Mesozoic atmospheric CO₂ concentrations reconstructed from dinosaur tooth enamel

Significance

Paleoclimate is closely linked to atmospheric pCO₂. Quantifying ancient CO₂ levels, however, is challenging. Air-breathing vertebrates respire air O₂ and incorporate its isotope signature via body water into their hard tissues. Fossil tooth enamel can thus serve as a robust time capsule for ancient air O₂ isotope compositions. Air O₂ has an ¹⁷O-anomaly that increases with increasing atmospheric pCO₂ and decreases with increasing gross primary productivity (GPP). Therefore, paleo-pCO₂ or paleo-GPP, respectively, can be determined by oxygen isotope measurements of fossil tooth enamel. Here, we reconstruct Mesozoic paleo-pCO₂ levels from the triple oxygen isotope composition of dinosaur teeth and obtain paleo-pCO₂ levels 2.5 to 4 times higher than preindustrial values. In addition, changes in the ¹⁷O-anomaly could also point to substantial fluctuations in GPP of the biosphere.

Abstract

Air-breathing vertebrates incorporate a fraction of isotopically anomalous air O₂ in their body water. The ¹⁷O isotope anomaly of air O₂ (expressed as Δ’¹⁷O_air) is related to atmospheric CO₂ concentrations (pCO₂) and gross primary production (GPP). Tooth enamel records the Δ’¹⁷O of body water and can thus preserve such paleo-pCO₂ or paleo-GPP information over geological time periods. Here, we demonstrate the potential of respective reconstructions of atmospheric pCO₂ or GPP from the triple oxygen isotope composition of fossil dinosaur tooth enamel. The data from unaltered enamel samples, along with an assumed modern GPP_t/GPP₀ ratio of 1 for the Mesozoic, suggest a mean Late Jurassic pCO₂ = 1,200 ± 150 ppmv and Late Cretaceous pCO₂ = 750 ± 200 ppmv. These estimates are in good agreement with other pCO₂ proxy data for the same time intervals. When utilizing a pCO₂ inferred from other proxies, tooth enamel Δ’¹⁷O_PO4 may also serve as a proxy for GPP. Using published pCO₂ data, we reconstructed GPP_t/GPP₀ ratios with 1.20 ± 0.17 for the Late Jurassic and 2.24 ± 0.96 for the Late Cretaceous, which would imply a 20 to 120% higher GPP in the Mesozoic than today. Overall, triple oxygen isotope analysis of fossil teeth of terrestrial amniotes can provide insights into past atmospheric greenhouse gas content and global primary productivity.

Atmospheric carbon dioxide plays an important role in global climate throughout Earth’s history. Arrhenius (1896) (1) first discussed that variations in atmospheric CO₂ concentrations can be related to climate changes in geological history. A close correlation between atmospheric CO₂ partial pressure (pCO₂) and air temperature is well established for the past ≈1 Ma from the ice core record of the Quaternary (2–4). Pre-Quaternary atmospheric CO₂ levels are estimated from mass balance models (5, 6) or various geological and geochemical proxies, such as paleosols, phytoplankton, stomata density, boron and calcium isotopes in marine carbonates, liverworts, and nahcolite (7, 8). Each of these proxies is associated with its individual limitations; thus, combining pCO₂ proxies and models is a viable means for refined pCO₂ reconstructions (7, 8). A new proxy is the triple oxygen isotope composition of bioapatite from air-breathing vertebrates (9–12).

Pioneering studies suggested using triple oxygen isotopes in sulfates (13), carbonates (14), and tooth enamel (9, 10) as proxy for the Δ’¹⁷O of air. The latter can be used to obtain information about past atmospheric pCO₂ if gross primary productivity (GPP) is known, or conversely, if pCO₂ is known, to infer GPP (14). The negative Δ’¹⁷O of air O₂ counterbalances the positive Δ’¹⁷O of the stratospheric CO₂ and ozone reservoirs (15). Therefore, Δ’¹⁷O of air O₂ decreases with increasing pCO₂. The negative anomaly of air O₂, in turn, is diluted by the flux of isotopically normal photosynthetic O₂, and hence, Δ’¹⁷O_air increases with increasing GPP. Therefore, Δ’¹⁷O of air O₂ can be used as proxy for either pCO₂ or GPP.

