Margaret L. Fraiser, Matthew E. Clapham, and David J. Bottjer

Transcription

1 Chapter 16 Mass Extinctions and Changing Taphonomic Processes Fidelity of the Guadalupian, Lopingian, and Early Triassic Fossil Records Margaret L. Fraiser, Matthew E. Clapham, and David J. Bottjer 6 Contents 1 Introduction Previous Understanding of Biases in the Middle Permian to Early Triassic Fossil Record End-Guadalupian Extinction and Lopingian Aftermath End-Permian Mass Extinction and Early Triassic Aftermath Methods Results Guadalupian Lopingian Lazarus Effect Patterns in Permian Silicification Early Triassic Lazarus Effect Patterns in Early Triassic Silicification Conclusions References Abstract The biotic crisis of the Middle Permian through Early Triassic is unmatched in the Phanerozoic in terms of taxonomic diversity losses and paleoecological reorganization. However, the potential taphonomic bias from post-mortem diagenesis for this crucial time has not been evaluated. We assessed the quality of the fossil record during this interval by quantifying the number of Lazarus taxa using our own database, data available in the Paleobiology Database and previous compilations. We also quantitatively tested for paleoecological differences between silicified versus M.L. Fraiser ( ) Department of Geosciences, University of Wisconsin-Milwaukee, Milwaukee, WI 53203, USA mfraiser@uwm.edu M.E. Clapham Department of Earth and Planetary Sciences, University of California Santa Cruz, 1156 High Street, Santa Cruz, CA 95064, USA mclapham@es.ucsc.edu D.J. Bottjer Department of Earth Sciences, University of Southern California, Los Angeles, CA , USA dbottjer@usc.edu P.A. Allison and D.J. Bottjer (eds.), Taphonomy, Second Edition: Process and Bias Through Time, Topics in Geobiology 32, DOI / _16, Springer Science+Business Media B.V. 2010

2 M.L. Fraiser et al non-silicified faunas. Herein we report that there is no major taphonomic bias due to skeletal mineralogy or fossil preservation affecting the Middle and Late Permian fossil record, but that aragonite-shelled molluscs may exhibit a significant Lazarus effect during the Induan. We propose that a variety of mechanisms affected the fossil record of the Paleozoic/Mesozoic transition, including ocean chemistry, paleobiology of the examined groups, and human influences on taxonomic and sampling practices Introduction Mass extinctions are geologically short intervals of time when biodiversity losses are significantly elevated above background rates of extinction (e.g. Jablonski 1986a; Sepkoski 1986; Flessa 1990). They are a prominent feature of the fossil record and, along with the rise and fall of the three great evolutionary faunas, shaped the Phanerozoic biodiversity curve (Raup and Sepkoski 1982; Sepkoski 1981, 1984; Courtillot and Gaudemer 1996). Mass extinctions are also important agents of macroevolutionary change because they eliminate successful groups of organisms and create new evolutionary opportunities for previously minor groups (Gould and Calloway 1980; Jablonski 1986a, b, 2001, 2005; Raup 1986, 1994; Erwin 2001; Bambach et al. 2002). A complete understanding of the evolutionary role of a mass extinction must include more than just an analysis of the taxonomic crisis because the effects of mass extinctions extend beyond the biodiversity losses: the aftermaths may be as important as the extinctions themselves because the new ecological patterns arising from survivors that interact in new ways in less crowded ecological niches (Droser et al. 1997, 2000; Erwin 2001; Bambach et al. 2002; Jablonski 2001, 2002). Proper interpretation of the duration, magnitude, and causes of mass extinctions and the nature of the survival and recovery of organisms during their aftermaths is contingent upon accurate reconstruction of taxonomic and ecological changes. Artifacts of sampling methods or taxonomic practice can obscure the real trends (Sepkoski 1986; Flessa 1990), whereas taphonomic biases inherent in the geologic record, such as mode of organism preservation (Schubert et al. 1997), rock volume (e.g., Crampton et al. 2003), and preferential loss of organisms with aragonitic shell mineralogy (e.g. Cherns and Wright 2000) may also influence observed patterns. Such taphonomic biases may have obscured the true biotic patterns during the Permian Triassic extinction and its aftermath. These potentially confounding effects have been inferred from the abundance of Lazarus taxa taxa that temporarily disappear from the fossil record but reappear later unchanged (Flessa and Jablonski 1983) and from a decrease in preservation by silicification (Erwin and Pan 1996; Schubert et al. 1997; Twitchett 2001). Lazarus taxa may be an indicator of the quality of the fossil record if the phenomenon reflects a failure of certain organisms to be preserved through taphonomic effects such as the Signor Lipps effect, outcrop area bias, paleolatitudinal sampling bias, or reduced preservation quality (Signor and Lipps 1982; Allison and Briggs 1993; Erwin and Pan 1996;

