End-Cretaceous extinction in Antarctica linked to both Deccan volcanism and meteorite impact via climate change

The cause of the end-Cretaceous (KPg) mass extinction is still debated due to difficulty separating the influences of two closely timed potential causal events: eruption of the Deccan Traps volcanic province and impact of the Chicxulub meteorite. Here we combine published extinction patterns with a new clumped isotope temperature record from a hiatus-free, expanded KPg boundary section from Seymour Island, Antarctica. We document a 7.8±3.3 °C warming synchronous with the onset of Deccan Traps volcanism and a second, smaller warming at the time of meteorite impact. Local warming may have been amplified due to simultaneous disappearance of continental or sea ice. Intra-shell variability indicates a possible reduction in seasonality after Deccan eruptions began, continuing through the meteorite event. Species extinction at Seymour Island occurred in two pulses that coincide with the two observed warming events, directly linking the end-Cretaceous extinction at this site to both volcanic and meteorite events via climate change.


Age (Ma)
. Stratigraphic heights have been adjusted based on the plane-projection method 6 . Age calculated based on stratigraphic position using age model of this study.

Supplementary Discussion
Correlation between Temperature and δ 18 O w Supplementary Fig. 8 shows a very strong correlation (R 2 = 0.844, p-value < 2.2E-16) between temperature and δ 18 O w among individual samples. We have interpreted this correlation as reflecting climatologically-controlled, seasonal delivery of continental runoff to the near-shore environment. In colder climates, more snowpack accumulates on land in winter, δ 18 O of the snow is lower, and meltwater is delivered in a concentrated spring thaw, reducing near-shore δ 18 O w significantly. In warmer climates, less snowpack accumulates, δ 18 O of the snow is higher, and delivery is less concentrated, causing higher near-shore δ 18 O w values than during cold climates. Here we discuss non-environmental ways to produce a correlation between temperature and δ 18 O w and enumerate why the climatological mechanism is the most likely explanation.
One possible explanation for the observed correlation is that variability in Δ 47 -derived temperature caused by analytical noise or alteration of the clumped isotope signal during sample drilling, propagated through the δ 18 O/Temperature/δ 18 O w relationship 7 to produce correlated variability in δ 18 O w . With regards to drilling, this preparation method has been used for conventional stable isotope studies for decades and does not affect carbonate δ 18 O 8 . However, drilling at high speed has the potential to increase the Δ 47 -derived temperature through frictional heating and resetting. If drilling increased the measured Δ 47 -derived temperature on a particular sample, it would result in an artificial increase in the calculated δ 18 O w value for that sample. The magnitude of the increase in δ 18 O w would be positively correlated to the magnitude of the increase in temperature according to the δ 18 O/Temperature/δ 18 O w relationship 7 , corresponding to roughly a 0.25‰ increase in δ 18 O w for every 1°C increase in temperature. Similarly, large analytical noise could cause either a positive or negative anomaly in the measured clumped isotope temperature. In this case, the resulting positive or negative perturbation in δ 18 O w would follow the same ratio derived from the δ 18 O/Temperature/δ 18 O w relationship 7 . If each sample was affected by drilling-induced alteration or analytical noise to a different degree, beginning from a similar starting temperature, the resulting Temperature vs. δ 18 O w scatter plot would show a positive correlation with a slope of ~0.25, similar to Supplementary Fig. 8, which has a slope of 0.242.
There are multiple reasons that suggest the correlation in Supplementary Fig. 8 is not due to drilling-induced alteration. First of all, in our experience, this effect has only been seen at drill speeds much higher than the setting used in this study (15,000 rpm vs. 1,000 rpm used here). Foster et al. 8 found 6% conversion to calcite and no change in carbonate δ 18 O at drill speeds of ~16,000-21,000 rpm (Δ 47 was not measured). Although this drilling effect on Δ 47 has been hypothesized based on a few observations, it has yet to be rigorously analyzed in any peerreviewed literature. Additionally, the proposed drilling artifact can only increase Δ 47 -derived temperatures and, therefore, δ 18 O w values. If this were occurring, we would see higher temperatures and δ 18 O w values than expected. Instead, the mean δ 18 O w value for the entire study interval (-1.3±0.8‰ (1sd), mean of horizon averages) agrees with (or is even slightly lower than) the predicted mean value for an ice-free world (-1.0‰) (ref. 9) or for an ice-free world adjusted for the modern latitudinal isotopic gradient (-1.2‰) (ref. 10). Also, the mean ocean temperature for the entire section is 7.7 ± 3.4°C (1sd) (mean of horizon averages), consistent with an independent terrestrial temperature estimate of 7°C from fossil wood 11 and a soil temperature estimate of 10-13°C from branched tetraethers 12 . These observations combine to suggest that broad resetting through drilling has not occurred.
