Terrestrial evidence for volcanogenic sulfate-driven cooling event ~30 kyr before the Cretaceous–Paleogene mass extinction

Abstract

Alongside the Chicxulub meteorite impact, Deccan volcanism is considered a primary trigger for the Cretaceous–Paleogene (K–Pg) mass extinction. Models suggest that volcanic outgassing of carbon and sulfur—potent environmental stressors—drove global temperature change, but the relative timing, duration, and magnitude of such change remains uncertain. Here, we use the organic paleothermometer MBT′_5me and the carbon-isotope composition of two K–Pg-spanning lignites from the western Unites States, to test models of volcanogenic air temperature change in the ~100 kyr before the mass extinction. Our records show long-term warming of ~3°C, probably driven by Deccan CO₂ emissions, and reveal a transient (<10 kyr) ~5°C cooling event, coinciding with the peak of the Poladpur “pulse” of Deccan eruption ~30 kyr before the K–Pg boundary. This cooling was likely caused by the aerosolization of volcanogenic sulfur. Temperatures returned to pre-event values before the mass extinction, suggesting that, from the terrestrial perspective, volcanogenic climate change was not the primary cause of K–Pg extinction.

Terrestrial proxy evidence for sulfur-induced cooling event linked to Deccan Trap eruptions 30 ka before K–Pg boundary.

INTRODUCTION

The Cretaceous–Paleogene (K–Pg) boundary [~66 million years ago (Myr)] represents the most recent mass extinction event; an estimated 75% of all species were extinguished (1), including all nonavian dinosaurs. This event changed the trajectory of the evolutionary tree of life (2, 3) and resulted in a complete rebuilding of ecosystems from dinosaur- to mammal-dominated communities. The Chicxulub meteorite impact [Mexico (4–6)] and eruption of the Deccan Traps [India (7–12)] have emerged as the primary—but fiercely contested—trigger mechanisms for the mass extinction and global climate change. Models of the climate response to the meteorite impact include an “impact winter” lasting months to millennia due to atmospheric loading of dust, soot, and sulfate aerosols (13–17), and longer-term warming caused by CO₂ released by wildfires and/or impact-volatilized carbonates (18). Two principal climate models are associated with Deccan volcanism: first, global warming, caused by eruption-, venting-, and contact-metamorphism-derived CO₂ (19) and sustained over thousands to hundreds of thousands of years [e.g., (20)]; and second, global cooling driven by the conversion of SO₂ into sulfate aerosols, but lasting only for the duration of the eruption (21–23).

Recent high-precision radiometric dating has established synchronicity between meteorite impact and extinction (24, 25) and has shown that the most major phase of Deccan volcanism—the Poladpur “pulse” [sensu Schoene et al. (26)]—erupted from 66.10 to 66.00 Myr, peaking 30 thousand years ago (kyr) before the K–Pg boundary (26, 27). These data do not exclude Deccan volcanism as a contributing or primary cause of extinction, but rather provide a high-precision geochronology of events around the K–Pg boundary against which competing hypotheses of climate change can be tested using age-constrained proxy reconstructions.

Here, we reconstruct mean annual air temperatures (MAATs) at a millennial resolution leading up to the K–Pg boundary based on branched glycerol dialkyl glycerol tetraethers (brGDGTs) in fossil peats (lignites) from two mid-paleolatitude sites in the Western Interior of the United States (Pyramid Butte, North Dakota, and West Bijou, Colorado; Fig. 1). BrGDGTs are membrane lipids of bacteria living in soils and peats that adjust the number of methyl branches attached to the alkyl backbone to changes in temperature. The degree of methylation is quantified in the Methylation index of 5-methyl Branched glycerol dialkyl glycerol Tetraethers [MBT′_5me index (28, 29)] and can be translated into MAAT using a peat-specific transfer function (30). MBT′_5me is an established proxy that has been successfully used for deep-time paleotemperature reconstruction (31), though confounding factors may influence the temperature signal reflected by the brGDGT distributions stored in sedimentary archives. Hence, brGDGT distributions need to be carefully assessed before their interpretation as a temperature signal. We have, therefore, developed a framework that allowed us to identify samples with potential non-thermal influences on brGDGT distributions (see the Supplementary Materials).

Fig. 1. Paleogeographic context of North America highlighting the study locations, alongside physical stratigraphy, MBT′_5me, and bulk-organic δ¹³C records for West Bijou (blue) and Pyramid Butte (orange).