3 16 Mass Extinctions and Changing Taphonomic Processes Smith and McGowan 2007). Lazarus taxa abundance may also be due to biological factors such as reduced population size (which may also affect the chance of sampling a taxon) or reduced geographic range and migration to refugia. (Jablonski 1986a,b; Kauffman and Harries 1996; Wignall and Benton 1999; Twitchett 2001; Rickards and Wright 2002). Taxonomic uncertainty can cause an apparent Lazarus effect (Wheeley and Twitchett 2005). Herein we test two aspects of the quality of the fossil record during the Guadalupian, Lopingian, and Early Triassic. The end-guadalupian and end-permian extinctions marked the end of the Paleozoic (Fig. 1) and heralded major changes in benthic marine ecology (e.g. Fraiser and Bottjer 2007; Clapham and Bottjer 2007a, b), but several studies have proposed that taphonomic processes make it difficult to extract real ecological patterns during these key intervals in evolutionary history (e.g. Erwin and Pan 1996; Twitchett 2001). First, we quantified the number of Lazarus taxa among several key taxonomic groups, as an increased number of Lazarus taxa may indicate reduced preservation quality. Second, a potential source of bias in the fossil record for this interval, changes in preservation via silicification, was tested by quantifying the proportion of silicified fossil collections, comparing the alpha diversity of silicified and non-silicified (preserved as molds and casts) collections, and assessing the number of taxa exclusive to silicified collections. Silicification is important because it allows fossils to be acid-etched and freed from calcareous matrix, often preserving very fine morphological details and improving ease of identification by taxonomists (e.g. Holdaway and Clayton 1982). It can also preserve a more faithful record of the original diversity and abundance within an assemblage (Cherns and Wright 2000; Wright et al. 2003, Butts and Briggs, this volume). Results of this test will reveal any temporal trends in silicification and the extent to which silicified faunas preserve a higher fidelity record. Together these tests document the Fig. 1 Geologic timescale of Middle Permian (Guadalupian), Late Permian (Lopingian), and Early Triassic stages. Ch = Changhsingian, Ind = Induan. The lower panel shows the per-capita extinction rates (Foote 2000) for rhynchonelliform brachiopods, bivalves, and gastropods in each stage based on data from Clapham et al. (2009) (Permian invertebrates), Chen et al. (2005) (Early Triassic brachiopods), Gastrobase (Early Triassic gastropods), and the Paleobiology Database (Early Triassic bivalves and gastropods). The per-capita extinction for rhynchonelliform brachiopods is undefined in the Induan because no genera cross both bottom and top boundaries of the stage

4 M.L. Fraiser et al taphonomic quality of the Permian Triassic fossil record in greater detail and elucidate the impact of temporal trends of taphonomic bias on the records of the end-guadalupian extinction, the end-permian extinction, and their aftermaths Previous Understanding of Biases in the Middle Permian to Early Triassic Fossil Record 2.1 End-Guadalupian Extinction and Lopingian Aftermath The end-guadalupian extinction, at the end of the Middle Permian (Guadalupian Series), was the first phase of the two-stage taxonomic crisis during the Permian Triassic interval (Jin et al. 1994; Stanley and Yang 1994). Initial estimates suggested that biodiversity loss during the end-guadalupian event was severe with as many as 55 60% of all marine invertebrate and fusulinid genera going extinct (Stanley and Yang 1994). However, more recent studies have shown that extinction rates were actually not elevated during the end-guadalupian interval among most marine invertebrate groups, including brachiopods, bivalves, and gastropods (Shen et al. 2006; Clapham et al. 2009). Nevertheless, the end-guadalupian extinction remains an especially severe event for fusulinids (Stanley and Yang 1994; Yang et al. 2004). The potential causes of the extinction are unclear, and there may not be a need to invoke serious perturbations given the negligible invertebrate extinctions. Environmental changes during the Guadalupian Lopingian interval include the Emeishan flood basalts (Wignall 2001), possible climate cooling (Isozaki et al. 2007), and the onset of deep-marine anoxia (Isozaki 1997). Despite the minimal effects on global invertebrate biodiversity, environmental stress during the Guadalupian Lopingian interval caused profound changes in the habitat distribution of bryozoans during the Lopingian (Powers and Bottjer 2007), shifts in the relative abundance of rhynchonelliform brachiopods and molluscs in offshore habitats (Clapham and Bottjer 2007a, b), and a dramatic reduction in the number and size of reefs in the Wuchiapingian (Weidlich 2002; Weidlich et al. 2003). The potential influences of taphonomic bias on the apparent severity of the end- Guadalupian extinction were first investigated by Stanley and Yang (1994). They applied three tests and concluded that the end-guadalupian extinction peak did not result from a poor Lopingian fossil record. The preferential extinction of large, complex fusulinid genera, the inconsistency between observed patterns of extinction and predicted Signor Lipps effects, and a strong excess of originations relative to extinctions in the Wuchiapingian and early Changhsingian all suggest that taphonomic biases only had minor effects on the end-guadalupian extinction (Stanley and Yang 1994). However, other taphonomic effects, including changes in the abundance of silicified fossil collections or reductions in preserved rock volume, may have exacerbated the severity of the end-guadalupian extinction without producing a spurious Signor Lipps effect or completely masking the Lopingian radiation. [AU1]

