Figure 6. Results of geochemical box modeling showing changes in pyrite burial (Fpry), weathering fluxes (Fw) needed to reproduce seen CAS trends in Monitor Range for a 3mM SO4-2 scenario (panels A and B) and a 5mM SO4-2 scenario (panels C and D). Also shown are changes in sulfate concentrations resulting from the associated weathering and pyrite burial fluxes (panels B and D). Grey interval in all panels represents the HICE interval. For additional sensitivity tests, see supplemental materials.
Changes to any single model parameter were unable to reproduce the fall in δ34SCAS in the required timeframe (supplemental material). However, model runs assuming moderate decreases in pyrite burial in concert with an increase in weathering can generate the observed δ34SCAS drop (Fig. 6). A 50–75% reduction in pyrite burial (i.e. 50% Fpyr and 25% Fpyr of initial burial rate, respectively) combined with a 25­–60% increase in the weathering flux (i.e., 125% Fw and 160% Fw, respectively) produces a ~10‰ drop using an initial marine sulfate concentration of 5 mM. This scenario is the most parsimonious with the initiation of sea-level fall, which would reduce shelf area and thus the total aerial extent of pyrite burial, as well as increase the weathering inputs. This change in sea level is a result of changes in global climate and associated changes in thermohaline circulation at this time (see section 5.3 below) may have increased marine oxygenation of the Late Ordovician oceans, and further reduced global pyrite burial. These combined effects would both contribute to the observed δ34SCAS drop.
A starting marine sulfate concentration of 5 mM is consistent with previous estimates for Late Ordovician seawater (Horita et al., 2002; Jones and Fike, 2013). Model simulations starting with 3 mM marine sulfate concentrations require smaller changes in the weathering and pyrite burial fluxes (i.e., 25% reduction yields a value of 75% Fpyr) to simulate the negative excursion (Fig. 6A). However, we do not favor an initial marine sulfate concentration of 3 mM or less due since it is the low end of estimates based on fluid inclusions and previous Late Ordovician sulfur isotope modeling (Horita et al., 2002; Hammarlund et al., 2012; Jones and Fike, 2013). Furthermore, sensitivity tests with an initial 3 mM oceanic reservoir show changes in seawater sulfate sulfur isotopes that are faster than those documented from Upper Ordovician records in terms of reaching the minimum value and the later return to baseline. Additionally, our model places an upper constraint on Late Ordovician marine sulfate concentrations, as our simulations with initial values of 10 mM or greater cannot reproduce the observed sulfur isotope records unless unreasonable changes in weathering and pyrite burial fluxes are prescribed (supplemental material).
The observed negative δ34SCASperturbation requires a major decrease in pyrite burial, thus potentially requiring a reduction in the global extent of euxinic conditions (since pyrite burial is highly efficient under such conditions). As a thought experiment, if we assume that most pyrite is buried in euxinic settings, we can calculate a maximum estimate of euxinic seafloor area. This approach is an oversimplification since pyrite is also formed in reducing sediments overlain by oxic and anoxic non-sulfidic waters. However, since pyrite burial in these settings is less efficient, an even greater reduction in the area of the seafloor subject to reducing conditions is required. Initial pyrite burial flux for the late Katian required an Fpyr of 1.1 × 1018 mol of S/Myr, compared to the modern global rate of 0.67 × 1018 mol of S/Myr (Kurtz et al., 2003). The extent of euxinic conditions in the modern oceans is estimated at ~0.15% of the global seafloor (Reinhard et al., 2013), with reduced sulfur burial flux that would equate to ~3.1×1016 mol of S/Myr, most of which occurs in the Black Sea (Neretin et al., 2001). This reduced sulfur burial flux includes pyrite burial and any intermediate valence reduced S species, as well as organically bound S, and is thus a maximum estimate for reduced sulfur burial. Given this data from modern oceans and the estimated late Katian pyrite burial rates, we can estimate the extent of euxinic conditions in the late Katian at most was ~35× more than the modern, equating to an aerial extent of approximately 5.3% (35 × 0.15% of the modern). We estimate that a subsequent 75 to 50% reduction in pyrite burial—corresponding to the minimum δ34SCAS values—would reduce the maximum global estimate for the extent of euxinia to ~1.3 - 2.7% within the Hirnantian. We observe a shift back to heavier δ34SCAS­ values in the late Hirnantian-early Silurian, likely signaling a return to more reducing conditions with the oceans.
