Rapid short-term cooling following the Chicxulub impact at the Cretaceous–Paleogene boundary

Significance

Here, for the first time (to our knowledge), we are able to demonstrate unambiguously that the impact at the Cretaceous–Paleogene boundary (K–Pg, ∼66 Mya) was followed by a so-called “impact winter.” This impact winter was the result of the injection of large amounts of dust and aerosols into the stratosphere and significantly reduced incoming solar radiation for decades. Therefore, this phase will have been a key contributory element in the extinctions of many biological clades, including the dinosaurs. The K–Pg boundary impact presents a unique event in Earth history because it caused global change at an unparalleled rate. This detailed portrayal of the environmental consequences of the K–Pg impact and aftermath aids in our understanding of truly rapid climate change.

Abstract

The mass extinction at the Cretaceous–Paleogene boundary, ∼66 Ma, is thought to be caused by the impact of an asteroid at Chicxulub, present-day Mexico. Although the precise mechanisms that led to this mass extinction remain enigmatic, most postulated scenarios involve a short-lived global cooling, a so-called “impact winter” phase. Here we document a major decline in sea surface temperature during the first months to decades following the impact event, using TEX₈₆ paleothermometry of sediments from the Brazos River section, Texas. We interpret this cold spell to reflect, to our knowledge, the first direct evidence for the effects of the formation of dust and aerosols by the impact and their injection in the stratosphere, blocking incoming solar radiation. This impact winter was likely a major driver of mass extinction because of the resulting global decimation of marine and continental photosynthesis.

The Cretaceous–Paleogene (K–Pg) boundary mass extinction was one of the most devastating events in the Phanerozoic history of life, both on land and in the oceans (1, 2). It is widely acknowledged to be related to the impact of an asteroid with an estimated diameter of ∼10 km at Chicxulub, Yucatan Peninsula, Mexico (3–6). Impact models suggest that the first hours after the impact were characterized by earthquakes and tsunamis and the so-called “fireball stage,” including an intense heat pulse resulting from the return flux of larger ejecta, in turn resulting in global wildfires (7). Next, the dust and sulfur aerosols, originating from the anhydrite target rocks, are predicted to have partially blocked incoming solar radiation, leading to an “impact winter” (8, 9), potentially further amplified by soot derived from burning of fossil organic matter in targeted sediments, a strong absorber of short-wave radiation (10). This dark phase is proposed to have temporarily inhibited photosynthesis, causing a global collapse of terrestrial and marine food webs (5, 6).

Model simulations suggest that the amount of sunlight reaching Earth’s surface was potentially reduced to ∼20% (9). This implies a ∼300 W⋅m⁻² reduction in energy supply, which should have resulted in a severe but short-lived drop in global surface temperature (6, 11). The resulting enhanced contrast between relatively warm oceans and cold atmosphere likely fueled large storms and hurricanes (12, 13), increasing the residence time of dust in the atmosphere. In the months to decades following the impact, the atmosphere probably stabilized, and dust began to rain down and accumulate in depositional settings. This included asteroid-derived trace elements, globally recognized as a peak in platinum group element (PGE; including iridium) concentrations in complete marine and terrestrial successions (14). Crucially, fossil evidence for this impact winter scenario is still missing because this period of reduced solar radiation may only have lasted several months to decades (8, 10, 15, 16), generally too short to be captured in the ancient sedimentary record. In case of the K–Pg boundary this is even more complicated because the traditional proxy carriers for the surface ocean conditions, calcareous microfossils, experienced major extinction (17). Furthermore, diagenetic alteration is commonly noted in postextinction biotic carbonates, inhibiting accurate temperature reconstructions (18).

Among the few sections with potentially sufficient temporal resolution across the K–Pg boundary are the exceptionally well-preserved and well-studied outcrops exposed along the Brazos River between Waco and Hearne, Texas (31° 7′53.59″N, 96°49′26.08″W; Fig. 1). In the Late Cretaceous and early Paleogene, the Brazos area was characterized by nearly continuous and predominantly siliciclastic sedimentation on the shallow shelf of the northern Gulf of Mexico, close to the entrance of the Western Interior Seaway (19, 20), at estimated depths of 75–200 m (21, 22). The sedimentary successions in this region comprise the Maastrichtian Corsicana (Kemp Clay) Formation and the Paleocene basal and upper Littig members of the Kincaid Formation.

