Five decades ago, Newell (1967) suggested a strong relationship between marine mass extinction and eustatic sea-level fall, and proposed loss of habitable shelf area as the primary driver. However, many extinctions are now linked with global warming and marine anoxia, which go hand in hand with marine transgressions (Hallam and Wignall, 1999, and references therein; Bond and Grasby, 2017). In contrast, Zou et al. (2018, p. 535 in this issue of Geology), now offer a much more nuanced picture, with expanding toxic sulfide conditions during cooling, and falling sea level during the first of the big five mass extinctions, the Late Ordovician mass extinction (LOME; ca. 445 Ma). Paradoxically, the LOME is the only one of the big five extinctions that seems to coincide with glaciations, with a double blow to metazoan-based communities. Excellent reviews have mapped out the taxonomic extinctions and ecological severity for the LOME event (Harper et al., 2014, and references therein).

Two main mechanisms have previously been proposed for the two extinction intervals of the LOME (Harper et al., 2014; Bond and Grasby, 2017), where the least controversial invokes marine anoxia during sea-level rise and global warming during the latter extinction interval (LOMEI-2). The first extinction interval (LOMEI-1) is, however, more controversial, particularly with respect to oxygenation. Many research groups link LOMEI-1 to rapid glacial sea-level lowering, cooling, and increased oxygenation of the open ocean. These models build, in part, on the widespread facies change from black to gray shales, as well as on nitrogen isotope data (Melchin et al., 2013). While possibly explaining the disappearance of the main extinction victims on the outer shelf-slope and in the upper oxygen minimum zone (OMZ), oxygenation is a difficult kill mechanism for high-latitude, cool-water taxa (Harper et al., 2014). Other research groups have proposed expansion of anoxic waters onto the outer shelf and slope during the cooling as the prime killer, based on redox geochemistry and pyrite sulfur isotopes (Hammarlund et al., 2012; Harper et al., 2014). In this scenario, the widespread change from black to gray shales reflects glacially lowered sea level, which pushed the chemocline deeper than the continental shelf observation window. Invariant sulfate isotope data, together with super heavy sulfide, could be interpreted as oxidation of a deep ocean sulfide reservoir (Jones and Fike, 2013; Kah et al., 2016). Alternatively, the sulfur isotope anomaly could be interpreted as a result of decreased fractionation during increased microbial sulfate reduction prompted by increased glacial ocean nutrients and productivity (Jones and Fike, 2013). Molybdenum and uranium isotopes suggest extensive ocean anoxia prior to LOMEI-1, and subsequent intensified oxygenation during the glacial interval (Lu et al., 2017). The latter interpretation is controversial, as it stems from a highly variable redox setting in the Yangtze Sea, China. Adding to the controversy, sequence stratigraphic concepts and a Pliocene-Pleistocene glacial analogy has led to the proposal that known LOME sections are incomplete, and missing interglacial high stands (Ghienne et al., 2014). The implication is that, perhaps, the LOMEI-1 also occurred during global warming and rising sea level, just like LOMEI-2 and other mass extinctions.

A problem with most studies to date is that each only gives a snapshot for a given basin or climate zone, and that sea-level change results in a “sliding” observation window in terms of paleo–water depth. A better understanding of the kill mechanism that caused the LOME can come from studies using multiple high-resolution sections in basin paleo-depth transects. Precisely such an undertaking is presented by Zou et al. They identify a reduction in chemical weathering interpreted as a response to glacial cooling. In addition, they use iron speciation and redox-sensitive metals to map changes in the extent of oxygen and anoxia, as well as euxinic conditions associated with toxic sulfide in the water column.

Perhaps not surprisingly, the authors find extensive ferruginous anoxia prior to the LOME in the Yangtze Sea, where the subtropical tradewind belt caused upwelling. Zou et al. observed a weathering decrease during LOMEI-1, suggesting the onset of a glacial weathering regime. They thus tie the first extinction phase to increased cooling and sea-level lowering, resulting in a decreased warm-water eco space when, in particular, low-latitude plankton went extinct. Then, what about the other victims?

Concomitant with cooling, Zou et al. found that the inner shelf was oxygenated as expected when the chemocline was pushed down during glacial sea-level fall. Interestingly, the decrease in weathering (cooling) is also associated with increasing water-column sulfide in mid-shelf and deeper settings. The latter explains the observed dying off of outer shelf-slope benthic faunas. However, increasing sulfide in an upwelling setting is surprising during cooling and sea-level lowering.

During the last glacial lowstand (ca. 20 ka) the OMZs of upwelling systems around the world seem to have been more oxygenated than during the interglacial (Jaccard and Galbraith, 2012; Scholz et al., 2014). In contrast, the deeper slope and abyss was more dysoxic, with a higher accumulation of organic carbon (Cartapanis et al., 2016). These changes presumably resulted from a strengthened oxygenated, equatorward undercurrent and reduced biological respiration in the upper water column, in response to lower metabolic activity at lower temperatures.

Ongoing research offers further insight into the oxygenation and biogeochemical response to variable shelf width during sea-level change (Fig. 1). Shelf and open ocean systems respond oppositely in the steady state with fixed input flux boundary conditions. Consistent with these results, three-dimensional regional ocean model shows that upwelling systems respond much more readily to shelf width changes than the average global shelf (cf. Al Azhar et al., 2013; M. Al Azhar and C.J. Bjerrum, our data.). These results contrast with Zou et al.’s observation of increased shelf anoxia during cooling and sea-level fall, and not during sea-level rise. The contrast indicates that other feedback relationships were at work. Reduced glacial weathering and limited increase in tradewind intensity result in no enhancement of nutrient delivery to the shelf system, underlining the observed increase in sulfide levels (Pohl et al., 2017; Pogge von Strandmann et al., 2017; Zou et al., 2018). Shelf exposure in a system with no rooted vegetation may result in further “shelf unloading”, increasing the oxygen demand in the deep ocean (Broecker, 1982). Lower temperatures further act to decrease metabolic activity, pushing the oxygen demand deeper into the ocean. When mapping the Earth system feedback, it is seen that sea-level change in the short term results in a negative feedback centered on the shelf, whereas sea-level change on the long term may be a positive feedback (see Broecker, 1982) (Fig. 2). From the results presented by Zou et al., it seems that long-term open ocean response “won” in expanding the OMZ with toxic sulfide during cooling during LOMEI-1. But perhaps the picture is more complicated.

Remember that LOMEI-1 could have occurred during global warming and rising sea level, just as LOMEI-2 did (Ghienne et al., 2014). Inspecting in detail the data of Zou et al., we see a few more intense weathering points just at the base of LOMEI-1. A short-lived interglacial could have resulted in intense weathering and sea-level rise, as also shown in temperature and glacial volume proxies (Finnegan et al., 2011). If so, it may not be the glacial interval that drove the first of the big five mass extinctions, but rather short-lived increasing temperature and sea-level rise.

Regardless, the thorough data set presented by Zou et al. presents several valuable observations. It confirms records from an upwelling location in western Laurentia, where sustained anoxia persisted though LOMEI-1 (Ahm et al., 2017). And their data set creates a paradox in how upwelling systems respond to glacial conditions during the Late Ordovician Mass Extinction.