Pack et al. (9) demonstrated that air-breathing vertebrates record a fraction of the ¹⁷O anomaly of air oxygen in their skeletal bioapatite. This is because the inhaled, isotopically anomalous O₂ is utilized for aerobic metabolism of food, and the respective ¹⁷O anomaly is thus first transferred to the body water (BW) and subsequently incorporated from there into the bioapatite. The diagenesis-resistant tooth enamel thus provides a time capsule of the in vivo incorporated ¹⁷O anomaly in ancient air O₂. Using this approach, Gehler et al. (10) constrained paleo-pCO₂ levels during the Paleocene–Eocene Thermal Maximum; assuming different scenarios for the GPP_PETM. Feng et al. (11, 12) reported data on bioapatite of modern terrestrial and marine mammals as well as of sharks and calculated taxon-specific oxygen mass balance models across a large body mass range, providing a thorough empirical baseline for the pCO₂ reconstruction from bioapatite of air-breathing vertebrates.

In this study, we further extend the approach to modern birds and a reptile (i.e., Sauropsida), compare data to those of modern mammals and apply the results to extinct dinosaurs. The dinosaur teeth analyzed herein come from five different fossil sites in the United States, Canada, Morocco, Germany, and Tanzania, representing the two geological periods, the Late Jurassic (latest Oxfordian to Tithonian) as well as the Late Cretaceous (middle Cenomanian to late Maastrichtian). The specimens include genera such as sauropods (Camarasaurus, Giraffatitan, Kaatedocus, and Europasaurus), theropods (Tyrannosaurus, Albertosaurus, Carcharodontosaurus, and Torvosaurus), and ornithischians (Edmontosaurus) (see Dataset S1 for further details). We present high precision phosphate triple oxygen isotope data of dinosaur teeth, both enamel and dentine, which we first screen for diagenetic alteration. We use only the pristine enamel samples to reconstruct pCO₂ during the Mesozoic greenhouse climate, specifically for the Late Jurassic (Kimmeridgian to Tithonian) and Late Cretaceous (Campanian to Maastrichtian).

Our approach has a range of inherent uncertainties mainly originating from: i) the unknown Jurassic and Cretaceous GPP; and ii) errors related to physiological and environmental parameters. We demonstrate how these challenges can be overcome and how triple oxygen isotope data may also serve as a tool to study vertebrate paleophysiology.

1. Results

A list with detailed information on the taxonomy, provenance, body mass, and geological age (for the dinosaur specimens) of all analyzed modern and fossil Sauropsida bone and tooth samples is provided in Dataset S1.

The triple oxygen isotope data (Δ’¹⁷O_PO4 vs. δ¹⁸O_PO4) for bone bioapatite of modern birds are listed in Dataset S2. The bird data are plotted in Fig. 1 and are fully overlapping with those of modern mammals from ref. 12. The data point for the alligator falls at the upper end of the field covered by modern mammals and birds (Fig. 1).

Fig. 1.

Modern bird bone and alligator tooth enamel δ¹⁸O_PO4 vs. Δ’¹⁷O_PO4 data. Modern terrestrial mammals’ enamel data (green shaded region) is from ref. 12 and represent the distribution of modern herbivores, omnivores, and carnivores from Central Europe (Germany), Africa, Australia, North America, and South America, with darker colors indicating a higher concentration of overlapping data. The 1σ SD of Δ’¹⁷O_PO4 values from modern mammal data (12) and the bird and reptile data from this study are ±34 ppm. Meteoric water (blue shaded region; for details and references, see Feng ref. 12) is displayed for comparison. The data displayed as orange circles are calculated bioapatite δ¹⁸O_PO4 and Δ’¹⁷O_PO4 values based on data from ref. 16 (carbonates in eggshells), ref. 17 (blood), and ref. 18 (carbonates in eggshells) (details about the calculation are provided in Datasets S4–S6, respectively).

The triple oxygen isotope data and the δ¹³C and δ¹⁸O_CO3 values of the structural carbonate in the bioapatite of the dinosaur teeth are listed in Dataset S3. The offset between phosphate and carbonate oxygen isotope composition (δ¹⁸O_CO3-δ¹⁸O_PO4) typical for modern bioapatite is taken as an indicator to assess diagenetic alteration of the oxygen isotope composition of the fossil dinosaur teeth (12).

