- © 2014 The Paleontological Society. All rights reserved.
Studies of the end-Permian mass extinction have suggested a variety of patterns from a single catastrophic event to multiple phases. But most of these analyses have been based on fossil distributions from single localities. Although single sections may simplify the interpretation of species diversity, they are susceptible to bias from stratigraphic incompleteness and facies control of preservation. Here we use a data set of 1450 species from 18 fossiliferous sections in different paleoenvironmental settings across South China and the northern peri-Gondwanan region, and integrate it with high-precision geochronologic data to evaluate the rapidity of the largest Phanerozoic mass extinction. To reduce the Signor-Lipps effect, we applied constrained optimization (CONOP) to search for an optimal sequence of first and last occurrence datums for all species and generate a composite biodiversity pattern based on multiple sections. This analysis indicates that an abrupt extinction of 62% of species took place within 200 Kyr. The onset of the sudden extinction is around 252.3 Ma, just below Bed 25 at the Meishan section. Taxon turnover and diversification rates suggest a deterioration of the living conditions nearly 1.2 Myr before the sudden extinction. The magnitude of the extinction was such that there was no immediate biotic recovery. Prior suggestions of highly variable, multi-phased extinction patterns reflect the impact of the Signor-Lipps effect and facies-dependent occurrences, and are not supported following appropriate statistical treatment of this larger data set.
The Permian/Triassic boundary (PTB) marks a milestone in biospheric history, with the transition from the Paleozoic Evolutionary Fauna to the Modern Evolutionary Fauna (Sepkoski et al. 1981). Although the end-Permian mass extinction has been studied intensively (Retallack 1995; Wignall and Twitchett 1996; Jin et al. 2000; Ward et al. 2005; Erwin 2006; Yin et al. 2000, 2007; Shen et al. 2011; Song et al. 2013), the cause and killing mechanism remain controversial. Some potential triggers include the eruption of the Siberian Traps (Bowring et al. 1998; Kamo et al. 2003; Shen et al. 2011; Payne and Clapham 2012), climate change due to global warming (Knoll et al. 2007; Svensen et al. 2009; Shen et al. 2011; Joachimski et al. 2012; Sun et al. 2012), and development of global oceanic anoxia to euxinia (Wignall and Twitchett 1996; Isozaki 1997; Grice et al. 2005; Algeo et al. 2008; Sengör and Atayman 2009).
Fundamental to understanding the end-Permian mass extinction is unraveling extinction patterns. However, different approaches have long yielded different results. Previous analyses of biodiversity across the PTB are either global compilations with coarse time bins (many at stage level) (e.g., Sepkoski et al. 1981; Wang and Sugiyama 2000; Shen et al. 2006a; Alroy et al. 2008) or studies of single sequences of strata. Several decades ago, coarse time resolution (bins of nearly 10 Myr) suggested that the end-Permian mass extinction was a prolonged event that peaked at the end of Guadalupian and lasted until the end of Permian (Sepkoski 1986). As more detailed fossil records from South China became available, the data favored two-pulses: a single extinction at the end-Guadalupian with a distinct loss of benthic faunas, and a second, much more severe crisis, at the end of Permian (Jin 1993; Erwin 1993; Jin et al. 1994; Stanley and Yang 1994; Shen and Shi 1996, 2002; Shi et al. 1999; Wang and Sugiyama 2000; Clapham et al. 2009). However, more-detailed studies of the pattern of extinction in the very latest Permian, even from the same sections, have produced contradictory results. Jin et al. (2000) documented an abrupt extinction pattern based on records from the Meishan section over an interval of less than a half-million years. On the other hand, another group documented two- or multiple-phased extinctions in different sections, including Meishan (e.g., Feng et al. 2007; Yin et al. 2000, 2007; Chen et al. 2009; Song et al. 2009, 2013). Studies of single localities avoid the problems of correlation and simplify the accounting of species richness in different stratigraphic units, but severely compromise the scope of the study and are susceptible to bias from stratigraphic incompleteness and facies control of preservation (e.g., Feng et al. 2007; Chen et al. 2009; Song et al. 2009). These discrepancies in extinction patterns reflect varying inventories of fossils in collections from different sections and different analytical approaches (Mark et al. 2003).
