Calcium isotope constraints on the end-Permian mass extinction

Abstract

The end-Permian mass extinction horizon is marked by an abrupt shift in style of carbonate sedimentation and a negative excursion in the carbon isotope (δ¹³C) composition of carbonate minerals. Several extinction scenarios consistent with these observations have been put forward. Secular variation in the calcium isotope (δ^44/40Ca) composition of marine sediments provides a tool for distinguishing among these possibilities and thereby constraining the causes of mass extinction. Here we report δ^44/40Ca across the Permian-Triassic boundary from marine limestone in south China. The δ^44/40Ca exhibits a transient negative excursion of ∼0.3‰ over a few hundred thousand years or less, which we interpret to reflect a change in the global δ^44/40Ca composition of seawater. CO₂-driven ocean acidification best explains the coincidence of the δ^44/40Ca excursion with negative excursions in the δ¹³C of carbonates and organic matter and the preferential extinction of heavily calcified marine animals. Calcium isotope constraints on carbon cycle calculations suggest that the average δ¹³C of CO₂ released was heavier than -28‰ and more likely near -15‰; these values indicate a source containing substantial amounts of mantle- or carbonate-derived carbon. Collectively, the results point toward Siberian Trap volcanism as the trigger of mass extinction.

An abrupt shift in style of carbonate sedimentation occurs across the end-Permian extinction horizon. Microbialites and oolites overlie diverse, fossiliferous limestones of the latest Permian age in carbonate strata deposited across the tropical Tethys (1–8) and in the Panthalassa Ocean (8, 9) (Fig. S1). The mass extinction and facies shift are associated with a large negative excursion in the carbon isotope (δ¹³C) composition of carbonate minerals.

Geochemical and sedimentary observations have been used to support various causal mechanisms for the mass extinction. Three different scenarios have been put forth. First, the “Strangelove Ocean” scenario links mass extinction to collapse of the biological pump—the vertical separation in the water column of carbon fixation and respiration, which results from the sinking of organic matter out of the surface ocean (10). Under this scenario, an initial decrease in carbonate deposition would occur because of mixing of surface waters with CaCO₃-undersaturated deeper waters (assuming an oxygenated deep ocean); subsequent increase in alkalinity because of continental weathering would lead to enhanced carbonate deposition, explaining the deposition of microbialites and oolites (10). Second, the ocean overturn model proposes that extensive sulfate reduction in anoxic deep waters of the Permian oceans resulted in a buildup of carbonate alkalinity and hydrogen sulfide in deep water prior to the extinction event (11–14). Upwelling of these alkaline deep waters would have triggered carbonate precipitation on the shelves (2, 5, 15, 16) and caused mass extinction through combined stresses of hypercapnia, anoxia, and hydrogen sulfide poisoning (14, 15, 17). Third, the ocean acidification model proposes that massive release of ¹³C-depleted carbon from a reservoir in the crust (e.g., methane clathrates, coal, and magma) (18–22) acidified the ocean, reducing carbonate sedimentation and potentially leading to dissolution of carbonate sediments (8). Subsequently enhanced continental weathering and consequent delivery of carbonate alkalinity to the oceans would account for the widespread deposition of microbialites and oolites above the extinction horizon (8).

One avenue for distinguishing among these hypotheses lies in their differing implications for the global cycling of calcium, an element with a residence time of approximately 600–1,000 ky and uniform isotope composition in the modern oceans (23, 24). Isotopes of calcium are fractionated during the precipitation of calcium carbonate (25–27): ⁴⁰Ca is preferentially incorporated into the solid phase, leaving seawater enriched in ⁴⁴Ca at steady state relative to the delivery and burial fluxes (24, 28). Consequently, scenarios that require imbalances between the delivery and burial fluxes of calcium in the oceans should impart changes in the calcium isotope composition in the oceans and associated sediments.