Dinosaur dentine and enamel Δ’¹⁷O_PO4 values show a large scatter with unaltered enamel (see Discussion for details) plotting toward the lower range of the field of modern mammals and birds, but reaching down to values as low as -300 ppm (Fig. 2), which represent the lowest values reported for vertebrates so far.

Fig. 2.

Plots A and B illustrate the difference between the δ¹⁸O values of the structural bioapatite carbonate (δ¹⁸O_CO3) and phosphate (δ¹⁸O_PO4) for dentine (open symbols) and enamel (filled symbols) of dinosaur tooth samples. The gray shaded areas represent the range from 7.4 ≤ (δ¹⁸O_CO3-δ¹⁸O_PO4) ≤ 13.0‰ obtained from modern mammal bioapatite analyses (n = 96) (12). Filled black symbols indicate apparently diagenetically altered enamel. The majority of analyzed dentine samples (open symbols) show a clear sign of diagenetic alteration (i.e., δ¹⁸O_CO3-δ¹⁸O_PO4 offset > 13‰). Plots C and D illustrate the triple oxygen isotope composition (Δ’¹⁷O_PO4 vs. δ¹⁸O_PO4) of tooth enamel and dentine samples from this study. The green-shaded areas represent the combined data of modern birds (this study) and mammals from ref. 12. Note that diagenesis tends to shift Δ’¹⁷O_PO4 values toward more positive Δ’¹⁷O values and more negative δ¹⁸O values of meteoric water. Thus, Δ’¹⁷O_PO4-based pCO₂ reconstructions represent minimum estimates.

2. Discussion

2.1. Comparison of Δ’¹⁷O_PO4 of Modern Birds and Terrestrial Mammals.

As illustrated in Fig. 1, triple oxygen isotope ratios of modern bird bioapatite plot in the same range as those of modern terrestrial mammals (12). Birds of similar size and feeding ecology typically have higher basal metabolic rates than terrestrial mammals (17). However, a higher metabolic rate alone does not necessarily lead to systematically lower Δ’¹⁷O_PO4 in birds. Instead, the relation between Δ’¹⁷O_PO4 of bioapatite and Δ’¹⁷O_air of inhaled air is affected by a number of factors, including body mass and water intake. This relationship has been modeled as a function of body mass (allometric scaling) using a mass balance model (9, 12). The observed similarity in Δ’¹⁷O_PO4 between birds and mammals is likely due to the higher specific water flux of birds compared to mammals (19, 20), which counterbalances the effect of higher basal metabolic rates of birds.

The physiology of the studied dinosaurs is a priori unknown. However, birds and crocodiles are the closest living relatives of dinosaurs forming their extant phylogenetic bracket (21, 22). Birds are warm-blooded descendants of nonavian dinosaurs that evolved from small theropod ancestors and are considered the only surviving lineage of dinosaurs, making them “living dinosaurs” in a phylogenetic sense. Therefore, birds (i.e., avian dinosaurs) may be suitable analogs here. The alligator falls at the upper end of the field covered by modern mammals and birds (Fig. 1). Our triple oxygen isotope data of recent birds demonstrate that their oxygen mass balance resembles that of mammals (Fig. 1). As a first approximation, we thus apply the Δ’¹⁷O_PO4 vs. body mass relationship obtained from modern mammals (12). This allows reconstructing the Δ’¹⁷O of air O₂ that was inhaled by the Late Jurassic and Late Cretaceous dinosaurs during their lifetime (see caption of Fig. 3 for details). The Δ’¹⁷O of air can then be used to estimate either pCO₂ or GPP.

Fig. 3.