5 16 Mass Extinctions and Changing Taphonomic Processes 2.2 End-Permian Mass Extinction and Early Triassic Aftermath The end-permian mass extinction, approximately 252 million years ago, was the largest biotic crisis of the Phanerozoic (Bambach et al. 2004; Henderson 2005) with 78% of marine genera going extinct (Clapham et al. 2009). For up to 5 million years during the Early Triassic aftermath of the end-permian mass extinction, benthic marine paleocommunities were characterized by low biodiversity and low ecological complexity compared to pre-extinction Permian and later Triassic paleocommunities (e.g. Fraiser and Bottjer 2005b; Lehrmann et al. 2006). Macroevolutionary changes in benthic marine ecology, such as a shift from primarily non-motile organisms to self-mobile taxa and a switch from rhynchonelliform brachiopod-dominated to bivalve-dominated paleocommunities, were triggered by the end-permian mass extinction (Bambach et al. 2002; Wagner et al. 2006; Fraiser and Bottjer 2007). Sedimentological and geochemical evidence indicate that much of the latest Permian through the Early Triassic had an atmosphere with elevated CO 2 and low O 2, and an ocean rich in H 2 S and depleted in O 2 ; these conditions were ultimately linked to extensive volcanism and the supercontinent configuration of Pangea (e.g. Wignall and Twitchett 1996; Wignall 2001; Berner 2004; Grice et al. 2005; Huey and Ward 2005; Sephton et al. 2005). It has been reported that a large portion of Early Triassic taxa are Lazarus taxa (Batten 1973; Erwin and Pan 1996; Twitchett 2001). For example, there are estimates that 69% of gastropod genera are Lazarus genera during the Griesbachian (Erwin 1996), and that 90% of sponge families are Lazarus taxa during all stages of the Early Triassic (Twitchett 2001). Though the Early Triassic Lazarus phenomenon heretofore had been examined for gastropods and sponges only (e.g. Erwin 1996; Erwin and Pan 1996; Twitchett 2001, Wheeley and Twitchett 2005), it has been implied that the Lazarus effect was very large for all groups of skeletoned benthic marine invertebrates during the Early Triassic (e.g. Twitchett 2001; Erwin 2006). An absence of faunas preserved by silicification has been proposed as a major cause of the Early Triassic Lazarus phenomenon and for the apparent delayed biotic recovery following the end-permian mass extinction (Erwin 1996, 2006; Erwin and Pan 1996; Kidder and Erwin 2001). This hypothesis is based on studies indicating that silicified faunas have a higher fidelity of fossil preservation than non-silicified faunas preserved as casts and molds (Cherns and Wright 2000; Wright et al. 2003). Furthermore, it has been proposed that the post-paleozoic fossil record suffers from a taphonomic megabias because of low numbers of silicified faunas compared to the Paleozoic (Schubert et al. 1997). Previous studies of the fidelity of the fossil record following the end-permian mass extinction have focused on only one group of benthic marine organisms (e.g. gastropods, Erwin and Pan 1996; or echinoids, Smith 2007), or have examined data from the Triassic period as a whole (Smith 2007), obscuring any processes that may have been unique to the aftermath of the end-permian mass extinction

6 M.L. Fraiser et al The extent of silicification during the Early Triassic has not been quantified previously, and the characteristics of silicified faunas have not been statistically compared to those of non-silicified ones Methods We compiled a database of Roadian (Middle Permian) through Anisian (Middle Triassic) rhynchonelliform brachiopod, bivalve, gastropod, and demosponge fossil occurrences. This was used to examine two additional taphonomic metrics that test the fidelity of the Permian Triassic fossil record and its potential influence on the end-guadalupian and end-permian extinctions. The dataset includes (1) more than 53,321 Permian marine invertebrate fossil occurrences, including records of all marine invertebrate groups, from 9863 collections (the database used in Clapham et al. 2009); (2) Triassic gastropods modified and updated from Gastrobase, a database of published occurrences of gastropod genera at the stage and substage levels for the Permian and Triassic periods ( (3) Triassic bivalves and sponges, and Anisian brachiopods, from the Paleobiology Database ( and (4) Early Triassic rhynchonelliform brachiopods (Chen et al. 2005). Though the PBDB is not flawless, it is the most complete database available for comparing benthic marine organisms from the Lopingian, Early Triassic, and Middle Triassic. Taxonomic assignments were corrected when necessary and possible. First, the number of rhynchonelliform brachiopod, bivalve, gastropod, and sponge Lazarus taxa in each stage from the Roadian to Anisian was quantified to test for poor preservation, especially in the Wuchiapingian stage immediately following the traditional end-guadalupian extinction interval and the Induan and Olenekian stages following the end-permian extinction (Appendix A). The significance of differences between the proportions of Lazarus taxa between stages was determined using a two-tailed t-test. Second, the number of Permian and Early Triassic silicified collections was tallied using the Clapham et al. (2009) database, 211 Paleobiology Database collections, and 358 additional Induan and Olenekian collections culled from the primary literature to determine whether a reduction in silicification, particularly due to the loss of the rich record from western North America, affected diversity and extinction (Appendix B). Included in the analyses were benthic marine invertebrates from level-bottom marine communities; planktonic, nektonic, and reef collections were excluded in the Triassic but not in the Permian data. Species richness (alpha diversity) of each silicified collection was determined and compared to the richness of non-silicified collections. The number of brachiopod, bivalve, and gastropod genera unique to silicified collections was also quantified to determine the influence of silicification on large-scale compilations of taxonomic diversity. [AU2]

