Isotopic Evidence for Glaciation During the Cretaceous Supergreenhouse

See allHide authors and affiliations

Science  11 Jan 2008:
Vol. 319, Issue 5860, pp. 189-192
DOI: 10.1126/science.1148777


The Turonian (93.5 to 89.3 million years ago) was one of the warmest periods of the Phanerozoic eon, with tropical sea surface temperatures over 35°C. High-amplitude sea-level changes and positive δ18O excursions in marine limestones suggest that glaciation events may have punctuated this episode of extreme warmth. New δ18O data from the tropical Atlantic show synchronous shifts ∼91.2 million years ago for both the surface and deep ocean that are consistent with an approximately 200,000-year period of glaciation, with ice sheets of about half the size of the modern Antarctic ice cap. Even the prevailing supergreenhouse climate was not a barrier to the formation of large ice sheets, calling into question the common assumption that the poles were always ice-free during past periods of intense global warming.

Despite the extreme warmth of the Turonian (13) [93.5 to 89.3 million years ago (Ma) (4)], it has been argued that there may have been several stages of continental ice growth during the period, reflected in both erosional surfaces and geochemical records associated with possible glaciation-induced sea-level falls (57). Rapid decreases (<1 million years) in sea level are known from diverse locations in the Turonian of northern Europe, North America, and the Russian Platform and are estimated at magnitudes of 25 to 40 m (7, 8) or even more (9). These rapid changes in sea level are too fast and too widespread to be accounted for by tectonic processes and were therefore plausibly triggered by glacioeustasy (7, 10). Further evidence comes from positive δ18O excursions in marine, but diagenetically altered, limestone sequences (5, 11) and brachiopod isotope data (6). However, evidence from sedimentological findings such as ice-rafted debris is still lacking, and unequivocal δ18O records of well-preserved open-ocean foraminifera are rare, are of low resolution, or do not support the idea of Late Cretaceous ice sheets (12). Moreover, there is only a poor understanding of how large ice sheets might grow in a period when tropical sea surface temperatures (SSTs) exceed 35°C (3, 13) and high-latitude temperatures are in excess of 20°C (2, 14).

We used two independent techniques to estimate SSTs during the early Late Cretaceous. One proxy is the δ18O paleothermometer, which we applied to monospecific planktic foraminiferal samples from a 40-m-thick, organic carbon–rich, laminated marlstone succession of Turonian to Santonian age from the western equatorial Atlantic at Demerara Rise [Ocean Drilling Program (ODP) Site 1259]. These sediments contain planktic foraminifera with a “glassy” appearance and pristine, well-preserved wall textures (15). This state of preservation is ideal for an extensive geochemical investigation in order to reconstruct past SSTs (16). Comparison of the δ18O values of planktic and benthic foraminifera was also used to provide information on the global isotopic composition of oceans when compared against a salinity-independent temperature proxy: the tetraether index of lipids with 86 carbon atoms (TEX86), which is based on the distribution of crenarchaeol membrane lipids (17, 18). Because the growth of continental ice enriches seawater in 18O, the δ18O chemistry, when constrained by TEX86 temperature estimates, can be further used to estimate the size of continental ice sheets.

Our tropical δ18O-derived SST estimates range from ∼34° to ∼37°C [–4.2 to –4.9 per mil (‰), Vienna Pee Dee belemnite (VPDB) standard] in the Turonian and from ∼31.5° to ∼35°C (–3.65 to –4.4‰ VPDB) in the late Coniacian and Santonian (Fig. 1A). These values are in good agreement with earlier estimates from spot measurements from the western Atlantic (1, 3) and with low-resolution data from the organic paleothermometer TEX86 (13). Newly generated TEX86 data indicate high SSTs of up to 36°C during the Turonian, followed by a shift to cooler temperatures during the Coniacian and Santonian (Fig. 1B). Today, temperatures in the western tropical Atlantic range from 28° to 29°C (19), so our data suggest that the Turonian surface ocean was 5° to 9°C warmer than at present.

Fig. 1.

Stratigraphy and calculated SSTs based on planktic foraminiferal δ18O and TEX86 data for the Turonian to Santonian interval at ODP Site 1259 (33), Demerara Rise, western equatorial Atlantic. (A) δ18O data based on monospecific planktic foraminiferal and corresponding conservative SST estimates. (B) TEX86 values and the calculated SSTs. The late Turonian δ18O peak interval is estimated to represent ∼200,000 years (200 ky). T, temperature.