We constructed a one-box, isotope mass-balance model of the global calcium cycle (see Methods and SI Discussion) to generate quantitative predictions for marine calcium isotopes and concentrations associated with the various Permian-Triassic (P/T) boundary scenarios. Fig. 1 illustrates the predictions for these scenarios as well as two alternative possibilities requiring a shift in calcium isotopes only. Both the Strangelove Ocean and acidification scenarios predict an initial decrease in the carbonate depositional flux followed by an increase in carbonate deposition to return the system to steady state, but the perturbation is potentially much larger under the acidification scenario. In contrast, under the ocean overturn scenario one would expect a positive excursion in δ^44/40Ca because of the stimulation of carbonate precipitation by upwelling of alkaline deep water (2, 5, 15). Oceanographic mechanisms allowing for the overturn scenario have been debated (29–31); here we consider its implications for calcium isotopes irrespective of its feasibility from an oceanographic standpoint. A permanent global shift from calcite- to aragonite-dominated carbonate deposition across the extinction horizon is capable of producing δ^44/40Ca variation similar to that predicted under the acidification scenario; it causes no change in the calcium concentration of the ocean. A global shift in the isotope composition of the calcium delivery flux within the range of likely values would cause a much smaller and more gradual change in δ^44/40Ca and no change in the marine calcium concentration (Fig. 1 and Fig. S2).

Fig. 1.

Output from calcium cycle model under scenarios proposed to explain the P/T boundary δ¹³C excursion and associated deposition of carbonate microbialites and oolites. (A) δ^44/40Ca in carbonate rocks. (B) Calcium concentration in seawater. The overturn and acidification scenarios are shown under an assumption of a 100-ky perturbation. Also shown are δ^44/40Ca predictions for a permanent increase by 50% in the proportion of aragonite in carbonate sediment and a 100-ky decrease in the δ^44/40Ca of the river calcium flux by 0.3‰. Modeled calcium concentrations do not change under these last two scenarios. Additional output and sensitivity tests are presented in Fig. S2.

To constrain changes in global calcium cycling across the P/T transition, we analyzed the δ^44/40Ca of the micritic fraction of limestone samples from a P/T boundary section at Dajiang, in Guizhou Province, China. P/T boundary strata at Dajiang were deposited above the storm wave base on the Great Bank of Guizhou (GBG), an isolated carbonate platform within the Nanpanjiang Basin (32). The Nanpanjiang Basin is a deep-marine embayment in the Yangtze Block, which was located at approximately 12°N in the eastern Tethys during Early Triassic time (33). The section contains more than 50 m of diverse, fossiliferous packstone and grainstone of the Upper Permian Wujiaping Formation, which contains fusulinid and nonfusulinid foraminifers, calcareous green and red algae, rugose corals, crinoids, brachiopods, calcareous sponges, and gastropods (5, 34). The Wujiaping Formation is overlain by a 15-m-thick thrombolitic microbialite deposited in the immediate aftermath of the mass extinction (Hindeodus parvus conodont zone) (35), which contains a low-diversity assemblage of foraminifers, gastropods, and bivalves with rare echinoderms and calcitic and phosphatic brachiopods (4, 5). The microbialite is overlain by 1–2 m of molluscan and brachiopod packstone with rare echinoderms. Above the thin packstone interval are 47 m of thinly bedded, poorly bioturbated micritic limestone. The Lower Triassic section continues with 95 m of dolomite and dolomitized ooid-bearing cryptalgal laminate overlain by 225 m of peritidal limestone cycles (5, 36). Carbonate sediments continued to accumulate on the GBG through Middle Triassic time, reaching a total thickness of nearly 2 km before the platform drowned early in the Late Triassic (32). The GBG was buried in siliciclastic sediments during Late Triassic time, reaching a maximum burial depth of 2.5–3 km (37). Representative P/T boundary facies are illustrated in Fig. S1.

Results

Fig. 2 illustrates the P/T δ^44/40Ca and δ¹³C records from the Dajiang section, along with age constraints from conodont biostratigraphy and geochronology and model output for the acidification scenario (our preferred scenario; see Discussion below) and a scenario involving a global shift from calcite- to aragonite-dominated carbonate deposition. The δ^44/40Ca averages approximately -0.6‰ in Permian strata. It decreases by ∼0.2‰ across the extinction horizon and by ∼0.3‰ from mean uppermost Permian values to the minimum in the basal Triassic (H. parvus zone), 8 m above the extinction horizon (Fig. 2). Although many individual pre- and postexcursion samples exhibit values within error of one another, the average value among samples within the excursion is significantly different from that for the preexcursion samples (Mann–Whitney test; p = 0.0006).

