An abrupt extinction in the Middle Permian (Capitanian) of the Boreal Realm (Spitsbergen) and its link to anoxia and acidification

The Middle Permian (Capitanian Stage) mass extinction is among the least understood of all mass extinction events; it is regarded as either one of the greatest of all Phanerozoic crises, ranking alongside the “Big 5” (Stanley and Yang, 1994; Bond et al., 2010a), or, in a fundamentally different appraisal, it is viewed not as a mass extinction but as a protracted and gradually attained low point in Permian diversity (Yang et al., 2000; Clapham et al., 2009; Groves and Wang, 2013). Until now, all detailed studies have focused on equatorial sections, especially those of South China, where the extinction was first recognized as an especially severe crisis. There, fusulinacean foraminifers and brachiopods lost 82% and 87% of species, respectively (Jin et al., 1994; Shen and Shi, 1996; Bond et al., 2010a). The South China marine losses coincided with the eruption of the Emeishan flood basalt province in the southwest of the country (Fig. 1; Wignall et al., 2009a), supporting a causal link between the two phenomena, but it is unclear if this regional crisis was also a global event. There are several purported kill mechanisms but no consensus, and there are also questions about the nature of the subsequent Late Permian postextinction interval (Bond et al., 2010a). Did marine biota recover from a Capitanian mass extinction (Bond and Wignall, 2009; Wignall et al., 2012) before their final coup-de-grâce in the latest Permian, or did they undergo gradual diversity decline in the run-up to the Permian-Triassic mass extinction, characterized by a slow transition to mixed faunas with Paleozoic and Mesozoic affinities (Clapham and Bottjer, 2007)?

The status and cause of the Capitanian crisis are based on evidence from tropical (Tethyan) latitudes, while little data exist from higher latitudes. Here, we redress this imbalance with a detailed evaluation of Boreal marine fossil ranges in the Kapp Starostin Formation of Spitsbergen, a Permian-aged mixed spiculite chert-carbonate unit that formed in cool, shelf seas of the Boreal Ocean (Ehrenberg et al., 2001; Stemmerik and Worsley, 2005; Blomeier et al., 2013; Dustira et al., 2013). We assess the role of redox changes during this interval using a combination of petrographic and geochemical approaches. The precise age of the formation is unclear, and we investigate this issue with a carbon and strontium isotope chemostratigraphic study and through comparison with better-dated East Greenland successions.

We logged and sampled the Kapp Starostin Formation over three field seasons at its eponymous type section near Festningen in central Spitsbergen, and three further sections at Van Keulenhamna, Eholmen, and Forkastningsdalen in southern Spitsbergen (Fig. 2). We examined the Kapp Starostin Formation in the field for its facies and fossil content, identifying the brachiopod and bivalves to genus or species level (see supplemental text: taxonomy¹), and we took several hundred samples for thin section and geochemical analyses.

The Kapp Starostin Formation consists of Kungurian to latest Permian strata that were deposited on an epicontinental shelf at the northeastern margin of Pangea (Blomeier et al., 2013). This was part of an extensive sea that stretched from Arctic Canada (Sverdrup Basin) and Greenland (Wandel Sea Basin) in the west, through Svalbard and the Barents Sea (Finnmark Platform, Stappen High), to Russia (Timan-Pechora Basin) in the east (Stemmerik and Worsley, 2005). The shelf was located at ∼45°N in the Middle to Late Permian. The Kapp Starostin Formation contains cool-water Boreal faunas that include abundant siliceous sponges, brachiopods, and bryozoans (e.g., Gobbett, 1964; Kobayashi, 1997; Shen and Shi, 2004). The formation attains its greatest thickness in central Spitsbergen, where it reaches >400 m thick, and thins southwards, eventually pinching out at the Sørkapp-Hornsund High (Dallmann, 1999). We logged and sampled the upper ∼220 m in the southern locations, and the upper 90 m at Kapp Starostin (Figs. 3 and 4; supplemental Fig. DR1 [see ^{footnote 1}]).