Due to the susceptibility of carbonate-associated sulfate data to be possibly compromised from various diagenetic processes (see above section 5.1) that can lead to some variability within the recorded δ34SCAS values, we performed a series of sensitivity tests allowing for some degree of overprinting of primary δ34SCAS­ values (Fig. S4–S6). These tests reveal that our fundamental conclusions of decreased seafloor euxinia do not change within a range of reasonable δ34SCAS variations and possible diagenetic overprinting, but simply affects estimated ranges of total seafloor euxinia. Through careful sample selection and laboratory treatment (i.e. preference of less permeable, low porosity micrite over pack/grainstone, careful extraction procedures to avoid pyrite oxidation) we have generated robust δ34SCAS­ datasets that are in good agreement with correlative and previously published δ34SCAS­ datasets (Jones and Fike, 2013; Present et al., 2015). However, it is important to acknowledge that diagenetic overprints are still possible but given the results of our sensitivity tests, agreement of trends among sections, and large-scale ( >5‰) trends within a section these secondary processes cannot be the primary mechanisms responsible for the major trends in data recorded from these study sites. Ultimately these model results have produced conservative estimates for the extent of seafloor euxinia in the Late Ordovician oceans.
5.3 Late Ordovician cascade of redox, environmental, and biotic change
The negative excursion in δ34SCASrecorded from the late Katian is interpreted to indicate a reduction in global euxinia, which is counter to the occurrence of undifferentiated “anoxic” black shales (i.e. anoxic, ferruginous, or euxinic) found at many locations across the globe during this interval (Melchin et al., 2013). Our I/(Ca+Mg) trends indicate that sub-oxic to anoxic conditions were locally pervasive at least in the sections we studied and by inference may have been widespread in the late Katian oceans and persisted into the Hirnantian. The explanation for this apparent contradiction may lie with the fact that iodine and sulfur respond to different types of reducing conditions, with iodine responding to changes in redox near O2 reduction (i.e., non-sulfidic anoxia), while sulfate reduction occurs in more reduced settings further down the redox ladder (Froelich et al., 1978; Rue et al., 1997; Lu et al., 2010). Additionally, these two proxies reflect different spatiotemporal relationships, with iodine reflecting local water-column conditions, while δ34SCAS values record changes in sulfur cycling in the global oceans. In the discussion that follows we focus on this new level of paleoredox specificity for the Late Ordovician oceans in the context of coincident changes in the environment, eustatic sea level, and the marine biosphere.
Significant changes in local and global marine redox conditions began in the late Katian and were coincident with high sea level, elevated sea surface temperatures (SSTs), and generally high levels of marine biodiversity (Finney et al., 1999; Haq and Schutter, 2008; Trotter et al., 2008; Rasmussen and Harper, 2011; Finnegan et al., 2011). There is growing evidence from clumped oxygen isotopes and conodont palaeothermometry that global average SSTs began declining in the latest Katian with the initiation of Gondwanan ice sheet expansion (Trotter et al., 2008; Finnegan et al., 2011). This relationship suggests that the negative δ34SCAS excursion and the implied changes in global average temperature were initiated by the intensification of thermohaline circulation as a result of increased deep-water formation around Gondwanan margins, consistent with sedimentary indicators of upwelling in the Monitor Range section (i.e., bedded cherts and phosphates; Fig. 2) (Pope and Steffen, 2003). Increased thermohaline circulation would have led to cooler globally averaged SSTs and increased renewal of dissolved deep marine O2, thus ventilating portions of previously euxinic environments along continental margins, shifting the sulfidic chemocline deeper and likely into the sediments in many regions. The ultimate result was a reduced global pyrite burial flux. Thus, euxinic water column conditions may have decreased globally in the latest Katian. However, widespread sulfidic sediment pore waters may help explain the mild enrichments in molybdenum concentrations and iron speciation records of anoxic, sulfide limited, water-column conditions leading into the Hirnantian as recorded in black shales/deep basinal settings on Laurentia, Baltica, Avalonia, and peri-Gondwana (Hammarlund et al., 2012; Hardisty et al., 2018).