(A and B) Medium and dark solid lines through the records show three- and five-point moving averages, respectively. Dashed lines show the correlation between sites. The stratigraphic levels of tuffs are shown, with their radiometric dates (25). The solid gray line denotes the position of the K–Pg boundary—the primary tiepoint—as defined by the presence of shocked minerals, iridium anomalies, and palynological extinctions at Pyramid Butte (32, 66) and West Bijou (33, 65). Note that data filtering (Supplementary Materials) precludes the generation of MBT′_5me records for the entirety of the lignites, and a short gap exists immediately below the K–Pg boundary.

Although the peats at both sites accumulated in discrete depocenters of the Western Interior Basin, the K–Pg boundary is clearly identifiable at each by the presence of diagnostic meteorite-impact indicators (shocked minerals and an iridium anomaly) and coincident palynomorph extinction (Fig. 1B) (32, 33). This datum forms the principal basis of correlation between our sites. In addition, a succession of primary volcanic tuffs interbedded with the lignites at West Bijou bracket the K–Pg boundary (25). Critically, these tuffs were dated (25) using the same method and calibration as Schoene et al.’s (26) study of Deccan eruptions [compare Sprain et al. (34)] and, for this reason, we focus on these dates. The lower tuff, KJ08157 (~10 cm thick), the top of which occurs 43 cm below the palynological K–Pg boundary and defines the base of the lignite seam, yielded an age of 66.082 ± 0.022 Myr (25). The base of tuff KJ0475 (~13 cm thick) occurs 52 cm above the palynological K–Pg boundary and yielded an age of 65.993 ± 0.019 Myr (25). Linear interpolation (of peat accumulation rates and peat-to-lignite compaction ratios) between these two dates implies 1 m of lignite = 97 kyr ± 45 kyr. Last, the bulk-organic carbon-isotope (δ¹³C) records at both sites provide a secondary correlation framework for our study. Our high-resolution geochronological model suggests that our filtered MAAT record spans the last ~50 kyr and ~100 kyr of the Cretaceous at West Bijou and Pyramid Butte, respectively. This interval encompasses the main eruptive phase of Deccan volcanism—the Poladpur phase (26).

We compare these reconstructions with bulk-organic carbon-isotope (δ¹³C) records generated from the same samples to determine relationships between climate and carbon cycling at this time. Our data, therefore, directly test hypotheses of terrestrial climate change driven by Deccan volcanism, particularly eruption of the Poladpur Formation, which overlaps temporally with our MAAT record (26).

RESULTS

We present our reconstructed MAATs using the MBT′_5me proxy calibration developed for peats [calibration error ±4.7°C (30)] and the abovementioned age model. Over the last ~100 kyr of the Cretaceous, MAATs ranged from ~24° to 27°C at Pyramid Butte and from ~21° to 27°C at West Bijou (Fig. 2). There is remarkable synchroneity between the two temperature records where they overlap stratigraphically. We highlight three distinct intervals (Fig. 2): (i) 100 to 30 kyr before the K–Pg boundary MAATs increased from 23° to 26°C, with a stepwise increase 70 kyr before the boundary (this trend is derived from the Pyramid Butte record); (ii) from 30 to 20 kyr before the K–Pg boundary, MAATs declined to a minimum of 23°C at Pyramid Butte and from 27° to 21°C at West Bijou; (iii) during the last ~20 kyr of the Cretaceous, MAATs at both sites returned to pre-event values and then stabilized at 27°C. Thus, both sites display a markedly symmetrical cooling event, with a magnitude of change of 2° to 5°C, beginning 30 kyr before the K–Pg boundary and lasting no more than 10 kyr. This trend is superimposed on a longer-term (at least 100 kyr) latest Maastrichtian warming of 3°C.

Fig. 2. Atmospheric-CO₂ δ¹³C and mean annual air temperatures (MAAT) plotted against time (kyr) relative to the K–Pg boundary.

The medium and dark solid lines through the records show three- and five-point moving averages, respectively. The 95% confidence interval (CI) of the K–Pg boundary (dashed line), as determined by Schoene et al. (26), is indicated in gray. The probabilistic volumetric eruption rate for the Deccan Traps (26) is shown on the right (red shading shows the 68% CI and the black line shows the mean value); the greatest volume was likely erupted between ~50 and 10 kyr before the K–Pg boundary, with the peak around 30 kyr before.

The δ¹³C values range from −26.8 to −23.6‰ and from −27.3 to −24.9‰ at Pyramid Butte and West Bijou, respectively. In the latest Cretaceous part of the record, the values are remarkably consistent, deviating little from a mean around −26‰ at West Bijou and −25‰ at Pyramid Butte (Fig. 2), until the distinct negative excursion that characterizes the K–Pg boundary worldwide (35–41).