Tethyan δ¹³C records reveal distinctive, probably global, trends in the carbon isotopic composition of Middle to Late Permian seawater. The Capitanian extinction is associated with a −6‰ δ¹³C_carb excursion in Tethys that coincides with the onset of Emeishan large igneous province volcanism in the Jinogondolella altudaensis conodont zone (Fig. 1; Wignall et al., 2009a). A younger and smaller negative shift in δ¹³C_carb is also known from the Capitanian-Wuchiapingian Stage boundary (the Guadalupian-Lopingian Series boundary; Fig. 1) in China and Japan (Fig. 1; Wang et al., 2004; Isozaki et al., 2007). The Middle to Late Permian δ¹³C chemostratigraphic record can be compared with the ⁸⁷Sr/⁸⁶Sr record of Permian seawater that derives from brachiopod shells (McArthur et al., 2001; Korte et al., 2006). The ⁸⁷Sr/⁸⁶Sr values decline to a Phanerozoic low point of ∼0.70685 in the late Capitanian (Fig. 1; Korte et al., 2006). This has been attributed to minimal runoff of radiogenic Sr from the highly arid Pangean supercontinent (Gibbs et al., 1999). Together, the C and Sr isotopic records provide a chemostratigraphic scheme to help better constrain the Permian strata of Spitsbergen.

We generated three independent carbon isotope (δ¹³C_org) curves through the Kapp Starostin Formation (Fig. 3) according to the following methods: In Calgary (curve of S. Grasby), δ¹³C_org was measured using continuous flow–elemental analysis–isotope ratio mass spectrometry, with a Finnigan Mat Delta+XL mass spectrometer interfaced with a Costech 4010 elemental analyzer, with standards run every fifth sample. Combined analytical and sampling error for δ¹³C_org is ±0.2‰ (1σ). In Erlangen (curves of M. Joachimski and Y. Sun), δ¹³C_org was determined with an elemental analyzer (CE 1110) connected online to a ThermoFisher Delta V Plus mass spectrometer, and accuracy and reproducibility of the analyses were checked by replicate analyses of international (USGS 40) or laboratory standards. Reproducibility of replicate standard analyses was ±0.07‰ (1σ). Values are reported in conventional δ notation in permil relative to Vienna Peedee belemnite (V-PDB).

Our Boreal strontium isotope record was generated from brachiopod shells screened using cathodoluminescence (Fig. 3; supplemental Fig. DR2 [see ^{footnote 1}]). Powdered samples were screened on the University of Leeds Philips PW1050 X-ray diffractometer (XRD), which revealed the presence of only pure calcite and quartz. Atomic absorption spectroscopy measurements for Ca, Mg, Sr, Mn, and Fe revealed elemental abundances resembling carbonates from shells in equilibrium with modern ocean waters. Subsequent ⁸⁷Sr/⁸⁶Sr measurements were performed by I. Savov at the Thermal Ionization Mass Spectrometry Laboratory in Leeds. Carbonate powders underwent the leaching procedure described by McArthur et al. (2000). Evaporated Sr extracts were re-dissolved in ultrapure weak HCl and added onto tungsten wire with a previously applied 5% TaCl₅ solution (to ease ionization). The ⁸⁷Sr/⁸⁶Sr ratios were measured on a Thermo-Scientific Triton-series thermal ionization mass spectrometer. Internal precision was maintained between 5 and 13 × 10⁻⁶ (at 2σ). Analytical precision was monitored by repeated analysis of the standard NBS-987. During analysis, the mean measured value obtained for NBS-987 was 0.710272 ± 0.000011 (2σ, n = 17). All reported ⁸⁷Sr/⁸⁶Sr data were normalized to the NBS-987 literature value of 0.710248. Total blanks were ∼300 pg Sr. The concentrations of Rb in calcite shells were too low to require correction for radiogenic (⁸⁷Sr) in-growth. The ⁸⁷Sr/⁸⁶Sr ratios were normalized for mass fractionation to ⁸⁶Sr/⁸⁸Sr = 0.1194.

Pyrite framboid analysis is a powerful tool for reconstructing ancient redox conditions. In modern environments, pyrite framboids form in the narrow iron-reduction zone developed at the redox boundary, but they cease growing in the more intensely anoxic conditions of the underlying sulfate-reduction zone (Wilkin et al., 1996; Wilkin and Barnes, 1997; Suits and Wilkin, 1998). If bottom waters become euxinic (i.e., free H₂S occurs within the water column), then framboids develop in the water column but are unable to achieve diameters much larger than 5–6 μm before they sink below the iron reduction zone and cease to grow (Wilkin et al., 1996). Thus, euxinic conditions are characterized by populations of tiny framboids with a narrow size range. In contrast, dysoxic or weakly oxygenated seafloors produce framboid populations that are larger and more variable in size, because their growth is restricted only by the local availability of reactants (Wilkin et al., 1996; Bond and Wignall, 2010).