The Turonian warmth was punctuated by a short-lived decrease in the δ18O content of planktic foraminifera ∼91.34 Ma [523 meters of composite depth (mcd)] and a pronounced positive δ18O excursion centered on the CC11/CC12 calcareous nannofossil biozone boundary ∼91.2 Ma (521.5 mcd) (Figs. 1A and 2A). The most negative δ18O ratios in our record are found ∼91.34 Ma, but the implied SST peak (>36°C) is not reflected in either the benthic foraminiferal δ18O or the TEX86 data (Fig. 2). Evidently, this δ18O event reflects an episode of acceleration of the hydrological cycle, causing a decrease in the local sea surface salinity (Fig. 2C), and was followed at ∼91.2 Ma by a synchronous positive shift in δ18O in both planktic and benthic foraminifera, which lasted for ∼200,000 years. The magnitude of the positive δ18O excursion in planktic foraminifera is >1‰, whereas it is up to 0.7‰ in benthic foraminifera.

Fig. 2.

Detailed data from the CC11 and CC12 calcareous nannofossil biozone interval of ODP Site 1259 (33). Data are plotted against absolute age (4) (table S1) and cover the interval from 518.02 to 530.78 mcd in Fig. 1. (A) The raw planktic foraminiferal δ18O data (see Fig. 1 for symbol explanation) and TEX86 data; both are plotted on the same scale with respect to the estimated SSTs in Fig. 1. (B) δ18O and TEX86 anomalies were calculated in reference to sample 1259B-22-4, 75–76.5 cm (530.08 mcd, 91.93 My). In (B) and (C), only those samples are shown from which both proxy data types (δ18O and TEX86) are available. These anomalies were then converted to expected δW anomalies (C) by application of a δ18O/T relationship of –0.208‰/°C. The resulting δW anomaly was then assumed to reflect changes corresponding to surface water salinity changes due to shifts in the precipitation/evaporation balance. The salinity anomaly field represents a δW/salinity relationship of 0.30‰/practical salinity units (p.s.u.) (33). (D) δ18O data from two benthic foraminiferal taxa show a positive δ18O shift of 0.3 to 0.7‰ within ∼200,000 years, which parallels the δW anomaly.

Because TEX86 temperature estimates are independent of changes in seawater δ18O (δW), we calculated temperature anomalies based on paired measurements of both δ18O and TEX86 to quantify the change in δW (Fig. 2C) produced by the possible growth of continental ice. The TEX86 data display only a 1°C decrease in surface ocean temperatures associated with the 91.2 Ma planktic oxygen isotope shift. Therefore the remaining 0.4 to 0.6‰ anomaly during the 91.2 Ma event in planktic foraminiferal δ18O, which is similar to the range suggested by the benthic foraminifera (0.3 to 0.7‰), must primarily reflect changes in δW rather than ocean temperature alone. Both the surface ocean and the deep sea floor (estimated to be at depths >1500 m) (20) show the positive δ18O anomaly, making it unlikely that this signal reflects local changes in surface ocean salinity. It is much more likely that the synchronous change in benthic and planktic foraminiferal δ18O was produced by sequestering 16O in glacial ice, causing a whole-ocean increase in δW.

The middle Turonian is characterized by a series of short-term relative sea-level changes (9, 21) that are consistent with our glacioeustatic interpretation of the isotopic record. Two major widespread unconformities, Tu-2 at 91.2 Ma and Tu-3 at 90.9 Ma (21, 22), correspond to sea-level falls of at least 25 m (9, 21) and occur close to the CC11/CC12 nannofossil zone boundary (91.2 Ma) (4). Data from the Russian Platform (8) suggest a drop in sea level of up to 40 m at ∼91 Ma, and a further drop by ∼25 to 30 m is reported from the New Jersey margin at the CC11/CC12 boundary (7, 10). The sea-level record from the Russian Platform is particularly important because this region is generally regarded as tectonically stable, making it a particularly good place to estimate global sea-level changes. A large-scale unconformity has also been reported from the Western Interior Basin (23) and northwest Europe (24) in the late middle Turonian. Because of stratigraphic uncertainties and the lack of sophisticated supraregional stratigraphic concepts, it is not clear which of these unconformities correspond to the observed positive δ18O shift. However, the widespread occurrence of high-amplitude relative sea-level changes supports the hypothesis that continental ice may have formed in the middle Turonian. Our biostratigraphy and the δ13C record (fig. S2) suggest that the positive δ18O shift is synchronous with the Pewsey Event in western Europe (25, 26), which is considered to have been associated with regional cooling based on bulk-rock δ18O data and faunal patterns from western Europe (11).