Crucial to uncovering the process of the end-Permian extinction is a detailed record of the extinction and recovery from many species, across a variety of depositional environments and encompassing a broad geographic swath from low to high paleolatitude. We have expanded a previous species-level analysis based on the single locality at Meishan (Jin et al. 2000) and developed a detailed analysis of inter-regional, species-level diversity from 18 intensively studied sections in South China and the northern peri-Gondwanan region. This analysis spans ∼16 Myr with no temporal binning of taxon durations (Shen et al. 2011). We used Constrained Optimization (CONOP), a program developed by one of the authors (P. Sadler), to provide the correlation among multiple sections and establish the biodiversity pattern. This paper expands upon the results presented by Shen et al. (2011), in particular by quantifying the detailed process of the mass extinction, and identifying possible sampling effects as well as regional differences in extinction patterns.
Data and Biostratigraphic Constraints
We chose 18 sections from the Late Guadalupian to Early Triassic for this study. Twelve of the richly fossiliferous sections are from shallow-water carbonate platform, slope, and basinal facies in three depositional basins (the Upper Yangtze Basin, the Lower Yangtze Basin, and the Dian-Qian-Gui Basin) in South China. Six sections from Tibet, Kashmir, and the Salt Range of Pakistan were also included, all of which were deposited on a marginal shelf in the northern peri-Gondwanan region (Shen et al. 2006a). Besides our measured sections, we also incorporated the published fossil records from the Huangzhishan, Meishan, Chaohu, Shangsi, Matan, Tieqiao (Meili), Penglaitan, Gyanyima, Tulong, Selong, and Qubu sections (Li et al. 1989; Jin et al. 2000; Zhang et al. 2004; Shen et al. 2006a, 2007; Wang et al. 2006) as well as other well-studied sections (Yang et al. 1987; Shang et al. 2001; He et al. 2005; Tong et al. 2005; Nafi et al. 2006) (Table 1).
Conodonts provide the primary correlation between sections and received extensive study, especially at the Penglaitan section, the GSSP for the base-Wuchiapingian Stage (Jin et al. 2006a), Meishan section D, the GSSPs for both the base of the Changhsingian Stage (Jin et al. 2006b) and the base of the Triassic System (Yin et al. 2001), and Pingdingshan section, one of the GSSP candidate sections for the base of the Olenekian Stage (Tong et al. 2005). In total, 23 conodont zones covering the Late Guadalupian to Early Triassic were used in this study. We also used 19 high-precision radiometric isotopic ages from volcanic ash beds in the measured sections. These dates established a high-precision geochronologic framework for the Lopingian and Permian–Triassic transition (Shen et al. 2011). Included in this analysis are CA-TIMS U-Pb dates from Beds 6, 7, 15, 22, 25, and 28 at the Meishan section, eight dates from the Shangsi section, three from the Penglaitan section, one from the Tieqiao (Meili), and one from the Matan section. As discussed further in Shen et al. (2011), we used 252.17 Ma as the age of PTB, 254.14 Ma as the age of the Wuchiapingian/Changhsingian boundary (WCB) and 260 Ma as the age of the Guadalupian–Lopingian boundary.
We tabulated first occurrences (FOs) and last occurrences (LOs) of 1450 species from 16 fossil clades (cephalopods, brachiopods, foraminifers, fusulinids, conodonts, corals, bivalves, radiolarians, bryozoans, gastropods, ostracods, fishes, calcareous algae, spores and pollens, and plants) in the data compiling software CONMAN, which is packaged with CONOP.