Field samples from Kapp Starostin and Van Keulenhamna were examined by P. Wignall using gold-coated polished chips viewed in backscatter mode at 2500× magnification under an FEI Quanta 650 FEG-ESEM scanning electron microscope to determine pyrite content, and, where present, the size distribution of pyrite framboids (Fig. 3). Where possible, at least 100 framboids were measured from each sample. Framboid size distributions are plotted with mean diameter versus standard deviation within each sample, which allows comparison of framboid populations in ancient sediments with modern euxinic, anoxic, and dysoxic populations.

The concentrations of redox-sensitive trace metals in marine sediment provide an additional measure of ancient depositional conditions, because the reduced forms of many metals (e.g., molybdenum, uranium, and vanadium) are insoluble in seawater, with the result that these elements are scavenged by sediment under anoxic conditions.

Determinations of U, Mo, and V concentrations were performed by S. Grasby at the Isotope Science Laboratory, University of Calgary, using the same samples that were taken for δ¹³C_org analysis. Elemental determinations were conducted on powdered samples digested in a 2:2:1:1 acid solution of H₂O-HF-HClO₄-HNO₃ and subsequently analyzed using a PerkinElmer mass spectrometer with ±2% analytical error. Our normalized trace metal data (Mo/Al, V/Al, and U/Th) confirm that fluctuations in trace metal concentrations were due to redox changes and not related to variations in carbonate content, because the major shift in values occurs within a spiculitic chert unit. We found excellent accordance between the redox history interpreted from pyrite framboid distributions and that inferred from trace metal concentrations.

Our three independent organic carbon isotope curves for the Kapp Starostin Formation are in close accord and reveal relatively high δ¹³C_org values in the central part of the Kapp Starostin Formation, before an ∼3‰ negative shift at a level marked by three, distinctive brachiopod-rich yellow limestone beds (hereafter called the “3YLB” level) beginning ∼44 m beneath the top of the formation at Kapp Starostin. The δ¹³C_org low point is followed by a gradual, rising trend (Fig. 3). A δ¹³C_org record is not available from Tethyan locations for comparison, but several δ¹³C_carb studies in the region show a similar (albeit often larger, at around 6‰) negative excursion that began in the middle Capitanian Jinogondolella altudaensis conodont zone (e.g., Bond et al., 2010b; Fig. 1). We note that some studies do not find this negative excursion, perhaps due to the presence of hiatuses (Jost et al., 2014).

The marine Permian ⁸⁷Sr/⁸⁶Sr record shows a decline to a low of 0.70685 in the middle to late Capitanian, followed by a rapid rise to values of 0.7075 at the Permian-Triassic boundary (McArthur et al., 2001; Korte et al., 2006; Kani et al., 2008, 2013; Fig. 1). Our record from Spitsbergen shows a similar low point, followed by a rising trend: We see values of 0.7069 around 90 m below the top of the Kapp Starostin Formation before a two-stage rise, first to ∼0.7073 some 55 m below the formation top, and more substantially to values of 0.7080 in the topmost meters (Fig. 3). This pattern is closely comparable to the most detailed previously published Sr record from the Kapp Starostin Formation of Gruszczynski et al. (1992). Our uppermost Kapp Starostin Formation values are radiogenically enriched in comparison to usual Late Permian values, suggesting that our youngest brachiopod samples have undergone diagenetic alteration and do not yield reliable data. Nevertheless, the ⁸⁷Sr/⁸⁶Sr low point recorded in the lower part of the section (90–55 m below the top of Kapp Starostin Formation) shows a good match to global data sets and provides some support for a middle Capitanian age for this level in the formation. This interpretation is in accordance with recent attempts at Sr dating of the Kapp Starostin Formation by Ehrenberg et al. (2010), who assigned a Roadian to Capitanian age to its upper half.