The collective data suggest that a combination of cooling temperatures, reduction of habitable space on shelves and in epeiric seaways due to eustatic sea-level fall (Fig. S3), and our new evidence for possibly widespread non-sulfidic anoxic marine conditions in many local basins culminated in the first LOME pulse near the Katian-Hirnantian boundary. Consistent with this hypothesis, the first appearance of bedded chert in the Monitor Range section, suggesting an increase in local upwelling, coincides with indicators of local sea-level fall (Finney et al., 1997) and expansion of a local OMZ recorded in a drop in I/(Ca+Mg). There is also evidence within previously published bulk nitrogen data (LaPorte et al., 2009) for a shift toward more reducing conditions. This data shows a trend to lighter δ15N values after our I/(Ca+Mg) ratios drop to near 0 μmol/mol (Fig. 4D) where it is attributed to a local increase in denitrification. Denitrification occurs after iodine reduction on the redox ladder (Lu et al., 2010), consistent with the observed relationship to our iodine data. Together these local redox proxies suggest a progressive loss of oxygen in this local environment prior to the Katian-Hirnantian boundary. Although many Hirnantian localities show evidence of locally reducing conditions, the anomalously high I/(Ca+Mg) ratios recorded from western Anticosti Island are likely due to lowered local sea level allowing for changes in surface currents and nutrient dynamics. The net result was well-oxygenated conditions in very shallow waters that supported patch reef environments in this region. Given the similarities in the iodine data across multiple basins and our inferred global signatures in δ34SCAS, we suggest that euxinia likely decreased globally while, paradoxically, less severe anoxia expanded globally in shallow settings thus non-sulfidic anoxia impacted marine life leading into the first LOME pulse. This reduction in euxinic conditions may be attributed to the observed global cooling of surface waters and subsequent increased solubility of O2, combined with enhanced thermohaline circulation, thus ventilating previously euxinic portions of Late Ordovician oceans. This enhanced ocean circulation may have in turn intensified local upwelling around continental margins throughout the globe, thus leading to more local primary productivity, enhancing global carbon burial and local anoxia, as evidenced by I/(Ca+Mg) trends. These climatic and oceanographic conditions during the late Katian–Hirnantian may have provided a unique balance that resulted in expansion of anoxic non-sulfidic water masses, but the increased oxygen solubility and circulation may have prevented these water masses from being pervasively euxinic. Ultimately, these marine redox conditions would have had a major impact on marine life in productive continental margins and remaining shallow seaways.
Changes in global marine redox conditions associated with eustatic sea-level rise have been invoked as a causal mechanism for the second LOME pulse in the late Hirnantian (within the M. persculptusgraptolite biozone) (Harper et al., 2014). As Gondwanan ice sheets melted and the late Hirnantian climate warmed (Finnegan et al., 2011), marine stratification and chemocline migration during eustatic sea-level rise likely played an important role in the second LOME pulse. The δ238U records from carbonates on western Anticosti Island (Bartlett et al., 2018)—along with δ98Mo and δ238U data, Mo concentrations, and iron speciation from organic-rich shale successions (Hammarlund et al., 2012; Zhou et al., 2015; Zou et al., 2018; Stockey et al., 2020)—indicate a return to widespread reducing conditions in global oceans during this time. Our iodine and sulfur isotope datasets are consistent with a shift to more reducing conditions. Specifically, late Hirnantian–early Silurian I/(Ca+Mg) values indicate local anoxia at all three sections, and δ34SCAS profiles from all sites trend positively by ~ 7‰ (Fig. 5), indicating a return to increased global pyrite burial. Increased reducing conditions along continental margins during this time would have largely tracked eustatic sea-level rise, warming sea surface temperatures would have led to decreased O2 solubility and circulation. As OMZs expanded from deep shelf/slope to shallower areas on the continental shelf during the late Hirnantian–early Silurian this increased the overall areal extent of seafloor overlain by anoxic and euxinic bottom waters.