Two approaches are most commonly used to document fossil diversity patterns and their relationship to other geological events. The first avoids the difficulties of time correlation by analyzing the local sequence of events in a single richly fossiliferous section. The second simplifies correlation by subdividing earth history into coarse time bins (intervals of uneven duration such as stage, stratigraphic unit, or bed) and analyzes global or regional diversity within and between bins. The five most extensive Phanerozoic mass extinctions were recognized in a binned database of the first and last occurrences of marine families (Sepkoski 1978, 1982; Sepkoski et al. 1981; Raup and Sepkoski 1982). Understanding of the Phanerozoic biodiversity pattern has been greatly expanded recently by the Paleobiology Database group (Alroy 2000, 2003; Miller and Foote 2003, 2009; Bambach et al. 2004; Bambach 2006; Alroy et al. 2008; Kiessling et al. 2010), but this approach is still subject to relatively coarse time divisions and is focused mainly on general patterns during the Phanerozoic. It is not ideal for studies of a single event such as the end-Permian mass extinction that happened within a very short time. Studies of events with short time scales, such as mass extinctions, have focused on fossil range data from specific sections. However, this approach can be seriously biased by incomplete sampling, fossil preservation, and facies change (Signor and Lipps 1982; Strauss and Sadler 1989; Marshall 1990, 2010; Sadler 2004).
There have been numerous previous studies of biases in the fossil record (Strauss and Sadler 1989; Marshall and Ward 1996; Marshall 1997, 2010; Alroy et al. 2001; Peters and Foote 2001; Holland and Patzkowsky 2001; Crampton et al. 2003; Bush et al. 2004; Sadler 2004; Sadler and Sabado 2004; Kiessling 2006; Lu et al. 2006). Such studies have shown that the observed local range of a fossil taxon in a stratigraphic section is almost always a truncated version of the true local range (Strauss and Sadler 1989), biasing extinction patterns in single sections (Signor and Lipps 1982).
To address effects of incompleteness of the fossil record, Strauss and Sadler (1989) developed an approach for calculating confidence intervals on the end-points of stratigraphic ranges of fossils, based upon the frequency distribution of gaps within the observed range. This approach has been further developed by Marshall (1990, 1994, 1997) and Wang and Marshall (2004) and used for subsequent studies to locate the extinction intervals in single sections (e.g., MacLeod 1996; Jin et al. 2000; Wang and Everson 2007). However, the confidence interval approach is hard to adapt to multiple sections, and cannot satisfactorily distinguish between genuinely rare taxa and poorly sampled taxa. This distinction is readily made by examining multiple sections and building a composite time line of first and last appearances.
The Meishan section D in South China contains both the PTB and WCB GSSPs (Yin et al. 2001; Jin et al. 2006b) and is the most intensively studied section for the end-Permian mass extinction. Jin et al. (2000) used confidence interval analysis to document an abrupt extinction within a half-million years; this extinction interval has recently been narrowed to 200 Kyr or less (Shen et al. 2011). In contrast, Yin et al. (2000, 2007) and Song et al. (2009, 2013) proposed a two-phase extinction pattern based on bed-by-bed stratigraphic ranges of fossils from Meishan and some other sections, with extinction horizons at beds 24 and 28 with an intervening distinct recovery stage. To clarify the difference between these results, we developed a database based on 18 intensively studied sections in South China and the northern peri-Gondwanan region. We used CONOP to provide the correlation among multiple sections and establish the biodiversity pattern.
CONOP uses the first and last occurrence datums (FOs and LOs) of all species from all sections. Because of the random and incomplete nature of the fossil record, the observed sequences of FOs and LOs commonly differ between one section and another, resulting in a tangled-fence diagram (Fig. 1A), thus highlighting the inadequacies of single sections. Although the conodont taxa provide a stable framework for correlation, many more taxa are needed to establish macroevolutionary patterns and most of these have a very incomplete record in any one section. CONOP finds a sequence for all FOs and LOs that is optimal in the sense that all local sequences may be fit to it with the least net adjustment of all ranges. The permitted adjustments consist solely of extending local ranges. In this way, we compensate for the inadequacies of single sections and thus, in effect, at least partly correct for the incompleteness of fossil records (Sadler and Cooper 2003; Sadler 2004, 2006).