The combined δ¹³C_org and ⁸⁷Sr/⁸⁶Sr records both suggest a middle Capitanian age for a level high in the Kapp Starostin Formation near the 3YLB. Paleomagnetic evidence also supports this age assignment. The strata in western Isfjorden (Fig. 2) have undergone partial remagnetization, but the primary magnetostratigraphic signal shows a switch from normal to reversed polarities ∼40 m from the top of the formation—a change that suggests an end-Capitanian age (Nawrocki and Grabowski, 2000; Hounslow and Nawrocki, 2008). Taken together, the carbon isotope, strontium isotope, and paleomagnetic evidences all suggest that the 3YLB level in the Kapp Starostin Formation is of intra-Capitanian age. In the absence of biostratigraphically useful fossils within the formation, this age designation must remain a compelling but unproven hypothesis.

The age of the topmost 40 m of the Kapp Starostin Formation is also rather equivocal. Considered to be latest Permian by most authors, there is debate as to whether the latest Permian Changhsingian Stage is missing (Nawrocki and Grabowski, 2000; Hounslow and Nawrocki, 2008) or whether the entire Late Permian is complete, at least in the expanded sections of western Spitsbergen such as Kapp Starostin (Wignall et al., 1998).

The 3YLB level forms a distinctive, correlatable horizon in all of our studied locations, together with other brachiopod- and bryozoan-rich limestone beds present at lower levels (Fig. 4). This allows faunal ranges from individual sections to be compiled onto a single chart. We focused our analysis on brachiopods and bivalves of the uppermost two thirds of the Kapp Starostin Formation (supplemental information: Taxonomy [see ^{footnote 1}]; Figs. 5 and 6). This reveals an abundant and diverse brachiopod-dominated assemblage with stable diversity and composition throughout much of the Kapp Starostin Formation, as originally noted in the detailed study of Gobbett (1964). This stability was sharply terminated immediately above the 3YLB, when 87% (28 of 32) of brachiopod species and two out of four bivalve species disappear within a 30-cm-thick section of dark spiculitic chert (Figs. 3 and 7; supplemental Fig. DR1 [see ^{footnote 1}]).

The extinction removed many common Kapp Starostin Formation species (Figs. 3 and 7; supplemental Fig. DR1 [see ^{footnote 1}]), including Bruntonia maynci, Dyoros (Dyoros) spitzbergianus, Kuvelousia weyprechti, Paeckelmania sp., Paraspiriferina sp., Tumarina sp., Fasiculatia striatoplicatus, Krotovia sp., Anemonaria sp., Rhynchopora variabilis, Spiriferella spp., Svalbardoproductus sp., Chaoiella sp., Composita sp., Kochiproductus sp., Derbyia grandis, Thuleproductus sp., Cleiothyridinia sp., Bathymyonia sp., Cancrinella sp., and Dielasma cf. plica. Only Waagenoconcha sp., Liosetella hemispherica sp., Lingula freboldi, and Stenocisma sp. nov. survived the cull (Fig. 7).

Following the extinction, younger Permian fossils are found in a series of shale-dominated horizons interbedded with barren cherts (Figs. 3 and 7; supplemental Fig. DR1 [see ^{footnote 1}]). These shales are exposed at Kapp Starostin but covered in the other study sections. The youngest Kapp Starostin Formation fauna consists of a few pre-extinction and several new brachiopod species, which include Stenocisma sp. nov. 2, Spiriferella keilhavii, Dyoros. (D.) mucronata, an indet. echinoconchid, Lissochonetes superba, Rhynchopora kochi, and Cancrinella spitsbergiana, and new bivalve species including Grammatodon (Cosmetodon)? suzuki, Palaeolima sp., Atomodesma sp. 1, and Atomodesma sp. 2. Bivalves suffered a lower proportion of losses across the extinction horizon relative to brachiopods, and this group constitutes a greater proportion of postextinction fossil assemblages (7 bivalves out of 15 bivalve and brachiopod taxa present, in contrast to 8 out of 40 pre-extinction taxa). Bryozoans also are abundant at this level, although their diversity is considerably reduced compared with the fauna below the first extinction horizon (Ezaki et al., 1994).