Fig. 1.

A paleogeographic map of the Gulf of Mexico at the end of the Cretaceous. The Brazos-1 locality (red dot) and the Chicxulub crater are indicated, as well as other sites (green dots) with K–Pg impact-related tsunamites, slumping, faulting, and sliding compiled in ref. 22. Paleogeography based on ref. 45.

At Brazos River, the K–Pg boundary interval has been further subdivided in a series of the lithological units (Units A to J; Fig. 2) (21, 22). The upper Maastrichtian fossiliferous shales of the Corsicana Formation (Unit A) (22, 23) are overlain by the basal part of the Paleocene Kincaid Formation, consisting of a sequence of sandstone layers yielding multiple types of clasts and shell debris (Units B, C, and D) that have been interpreted as impact-triggered tsunami deposits (21–25). The top of this sandstone complex comprises abundant burrows. The overlying complex of organic-rich silts and mudstones (Units E, F, and G) includes the Ir anomaly, consisting of scattered and smeared-out peaks of up to 2 ppb (22, 23) (SI Age Model).

Fig. 2.

Sample positions of the 1995 sample set, plotted with lithological units, mean grain size on the Krumbein phi (φ) scale, and four different Iridium profiles, from refs. 46–49. A closely spaced sample set was obtained to acquire a high temporal resolution.

Traditionally, the age model of the Brazos River 1 (BR1) section is based on biostratigraphy, the Ir anomaly, and the identification of impact-related tsunami beds (Fig. 3). Here we update this age model with higher-resolution planktic foraminifer and organic-walled dinoflagellate cyst (dinocyst) biostratigraphy (SI Age Model). We also present a grain size record to further refine the temporal sequence of events following the impact. Finally, we generated TEX₈₆ paleo-sea surface temperature (SST) proxy record to assess temperature changes across the K–Pg boundary at Brazos River. The TEX₈₆ paleothermometer is based on glycerol dibiphytanyl glycerol tetraether (GDGT) lipids produced by marine Thaumarcheota (26). A full methodological description is available in SI Materials and Methods. Our study is the first to apply the TEX₈₆ SST proxy to reconstruct very fast (decadal) changes in deep time. Critically, mesocosm and sediment trap studies have shown that in the modern ocean, the Thaumarchaeota that produce the GDGTs adjust their membrane lipids to ocean water temperatures within weeks (27, 28). Furthermore, Thaumarchaeota have been shown to be able to grow chemoautotrophically in complete darkness (27). Therefore, strong changes in SST over months to decades, as projected to have occurred during an impact winter, can be recorded using TEX₈₆.

Fig. 3.

Age model, ${\mathrm{TEX}}_{86}^{\mathrm{H}}$, and grain size results at the Brazos-1 section. (A) Lithostratigraphy and biostratigraphy, including calcareous nannofossil, planktic foraminifera, and dinocyst biostratigraphy, at Brazos-1. (B) ${\mathrm{TEX}}_{86}^{\mathrm{H}}$-based paleo-SST reconstruction. (C) Zoom of the ${\mathrm{TEX}}_{86}^{\mathrm{H}}$ record across the K–Pg boundary interval. (D) Four different Ir records across the K–Pg boundary interval at Brazos-1, from refs. 46–49 (SI Age Model). (E) Lithology and grain size distribution on the Krumbein phi (φ) scale across the K–Pg boundary interval and first appearance datums of the first Danian dinocyst marker species.