DISCUSSION

Our two sites show remarkable similarity in absolute MAATs and temporal trends, despite being located 750 km apart, providing confidence that the individual sites are recording a regional climatic signal, and bolstering support for the reliability of the MBT′_5me proxy [e.g., (42)]. Furthermore, the absolute MAAT values are similar to those reconstructed for the latest Maastrichtian from lignites in nearby southern Saskatchewan [Canada; 20° to 25°C (31)], and clumped-isotope paleothermometry on terrestrial mollusks in Montana (USA; 23° to 30°C) (43). Likewise, the δ¹³C records at both our sites are notably similar to one another (Figs. 1 and 2) and to other terrestrial records (40). Our record therefore reflects a global signal of atmospheric-CO₂ δ¹³C (38, 44).

There are few single-proxy paleotemperature studies at the resolution of this study against which even the longer-term trend in our study—MAAT rise of 3°C over the last 100 kyr of the Cretaceous—can be compared. However, warming in the lead-up to the boundary is captured in both marine-carbonate δ¹⁸O (39) and terrestrial brGDGT (31) records. Certainly, this latest-Maastrichtian warming trend in our data postdates the Late Maastrichtian Warming Event (LMWE) of ~5°C dated to ~300 to 200 kyr before the K–Pg boundary (39, 41, 45).

The cause of the 3°C warming in our record can circumstantially be linked to Deccan volcanism. U–Pb radiometric dating reveals that the Deccan Traps were erupted as four discrete high-volume events separated by relative volcanic quiescence (26). These include an event 300 to 150 kyr before the K–Pg boundary (39, 44), and another—the Poladpur pulse—centered on the last 100 kyr of the Cretaceous (26). These events, or pulses, in fact comprised multiple eruptions that individually lasted decades to centuries (26), with peak eruptive volumes of 5 to 10 km³/year occurring 30 kyr before the K–Pg boundary (26). [We note, though, that ⁴⁰Ar/³⁹Ar dating undertaken by Sprain et al. (34) suggests a more continuous eruption style for Deccan volcanism.] Both the LMWE and the later warming trend observed in our data correspond to the onset and duration of these two eruptive events: the Kalsubai/Lonavala and Poladpur, respectively. Gilabert et al. (46) have previously attributed the LMWE to increased CO₂ associated with emplacement of the Kalsubai/Lonavala subgroup combined with background orbital forcing. The length of our combined record (~100 kyr) does not permit the testing of orbital cyclicity in a statistically significant way. However, the recognition of two distinct warming episodes—the LMWE and the later warming captured in our record—is inconsistent with the notion of quasicontinuous eruption (and CO₂ release) [e.g., (34)].

Models based on the understood volumes and rates of total Deccan lava eruption, and assumptions about associated CO₂ emissions imply that they resulted in net global warming of less than 1°C (47) to 4°C (19, 20, 48). Recent geochemical analyses of accompanying intruded (i.e., unerupted) magmas increase total CO₂ emission estimates and imply that up to 6°C warming may have been possible (49) but diminish the temporal connection between lava eruptions and the tempo and rates of CO₂ outgassing.

The time interval reconstructed in this study (~100 kyr) represents a fraction of the total Deccan eruptive time, and the Poladpur Formation represents less than a quarter of the volume of the Deccan Traps. Therefore, the ~3°C of warming we observe in our records appears possibly high compared to model outputs based on the entire Deccan Traps. We do not consider the paleolatitudes of our two sites (45°N and 51°N for West Bijou and Pyramid Butte, respectively) to be sufficiently high to invoke polar amplification as an explanation for the relatively large rise in MAAT that we observe. As an explanation, Hernandez Nava et al. (49) found that late Deccan magmas became depleted in CO₂, meaning that the earlier phases of outgassing were more efficient in driving global warming. Furthermore, combined biological and mineral reaction sequestration rates could not fully offset the volumes of CO₂ injected into the atmosphere at the timescales between individual Deccan events (20, 48), resulting in diminishing greenhouse effects of each new mole of CO₂ emitted (50). The 5°C of LMWE warming that has been ascribed to the first Deccan eruptive event 300 to 200 kyr before the K–Pg boundary is greater than the 3°C observed in our record, consistent with this hypothesis (45). We propose that subsequent eruptive phases would show decreasing effects on global temperatures.