7 16 Mass Extinctions and Changing Taphonomic Processes 4 Results 4.1 Guadalupian Lopingian Lazarus Effect During the Guadalupian and Lopingian, the prevalence of Lazarus taxa varied significantly among different taxonomic groups. At the genus level, a substantial percentage of gastropod taxa in a given stage, up to 38% of the total genus richness, are actually Lazarus taxa (Fig. 2a). In contrast, only 20 25% of all bivalve genera (Fig. 2b) and 18 20% of all rhynchonelliform brachiopods (Fig. 2c) are Lazarus taxa. Lazarus abundance is calculated by dividing the number of Lazarus taxa by the total diversity (Lazarus taxa plus taxa sampled within the stratigraphic interval) in each stage. Despite the pronounced difference between clades, the proportion of Lazarus taxa within most clades typically exhibits little variation (Fig. 2). There was no statistically significant change in the percentage of gastropod Lazarus taxa from the Roadian through Wuchiapingian stages (varying between 33.3% and 38.6%). Likewise, the number of bivalve Lazarus taxa remained statistically unchanged at % from the Roadian to the Wuchiapingian. Both aragonitic bivalves and those with a calcite shell layer (pterioids, pectinoids, and mytiloids) displayed statistically similar patterns and there is no systematic variation in the number of Lazarus genera between the two mineralogies, suggesting that the number of Lazarus taxa is most strongly controlled by the abundance of a group rather than its skeletal mineralogy. Lazarus taxa accounted for % of total rhynchonelliform brachiopod diversity in the Roadian Capitanian interval, with a significant decrease to 10% in the Wuchiapingian (Z = 3.03, p = 0.002). However, in notable contrast to the other groups, demosponges exhibit dramatic variation in the percentage of Lazarus taxa in a given stage (Fig. 2d). Lazarus genera account for 50.9% of all present or inferred sponges during the Wuchiapingian, but only 10.2% in the Wordian and 17.0% in the Changhsingian. Although there are few changes in the percentage of Lazarus taxa from the Roadian to Wuchiapingian, all investigated groups have substantially fewer Lazarus genera in the Changhsingian stage (Fig. 2). The percentage of gastropod Lazarus taxa decreased from more than 37% in the Wuchiapingian to only 18.8% in the Changhsingian (Z = 2.73, p = 0.006), bivalve Lazarus taxa decreased from 19.6% to only 4.9% (Z = 2.90, p = 0.003), rhynchonelliform brachiopods decreased from 10% to 0% in the Changhsingian (Z = 4.85, p < 0.001), and demosponges from 50.9% to 17.0% (Z = 3.69, p < 0.001). However, this dramatic reduction in the percentage of Lazarus taxa does not imply a pronounced increase in the quality of the fossil record or the fidelity of sampling during the Changhsingian. Rather, it reflects edge effects due to the severe taxonomic impact of the end-permian extinction. Because so many Permian genera became extinct (51% of gastropod genera, 65% of bivalves, and 96% of rhynchonelliform brachiopods, with the remaining brachiopods disappearing in the Griesbachian), the likelihood of Permian taxa occurring in the Triassic was greatly reduced and the latest Permian Changhsingian

8 M.L. Fraiser et al. a Genera Lazarus Genera (%) b Genera Lazarus Genera (%) c Aragonitic Calcitic Genera Lazarus Genera (%) Lazarus Genera (%) d Genera Roadian Wordian Capitanian Wuchiaping Changhsing Induan Olenekian Anisian Fig. 2 Total (within-bin and Lazarus) diversity and percentage of Lazarus taxa for gastropods (a), bivalves, with aragonitic and calcitic forms plotted separately (b), rhynchonelliform brachiopods (c), and sponges (d) in Middle and Late Permian and Early Triassic stages. Error bars indicate 95% confidence interval for Lazarus percentage

9 16 Mass Extinctions and Changing Taphonomic Processes Stage has anomalously low numbers of Lazarus taxa compared to more typical Permian values. For example, there are no rhynchonelliform brachiopod Lazarus genera in the Changhsingian stage due to the extreme severity of the end-permian extinction event. The striking stability in the percentage of rhynchonelliform brachiopod, bivalve, and gastropod genera represented by Lazarus taxa during the Roadian Wuchiapingian interval, and especially across the end-guadalupian extinction, implies that the quality of the benthic invertebrate fossil record remained consistent across the Guadalupian/Lopingian boundary. Demosponges may be an exception and the significant increase in Lazarus taxa across the end-guadalupian extinction, from 22.5% in the Capitanian to 50.9% in the Wuchiapingian (Z = 3.49, p < 0.001), could either reflect poor preservation of sponges or small sponge population sizes in the Wuchiapingian. The Wuchiapingian has few demosponge occurrences compared to the well-sampled surrounding intervals; only 103 generic occurrences of sponges compared to 1,513 in the Capitanian and 223 in the Changhsingian. In addition, the Wuchiapingian is a time of turnover or crisis in the reef ecosystem, and the number of preserved reefs is low compared to the Wordian, Capitanian, or Changhsingian (Weidlich 2002). Thus, the high number of sponge Lazarus taxa is primarily a result of actual decreases in population size rather than taphonomic biases due to poor preservation. The overall lack of substantial variation in the abundance of Lazarus taxa across the end-guadalupian extinction is consistent with the conclusions of Stanley and Yang (1994) that taphonomic biases did not substantially influence the observed pattern of extinction during the end- Guadalupian crisis Patterns in Permian Silicification Although variations in Lazarus taxa abundance are not consistent with major taphonomic biases during the Guadalupian Lopingian interval, shifts in the amount of silicification may have independently affected diversity patterns. Early diagenetic silicification can preserve a higher fidelity record of a fossil assemblage because of enhanced aragonite preservation (Cherns and Wright 2000; Wright et al. 2003, Butts and Briggs, this volume) and temporal variations in the amount of silicareplaced fossils have been argued to influence diversity patterns and extinction estimates (Schubert et al. 1997). To evaluate the potential effects of silicification on the end-guadalupian extinction, we calculated the percentage of collections with silica-replaced fossils in each Permian stage, quantified the difference in alpha diversity between silicified and non-silicified assemblages, and counted the number of genera that are uniquely found in silicified collections. Diagenetic silica replacement is thought to be a common phenomenon during much of the Permian, as exemplified by famous silicified localities from Thailand, the Salt Range in Pakistan, and especially from the Glass and Guadalupe Mountains