Fig. 2.

Lithostratigraphy and calcium and carbon isotope data from the Dajiang section. Error bars indicate one standard deviation between replicate measurements. Carbon isotope data from ref. 65. Conodont zones from refs. 66 and 67. A constant depositional rate of 160 m/My is assumed for the numerical time scale in order to facilitate comparison to model output; the assumed rate is compatible with constraints on the platform (37) and an age of 252.6 My for the extinction horizon (68). Model output for the acidification scenario and the change in mineralogy are shown. See Table S4 for raw data.

Several lines of evidence suggest that the δ^44/40Ca measurements are primary. The δ^44/40Ca shift coincides with the globally recognized negative δ¹³C excursion (38), and the δ^44/40Ca values are significantly correlated with the δ¹³C values (Pearson’s r = 0.62; p = 0.002) but not with indicators of precipitation rate or diagenesis such as Sr/Ca, Mn/Sr, or δ¹⁸O (Fig. S3). The preexcursion isotope composition (approximately -0.6‰ versus bulk Earth or +0.6‰ versus SRM-915a) is in agreement with Permian values from a Phanerozoic compilation of δ^44/40Ca values for carbonate rocks and well-preserved shell calcite (39). The light values in δ^44/40Ca continue from the microbialite facies into overlying micritic limestones before trending back toward preexcursion values (Fig. 2), suggesting that variation in δ^44/40Ca does not simply reflect differences among facies in either primary fractionation of calcium isotopes or later diagenetic alteration. Finally, similar to the δ¹³C of carbonate rocks, there is far more calcium in the mineral than the pore fluids, helping buffer the δ^44/40Ca of carbonate minerals against alteration during burial diagenesis (40).

Discussion

The P/T boundary δ^44/40Ca excursion is inconsistent in direction with the prediction for ocean overturn and is larger in magnitude than the excursion predicted under the Strangelove Ocean or through a change in the isotope composition of weathered calcium. The δ^44/40Ca excursion is consistent in direction, magnitude, and time scale with ocean acidification or a large shift in the proportional deposition of aragonite versus calcite. The δ^44/40Ca excursion could also result from a local (not global) change in isotope fractionation during carbonate precipitation or a change in the local δ^44/40Ca composition of seawater. Below, we consider the support for these local factors before addressing the global scenarios.

One possibility is that there is a temperature component to the δ^44/40Ca excursion. Calcium isotope fractionation decreases with increasing temperature (41). Because temperatures increased across the P/T boundary (42), the temperature effect tends in the opposite direction from the observed excursion, reducing its apparent magnitude. The amount of warming across the boundary is poorly constrained, but an increase of 5–10 °C (43, 44) could reduce the fractionation by as much as 0.1–0.2‰ (26, 41). Thus, the negative excursion in seawater δ^44/40Ca may be somewhat larger than reflected in the raw data.

An alternative mechanism for locally changing the δ^44/40Ca of carbonate minerals is through a change in the local precipitation rate. Sr/Ca can serve as a proxy for precipitation rate (41); it is not correlated with δ^44/40Ca in our samples (Fig. S3), suggesting that variation in precipitation rate does not explain the calcium isotope excursion. Moreover, the range of fractionation associated with typical modern carbonate precipitation rates is small (40 and references therein).

The isotope excursion could also reflect a change in the local, and not global, δ^44/40Ca composition of seawater. There is no evidence for a significant δ^44/40Ca gradient in the modern oceans (23), but stratigraphic variability in the isotope composition of ancient carbonates has been suggested to result from limited mixing between the open ocean and tectonic basins or epeiric seaways (45). The δ¹³C excursion on the GBG, however, has been observed in carbonate sections across the globe (46, 47) and appears to be representative in terms of relative change in isotope composition as well as in absolute value. Any oceanographic gradient in δ^44/40Ca would therefore need to have existed in the absence of a similar gradient in δ¹³C. Consequently, we view a changing local δ^44/40Ca gradient as unlikely.