This new postextinction fauna persists through only ∼10 m of strata before it disappears 10 m below the top of the Kapp Starostin Formation (Fig. 7). This level may occur in either the early Late Permian Wuchiapingian Stage or within the subsequent Changhsingian Stage. Further work is needed to better date this second extinction. A third extinction occurs at the top of the Kapp Starostin Formation, where siliceous sponges together with the soft-bodied infauna responsible for abundant trace fossils disappear. This final extinction is associated with a major negative carbon isotope excursion and is considered to be the end-Permian mass extinction (Wignall et al., 1998; Dustira et al., 2013).

Elemental concentrations of U, V, and Mo as well as normalized values (U/Th, Mo/Al, and V/Al) are presented in Figure 3. Concentrations of these redox-sensitive trace metals are low throughout the Kapp Starostin Formation but rise at the first extinction level and remain high in the succeeding barren interval (Fig. 3). The diverse recovery fauna coincides with a decline of Mo values, but other trace metal concentrations remain stable—perhaps a subtle indicator of improved ventilation and a function of the greater enrichment of U in dysoxic waters relative to Mo, which only becomes enriched in sulfidic waters (Zhou et al., 2012).

Pyrite framboid analysis supports the redox history inferred from the trace metal trends. In the Kapp Starostin section, larger framboids with a wider size distribution (greater standard deviation) reveal weakly dysoxic conditions prior to the extinction. This is followed by a reduction in mean framboid diameters (and standard deviation) at the extinction level, indicating euxinic conditions (Table 1; Figs. 3 and 8). Framboids remain small until near the top of the Kapp Starostin Formation at Kapp Starostin. In contrast, the Kapp Starostin Formation strata from southern Spitsbergen (e.g., Van Keulenhamna) contain few pyrite framboids, suggesting development of much weaker dysoxia in this area.

Our Spitsbergen study reveals two shelly invertebrate extinction levels in the upper Kapp Starostin Formation. Most losses occurred during the first extinction, which was especially severe for the dominant brachiopod fauna, of which 87% of species disappear within a few tens of centimeters of strata. This extinction level is considered to occur within the Capitanian Stage, based on the multiple lines of evidence discussed herein, and so it provides the first evidence for a mass extinction of this age in higher latitudes. It is noteworthy that the brachiopod losses are similar to those suffered by brachiopods in South China, where 87% of species were also lost (Shen and Shi, 1996). Published range data on bryozoans and rugose corals in the Kapp Starostin Formation are of low resolution (Ezaki and Kawamura, 1992; Ezaki et al., 1994; Nakrem, 1994), but their diversity is much reduced in the upper 40 m of the formation, suggesting that these groups also suffered a Capitanian extinction.

The occurrence of a Boreal mass extinction in the Capitanian is also supported by comparison with the published marine fossil record from East Greenland, which was situated ∼1500 km to the south of Spitsbergen during the Permian, although here too the age assignments are controversial. The youngest Permian strata in East Greenland consist of the Wegener Halvø Formation and the overlying Schuchert Dal Formation (Stemmerik, 2001). The Wegener Halvø Formation contains abundant and diverse brachiopods and bryozoans that share many taxa with the Kapp Starostin Formation (Dunbar, 1955). The majority of these taxa disappear at the top of the formation, and the overlying Schuchert Dal Formation fauna is both rare and of low diversity (Sørensen et al., 2007). However, there are reworked blocks of white limestone present in the lowest Triassic that have been eroded from a latest Permian unit not seen in outcrop (Dunbar, 1955). The white blocks contain a diverse and distinct fauna of brachiopods and bivalves that is remarkably similar to the distinctive postextinction assemblages of the Kapp Starostin Formation in Spitsbergen, and includes Rhynchopora kochi, Grammatodon spp., and atomodesmatids (Dunbar, 1955). Thus, the East Greenland fossils reveal a very similar history to those from Spitsbergen, with major diversity losses followed by recovery of a mixed brachiopod-bivalve assemblage (only found in reworked blocks in Greenland). The age of the boundary between the Wegener Halvø and Schuchert Dal Formations therefore becomes critical to determining whether these seemingly Boreal-wide extinction losses were contemporaneous across the region.