10 M.L. Fraiser et al in the United States, among others (e.g. Cooper and Grant 1972; Grant 1968, 1976). However, the number of silicified fossil collections in each stage is actually quite variable (Fig. 3); for example, 7.5% of the 1,280 Wordian collections contain silicareplaced fossils whereas 18.1% of the 1,444 collections in the Capitanian have been silicified. In contrast to the Guadalupian, silicification is much less widespread in the Lopingian. Only 3.3% of the 1,241 Wuchiapingian collections and 1.5% of the 983 Changhsingian collections have been silicified, suggesting that the substantial decline in the proportion of silicified fossils across the end-guadalupian boundary may contribute to apparent elevated extinction rates. Collections with silica-replaced fossils also have consistently higher sampled alpha diversity than non-silicified collections (Fig. 4). Note that overall mean alpha Silicified Collections (% of total) N = 455 N = 950 N = 1211 N = 1419 N = 880 N = 1280 N = 1444 N = 1241 N = 983 N = 162 N = 354 Assel Sak Art Kung Road Word Cap Wuch Chang Induan Olenek Fig. 3 Percentage of fossil collections containing silicified fossils in each Permian and Early Triassic stage. Assel: Asselian; Sak: Sakmarian; Art: Artinskian; Kung: Kungurian; Road: Roadian; Word: Wordian; Cap: Capitanian; Wuch: Wuchiapingian; Chang: Changhsingian; Ind: Induan; Ole: Olenekian. The n values indicate the total number of collections from each stage Silicified Collections Non-Silicified Collections Mean Species Richness Roadian Wordian Capitanian Wuchiapingian Changhsingian Induan Olenekian Fig. 4 Mean species richness for collections containing silicified fossils (solid line and square symbols) and non-silicified fossils (dashed line and open circle symbols) in Middle Permian, Late Permian, and Early Triassic stages. Error bars are 95% confidence intervals

11 16 Mass Extinctions and Changing Taphonomic Processes diversity values are a function of the nature of reporting in the published literature and are not representative of actual alpha diversity; many papers are taxonomic descriptions and only consider a single taxonomic group and record one or a few new species of interest from a given locality. In particular, the large discrepancy between silicified and non-silicified alpha diversity in the Middle Permian is primarily a result of the large taxonomic lists reported from the extraordinarily large silicified collections from west Texas. Nevertheless, apparent changes in sampled alpha diversity, whether real biological phenomena or due to changes in the number of taxa actually reported for a collection in published papers, still affect our perception of diversity and extinction in the fossil record. During the Roadian and Wordian, mean silicified alpha diversity is 24.4 species and 14.6 species per collection, compared to only 3.0 and 3.65 species in non-silicified collections from the same stages. The difference between silicified and non-silicified alpha diversity is statistically significant during the Guadalupian, but not in the Lopingian stages (4.5 vs 3.95 species in the Wuchiapingian, p = 0.51; 5.95 vs 4.4 species in the Changhsingian, p = 0.18). Although the difference in alpha diversity is not always statistically significant, the consistently higher values in silicified collections may have acted in conjunction with the significant decrease in the amount of silicification to exacerbate apparent diversity loss and increase calculated extinction rates. However, the major decrease in alpha diversity in silicified collections occurs from the Roadian to Capitanian stages (Fig. 4), earlier than the traditionally recognized end-guadalupian extinction. There is a minor but significant decrease in silicified alpha diversity across the Guadalupian/Lopingian boundary ( species; p = 0.05) but non-silicified alpha diversity actually increases significantly ( species, p = 0.02). Although there were substantial changes in the extent of silicification during the Permian, and silicification may preserve a better record of alpha diversity and relative abundance (Cherns and Wright 2000; Wright et al. 2003), it is not clear to what extent it affects global diversity patterns. If many genera are known exclusively from silicified collections, silica-replacement may exert an important control on global diversity patterns. In contrast, if most genera from silicified assemblages are also found in non-silicified assemblages, the implication is that silicification itself is not important for reconstructing diversity. During the Permian, the percentage of genera uniquely known from silicified assemblages in a given stage is influenced by the percentage of collections that are silicified, and can be as high as 23% of bivalves, 33% of brachiopods, and 60% of gastropods, all during the Roadian Stage. However, overall only a small number of genera are known only from silicified specimens, as many found in silicified collections from one stage are then recorded from non-silicified assemblages at another time. Only 3.2% of Permian bivalves (6 of 190 genera) are exclusive to silicified collections, while 17.4% of gastropods (31 of 178 genera, although several of those may be known from non-silicified collections in the Carboniferous) and 12.9% of brachiopods (94 of 727 genera) are uniquely found in silicified assemblages. Total genus richness is only 5 25% higher when silicified collections are included, compared to the value obtained solely from non-silicified fossils. However, calculated

12 M.L. Fraiser et al extinction and origination rates vary by no more than 5% if silicified collections are excluded. These results indicate that silicification itself is not necessary for preserving a good record of diversity during the Guadalupian Lopingian interval and that the severity of the end-guadalupian extinction is not biased by changes in silicification Early Triassic Lazarus Effect Only 18.8% of gastropod genera during the Changhsingian and 26.9% during the Anisian stage are Lazarus taxa, while 34.9% of Induan and 37.2% of Olenekian gastropod diversity are Lazarus taxa (Fig. 2a). The differences between the proportion of Lazarus gastropod genera from the Changhsingian to Induan is statistically significant (Z = 2.00; p = 0.045) but the other differences are not significant at p = The proportions of gastropod Lazarus taxa are also lower than those previously published (Erwin and Pan 1996). Though the proportions of gastropod Lazarus taxa during the Induan and Olenekian are similar to those of the Middle Permian (Fig. 2), the predicted proportion of Lazarus taxa in the Induan would be lower because of extinction edge effects, as in the Changhsingian when less than 20% of taxa were Lazarus genera. The proportion of bivalve Lazarus taxa was 22% in the Induan and 11.3% in the Olenekian. The differences in the proportion of bivalve taxa are not statistically significant between the Induan and Olenekian or from the Olenekian to Anisian, but the change from the Changhsingian to Induan, is (Z = 2.88, p = 0.004: Fig. 2b). Aragonitic bivalves had a higher proportion of Lazarus taxa compared to calcitic taxa during the Induan (40% versus 13.3%), but the difference is not significant due to the small sample size, especially of aragonitic taxa (Z = 1.82, p = 0.07). There was also little difference between aragonitic and calcitic mineralogy during the Olenekian. The proportions of Lazarus aragonitic genera were significantly different between the Changhsingian and Induan and the Induan and Olenekian (Z = 3.65, p < 0.001; Z = 2.01, p = 0.04). The proportions of calcitic Lazarus taxa did not differ significantly between the Changhsingian through Anisian stages. There are no rhynchonelliform brachiopod Lazarus genera during the Early Triassic stages. One hundred percent and 95.8% of demosponge genera were Lazarus genera during the Induan and the Olenekian, respectively, while many sponges remained known only as Lazarus taxa in the Anisian Controls on Early Triassic Lazarus Taxa Potential controls on the occurrence of Lazarus taxa during the Early Triassic include taphonomic processes, sampling, environmental conditions, paleobiology of the organisms, and taxonomic practices, or a combination thereof. The Early Triassic Lazarus phenomenon among bivalves and gastropods may have resulted from taphonomic bias related to their aragonitic composition.