In optimizing the sequence of species FOs and LOs, the only permissible adjustments are extensions of the locally observed species ranges. The goal is to find the smallest set of adjustments necessary to bring all the local range charts into agreement with a single time line. Acceptable sequences are constrained to include all observed coexistences of pairs of taxa, and to place dated ash fall events in an order compatible with their analytical uncertainties. The search process begins with a sequence of all FOs in random order at the start, all LOs in random order at the end, and all dated ash falls in between. This sequence is subject to a series of mutations in which one event moves to a new position; both the event and position are chosen at random. Mutations are retained or removed according to the fit of the new sequence to the field data. Mutations that improve the fit are always retained; others are retained with a probability determined by the simulated annealing algorithm. This probability decreases with the size of overall misfit and throughout the progress of the search toward the best-fit sequence. The best solution, as a result, can be shown in the fence diagram, in which the expected FOs and LOs of all species are plotted as wire-lines between section columns (Fig. 1B). In contrast with the raw observed range ends (Fig. 1A), the adjusted range ends generate time lines that do not cross.
CONOP finds the purely ordinal solution for the sequence of all the event horizons. The conodont zonation and U-Pb CA-TIMS dates provide a timescale to evaluate the rate of change of the fossil data. Fortunately, the broad environmental and geographic distribution and good preservation of conodonts allow their ranges to be consistently recorded across all sections. An initial time line was built using only conodont species and geochronologic data. The resulting conodont sequence and geochronologic data were then included in a “pseudosection” (a section without a true thickness scale), which was weighted 100 times more heavily than the individual sections, and the remaining taxa were added to the optimization. This weighting scheme prevented the conodont framework from being distorted by the more numerous, but less completely preserved, range information for the other taxa. Thus, the conodonts provide a biostratigraphic constraint in the optimized sequence, while the levels of origination and extinction events are dated using piece-wise linear interpolation of geochronologic ages between placement of the dated levels. Richness is then a simple unbinned running sum with +1 for each origination and −1 for each extinction in the time line.
CONOP minimizes the sum total of all range extensions required to make all local sections fit a single time line of species origination and extinction events. The optimal or best-fit time line requires the least net range extension. Although this approach yields a better result than the information from any one section, the total information is still inadequate to yield a unique solution. The best-fit solution is not unique. We deliberately illustrate not a single best-fit solution but a set of solutions all falling within 2% of the best-known fit (2% more net range extension than the minimum). The 2% threshold allows for the likelihood that the optimization routines overfit the data in the sense that the final small adjustments to ranges may be finer than the precision of the field methods. The result is similar to showing observed ranges in one section with confidence intervals, and supplies an estimate of uncertainty bounds on the time line and derived macroevolutionary time series. The process of finding the optimal sequence was repeated many times in this study to generate a set of 17 equally best-fit and nearly best-fit sequences that together place uncertainty bounds on the pattern of taxon richness (Fig. 2).
Results: Abruptness of the end-Permian Mass Extinction
We generated a composite species richness curve from the 18 sections from South China and the northern peri-Gondwanan region across more than 16 million years (Fig. 2). A set of 17 solutions are displayed; all can be fit to the field data with adjustments to the observed ranges that are within 2% of the best known fit—a level of fit at which differences are primarily due to random differences within irresolvable clusters of first and last events. Other small differences arise from the piece-wise linear interpolation of ages between placement of the dated levels in the optimized sequences of origination and extinction events.
Species richness shows a steady increase through the Wuchiapingian to early Changhsingian, followed by a rapid rise in the middle and late Changhsingian. Early Triassic diversity is much lower than in the Late Permian in all measures (Fig. 2), with the drop in species richness beginning about 500 Kyr before the PTB to background levels for the Wuchiapingian and early Changhsingian. This is followed by a rapid extinction beginning at about 252.3 Ma (slightly before the PTB) during which 62% of the 408 species disappeared in less than 200 Kyr. When projected onto the Meishan section D, the extinction starts just below Bed 25 (Fig. 3), dated at 252.28 ± 0.08 Ma, and continued no later than Bed 28 dated at 252.10 ± 0.06 Ma (Shen et al. 2011). This result sharpens the suddenness of the extinction from the single Meishan section (Jin et al. 2000) at a wider geographic scope. Increasing the size of the data set can narrow the extinction interval by improving the resolution of range ends and adding taxa not present at Meishan. The decrease in diversity during the Early Triassic indicates continuing environmental deterioration for at least 300 Kyr and possibly longer (Fig. 2).