Transfer of CO₂ or CH₄ to the atmospheric reservoir may have also occurred through the metamorphism and/or combustion of hydrocarbons associated with the emplacement of the Deccan Traps [e.g., (51–53)]. Further, CO₂ may have been released from the terrestrial and marine biogenic reservoirs, caused by a latest Maastrichtian ecological stress and reduction in biomass (54, 55). These organic, isotopically light [−29‰ (56)] carbon sources would have resulted in a distinct imprint on our atmospheric δ¹³C record in the form of a negative excursion. The only notable negative excursion in our data corresponds to the K–Pg boundary itself (Figs. 1 and 2) and postdates our temperature record. There is no correspondence between our MAAT and atmospheric δ¹³C record (r² of regression = 0.005), so we preclude a major organic source of CO₂ as the driver of this longer-term warming phase. This finding supports the argument that organic carbon played only a minor role in the earlier Deccan-associated LMWE (45), despite assertions that warming in the absence of a negative δ¹³C excursion does not preclude protracted addition of isotopically light carbon (53). By comparison, volcanogenic carbon has a comparatively heavy δ¹³C signature [−5‰ (56)] requiring much greater carbon emissions to affect change in atmospheric δ¹³C (56), and to be detectable in the sedimentary geochemical record (57, 58). The ~3°C of warming observed in our records is therefore consistent with the greenhouse effects of CO₂ emissions associated with the onset of the Poladpur pulse.

The prominent, ~10 kyr duration, 2° to 5°C cooling event observed in both of our records has not previously been recognized in terrestrial paleotemperature reconstructions. Investigations of any such cooling linked with Deccan volcanism have—until now—been limited by the paucity of well-dated high-resolution temperature records [e.g., (43)]. The well-dated marine δ¹⁸O record of Keller et al. (12) does recognizes a cooling event ~45 to 25 kyr before the K–Pg boundary (12), which overlaps with our event within error of the respective age models, though the authors attributed the cooling instead to a cessation of volcanogenic CO₂ release. Nonetheless, several lines of evidence imply that our record is the first, to our knowledge, to test and support recent models of global cooling driven by the conversion of erupted SO₂ into sulfate aerosols for short periods [e.g., (21–23, 59, 60)].

First, the timing of the cooling event, beginning ~30 kyr before the K–Pg boundary, coincides with peak rates of Poladpur Formation lava emplacement (26), using the same U–Pb dating methods and calibration technique as the tuff ages at West Bijou (25) (Fig. 2). Sulfur outgassing occurs at, or close to, the extrusive surface (49); thus—unlike CO₂ emissions associated with a combination of intrusive and extrusive processes—SO₂ emissions must be shown to be contemporaneous with the Deccan eruptive phases. Studies of stratigraphic mercury (Hg) in combination with high-precision ⁴⁰Ar/³⁹Ar tuff dates from lignite-bearing K–Pg successions in the Williston Basin (Montana) have, additionally, revealed elevated concentrations ~30 kyr pre-boundary (61). Fendley et al. (61) attributed the Hg spike to a pulse of Deccan volcanism, which they determined from overall Hg concentrations to have lasted on the order of several centuries.

Individual Deccan lavas were erupted over decades to centuries (26, 62). Volcanogenic sulfur emissions are modeled to have driven cooling of 5° to 8°C, sustained for the duration of individual eruptions, after which the aerosols would be removed within ~50 years as acid rain (21). The cooling event observed at both West Bijou (5°C) and in particular Pyramid Butte (2°C) is of a slightly lesser magnitude than modeled by Schmidt et al. (21), and at least an order of magnitude longer in duration. These discrepancies can both be explained by (i) the higher heat capacity of water in the saturated peat compared with air, introducing a dampened signal in the brGDGT record, or (ii) the effect of mixing of organic material in the peat profile, as younger plants grow roots into and remobilize older peat. As such, our data imply that there is a limit to the temporal resolution of lignite records such as these. The notion that the climatic signal has been overprinted by these processes is supported by the difference in reported magnitude of the negative temperature excursion between the two sites; it is unlikely that localities separated by 750 km and showing such similar trends and absolute values of MAAT over the longer term should experience such a substantial difference in SO₂ aerosol-induced cooling. Although the markedly symmetrical nature of the cooling and then warming could be interpreted as the onset, acme, and waning phases of volcanic activity, we posit that it reflects the smoothed expression of a more discrete, shorter-duration and higher-magnitude cooling event. We consider the expression of the transient cooling event in our data to represent an absolute minimum in terms of magnitude, a maximum in terms of duration, and consistent with models of the Deccan-sourced SO₂-induced cooling (21–23).