The magnitude of the sea-level fall associated with the 91.2 Ma δ18O shift is constrained by four variables: (i) our 0.3 to 0.7‰ δW anomaly, (ii) the magnitude of the late Turonian sea-level fall [∼25 to 40 m (7, 8)], (iii) an estimate of the relationship between sea level and δ18O [0.11‰ VPDB δ18O per 10 m of sea-level fall in the Quaternary (27) and 0.075‰ VPDB δ18O per 10 m of sea-level fall in a warm climate scenario (7)], and (iv) the estimated δ18O composition of Cretaceous ice. The first three variables are well known, whereas the isotopic composition of Cretaceous ice can be constrained between the average modern value for the Antarctic ice cap [–44‰ Vienna standard mean ocean water (VSMOW)] (7) and the predicted ice composition for past warm climates (–30‰ VSMOW) (7, 28). Given the modern δ18O composition of the Antarctic ice cap, the observed δ18O anomaly is consistent with a sea-level fall of 27 to 64 m. Today, the Antarctic ice sheet stores sufficient water to change global sea level by 61 m (29). Therefore, our calculated volume of the Cretaceous ice sheet is equivalent to 44 to 105% that of the modern Antarctic ice sheet. If Cretaceous ice had a δ18O composition of –30‰ VSMOW, the 91.2 Ma event would be consistent with a sea-level fall of 40 to 93 m and an ice volume 66 to 152% that of modern Antarctica.

We propose that any large Turonian ice sheet was probably located on Antarctica, given the polar position of the continent and the widespread areas of elevated terrain (with altitudes of 1500 to 2500 m) when the modern ice cap is isostatically removed (30). However, the uplift history of the Transantarctic Mountains before the Cenozoic is very poorly known and may have commenced during the Cretaceous (31) or the Eocene (32). Warm tropical and subpolar SSTs in the Turonian (13) would seem to preclude substantial ice development at or near sea level, even on Antarctica, emphasizing the need for further work on the paleoelevation history of the continent.

It is unlikely that an ice sheet of the size of the modern Antarctic ice cap existed in the Cretaceous, both because of the warm surface temperatures noted above and because there is no evidence for the ice-rafted debris that should be present in the Southern Ocean if all of Antarctica had been glaciated. However, an ice cap of up to ∼60% the size of the modern Antarctic ice sheet is plausible given the constraints imposed by the sea-level record, as well as our estimate of the change in mean ocean δW, and our –44‰ VSMOW estimate for the isotopic composition of glacial ice. These results also imply that the δ18O composition of Turonian polar precipitation was not substantially heavier than today and contradict expectations that greenhouse warming should decrease fractionation during vapor transport from mid- to high latitudes (7, 28).

We are left with the apparent paradox that the prevailing extraordinarily high tropical temperatures during the Turonian were not a barrier to the initiation and growth of large continental ice sheets. The development of these ice sheets could be attributed to an increase in the activity of the hydrological cycle, which must have initiated more humid conditions and enhanced precipitation in the high latitudes. As with periods of Cenozoic ice growth, the initiation of Cretaceous ice expansion may have been triggered by orbital dynamics, because the ∼200,000-year duration of the 91.2 Ma event is similar to the half period of the 400,000-year eccentricity cycle. Our results suggest that fairly large ice sheets could grow and decay equally rapidly, which is very much the same pattern as during the Pleistocene. However, unlike the Pleistocene, Cretaceous ice sheets were apparently not a regularly recurring phenomenon, possibly because the extreme warmth of the Turonian, the paleoelevation of Antarctica, and the orbital configuration allowed the initiation of ice sheet development only under certain rare conditions. Our results further suggest that the common assumption that ice sheets did not exist during periods of past supergreenhouse climates should be reexamined, with implications for paleotemperature estimation, the determination of the past isotopic composition of seawater, and high-latitude terrestrial climate reconstruction.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S4

References and Notes

References and Notes

View Abstract

Navigate This Article