Before the mass extinction, background extinction during the late Guadalupian to the latest Changhsingian (about 10 Myr) removed 865 of 1033 species. The taxon turnover rate was low and stable during the Wuchiapingian and through the Changhsingian until the very late Permian (Fig. 4). Similarly, the taxon diversification rate has very low fluctuations, but is similar to the turnover rate. The turnover rate starts to increase ca. 253.5 Ma, indicating an origination and extinction rate higher than the background. However, the taxon diversity goes up at the same time (Fig. 2). As will be demonstrated later, the high species richness in the latest Changhsingian is largely related to sampling effort, but it does reveal some points of interest. The combination of high diversity and high origination and extinction rates requires that average lineage durations must have been shorter than they were previously. A similar pattern has been observed in foraminifera during the Late Paleozoic Ice Age, when they suffered from glacioeustatically induced variability in neritic environments (Groves and Lee 2008). Our data set shows the decreasing tendency of mean longevity at the turning point of the diversity (Fig. 4). This pattern could indicate that the biota was under higher stress from a deteriorating environment nearly 1.2 Myr before the extinction episode. Consistent with this view, the values of isorenieratene and aryl isoprenoids were high in Beds 11–14 (Cao et al. 2009: Fig. 3, columns D and E). These beds are tuffaceous limestone intercalated with black laminae, deposited during an interval of marine highstand (Zhang et al. 1996; Cao and Zheng 2009). A subtle negative excursion can also be seen in the δ13Ccarb record (Cao et al. 2009: Fig. 2, column B). These values are interpreted to reflect elevated sulfide concentrations within the photic zone. Thus environmental deterioration probably began as much as 1.2 Myr prior to the peak extinction, and is correlated with maximum transgression as well as intensity of volcanism. Further study of this interval will be needed to identify the nature and duration of the environmental deterioration and its effect on biotic diversity at a global scale.
The decline in the diversification rate began at 252.5 Ma (Fig. 4). The turnover rate then accelerated, reaching a peak at 252.2 Ma. This was followed by a sharp drop in the extinction rate until 252.12 Ma as well as a rebound of the diversification rate. The combination of these two curves further demonstrates the suddenness of the end-Permian mass extinction (Shen et al. 2011).
To investigate the possibility that extinction rates and the duration of extinction may have differed in high-latitude regions from those in tropical regions (Shen and Shi 2002; Algeo et al. 2012), we examined the latitudinal structure of extinction within our data. In Figure 5, we compare extinction patterns from South China with those of the northern peri-Gondwanan region. In general, Late Permian diversity was greater in South China than in the northern peri-Gondwanan region. However, the rapid single mass extinction is apparent in both regions and evidently began at the same time. In South China, 71% of the 304 species went extinct within 200 Kyr by the end of the Permian. Similarly, 67% of the 75 species disappeared within 60 Kyr in the northern peri-Gondwanan region. The extinction rate in the northern peri-Gondwanan region seems to be higher than in South China. Because the northern peri-Gondwanan regions are all shallow-water facies, such rapid extinction may indicate increased susceptibility of shallow-water biota to environmental conditions that led to the extinction.
Testing the Sampling Effect
Uneven sampling intensity is a potentially serious problem affecting fine-grained analysis of diversity patterns. Therefore, we considered the potential impact of higher collecting effort close to the extinction horizon, and applied rarefaction to the data. The relationship between sample size and raw richness can then be tested.
Sample Size and Raw Richness.—
As a first step in accounting for differences in sampling intensity, we compare the relationship between taxon richness and the number of sections investigated (Fig. 6). Most sections span the PTB, but the coverage deteriorates toward both ends of the study interval. Raw species richness varies with the number of sections. For the time-series estimates of diversity (Fig. 2), each species richness value was first calculated on a per-section basis and the y-axis values rescaled so that the per-section richness and raw richness have the same average values. This treatment does not remove the increased richness in the latest Permian, which we found to result from greater within-section collection efficiency near the extinction level.