Conodonts extracted from the Ravnefjeld Formation (a black shale unit that is laterally equivalent to the Wegener Halvø Formation) include Neogondolella rosenkrantzi, Xaniognathus abstractus, and Merrillina divergens. Rasmussen et al. (1990) considered this assemblage to be no younger than earliest Capitanian in age, and equivalent to conodonts from the first cycle of the Zechstein Seaway infill of northern Europe. If this is the case, then the top Wegener Halvø level is much older than the Capitanian extinction in Spitsbergen. In contrast, some subsequent studies consider the conodonts (and the basal Zechstein rocks) to be much younger—Wuchiapingian in age—based on the presence of rosenkrantzi, now considered a Mesogondolella species (e.g., Nielsen and Shen, 2004; Sørensen et al., 2007). This revised age assignment places the top of the Wegener Halvø Formation in the Late Permian, at a level considerably younger than the Capitanian extinction. However, the most recent taxonomic reassessment suggests that the Zechstein Mesogondolella species is not rosenkrantzi but a younger, derived species (M. brittanica), and the true Mesogondolella rosenkrantzi of the Ravnefjeld Formation indicates an intra-Capitanian age (Legler and Schneider, 2008). This re-revised age assessment indicates that the brachiopod and bryozoan losses at the top of the Wegener Halvø Formation record a mid- to late Capitanian crisis that can be correlated with the extinction near the 3YLB level in Spitsbergen. This points to a Boreal-wide intra-Capitanian crisis.

There is a similar loss of marine benthic faunal diversity around the Capitanian-Wuchiapingian boundary in the Sverdrup Basin of Arctic Canada. In that location, faunal losses are associated with the near-total disappearance of shallow-marine carbonates (Beauchamp and Grasby, 2012). The low-diversity survivors, consisting of brachiopods and bryozoans, persisted into the younger Permian, only to disappear in the early Changhsingian (Beauchamp, 1994; Algeo et al., 2012). This left only siliceous sponges in the Sverdrup Basin, but they too soon disappeared (in the late Changhsingian Stage). Thus, the Sverdrup Basin record reveals a very similar three-step diversity decline to that seen in Spitsbergen, with the added benefit of conodont-age control for the younger extinctions. However, the post-Capitanian extinction recovery of brachiopods and bivalves seen in Spitsbergen is not seen in the Sverdrup Basin.

The role of marine anoxia in the Capitanian extinction has been put forward (e.g., Clapham and Bottjer, 2007; Clapham et al., 2009). While indeed Isozaki (1997) suggested that a 10-m.y.-long “superanoxic event” in the Panthalassa Ocean began at the Guadalupian-Lopingian boundary, the development of open-ocean anoxia is now known to postdate the intra-Capitanian extinction (Wignall et al., 2010).

There is clear evidence from our combined pyrite petrographic and trace metal study (Figs. 3 and 8) that the intra-Capitanian extinction at Kapp Starostin coincided with a deterioration of seafloor oxygenation, lending support to the anoxia kill hypothesis. However, this redox change is more muted in the shallower-water sections of southern Spitsbergen, and further afield there is little evidence for middle Capitanian anoxia in either the South China sections (Wignall et al., 2009b) or in the Panthalassan accreted terranes of Japan (Wignall et al., 2010). The role of anoxia in the wider Capitanian extinction scenario remains enigmatic, and although it might have played a part at Kapp Starostin, it was probably not the only killer—some other mechanism must have been responsible for losses in southern Spitsbergen where anoxia did not develop.

The extinction level also coincides with the disappearance of carbonates throughout Boreal seas. No limestones are seen above 3YLB in Spitsbergen, nor above the Wegener Halvø Formation in East Greenland, and neither are they developed in the Upper Permian of the Sverdrup Basin. This observation potentially supports ocean acidification as a cause of the Capitanian crisis, at least in higher-latitude regions (e.g., Beauchamp and Grasby, 2012). Due to the greater solubility of CO₂ in colder waters, ocean acidification stresses are more intense at higher latitudes. The observation that extinction losses are very similar amongst brachiopods in China and Spitsbergen would appear to rule out an extinction cause related to temperature alone.