13 16 Mass Extinctions and Changing Taphonomic Processes Aragonitic shells typically dissolve during meteoric or burial diagenetic processes during early diagenesis because aragonite is less stable than calcite at surface temperatures and pressures, even during aragonite seas (e.g. Tucker and Wright 1990). These diagenetic processes may have been compounded by increased acidity and CaCO 3 undersaturation of seawater caused by elevated atmospheric CO 2 levels during the Early Triassic (Berner 2004). Thus, the large proportion of aragonitic Lazarus taxa could reflect post-mortem taphonomic processes in a high CO 2 world. Similar processes have been proposed to explain Triassic Jurassic boundary patterns (e.g. Hautmann 2004; Hautmann et al. 2008a), and are observed in the modern ocean and predicted for the future (e.g. Feely et al. 2004). Alternatively, low levels and depth of bioturbation during the Early Triassic (e.g. Twitchett 1999; Pruss and Bottjer 2004; Fraiser and Bottjer in press) may have buffered some shell dissolution during the Early Triassic. Shell dissolution is high in areas with well-developed infaunal benthic communities because biogenic reworking of sediments increases oxygen levels in the mixed layer and promotes oxidative decay of organic matter, thereby increasing acidity near the sediment water interface and causing pore waters to become undersaturated with respect to both aragonite and calcite (Aller 1982; Walter and Burton 1990). Sediment reworking by infaunal organisms also disrupts mold space left after shells dissolve (Cherns and Wright 2000). Even if after death benthic calcareous shells were dissolved on the seafloor in some regions due to a lowered carbonate saturation state of seawater (Berner 2004; Feely et al. 2004), the reduction in bioturbating activity that characterized much of the Early Triassic may have prevented molds from being disturbed. Sample size, either as a result of actual sampling effort or of true population size, is another potential contributor to the Early Triassic Lazarus phenomenon. A good correlation between a group s abundance, the number of occurrences, and the number of Lazarus taxa can be observed in Middle Permian rhynchonelliform brachiopods, bivalves, and gastropods, confirming the importance of sampling on Lazarus abundance. Reduced Early Triassic sampling between two well-sampled stages tends to increase the number of Lazarus taxa. Very low levels of sampling in the Early Triassic (378 Induan occurrences and 673 Olenekian occurrences of rhynchonelliform brachiopods, gastropods, and bivalves) relative to the Changhsingian (4,755 occurrences) and Anisian (2,439 occurrences) may be influenced by sampling effort but more likely reflect reduced marine invertebrate abundance due to environmental stress following the end-permian mass extinction (e.g. Fraiser and Bottjer 2007). A specimen of a Griesbachian Lazarus gastropod taxon was found recently in a newly discovered Tethyan section (Wheeley and Twitchett 2005), further highlighting that low numbers of some gastropod taxa contributed to the Lazarus phenomenon (sensu Wignall and Benton 1999) and suggesting that more sampling of Lower Triassic strata, especially in largely ignored regions, could aid in finding missing gastropod taxa that migrated to refugia or that were low in abundance. Future work on physiological and ecological characteristics of Lazarus gastropod genera could determine the extent to which Early Triassic environmental conditions contributed to the low numbers of certain gastropod taxa. The prevalence

14 M.L. Fraiser et al of sponge Lazarus taxa is also likely tied to sampling; namely, the lack of reef sites during the Early Triassic (a metazoan reef gap, Flügel and Stanley 1984). Permian Triassic sponge occurrences are strongly covariant with times of widespread reef-building in the Wordian Capitanian, Changhsingian, and Late Triassic. Stages with low reef abundance between those reef episodes have many demosponge Lazarus taxa e.g. the Wuchiapingian (50.9%), Anisian (the beginning of the Triassic reef recovery, but still with 53.3% Lazarus taxa), and to an extreme degree the Induan and Olenekian. As the Early Triassic reef gap may have been due to elevated atmospheric CO 2 and ocean acidification that prevented metazoan reef organisms from forming skeletons (Stanley et al. 2007), extinction-related environmental factors may have contributed to the reduced sampling through reduced population size. Biological factors may have facilitated the Lazarus effect among some Early Triassic taxa. Most of the aragonitic Induan Lazarus genera had infaunal or semi-faunal lifestyles. During the end-triassic mass extinction, aragonitic infaunal bivalves suffered greater extinctions than epifaunal bivalves (Hautmann et al. 2008b), and it has been proposed that this pattern indicates a reduction in primary productivity as epifaunal bivalves have physiological characteristics that enabled them to fare better during conditions of reduced food availability (McRoberts and Newton 1995). A decrease in primary productivity has also been proposed for the Early Triassic (Twitchett 2001). It is unclear whether the Lazarus pattern among Early Triassic bivalves resulted more from diagenetic processes or from biological reasons, but both mechanisms were linked to Early Triassic environmental conditions (elevated CO 2 ). Furthermore, the reason that the Lazarus effect among aragonitic bivalve taxa is more pronounced in Induan versus Olenekian age strata is unknown. However, this temporal pattern supports the argument for environmental conditions contributing to the Lazarus effect because it could reflect an amelioration of some aspect of the global environment later in the aftermath. Poor taxonomic practice is a plausible hypothesis to part explain the Early Triassic Lazarus phenomenon among some groups. For example, partial preservation has made it difficult to definitively identify some Early Triassic gastropod taxa (Wheeley and Twitchett 2005). The small size of Early Triassic gastropods could make it difficult to determine gastropod taxonomy; many Early Triassic gastropods are microgastropods <1cm in height (Fraiser and Bottjer 2004; Fraiser et al. 2005), and needles have been required to prepare them to expose areas for proper identification (Batten and Stokes 1986). Some Middle Triassic gastropods have been incorrectly identified as Elvis taxa (Wheeley and Twitchett 2005), taxa that were misidentified as having re-emerged after their presumed extinction, but are not actually descendants of the original taxa (Erwin and Droser 1993). More accurate gastropod taxonomy across the P/T boundary and into the Middle Triassic could determine the extent to which the Lazarus effect among gastropods is real and significant.