The plot shows models of bioapatite Δ’¹⁷O vs. logarithmic body mass (M_b, in kg) as calculated by ref. 12 for 42 recent taxa (“modern terrestrial herbivorous mammal line”), but solved for variable Δ’¹⁷O_air to fit measured Δ’¹⁷O_PO4 values from well-preserved dinosaur enamel samples (see SI Appendix, Table S1 for details regarding the fossil samples). (A) Different theropods from the Late Cretaceous. The green curve represents the average Δ’¹⁷O_air = –0.63‰ reconstructed from Tyrannosaurus rex (No. 2), theropod indet. (No. 4), and Albertosaurus sarcophagus (No. 5). The blue curve represents the Δ’¹⁷O_air = –1.10‰ reconstructed from T. rex (No. 1). (B) Different sauropods from the Late Jurassic. The purple curve represents the average Δ’¹⁷O_air = –0.85‰ reconstructed from Europasaurus holgeri (No. 11), Giraffatitan brancai (No. 12), and three other undetermined sauropod specimens (No. 13, 14, 15). The blue curve represents the extremely low Δ’¹⁷O_air = –1.60‰ reconstructed from Kaatedocus siberi (No. 10). Dashed lines around modeled curves reflect a 1σ (SD) CI from the modern data from ref. 12 and the birds and reptile data from this study to ± 34 ppm in Δ’¹⁷O_PO4. The gray curve is the calibration line for modern terrestrial herbivorous mammals after ref. 12. Data are listed in Dataset S3.

2.2. Preservation of Primary Oxygen Isotope Compositions in Fossil Teeth.

The oxygen isotope composition of fossil bioapatite may undergo diagenetic alteration (23–27), which would obliterate the original isotope information. We have thus screened the studied fossil tooth samples for diagenetic alteration. In unaltered bioapatite, the difference between δ¹⁸O_CO3 of structural carbonate and δ¹⁸O_PO4 of the main phosphate component is 9.8 ± 1.5‰ (12, 23, 28–30). This difference can change during fossilization processes as the oxygen bound in the carbonate and phosphate group has different sensitivities against alteration processes, and hence, the δ¹⁸O_CO3-δ¹⁸O_PO4 offset is a good indicator for the diagenesis of the bioapatite oxygen isotope composition. As expected, most dentine samples show elevated δ¹⁸O_CO3-δ¹⁸O_PO4, reflecting their greater susceptibility to diagenesis compared to well-crystallized and more mineralized tooth enamel, for which many samples suggest that their oxygen isotope composition was not affected by alteration (Fig. 2 A and B).

Based on the δ¹⁸O_CO3-δ¹⁸O_PO4 criterion, we conclude that the tooth enamel samples of the Late Jurassic sauropods Kaatedocus siberi, Europasaurus holgeri, Giraffatitan brancai (No. 12), and three other sauropods indet. (Nos. 13, 14, and 15) as well as of the Late Cretaceous Tyrannosaurus rex (Nos. 1 and 2), theropod indet. (No. 4), and the Albertosaurus sarcophagus have not experienced a significant diagenetic alteration (Fig. 2) and thus can be used as a proxy archive for the Δ’¹⁷O of air O₂ at the time when the dinosaur was alive. The dentine and altered enamel from dinosaurs exhibit higher Δ’¹⁷O values than the unaltered enamel, as shown in Fig. 2 C and D, which suggests diagenesis tends to shift bioapatite Δ’¹⁷O values toward more positive values. Hence, the Δ’¹⁷O_PO4 values of unaltered enamel represent an upper limit of the original in vivo compositions, and Δ’¹⁷O_PO4-based pCO₂ reconstructions should thus be regarded as a minimum estimate. In this contribution, we discarded all samples showing any sign of diagenesis from our discussion.

2.3. Triple Oxygen Isotope Ratios of Tooth Enamel as Paleo-Δ’¹⁷O_air Proxy.

Bone and tooth bioapatite (PO₄) precipitates from BW in oxygen isotope equilibrium (24). The isotope composition of BW is a function of in- (F_i) and outfluxes (F_j) and their respective isotope compositions (δ^17,18O_{i, j}) (SI Appendix, section 2). For this study, the Δ’¹⁷O of the ambient air O₂ is of particular interest. Oxygen fluxes can be modeled on the basis of allometric scaling laws (31) or physiological and environmental data for individual species (32). Isotopic compositions and fractionations are known from published datasets (9, 12).