For both the Late Permian and Early Triassic, observed richness increases with the number of sections that span the moment of richness estimation. The relationship is strikingly different between the Permian and Triassic. Richness varies less strongly with the number of sections sampled in the Triassic. In both the pre- and post-extinction intervals, the sampling intensity, as measured by the number of sections, decreases away from the mass extinction interval. In the mass extinction interval, raw richness falls abruptly without any significant change in sampling intensity. The difference between pre- and post-extinction dependence of richness on sampling illustrates the need to perform rarefaction corrections to different levels in the Permian and Triassic intervals. The Permian can be effectively rarefied to an even sampling level that is higher than for the Triassic (see Fig. 7).
Rarefaction of the Richness Data.—
The rarefied diversity curve (Fig. 7) demonstrates the stable Late Permian diversity plateau and that the richness peak in the middle and late Changhsingian is a sampling artifact. The Late Permian richness spike disappears into the 95% confidence band once the longer interval is corrected for differences in sampling intensity. Similarly, the Early Triassic richness spike is also an artifact of uneven sampling. A careful check of the fossil list indicates that the composite Meishan section includes 50 species of 47 genera of spores and pollen from the lowermost 14 m of the Lower Triassic Yinkeng Formation, which largely accounts for the spike in the Early Triassic. Some ammonoid, bivalve (Fang 2004), small foraminifer (Song et al. 2009), and brachiopod (Shen et al. 2006b) species survived the extinction (Zhao et al. 1981; Sheng et al. 1984) but disappeared soon afterwards, accounting for the decreasing diversity during the earliest Triassic. Some new species, such as Claraia wangi and Hypophiceras sp., produced a trivial rise in diversity immediately above the PTB, but these new additions do not alter the general trend of decreasing diversity. The extinction rate remained high from the latest Permian into the earliest Triassic (Fig. 2).
The abrupt step at the mass extinction survives this harsh degradation of the data. The apparent two- or multi-stepped extinction (Yin et al. 2007; Song et al. 2009, 2013) similarly disappears below the 95% confidence threshold; only one significant step remains and this would be even more evident if the Permian data were rarefied to the low sampling level of the Triassic.
Discussion: a Single or a Multi-Phased Extinction?
The patterns of extinction at the close of the Permian have been the subject of considerable discussion and controversy. Numerical analysis of the fossil records at Meishan revealed a sudden extinction spanning less than 500 Kyr (Jin et al. 2000). Analyses from western Austria (Rampino et al. 2000), Iran (Angiolini et al. 2010), and East Greenland (Twitchett et al. 2001) are consistent with a single short-lived event. In contrast, a three-phase biotic extinction was proposed on the basis of investigations of tens of Permian–Triassic transitional sections in South China (Yang et al. 1991; Wu 1993; Yin and Tong 1998) (Fig. 8). These authors pointed out that the extinction surface was different from place to place, but was consistent with the lithological boundary that was widely observed in South China. They placed the major biotic extinction horizon a few decimeters to less than 4 m below the PTB (Wu 1993), which corresponds to the level between Beds 24d and 24e at Meishan, and the second and the third phase at the top of Bed 26 and base of Bed 29 respectively (Yin and Tong 1998).
This hypothesis was later modified to two pulses of extinction, one at Bed 25 and the other at Beds 28–29, based on the fossil record and the dramatic change of 2-methylhopane index at both beds 25–26 and 28–29 (Xie et al. 2005); the previously identified major extinction horizon (24e) was recognized as a prelude to the extinction (Yin et al. 2007). The position of the extinction horizon and the number of such horizons have differed among observers (Yang and Wang 2000; Fang 2004; Chen et al. 2005; Kaiho et al. 2006; Wu 2006; Feng et al. 2007; Song et al. 2009, 2013) (Fig. 8). The variations of these episodes are high, and very likely to be controlled by lithofacies dependence of the different fossil groups.