15 16 Mass Extinctions and Changing Taphonomic Processes 4.4 Patterns in Early Triassic Silicification Contrary to previous reports that shell replacement by silica is absent among Lower Triassic faunas (e.g. Erwin and Pan 1996; Kidder and Erwin 2001; Twitchett 2001), 5% of Early Triassic collections contain fossils preserved via silicification (Fig. 3). Silicified Early Triassic faunas have been reported from Oman, China, and the U.S.A. Although this value likely represents a maximum estimate due to the easilyaccessible literature on silicified faunas, the value is broadly similar to the proportion of silicification in Permian stages, in the Lopingian in particular. Of the Early Triassic benthic collections with preservation via silica replacement, 30% are from Induan and 70% are from Olenekian age strata (Fig. 3). Olenekian U.S.A. collections comprise 70% of the silicified collections. The mean alpha diversity of PBDB Early Triassic collections preserved via silicification is 7.86 for Induan collections, 2.63 for Olenekian collections, and 4.19 for the series. An independent groups t-test of means indicates that the difference between the mean alpha diversities of silicified Induan and Olenekian collections is significant (t = 5.20, p < ). Non-silicified Induan collections have a mean alpha diversity of 3.18, and Olenekian ones have a mean alpha diversity of 4.14 (Fig. 4). The mean alpha diversity for all Early Triassic collections preserved as casts and molds is The difference between the means for non-silicified Induan and Olenekian collections is statistically significant (t = 2.50, p = 0.013). The difference between means of silicified and non-silicified Induan collections is statistically significant (t = 4.87, p < ), but there is no significant difference between the means for silicified and non-silicified Olenekian collections (t = 1.49, p = 0.14), or between the means for silicified and non-silicified Early Triassic collections (t = 0.696, p = 0.49). Fourteen gastropod genera (Ananias, Anomphalus, Bellerophon, Chartronella, Coelostylina, Donaldina, Jiangxispira, Laxella, Naticopsis, Omphaloptycha, Platyzona, Streptacis, Strobeus, and Worthenia) are found in Induan silicified collections, but all but two of those (Jiangxispira, Laxella) also occur in non-silicified collections at another time. No Olenekian gastropods are known exclusively from silicified faunas. No Induan bivalve genera are known exclusively from silicified faunas. At least eight bivalve genera (Elegantinia, Entolioides, Eumorphotis, Leptochondria, Neoschizodus, Pegmavalvula, Pleuronectites, Placunopsis) are known from Olenekian silicified collections, but Pegmavalvula is the only one of 46 Olenekian bivalves exclusively found in silicified localities throughout its entire range. No rhynchonelliform brachiopods are known only from shells preserved via silicification. Therefore, most Early Triassic genera are known from fossil casts and molds. Documentation of silicified fossil collections from Lower Triassic strata refutes the hypothesis that there is a complete absence of silicified faunas during the Early Triassic (Twitchett 2001). Silicified Induan and Olenekian collections are rare in comparison to some Permian stages, but actually occur as frequently as