A shift in mineralogy of carbonate sediment from calcite to aragonite could produce a negative excursion in δ^44/40Ca of carbonate rocks because aragonite is ∼0.6‰ lighter than calcite relative to source fluid (26). The observed δ^44/40Ca excursion of 0.3‰ would thus require that half of the sediment pool in the ocean switched from calcite to aragonite across the extinction horizon. If the shift in mineralogy was only local, then the isotope recovery would require a local return to calcite-dominated sedimentation; if it was global and permanent, then the excursion decays because the global ocean approached a new steady state with a heavier δ^44/40Ca of seawater because of the globally lighter δ^44/40Ca in carbonate sediments (Fig. S2). The global scenario is consistent with the greater relative abundance of aragonite- versus calcite-producing skeletal animals globally in the Lower Triassic (48), with petrographic evidence for an originally aragonitic mineralogy for Early Triassic ooids and carbonate crystal fans (5), and with the similarity of Lower Triassic carbonate strata across the global tropics (8).

Finally, the excursion could reflect a decrease in the global δ^44/40Ca of seawater. Such a decrease is most plausibly generated by ocean acidification (Fig. 1), which would reduce the carbonate ion concentration and carbonate saturation level of seawater and therefore allow calcium to accumulate. The calcium delivery flux is coupled to the dominant burial flux of carbonate sedimentation (F_carb) through the carbonate saturation state of seawater:

[1]

where k is a scaling constant, η is the order of the reaction [approximately 2 (refs. 10 and 49)], Ω is the saturation state with respect to calcite or aragonite (), and K_sp is the solubility product of the relevant carbonate mineral (50). Therefore, the effect of elevated calcium delivery via rivers, for example, is mitigated by an increase in Ω, which causes F_carb to increase concomitantly. An increase in seawater [Ca²⁺] is most likely to result from decreases in k and/or . The loss of skeletal carbonate sinks following the end-Permian mass extinction (34) may have caused k to decrease because skeletal animals and algae had been expending metabolic energy to drive carbonate precipitation prior to the extinction. A decrease in and an increase in the calcium river flux are expected under scenarios of CO₂ or methane release previously proposed to account for the negative excursion in δ¹³C (51–53): Because CO₂ is a weak acid (and methane oxidizes to CO₂ within years), rapid CO₂ input would lower ocean pH and , reduce Ω and F_carb, and thereby allow calcium to accumulate in seawater while simultaneously warming the climate and accelerating the weathering of calcium-bearing rocks. If acidification caused the oceans to become undersaturated with respect to calcite, calcium could also have been released into seawater through dissolution of marine carbonate sediments.

The magnitude of excess calcium delivery required to explain our measured δ^44/40Ca excursion under this acidification scenario can be calculated by mass balance, assuming a known isotopic composition of the calcium input and the starting size of the marine calcium reservoir. Because the mass extinction event occurred over less than 500 ky and likely less than 150 ky (54), we can approximate the event as instantaneous (see SI Discussion, Tables S1 and S2, and Fig. S4). Assuming that the input δ^44/40Ca is isotopically similar to the river calcium flux and that the ocean δ^44/40Ca was 1.4‰ heavier than our measured preextinction δ^44/40Ca (i.e., fractionation similar to modern), we calculate an increase in seawater [Ca²⁺] of 21% (2.15 mM) for a 0.2‰ excursion or 36% (3.6 mM) for a 0.3‰ excursion (see SI Discussion). Here we have ignored the potential effects of changing temperature and mineralogy in order to investigate the effects of acidification alone. Because expected effects of temperature and mineralogy work in opposing directions (given increases in temperature and prevalence of aragonite) and the likely magnitudes of their effects are similar, in aggregate they may have mostly offset one another.