Because animal physiology and environmental parameters show considerable variation even within populations of the same species in the same habitat (12), we follow a different approach rather than strictly applying allometric scaling of fluxes. Feng et al. (12) have modeled the composition of bioapatite for a set of 42 modern taxa of land-living herbivorous mammals comprising a broad range of body masses (M_b) (0.002 ≤ M_b ≤ 6,000 kg) as well as covering various environments and different physiologies. The 42 taxon-specific mass balance models were fitted to measured bioapatite δ¹⁸O_PO4 and Δ’¹⁷O_PO4 data of 42 modern herbivorous mammal taxa, living and breathing in today’s atmosphere. Each mass balance was based on the model of Pack et al. (9). A value of modern Δ’¹⁷O_air = –0.432 ± 0.015‰ (33) is applied in all these models, which are used to define the modern “herbivorous mammal curve” in a plot of Δ’¹⁷O_bioapatite vs. body mass M_b (Fig. 3).

In certain periods of the Earth history, Δ’¹⁷O_air may have been much lower than today (13). To simulate such scenarios of changing Δ’¹⁷O_air, we simultaneously lowered the Δ’¹⁷O_air in all 42 mammal models and obtained characteristic Δ’¹⁷O_bioapatite vs. M_b curves for different Δ’¹⁷O_air (Fig. 3). Such simulations of the entire ensemble of these 42 BW models are used to infer ancient air Δ’¹⁷O_air for a given Δ’¹⁷O_PO4. Specifically, for each well-preserved dinosaur enamel sample the measured Δ’¹⁷O_PO4 can be fitted onto such a curve by tuning Δ’¹⁷O_air accordingly.

From the dataset of modern terrestrial mammals, birds, and the reptile, we deduce an overall uncertainty intrinsic to the Δ’¹⁷O_PO4 of 34 ppm. The uncertainty in reconstructed Δ’¹⁷O_air is obtained by fitting the measured Δ’¹⁷O_PO4 to the upper and lower Δ’¹⁷O_air error envelope of the respective Δ’¹⁷O_bioapatite vs. M_b curve. The uncertainty in reconstructed Δ’¹⁷O_air then is used to assess the respective uncertainties in paleo-CO₂ and/or GPP_t/GPP₀ (SI Appendix, Eq. S4). This practical approach to approximate an error for Δ’¹⁷O_air is complemented by a comprehensive sensitivity analysis of the mass balance model provided in the Supplementary material.

2.4. Δ’¹⁷O_PO4 of Tooth Enamel: A Proxy for Δ’¹⁷O_air and Thus Paleo-pCO₂ or Paleo-GPP.

The enamel triple oxygen isotope data from Late Cretaceous theropod teeth of T. rex (No. 2), an unidentified theropod (No. 4), and A. sarcophagus suggest an overall common Δ’¹⁷O_air = –0.63 ± 0.09‰, which is significantly lower than today (Fig. 3). Assuming a modern-day GPP_t/GPP₀ = 1, this transforms to a pCO₂ = 750 ± 200 ppmv (SI Appendix, Eq. S4). Varying the GPP_t/GPP₀ to 0.67 and 1.5 would yield Late Cretaceous pCO₂ of 500 ± 135 and 1,100 ± 300 ppmv (SI Appendix, Fig. S2), respectively. The data point of T. rex (No. 1) plots at a very low Δ’¹⁷O_PO4. Such a low Δ’¹⁷O_PO4 would imply a Δ’¹⁷O_air as low as –1.10 ± 0.07‰, which would suggest a high pCO₂ = 1,800 ± 250 ppmv (at GPP_t/GPP₀ = 1).

The enamel triple oxygen isotope data from Late Jurassic sauropod teeth of E. holgeri, G. brancai, and the three undetermined sauropod specimens (No. 13, 14, 15) can be well-explained by a common Δ’¹⁷O_air of –0.85 ± 0.07‰. Assuming a modern-day GPP_t/GPP₀ = 1, this transforms to a pCO₂ = 1,200 ± 150 ppmv (SI Appendix, Eq. S4). Setting the GPP_t/GPP₀ to 0.67 and 1.5 would yield Late Jurassic pCO₂ of 830 ± 100 and 1,800 ± 230 ppmv, respectively (SI Appendix, Fig. S2). The data for the diplodocid K. siberi show significantly lower Δ’¹⁷O_PO4 than all the other dinosaurs from the Late Jurassic, which suggests Δ’¹⁷O_air of –1.60 ± 0.11‰. This would correspond to a high pCO₂ = 2,900 ± 250 ppmv (at GPP_t/GPP₀ = 1).