Conclusions
Paired iodine and sulfur isotope geochemistry reveal new spatiotemporal relationships between marine non-sulfidic anoxia and euxinia associated with the Late Ordovician Mass Extinction. Our I/(Ca+Mg) ratios are low throughout this time interval in all sections, except for a set of high values (average of 6 μmol/mol) recorded from the shallow patch reef facies on western Anticosti Island. At the same time, our new δ34SCAS records show a large negative excursion of ~ 10‰ magnitude over the late Katian–Hirnantian. Modeling of these δ34SCAS records suggests that the negative excursion was driven by moderate decreases in the pyrite burial rates combined with modest increases in weathering. Our model also suggests reductions of global seafloor euxinic conditions by ~ 3% from the late Katan into the Hirnantian. This is a maximum estimate, but it is consistent with recent models of Hirnantian to early Silurian global redox-based on other proxy data (Bartlett et al., 2018; Stockey et al., 2020). Importantly, this transition does not preclude the possibility of increasing oxygen deficiency as recorded in iodine data in marginal settings due to enhanced upwelling as seen in the Monitor Range. Additional data from redox-sensitive elements are needed from multiple paleobasins to constrain the extent of these non-sulfidic reducing conditions.
In sum, our multiproxy data and modeling indicate widespread ventilation of marine environments followed by enhanced weathering during the late Katian–early Hirnantian. This sequence of events likely resulted from enhanced thermohaline circulation and growth of Gondwanan ice sheets that cooled sea surface temperatures and potentially increased deeper ocean oxygenation, therefore reducing euxinic conditions in the global oceans. However, non-sulfidic anoxic conditions remained pervasive throughout shallow shelf settings due to attendant increases in productivity resulting from increased upwelling and ocean circulation. These relationships indicate that a unique combination of reducing marine conditions, climatic cooling, and glacioeustacy led to the first LOME pulse. Subsequently, deglacial eustatic sea-level rise during the late Hirnantian that coincided with warming temperatures, deoxygenation, and decreased ocean circulation led to an expansion of global euxinic conditions, broadly coincident with the second LOME pulse. Our study sheds new light on the possibility of a complex and evolving redox landscape reflecting the interplay of multiple interrelated controls—with severe biotic turnover as a consequence. More generally, these results hint at the improved perspective that can come by integrating multiple local and global proxies from a wide distribution of locations.
Acknowledgments
N.P.K would like to thank Chelsie Bowman, Randall Funderburk, Anders Lindskog, and Westly Owings for their support and/or assistance in sample collection and preparation. B.C.G. would like to thank Charles Gill for assistance in sample collection in Nevada. We thank Dimitri Kaljo for access and assistance in sample collection of the Kardla drill core. This research was funded by the American Chemical Society Petroleum Research Fund (grant ACS-PRF# 57487-DNI2 to S.A.Y.) and the National Science Foundation (EAR-1748635 to S.A.Y. and J.D.O. and EAR-0418270 to T.W.L.). Additional funds were provided by the NASA Astrobiology Institute under Cooperative Agreement No. NNA15BB03A issued through the Science Mission Directorate (to T.W.L.). This work was in part performed at the National High Magnetic Field Laboratory, which is supported by the National Science Foundation Cooperative Agreement No. DMR-1157490 and the State of Florida. The authors declare no conflict of interest.