16 M.L. Fraiser et al. silicified collections in the Wuchiapingian or Changhsingian stages of the Late Permian. There is no change in silicification across the Permian Triassic boundary, indicating that changes in preservation style are not the primary contributor to the unusual Early Triassic record. The statistically significant differences between silicified Induan and Olenekian collections and between non-silicified Induan and Olenekian collections supports previous findings that diversity increased through the Early Triassic (e.g. Schubert and Bottjer 1995) and could be an indication of biotic recovery following the end- Permian mass extinction. That there is no statistically significant difference between silicified and non-silicified Olenekian collections does not support the hypothesis that silicified faunas record more information than non-silicified ones. Furthermore, Lazarus taxa have been found in non-silicified collections (Hautmann and Nützel 2004). The statistically significant difference between silicified and non-silicified Induan collections could reflect a real difference in the amount of data preserved in silicified and non-silicified collections. However, 22% of silicified faunas are from one recently discovered Griesbachian-age section in Oman, and these collections have a mean alpha diversity of 8.5 that skews the mean alpha diversity for Induan silicified faunas. When these collections are removed from the analysis, there is no statistically significant difference between the mean alpha diversities of silicified and non-silicified Induan collections. More sampling of Lower Triassic sections around the world would probably lead to the discovery of more silicified faunas. Only a very small percentage of gastropod, bivalve, or brachiopod genera are known exclusively from silicified collections, whether in the Permian or in the Early Triassic. The impact of silicification, or the lack thereof, on global diversity compilations is therefore minimal. Differences between silicified and non-silicified collections were likely influenced more by worker-introduced bias and not taphonomic processes. One source of bias could potentially be the focus of many Early Triassic studies. Publications of taxonomic lists from Lower Triassic strata commonly report only one higher taxonomic group, e.g., rhynchonelliform brachiopods (Perry and Chatterton 1979), ostracodes (Crasquin-Soleau and Kershaw 2005), or pectinoid bivalves (e.g. Newell and Boyd 1995), even when other taxa have been collected in the field (D.W. Boyd, pers. comm.). Furthermore, authors commonly report only higher taxa rather than detailed lists of genera or species. Incompletely reported taxonomic lists mean that the alpha diversities of both silicified and non-silicified Induan and Olenekian collections are likely underestimates. Bias also could be introduced by the methods in which data are collected (D.W. Boyd pers. comm.). If only silicified faunas are searched for and collected from the field for analysis, any data from fossils that are not silicified are omitted. It is important to note that the only known Smithian (early Olenekian) collections are from non-silicified collections. Though silicified shells are easily extracted from their encompassing matrix with weak acids (such as hydrochloric or acetic acid), acids also cause the dissolution of any non-silicified material in the sample. Incomplete and incorrectly entered data in the PBDB, i.e., preservation entered as silicification when fossils were not actually silicified, incorrect [AU3]

17 16 Mass Extinctions and Changing Taphonomic Processes and outdated taxonomy, and omitted publications, prevent many collections and taxa (e.g., sponges) from being downloaded when certain searches are performed. Outdated taxonomy, incomplete taxonomic lists, and under-sampling likely exert a stronger influence on understanding of Early Triassic ecology than taphonomic processes Conclusions 558 [AU4] A full accounting of the effects of taphonomic biases during mass extinction intervals is critical in any attempt to extract meaningful biological signals from the fossil record during a biotic crisis and its aftermath. Changes in the fidelity or dominant style of fossil preservation can have a substantial impact on the composition and diversity of marine fossil assemblages, and it has been proposed, based on an abundance of Lazarus taxa and a reduction in silicification, that such taphonomic changes irreparably bias the fossil record of the Permian Triassic mass extinction. A new tabulation of the proportion of Lazarus genera in the major Permian Triassic taxonomic groups suggests that there was no major bias or change in taphonomic style in the Late Permian but that aragonitic taxa (gastropods and some, primarily infaunal, bivalves) may have suffered from reduced preservation in the Induan. Comparisons between silicified and non-silicified faunas indicate there was little change in the amount of silicification across the Permian Triassic boundary; regardless, silicification is unlikely to be a major taphonomic bias in global compilations because few taxa are known exclusively from silicified collections. Documentation of Lazarus taxa does not necessarily indicate that the fossil record is biased. Indeed, taxon outages are a common phenomenon in the fossil record and are caused by a variety of mechanisms. Rickards and Wright (2002) suggest that the concept of a Lazarus taxon is not useful as a taphonomic indicator because it represents nothing more than a taxon s low abundance during a given interval. The Permian Triassic pattern of Lazarus taxa documented here is partially consistent with this concept; the Early Triassic Lazarus effect is a function of sampling effects, biological and environmental factors, and actual taphonomic degradation. Though more sampling and refined taxonomy will improve the reconstruction of the end-permian extinction and its unusual aftermath, our current understanding of the Early Triassic fossil record likely reflects a primary ecological signal (to the extent that any Paleozoic or Mesozoic fossil assemblage reflects a primary biological signal) (Fig. 5). Taphonomy remains an important factor that must be assessed in each fossil assemblage, but analysis of the Permian Triassic record of Lazarus taxa and silicification demonstrates that the fossil record of the end-permian extinction and the Early Triassic aftermath is not completely obscured by a taphonomic megabias due to skeletal mineralogy or fossil preservation. Instead of being solely an indication of the poor quality of the fossil record (e.g., Twitchett 2001; Smith 2007), Lazarus taxa could also provide clues about the environmental conditions during deposition

18 M.L. Fraiser et al. Fig. 5 Preservation of Early Triassic fossils. (a) Silicified microgastropods, Virgin Limestone Member, Moenkopi Formation. (b) Internal molds of microgastropods, Campil Member, Werfen Formation. (c) Silicified Promyalina, upper member, Thaynes Formation. Scale in mm. From D. Boyd collection. (d) Internal molds of bivalves (Unionites), Siusi Member, Werfen Formation (Modified from Fraiser and Bottjer (2005a) 594 References Aller, R. C. (1982). Carbonate dissolution in nearshore terrigenous muds: the role of physical and biological reworking. Journal of Geology, 90, Allison, P. A., & Briggs, D. E. G. (1993). Paleolatitudinal sampling bias, Phanerozoic species diversity, and the end-permian extinction. Geology, 21, Bambach, R. K., Knoll, A. H., & Sepkoski, J. J., Jr. (2002). Anatomical and ecological constraints on Phanerozoic animal diversity in the marine realm. Proceedings of the National Academy of Sciences of the United States of America, 99, Bambach, R. K., Knoll, A. H., & Wang, S. C. (2004). Origination, extinction, and mass depletions of marine diversity. Paleobiology, 30, Batten, R. L. (1973). The vicissitudes of the gastropods during the interval of Guadalupian- Ladinian time. In A. Logan & L. V. Hills (Eds.), The Permian and Triassic systems and their mutual boundary (Vol. 2, pp ). Boulder: Canadian Society of Petroleum Geologists Memoir. [AU5]