We favor the ocean acidification scenario over a change in sediment mineralogy as the explanation for the calcium isotope excursion because it also explains a wide range of geochemical, sedimentological, and paleontological observations associated with the P/T boundary not predicted by a shift in carbonate mineralogy alone. First, the synchronous deposition of microbialites and oolites across the global tropics within the H. parvus conodont zone (7, 8) could reflect rapid carbonate deposition resulting from enhanced silicate and carbonate weathering in the aftermath of a CO₂ release event. The weathering hypothesis can explain why synchronous, widespread deposition of these facies is confined to the earliest Triassic, despite the persistence of ocean anoxia (55) and skeleton-poor carbonate sediments (34) through much of Early Triassic time. Moreover, removing the calculated 2.2–3.6 × 10¹⁸ mol of calcium from the oceans as calcium carbonate would translated to roughly 8–13 m of carbonate sediment on the 10⁷ km² of Late Permian carbonate shelves (56), which is broadly consistent with the thickness of the microbialite and oolite deposits within the H. parvus zone (Table S3). Second, acidification can explain observations of submarine carbonate dissolution of beds immediately underlying the microbialite (8). Third, rapid CO₂ release can account for the negative excursion in δ¹³C at the P/T boundary (46). Finally, ocean acidification can account for the preferential extinction of heavily calcified marine animals with limited ability to buffer calcifying fluids against changes in ambient water chemistry (17).

If the primary acid added to the oceans was CO₂, then the δ^44/40Ca excursion can provide first-order constraints on the magnitude and C-isotope composition of carbon released because each mole of CO₂ released would allow approximately 1 mol of calcium to be stored in the oceans (57). Assessing the source, magnitude, and isotope composition of carbon released at the P/T boundary has been challenging because the δ¹³C excursion can potentially result from a smaller release from a ¹³C-depleted source (e.g., biogenic methane) or a larger release from a less depleted reservoir (e.g., coal and basaltic magma). Berner (51) argued that only biogenic methane could account for the magnitude of the δ¹³C excursion, but more recent studies of Siberian Traps intrusions suggest that the magnitude of carbon release may have been larger than previously suspected (22), opening the possibility of carbon release from less depleted reservoirs. One can estimate the δ¹³C of the carbon source as a function of the initial total carbon given the estimated excess calcium delivery (2.2–3.6 × 10¹⁸ mol) and the excursion in (-3.6‰; cf. Fig. 2). The likely δ¹³C composition of the CO₂ added on the basis of the above calculation is between -5 and -13‰ (Fig. 3). This calculation assumes that all acidifying power comes from CO₂. However, if emplacement of the Siberian Traps played a significant role (22), then sulfur release could account for as much as half of the acidification [volatile fluxes from flood basalts suggest a molar ratio of CO₂ to SO₂ release of 2.5∶1 (58)]. Because H₂SO₄ can acidify the ocean without contributing to the decrease in δ¹³C, including its effects in the calculation requires a smaller, and thus isotopically lighter, carbon source. In this more conservative case, only 1.1–1.8 × 10¹⁸ mol of CO₂ release would account for the δ¹³C excursion and the δ¹³C of the CO₂ would be between -11 and -28‰ (Fig. 3). Even under this more conservative scenario, the δ¹³C of methane clathrates (-60‰) is far too negative to have been the primary source of carbon; the range of likely values instead points toward a mixture of carbon sources, which appear to have included a substantial component from carbonate rocks and/or mantle-derived volatiles.

Fig. 3.

Carbon isotope composition of carbon added in a P/T boundary perturbation as a function of the initial dissolved inorganic carbon content of the oceans, assuming a δ¹³C excursion from +3.6 to 0‰. Results are shown for calculated upper (3.6 × 10¹⁸ mol; 43,200 GT) and lower (1.1 × 10¹⁸ mol; 13,200 GT) bounds on the amount of added carbon and intermediate values discussed in the text (1.8 × 10¹⁸ mol; 2.2 × 10¹⁸ mol).