Because diagenesis would unidirectionally tend to increase the Δ’¹⁷O_PO4, any diagenetic overprint would lead to an underestimation of the pCO₂ or GPP_t/GPP₀. Hence, low Δ’¹⁷O_PO4 values observed for the Late Jurassic K. siberi and Late Cretaceous T. rex (No. 1) are likely not the result of diagenetic alteration.

The reconstructed pCO₂ data for the Late Jurassic and Late Cretaceous agree well with other paleo-CO₂ proxy data for these periods. The apparently high pCO₂ values suggested by the low-Δ’¹⁷O_PO4 enamel samples from the Late Jurassic and Late Cretaceous are significantly higher than average values suggested by other proxies, but would still fall within the range of reconstructed values, in particular they fit to some high pCO₂ data inferred from soil carbonates (34, 35).

The Δ’¹⁷O of air O₂ does not only reflect variations in pCO₂ but also variations in GPP_t/GPP₀ ratios. Using published pCO₂ estimates of Royer (7) and Beerling and Royer (8) of 1,500 ppmv for the Late Jurassic and 1,000 ppmv for the Late Cretaceous from other proxy data, respective GPP_t/GPP₀ estimates are 1.20 ± 0.17 (Δ’¹⁷O_air = –0.85 ± 0.07‰, SI Appendix, Eq. S4) for the Late Jurassic and 2.24 ± 0.96 (Δ’¹⁷O_air = –0.63 ± 0.09‰, SI Appendix, Eq. S4) for the Late Cretaceous.

2.5. Dinosaur Enamel Samples with Low Δ’¹⁷O_PO4.

One explanation for the T. rex (No. 1) and K. siberi having lower Δ’¹⁷O_PO4 than other near-contemporaneous dinosaurs is fluctuations in Δ’¹⁷O_air at geologically short time scales, i.e., Δ’¹⁷O_air was low at the time during which these animals lived. For the Late Cretaceous (Maastrichtian), T. rex (No. 1) would then suggest a pCO₂ 1,800 ± 300 ppmv, which is 1,000 ppmv higher than what is obtained from other theropod dinosaurs (including other T. rex specimens) of the same time period (Fig. 4). For the Late Jurassic K. siberi, the reconstructed pCO₂ is 1,700 ppmv higher than from specimens of other sauropod taxa of the same time period (Fig. 4).

Fig. 4.

Plot of pCO₂ (ppmv) vs. time in millions of years (Ma) from a compilation of different available pCO₂ proxies (7, 8). Colored symbols reflect the pCO₂ concentrations reconstructed from well-preserved dinosaur enamel samples. Black stars: average pCO₂ during the Late Jurassic and Late Cretaceous reconstructed from the dinosaur tooth enamel assuming a GPP_t/GPP₀ = 1.

The Earth’s pCO₂ record is dynamic and comprises natural variations including spikes, for instance, during hyperthermals such as the Paleocene-Eocene thermal maximum. For this time interval, a sudden increase in pCO₂ by 1,130 ppmv (36) had been suggested. As shown in ref. 37, the Δ’¹⁷O_air of air O₂ reacts to changes in pCO₂ within a few thousands of years. The Late Cretaceous and the Late Jurassic specimens may well have lived within a time span covering a short-term pulse in pCO₂ (e.g., the end-Cretaceous Deccan trap volcanism, 38–41) that could, in principle, be recorded in Δ’¹⁷O_PO4 of dinosaur tooth enamel. Alternatively, these individuals represent outliers in terms of paleoenvironment, physiology, and/or behavior. We hereafter explore this second scenario.