Results and Discussion

The grain size data show that the complex of Units E, F, and G is fining upward (Fig. 2), suggesting that it was formed in a rapid depositional event (19). This material was probably initially deposited as the settling tail-end of the impact-induced tsunami/seiche (21, 22) and subsequently resuspended during postimpact storms (23, 29). Such storms might have been triggered by the enhanced contrast between warm oceans and cold atmosphere during an impact winter (13). Supporting evidence for these storms has also been found in the Geulhemmerberg section in The Netherlands, where the lowermost Danian is characterized by an alternating sequence of shell hashes and clays, interpreted to be related to episodic storm wave activity alternated with unusually low energy conditions (30). At Brazos River, the rapid deposition of the complex of Units E, F, and G would have occurred in the waning stages of such storms. Numerical model experiments have shown that the temperature contrast causing these storms lasted for less than a century (16). Therefore, we assume that Units E, F, and G were deposited less than 100 y after the impact. The occurrence of the Ir anomaly within this complex confirms our estimation of the maximum duration of the deposition of Units E, F, and G because the settling down of PGE-bearing impact dust likely occurred on a similar timescale (31). Modern-day examples of dust input by volcanic eruptions, burning oil wells, and nuclear bomb testing resulted in settling times of months to years (32–35). Various studies have indicated that with high-energy events such as nuclear explosions and asteroid impacts, black carbon particles and aerosols are knocked into the upper stratosphere, where they quickly spread globally and produce a long-lasting climate forcing (35). With these kinds of events, particles and aerosols end up much higher than, for example, volcanic aerosols, which typically end up just above the tropopause (32). As a result, aerosols produced by large impacts will have a residence time that is considerably longer that those produced by, for example, volcanos and burning oil wells.

Some of these estimates do not include the time it takes for particles to sink from the sea surface to the seafloor, which significantly prolongs the settling time for the very fine fraction to which the impact-derived PGEs are believed to be associated (14, 31). Moreover, the proposed extraordinary large storms during the initial impact winter phase might have temporarily resuspended material, both in the atmosphere and in the ocean, further delaying the deposition of PGE-bearing impact dust. As a result, previous studies have resulted in a wide range of different estimated settling times, ranging from <10 y up to >10 kyr, although these latter values seem improbable because mechanisms to keep particles suspended for such prolonged periods of time appear lacking (31). With estimated water depths of 75–200 m (21, 22), settling time of PGE-bearing impact dust will have been in the order of 1–100 y at the Brazos River site. The scattered nature of the Ir record, combined with the dinocyst assemblages, implies that the complex of Units E, F, and G reflects a mixture of earliest Paleocene and reworked uppermost Maastrichtian materials. Given the time required to deposit the iridium on the sea floor and given the presence of burrows in the top of the underlying tsunami deposited sandstone Unit D, the rapid deposition of this mixture eventually took place years to decades after the impact. Therefore, although mixed, the complex of Units E, F, and G provides a unique insight in the environmental conditions in the first decades following the K–Pg boundary impact.

Our TEX₈₆ proxy record can be subdivided in three phases/intervals (I–III; Fig. 3). Interval I shows that Uppermost Cretaceous SSTs were stable and high, with values of ∼30–31 °C using the ${\mathrm{TEX}}_{86}^{\mathrm{H}}$ calibration (36), in agreement with published proxy records and climate model simulations for the Upper Cretaceous (37, 38). Within the tsunami deposits, TEX₈₆ values cannot be used to reconstruct SST because of high concentrations of terrestrially derived GDGTs (39) (SI Application of TEX₈₆). However, in the section directly above the main tsunami deposits (Units E, F, and G; interval II) a distinct cooling phase is recorded, with average SSTs up to 2 °C lower than preimpact values and two significant drops of up to 7 °C below preimpact values. In the subsequent Interval III, SSTs are generally 1–2 °C higher than those for the preimpact deposits.