Large analytical noise, which could cause artificial increases or decreases in Δ 47 -derived temperatures, would be equally likely to affect any given sample. This would result in a larger range in temperatures seen over the study interval, caused by the addition of analytical-noiseinduced scatter on top of the 'true' temperature record. If common assumptions about the minimal variability of marine δ 18 O w are to be believed (no variability, constant δ 18 O w ), the δ 18 O data suggests a total temperature range of 8°C (from ~4-12°C). In comparison the clumped isotope record shows a much larger temperature range of ~20°C (from ~-3-15°C), consistent with expectations for the "noise plus 'true' variability" scenario described above. However, in this hypothetical noisy record, there should still remain a correlation between temperature and shell δ 18 O. In this case, a linear regression between δ 18 O and temperature gives an R 2 value of 0.021 -no correlation. R 2 values less than 0.1 are also achieved for all position-species subsets of the data (all Lahillia umbo positions, for example) for Lahillia, Cucullaea, and Eselaevitrigonia, with the exception of Cucullaea ventral margin, which has an R 2 value of 0.25.
Nordenskjolidia shows R 2 values of 0.67 and 0.80 for the umbo and ventral margin, although this is only based on three points. In fact, for all Cucullaea shells, δ 18 O values are equal or higher in ventral margin than in the umbo position, suggesting colder temperatures in the ventral margin position. Colder temperatures are also expected for Lahillia in its umbo position relative to ventral margin. Clumped isotope temperature reveal the opposite, finding warmer temperatures recorded in Cucullaea ventral margin and Lahillia umbo, where δ 18 O predicted colder. This extreme lack of correlation between δ 18 O and temperature goes against the fundamental assumption underlying all δ 18 O-based paleoclimate studies -that δ 18 O w should be fixed, and that, therefore, temperature must be the sole driving force setting δ 18 O values -and suggests that δ 18 O, temperature, and δ 18 O w are related in unusual ways in this data set.
The samples in this study were drilled, prepared, and analyzed over three separate measurement sessions, spanning a total of nine months. The order of analysis of individual replicates was scrambled on a daily, weekly, and monthly basis to avoid potential cumulative effects of measuring the same samples in the same order, or biases caused by unusual measurement conditions on a given day or week. Carbonate standards were continually monitored and showed consistency and reproducibility during the same measurement sessions. These practices are done to reduce biases caused by short-and long-term drift in instrument conditions and make it unlikely that analytical noise would produce artificial spikes or trends in the temperature record that show the agreement between samples through time and the positionspecific relationships we observe. Samples of similar stratigraphic age produce similar Δ 47derived temperatures, showing a clear coherence in the temperature record on both short and long timescales. We also see consistent position-specific patterns within species (Fig. 2). For example, all Lahillia shells, excluding the two at the KPB, show warmer temperatures in the umbo position relative to the ventral margin. These observations, combined with our laboratory practices designed to avoid bias make it unlikely that analytical noise produced the temperature record we observe.
Another possibility is that the observed correlation is the result of diagenetic alteration occurring in situ on Seymour Island, prior to collection and analysis. For diagenesis to produce the observed correlation between temperature and δ 18 O w but retain the lack of correlation between temperature and carbonate δ 18 O, alteration must change temperature without changing carbonate δ 18 O. This could occur by recrystallization in a rock-buffered system or thermal resetting. Our extensive consideration of diagenetic alteration though trace element screening and cathodoluminescence and petrographic observations of textural features, combined with other studies documenting a shallow burial history for Seymour Island 13,14 (see Methods) suggest that this correlation is unlikely to reflect post-depositional alteration in situ.
Temperature is not the dominant control on δ 18 O in this data set, yet evidence suggests that the temperature record is robust. Therefore, the only possible explanation that fits both the temperature and δ 18 O data is that temperature and δ 18 O w are varying in conjunction and in opposition. The combination of the large observed variability in Δ 47 -derived temperatures (~20°C range) and the small range in δ 18 O values forces δ 18 O w to vary significantly (a 5‰ range) and in tight coupling with temperature.
Prior to the development of the clumped isotope paleothermometer, studies universally assumed that δ 18 O w was constant and, therefore, that measured variability in δ 18 O was dominated by changes in temperature. As more and more clumped isotope measurements are made and δ 18 O w is directly assessed instead of assumed, we are finding that δ 18 O w doesn't always follow