For the Late Cretaceous T. rex (No. 1), the comparably high δ¹⁸O_PO4 (Fig. 2C) hints toward dryer climates with evaporatively ¹⁸O-enriched drinking water (14, 18). The sensitivity analysis demonstrates that drinking water composition and flux are dominant factors controlling the oxygen isotope composition, in particular for the large animals (SI Appendix, section 3.1.5). Surface water that is enriched in ¹⁸O also drives δ¹⁸O_FW of water in food and δ¹⁸O_F oxygen in carbohydrates, fats, and proteins to higher δ¹⁸O values. As the enrichment in ¹⁸O is related to evaporation, these oxygen sources come along with a lower Δ’¹⁷O (42). Additionally, the restricted availability of drinking water in such arid settings could also result in a lower flux of drinking water F_DW and hence a higher portion of metabolic water F_A with low Δ’¹⁷O_A. The respective effect on Δ’¹⁷O_BW in BW is demonstrated by Hu et al. (18), who present data (Dataset S6) from the Australian emu reflecting a wide range of drinking water compositions. The animals with high δ¹⁸O also comprise low Δ’¹⁷O with the total emu dataset covering an 80 ppm range in Δ’¹⁷O. From these observations, we caution that the reconstructed low Δ’¹⁷O_air from T. rex (No. 1) could at least partially reflect an artifact from a rather harsh paleoenvironment for T. rex (No. 1). Hence, this individual pCO₂ estimate may be inaccurate (i.e., represent an overestimation).

In contrast to the Late Cretaceous T. rex (No. 1), the low Δ’¹⁷O_PO4 value for the Late Jurassic sauropod K. siberi (No. 10) is not associated with elevated δ¹⁸O_PO4, but plot within the δ¹⁸O_PO4 range measured for the other Late Jurassic sauropod tooth enamel samples (Fig. 3). In order to explore whether physiological effects can explain the low Δ’¹⁷O_PO4 of this specimen, we provide detailed BW modeling in the SI Appendix. We outline a range of approaches to approximate δ¹⁸O_DW and drinking water amount as well as metabolic rates. We test these assumptions by comparing the results of the BW model to the actual data. The allometric scaling relations on metabolic rates vs. body mass of mammals and birds reveal rather similar estimates for K. siberi (No. 10) and the reptile scaling relationship calculates only a 50% lower rate (using data from ref. 43). Respective mass balance model fits can be achieved with drinking water flux of 70 L d^–1 (reptile, 3,885 mol O/d) to 150 L d^–1 (mammal, 8,324 mol O/d), corresponding to 1 to 3% of respective M_b (SI Appendix, Fig. S8). This modeling was accomplished using the Δ’¹⁷O_air reconstruction used herein and suggested a Δ’¹⁷O_air = −1.60 ± 0.11‰.

We have been unable to fit the mass balance model results to the analyzed tooth enamel data using the mean Δ’¹⁷O_air = −0.85 ± 0.07‰ determined for the other Late Jurassic sauropods (SI Appendix, Fig. S9). Therefore, we conclude that K. siberi indeed records an unusually low Δ’¹⁷O_air = −1.60 ± 0.11‰, possibly reflecting a spike in pCO₂ and/or a decline in GPP. More, well-stratified specimens from the Morrison Formation in North America need to be analyzed to further assess this hypothesis as other near-contemporaneous sauropods from the Tendaguru Formation in East Africa do not record such low Δ’¹⁷O_air.

3. Conclusions

The triple oxygen isotope data (Δ’¹⁷O_PO4, δ¹⁸O_PO4) of modern birds agree with data from a large variety of modern terrestrial mammals covering a large body mass range and a broad ecological spectrum (12). We conclude that the Δ’¹⁷O_PO4 vs. body mass relationship applies to air-breathing terrestrial amniotes in general and can thus also be used to reconstruct Mesozoic atmospheric Δ’¹⁷O_air values from well-preserved dinosaur tooth enamel. For the Late Jurassic (152 to 143 Ma), sauropod tooth enamel data suggest pCO₂ = 1,200 ± 150 ppmv (at GPP_t/GPP₀ = 1) or a GPP_t/GPP₀ = 1.20 ± 0.17 (fixing pCO₂ at 1,500 ppmv based on other proxies). For the Late Cretaceous (73 to 66 Ma), our theropod tooth enamel data suggest an average pCO₂ = 750 ± 200 ppmv (at GPP_t/GPP₀ = 1) or a GPP_t/GPP₀ = 2.24 ± 0.96 (fixing pCO₂ at 1,000 ppmv based on other proxies). For both periods, one specimen exhibits considerably lower Δ’¹⁷O_PO4. In the case of the Late Cretaceous T. rex from the Lance Formation, this could be related to an evaporatively influenced drinking water source, as indicated by the high δ¹⁸O_PO4 of this specimen. In the second case of the Late Jurassic sauropod K. siberi from the Morrison Formation, the low Δ’¹⁷O_PO4 is likely related to a short-term fluctuation in Δ’¹⁷O_air from −0.85 ± 0.07‰ to −1.60 ± 0.11‰, which may well be related to temporal high emission of CO₂ into the paleoatmosphere. Studies of larger sets of triple oxygen isotope data of fossil tooth enamel of land-living vertebrates can now be used to reconstruct the pCO₂ evolution of the atmosphere or (in concert with other pCO₂ proxy data) provide a tool to quantify changes in GPP. But this approach may also help to reveal differences in physiology and changing environmental parameters affecting these terrestrial vertebrates.