The most remarkable features in our data are the two prominent drops in SST in the postimpact, mixed tsunami–storm deposits. As indicated above, this interval, and likely also the GDGTs, represents a mixture of redeposited uppermost Maastrichtian and immediate postimpact materials that was eventually deposited within the settling tail of tsunami activity and in the waning stage of the postimpact storms. Variable amount of mixing of reworked uppermost Maastrichtian GDGTs with basal Paleocene postimpact GDGTs might explain the multiple peaks of both our TEX₈₆ record and the Ir profiles. The chaotic nature of the basal Paleocene TEX₈₆ record contrasts with relatively stable and warm uppermost Maastrichtian SSTs. Hence, the samples that yield the lowest SSTs probably represent a mixture of uppermost Maastrichtian and direct postimpact GDGTs with a relatively high abundance of postimpact materials, causing substantially lower TEX₈₆ values. Considering the stable and warm uppermost Maastrichtian, the immediate postimpact SSTs must have been substantially lower. Because rapid deposition of the complex of Units E, F, and G occurred within 100 y after the impact, the cooling recorded in these units likely happened in the first months to decades following the K–Pg impact. Our SST record thus provides, to our knowledge, the first evidence for a transient, global impact winter after the K–Pg boundary impact. The duration of this cold spell is in agreement with coupled ocean–atmosphere model results, suggesting that impact-induced dust and aerosol loading will result in lower SSTs for several decades, even after most of the dust has been removed from the atmosphere (40). Our results of short-term cooling following the K–Pg asteroid impact are supported by a migration of cool, boreal dinoflagellate species into the subtropic Tethyan realm directly across the K–Pg boundary interval (16, 41) and the ingression of boreal benthic foraminifera into the deeper parts of the Tethys Ocean, interpreted to reflect millennial timescale changes in ocean circulation following the impact, attributed to a hypothesized short-term cooling of 1–10 y (16).

The global impact winter, characterized by darkness and a dramatic cooling of ocean surface waters, perturbed a relatively stable, warm, latest Cretaceous climate (37) and likely represented a major stress factor for life on Earth. Therefore, it is expected to have been a key contributory element in the mass extinction at the K–Pg boundary. Additionally, when the large amount of aerosols injected into the atmosphere rained out, they might have resulted in acidification of the surface oceans (42), a further stressor for surface-dwelling organisms. The initial cooling was followed by a long-term warming trend (41), also observed in our TEX₈₆ record (Fig. 2) and in previously reported stable isotope analyses (43), that most likely is associated with greenhouse gasses released into the atmosphere from the vaporization of carbonate target rock, the mass mortality, and forest fires (6, 8, 10, 41). Our study reveals a combination of environmental and climatological events that is compatible with the pattern of extinction of many biological clades, including most species of planktic foraminifera and many coccolithophorids but also larger marine taxa like ammonites and marine reptiles, in addition to the dinosaurs and flying reptiles (1, 2, 16).

Materials and Methods

For TEX₈₆ analyses, freeze-dried, powdered samples (∼10 g dry mass) were extracted with an accelerated solvent extractor using a 9:1 (vol/vol) dichloromethane (DCM):methanol solvent mixture. The obtained extracts were separated over an activated Al₂O₃ column, using 9:1 (vol/vol) hexane:DCM, ethyl acetate (100%), 95:5 (vol/vol) DCM:MeOH, and 1:1 (vol/vol) DCM:methanol, into apolar, ethylacetate, and tethraether and polar fractions, respectively. The tetraether fractions were analyzed by HPLC/APCI-MS (HPLCy/atmospheric pressure positive ion chemical ionization mass spectrometry) using an Agilent 1100 series Liquid Chromatography/Mass Spectrometric Detector type SL. TEX₈₆ indices were calculated and converted into temperature estimates as described in SI Materials and Methods.

For palynological analyses, oven-dried samples (∼10–15 g dry mass) were spiked with Lycopodium spores to facilitate the calculation of absolute palynomorph abundances. Chemical processing comprised treatment with 10% HCl and 40% HF for carbonate and silica removal, respectively. Ultrasonication was used to disintegrate palynodebris. Residues were sieved over a 15-µm mesh and mounted on microscope slides, which were analyzed at 200× and 1,000× magnification to a minimum of 200 dinocysts. A detailed, step-by-step processing protocol is given in SI Materials and Methods. Taxonomy used follows that cited in the Lentin and Williams Index of Fossil Dinoflagellates, 2004 (44), unless stated otherwise. See SI Taxonomic Notes on Organic-Walled Dinoflagellate Cysts for taxonomical notes. All slides are stored in the collection of the Laboratory of Paleobotany and Palynology, Utrecht University.

For analyses of planktic foraminifera, standard micropaleontological techniques were applied (SI Materials and Methods).

The grain size distribution was determined on a Fritsch A-22 laser particle sizer (SI Materials and Methods).