The cooling event does not coincide with any excursions in our atmospheric δ¹³C record (Fig. 2), as would be expected if it was caused by orbital forcing and/or biogenic sequestration of CO₂ and a reduction in greenhouse effect. Consequently, as with the longer-term warming observed in our record, interpretations of cooling as generated by volcanism does not necessitate a perturbation to the carbon-isotope composition of atmospheric CO₂.

Overall, our data show a longer-term warming signal of ~3°C over the last 100 kyr of the Cretaceous, on which a transient cooling event of 2° to 5°C is superimposed, centered on ~30 kyr before the K–Pg boundary. To a degree, our data therefore support both models of climate change induced by Deccan volcanism, that is, longer-term (>100 kyr) warming caused by Deccan volcanogenic and magmatic outgassing of CO₂ (19, 49), and short-term (<10 kyr), high-magnitude (5°C) cooling caused by the conversion of volcanogenic SO₂ into sulfate aerosols [e.g., (21)]. This study is the first, to our knowledge, to test and support the latter model, the climate response of which has been asserted as being too short in duration to be detectable in the stratigraphic record (26).

Peat accumulation is relatively steady and at longer timescales, fast, compared to clastic floodplain sedimentation (63), and this study emphasizes the utility of lignite and coal records as sensitive archives of deep-time terrestrial climate change. Our record highlights the varying tempo of Deccan volcanism–induced climate change directly tied to the different timescales of carbon and sulfur cycling. The longer-term, CO₂-induced warming of the latest Maastrichtian can now be separated into two discrete events: the LMWE and that of our new record, both coinciding with pulses of Deccan eruption. The latter was punctuated by a brief cooling event linked to sulfur cycling, and similar events may have occurred during other eruptive events but have not yet been resolved in paleoclimate records. Further study of, for example, longer lignite records spanning the K–Pg boundary [e.g., (64)] with high-precision chronological constraint will test the notion that each of the main Deccan eruptive phases will be accompanied by concomitant short-duration cooling.

To what degree were these climatic phenomena implicated in the mass extinction at the K–Pg boundary? The 5°C cooling and subsequent return to pre-event warmth, which we attribute to the effect of aerosolized SO₂ emissions by the Poladpur eruption, predates the palynological extinction at the K–Pg boundary at both West Bijou (33, 65) and Pyramid Butte (32, 66) by several millennia. Despite a relatively coarse sampling density (12 samples for the whole of the lignite at Pyramid Butte, as opposed to the n = 51 data points for this section in our record), Nichols and Johnson (66) show that of the 17 angiosperm taxa that predate the transient cooling episode, 15 continue to the K–Pg boundary or into the Paleocene. Therefore, while this rapid climate change associated with volcanogenic aerosolized SO₂ emissions may have induced ecological stress, from a palynological perspective, it was not the primary cause of the extinction at the K–Pg boundary.

MATERIALS AND METHODS

The lignite seams at Pyramid Butte (46°25′03″N 103°58′33″W) and West Bijou (39°34′14″N 104°18′09″W) were sampled contiguously, and samples were freeze dried and powdered before geochemical analysis. Samples were solvent extracted using a MARS6 microwave extraction system, and the total lipid extracts were separated into polar and apolar fractions using aluminum oxide column chromatography. The polar fractions (containing the GDGTs) were filtered (0.45 μm polytetrafluoroethylene, PTFE) using hexane:isopropanol (99:1, v/v) before analysis by high-performance liquid chromatography/atmospheric pressure chemical ionization–mass spectrometry (HPLC/APCI-MS) on a TSQ Quantum Access Orbitrap HPLC-MS [University of Plymouth (31)]. Samples were screened for nonthermal influences on GDGTs using several GDGT indices: the Branched and Isoprenoid Tetraether index (67), relative abundance of 6-methyl versus 5-methyl brGDGTs (68), degree of cyclization (68), and community index (69). After the exclusion of samples biased by nonthermal influences (Supplementary Materials), the MBT′_5me index was calculated following De Jonge et al. (29). The peat-specific calibration [MAAT_peat (30)] was used to convert MBT′_5me values into MAATs. Bulk-organic δ¹³C analyses were conducted at Plymouth University using an Isoprime mass spectrometer connected to an Isoprime Microcube elemental analyzer.

Supplementary Materials

The PDF file includes:

Supplementary Text

Figs. S1 and S2

Legend for data S1

References

Other Supplementary Material for this manuscript includes the following:

Data S1