Fang (2004) claimed that bivalves displayed a two-phase extinction, with the main episodes between Beds 25 and 26 and a second by the base of Bed 28 (Fang 2004), but precise time controls on the two-phase extinction in different sections were not available. Two extinction phases for conodonts were placed at the base of Bed 27 and the base of Bed 28 (Wu 2006). Chen et al. (2005) suggested three survival phases for brachiopods, corresponding to Beds 25–26, 27, and 28–34, indicating a three-phase extinction, but only two extinction horizons were identified, with the main one at the base of Bed 25 and a second at the base of Bed 28. Most Permian brachiopods disappeared within beds 26 or 27. There are no stratigraphic data to support most disappearances of brachiopods at the top of Bed 27. Ammonoids experienced seven extinction episodes from the Late Permian to Early Triassic, with the largest one at “the terminal late-Changhsingian” followed by a subordinate one at the end of Changhsingian (in middle of Bed 27 at Meishan) (Yang and Wang 2000). The main extinction horizon was deduced to be at the base of Bed 25, on the basis of information from Yang and Wang (2000) that all Permian ammonoid genera except one were eliminated during the largest extinction; this was followed by the occurrences of early members of the Ophiceratidae. However, the two- or multiple-phase extinction pattern has rarely been tested in view of fossil incompleteness and collecting bias by analytical approach. The radiolarians at the Dongpan section show an earlier main extinction approximately correlative to the top of Bed 24d at Meishan and a second extinction at the base of Bed 25 (Feng et al. 2007). Both events are below the PTB and the interval between them is only 20 cm of Bed 24e. The foraminifers suffered the first extinction near the top of Bed 24e and the second phase at the top of Bed 27, with a recovery stage between them (Song et al. 2009); different results, however, were presented by Korchagin (2011). It is noteworthy that foraminifers reached high diversity in the limestones of both Bed 24 and Bed 27, but are very rare in the ash clay of Bed 25, and the black shales of Bed 26 and 29, with none in the ash clay of Bed 28 or the argillaceous limestone above. This shows a clear lithofacies control on fossil preservation. This facies control apparently also influences all other fossil groups. In addition, the distribution of bioclastic material seen in thin sections indicates that the extinction horizon is confined to a 12 mm thick interval with the top 19 mm below the top of Bed 24e (Kaiho et al. 2006), which places it at the base of Bed 24e.
The variations in apparent extinction phases and horizons demonstrate the extensive Signor-Lipps effects from single localities that record taxon richness in a particular community (alpha diversity). The Signor-Lipps effect smears abrupt extinctions into an interval of older strata as a result of incomplete and uneven recovery of fossils. The smeared range ends are not uniformly redistributed into older strata; they tend to concentrate (clot) at horizons that best preserve rare taxa. In single sections, therefore, false turnover peaks appear at horizons where sampling or preservation abruptly improves. This problem is mitigated by compositing multiple sections and further emphasizes the importance of statistical analysis in evaluating such claims.
In our test of the different data sets, the diversity patterns for the three South China basins are also analyzed separately. This analysis suggests minor differences in the timing of onset of extinction: 252.39 Ma in the Lower Yangtze Basin, 252.47 Ma in the Dian-Qian-Gui Basin, and 252.62 Ma in the Upper Yangtze Basin (Fig. 5). This variation could reflect different response time to environmental degradation, but it is more likely to reflect the different qualities of the data and time resolution, especially given the more intensive sampling in the Lower Yangtze Basin. The Signor-Lipps effect is evident from the less intensively studied fossil records of the Upper Yangtze Basin and the Dian-Qian-Gui Basin. Overall the agreement is remarkable and supports the rapid onset of a single extinction phase.
1. Multiple analyses of a data set comprising 1450 species from 18 fossiliferous sections across different paleoenvironmental settings, integrated with high-precision geochronologic data, demonstrate the rapidity of the extinction across South China and the northern peri-Gondwanan region. The statistical analysis of these data reveals a single catastrophic extinction event that occurred over a maximum interval of approximately 200 Kyr. The sudden extinction began at 252.30 Ma, just below Bed 25 at Meishan; its short duration points to a severe deterioration in ecosystem structure to a point where there was no immediate biotic recovery.
2. The taxon turnover rate and mean longevity indicated a possible earlier deterioration of the living condition for the biota nearly 1.2 Myr before the sudden extinction, but this needs testing further in different sections of different regions. During this interval, the biota displays a combination of high diversity and high turnover rate. The mean longevity of taxa declines toward the event.
The authors are grateful for the constructive suggestions and comments by three anonymous reviewers. This research was supported by National Natural Science Foundation of China and the National Basic Research Program of China (2011CB808905).
- Accepted 8 August 2013.