4. Materials and Methods

4.1. Definitions.

VSMOW2 denotes the international reference material Vienna Standard Mean Ocean Water 2. Oxygen isotope anomalies are expressed with the Δ’¹⁷O (Eq. 1) notation. Δ’¹⁷O is defined by the deviation of a sample with linearized forms of the δ-notations (44–46) from a given reference line (RL) with a particular slope (λRL) as follows:

$$\begin{array}{c}{{\mathrm{\Delta }}^{\prime }}^{17}{\mathrm{O}}_{0.528}=\mathrm{1,000} \times \ln \left(\frac{{\delta }^{17}{\mathrm{O}}_{\text{VSMOW}2}^{\text{sample}}}{\mathrm{1,000}} + 1\right)-{\lambda }_{\mathit{RL}}\hfill \\ \times \mathrm{1,000}\times \ln \left(\frac{{\delta }^{18}{\mathrm{O}}_{\text{VSMOW}2}^{\text{sample}}}{\mathrm{1,000}}+1\right).\hfill \end{array}$$ [1]

A λ_RL of 0.528 is used in this study. All errors are reported as 1σ SD.

4.2. Materials and Methods.

Oxygen isotopes of bones were analyzed from 10 different species of modern avian dinosaurs (i.e., birds) from Germany and Africa, as well as from 22 tooth enamel samples of fossil nonavian dinosaur taxa from the Late Cretaceous [T. rex (n = 3), theropod indet., A. sarcophagus, Edmontosaurus (n = 3), and Carcharodontosaurus], and from the Late Jurassic [K. siberi, E. holgeri, G. brancai, diplodocid sauropods (n = 5), theropods indet. (n = 2), Torvosaurus, Camarasaurus (n = 2)], as well as from one modern alligator for comparison. A detailed overview on the modern and fossil samples, including their taxonomy, body mass, provenance, stratigraphy, and geological age, is presented in the Dataset S1. We report 1σ errors throughout this study.

The bulk carbon (δ¹³C) and oxygen (δ¹⁸O_CO3) isotope compositions of the structural carbonate group were analyzed by means of phosphoric acid digestion and CO₂ gas source mass spectrometry with δ¹⁸O reported relative to VSMOW2 and δ¹³C relative to VPDB (28). The triple oxygen isotope analyses of the bulk phosphate follow the laser fluorination protocol described in detail in ref. 11. All phosphate oxygen isotope data reported herein are reported relative to VSMOW2 and normalized to San Carlos Olivine with δ¹⁸O = 5.3‰ and Δ’¹⁷O_0.528 = −52 ppm (47–49). The internal reference material AG-Lox (enamel of a molar from an African elephant, Loxodonta africana) was run several times in each analytical session, suggesting uncertainties of ±0.63‰ for δ¹⁸O_PO4 and ± 11 ppm for Δ’¹⁷O_PO4 (1SD, n = 162) (11, 12). All δ¹⁷O_PO4, δ¹⁸O_PO4, and Δ’¹⁷O_PO4 data are presented in Datasets S2 and S3, along with further details on the respective analytical procedures.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Dataset S11 (XLSX)

Dataset S12 (XLSX)

Dataset S13 (XLSX)

Dataset S14 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.