowledged. The geochronology was supported by NASA Astrobiology funding to D.E. and S.B.

    Acknowledgements

    We wish to thank Ken Angielczyk and Jörg Fröbisch for discussions concerning dicynodonts, Ellen de Kock at the Council for Geoscience and Sheena Kaal at the Iziko South African Museum for responding promptly to many requests for specimen information, and Charlton Dube and Sifelani Jirah for conscientious preparation of fossil specimens. We are grateful to Spencer Lucas and two anonymous reviewers for valuable comments on the original manuscript.

    Footnotes

    References