High δ¹³C values in the early Capitanian have been interpreted as evidence for a cold climatic episode that drove the Capitanian crisis. Support for this idea came from the observation that only warm-water taxa such as fusulinacean foraminifers were affected (e.g., Isozaki et al., 2007; Kossovaya and Kropatcheva, 2013). However, our discovery of extinction amongst Boreal brachiopods suggests that cold adaptation was of no benefit. Furthermore, a prolonged cooling could be expected to result in the migration of Boreal species into the warmer waters of Panthalassa and Tethys, but this is not observed. Extreme cooling of the magnitude required to cause extinctions would likely result in glacio-eustatic regression, but in Spitsbergen, there is no facies change across the extinction horizon.

Our new data suggest that the Capitanian extinction was of global extent. In South China, the extinction closely coincides with the onset of Emeishan volcanism, suggesting that the deleterious effects of this large igneous province played a key role in the extinction scenario (Wignall et al., 2009a). Volcanically induced effects are multiple and include acidification (from CO₂ release). More subtle volcanogenic killers might include poisoning by trace metals. For example, mercury has been implicated in the Permian-Triassic extinction (Sanei et al., 2012). Short-term cooling caused by volcanic aerosols is another possibility (e.g., Bond et al., 2010b). A proxy for flood basalt volcanism at the level of extinction in Spitsbergen would help to strengthen any volcanism-extinction links.

Bivalves suffered modest losses during the first Spitsbergen Permian crisis, and the postextinction interval witnessed a marked increase in molluskan dominance of the shelly benthos, resulting in assemblages more typical of the Mesozoic. This supports the notion of Clapham and Bottjer (2007), who suggested that the transition from brachiopod- to bivalve-dominated benthos was under way before the start of the Triassic. The aftermath of the Capitanian extinction clearly witnessed some bivalve successes, but the second extinction event in Spitsbergen removed all shelly taxa at some point in the Late Permian.

Detailed sample collection from Spitsbergen reveals a major brachiopod extinction event from near the top of the Kapp Starostin Formation. This extinction predates the end-Permian mass extinction, because a subsequent recovery of brachiopods and especially bivalves is seen in the Late Permian. This postextinction fauna disappears 10 m below the top of the Kapp Starostin Formation and thus fails to survive until the end of the Permian. The end-Permian extinction in Spitsbergen only eliminated the remaining fauna of siliceous sponges and trace-making organisms. Thus, there are three extinction levels in the Kapp Starostin Formation and one interval of radiation after the oldest extinction.

A combination of Sr isotope, chemostratigraphic, and magnetostratigraphic data suggests that the first and most severe extinction was contemporary with the middle Capitanian mass extinction of lower (Tethyan) latitudes, although further improvements in age dating are needed. Comparison with the record in East Greenland suggests that there was also a major extinction of shelly invertebrates at the top of the Wegener Halvø Formation. Based on conodont dating, this event was likely contemporaneous with the Spitsbergen losses and points to a Boreal-wide crisis in the Middle Permian.

Scenarios involving marine anoxia during the Capitanian extinction are supported by pyrite framboids and trace metal enrichment data from our deepest-water section at Kapp Starostin. However, the extinction is equally severe in the shallower-water sections of southern Spitsbergen, where there is little evidence for anoxia. The crisis also saw the permanent loss of carbonate shelf facies in the region, possibly supporting a role for ocean acidification during the Capitanian crisis, at least in higher, cooler latitudes. The loss of the cool-water invertebrates in Boreal Seas reveals that the Capitanian crisis did not, as previously thought, simply affect tropical marine biota: It was a truly global mass extinction.

This project was supported by funding from the Research Executive Agency (Marie Curie Intra-European Fellowship FP7-PEOPLE-2011-IEF-300455 to D. Bond), and the Natural Environment Research Council (grants NE/J01799X/1 to D. Bond and NE/I015817/1 to P. Wignall). We thank the Norwegian Polar Institute, and in particular Jørn Dybdahl and Monica Sund, for logistical support in the field. Richard Newport at the University of Manchester assisted in cathodoluminescence microscopy. The following were valuable members of the field-work teams: Dan Collins, Tom Goode, Christian Scheibner, and Tom Wignall. We thank Tom Algeo for his constructive review, and Associate Editor Ken MacLeod and Editor Christian Koeberl for their valuable input. This article has been published under the terms of the CC-BY 3.0 license thanks to Gold Open Access funding from the University of Hull.