Intrusion of Siberian Traps magmas through a thick sequence of carbonate sediments and coal (22) coeval with mass extinction (59) provides a mechanism for releasing sufficient quantities of CO₂ with the appropriate δ¹³C composition through the production of thermogenic methane during the heating of coal (22) and CO₂-rich volatile phases during the incorporation of carbonate rocks into basaltic magma (60). Although previous studies have emphasized the release of carbon from methane clathrates or coal (51–53), δ^44/40Ca constraints instead suggest that carbon release associated with the end-Permian extinction may have included a substantial contribution from the incorporation of carbonate country rock into Siberian Traps magmas.

The ocean acidification scenario that we favor for the P/T boundary is similar to that proposed for the Paleoecene-Eocene Thermal Maximum (PETM), where there is abundant evidence of a global decrease in the δ¹³C and dissolution of deep-marine carbonate sediments (61) and possibly a negative excursion in δ^44/40Ca (62). Unlike the end-Permian event, however, the PETM is not associated with global mass extinction—only deep-sea benthic foraminifera were severely affected (63). Differences in the distribution of carbonate sediments between 250 and 55 million years ago may account for much of the contrast. At the PETM, as today, the oceans were buffered against acidification by extensive, fine-grained, unlithified carbonate sediments on the deep-sea floor, which could relatively rapidly dissolve to counter acidification (64). By contrast, the Late Permian deep sea contained no such carbonate buffer because Permian oceans lacked abundant pelagic carbonate producers such as coccolithophorids and planktonic foraminifers. Consequently, any buffering against acidification via dissolution of carbonate sediments could only have occurred more slowly in the less extensive, coarse-grained, mostly lithified, shallow-marine carbonate platform sediments or via chemical weathering of silicate and carbonate rocks on land. Thus, carbonate sediments may not only record critical information constraining the causes of end-Permian mass extinction; their spatial distribution in the oceans may also have been a key control on its severity.

Methods

Notation.

Calculation of δ^44/40Ca follows standard delta notation:

In this study, δ^44/40Ca values are referenced to a bulk Earth standard. To compare our numbers to those referenced to the SRM915a scale, our measured value of SRM915a (-1.2‰) can be added to the numbers reported in this work.

Sample Preparation.

Samples collected from the outcrop were cut on a rock saw and subsequently ground and polished on a vibrating lapidary tray. The polished face was then drilled with a 1-mm dental bur to produce powder for analysis. Areas of micrite were selected for drilling in samples exhibiting macroscopic variation in carbonate phases. Early and late diagenetic cements, especially fracture-filling vein calcite, were avoided.

Calcium Isotope Analysis.

Carbonates were dissolved in SEASTAR acetic acid. Calcium concentrations were determined, and 30-μg Ca aliquots were removed. A 42–48 calcium double spike was added to the aliquots and equilibrated in 5–8 mL of 2N nitric acid. The samples were loaded in 1N nitric acid onto ion separation columns filled with DOWEX® AG 50W-X8 resin. The samples were eluted in 2N nitric acid and dried down. These chemically separated, spiked samples were loaded onto Re filaments with phosphoric acid and measured on a multicollector thermal ionization mass spectrometer. Two replicates of each sample were run, and the standard deviation between replicate measurements is presented.

Trace Metal Analysis.

For trace metal analysis, ∼2 mg of dried, powdered limestone was dissolved in 1M acetic acid. After centrifugation, the supernatant was removed and subjected to trace metal analysis (Mn and Sr) as well as Ca and Mg analysis by using a Thermo Jarrell Ash IRIS Advantage/1000 Radial Inductively Coupled Argon Plasma Spectrometer with a solid state charge induction device detector.

Oxygen Isotope Analysis.

For oxygen isotope analysis, samples were reacted with concentrated phosphoric acid at 90 °C in a common acid bath and then measured on a dual inlet mass spectrometer (VG Optima). Isotopic composition is reported in standard delta notation relative to the Vienna Pee Dee belemnite (VPDB) standard. Analytical precision was ± 0.1‰ (2σ) on the basis of replicate measurements of a laboratory standard.

Geochemical Model.

Coupled models of the global calcium and carbon cycles were constructed assuming a one-box ocean and steady-state conditions prior to perturbation.