A deep-time CO2 barometer based on triple oxygen isotope compositions of dinosaurian eggshell carbonate

Photochemical reactions in the stratosphere lead to mass independent fractionation of oxygen isotopes: oxygen exchange among O₂, O₃, and CO₂ produces ¹⁷O-enriched O₃ and CO₂, and ¹⁷O-depleted O₂. This effect increases with increasing atmospheric CO₂ concentration, and thus the ¹⁷O anomaly of O₂, Δ¹⁷O (O₂), is reflective of pCO₂. Animals incorporate this signal into body water via respiration, and minerals such as bioapatite and eggshell calcite forming in equilibrium with body water can preserve the signal for millions of years. We contribute to the development of this new pCO₂ barometer by developing analytical methods for high-precision triple oxygen isotope analysis of carbonates, by developing an ecophysiological model of body water triple oxygen isotopes, and by applying the method to eggshell from modern birds and late Cretaceous (Campanian and Maastrichtian) dinosaur eggshells. Our findings include the following: (1) If animal ecophysiology and climatic context are perfectly known, the sensitivity of Δ¹⁷O (body water) to atmospheric CO₂ is on the order of 0.01 ‰ per 100 ppm CO₂; our analytical precision is ~ 0.01 ‰, thus ultimately permitting sub -100 ppm - level pCO₂ reconstructions. (2) However, the effect of ecophysiology and climate can lead to a range in Δ¹⁷O (body water) of about 0.15 ‰ for animals living under the same Δ¹⁷O (O₂); this prediction, confirmed by analyses of eggshells and body water of modern birds, translates to an apparent pCO₂ range of about 1500 ppm. (3) Animals that are highly dependent on unevaporated free surface water ('drinking water') and live in humid climates have Δ¹⁷O (body water) signals that mimic low pCO₂, whereas animals that consume primarily evaporated water (e.g., leaf water) and living in arid environments have Δ¹⁷O (body water) signals that mimic high pCO₂. (4) There is an upper limit to this 'evaporation / aridity' effect mimicking high pCO₂, so Δ¹⁷O (fossil eggshell) can be modeled assuming such upper limits to produce conservative lower limits on estimates of past pCO₂. (5) We find that late Cretaceous Δ¹⁷O (fossil eggshell) is generally lower than modern Δ¹⁷O (eggshell), implying generally higher pCO₂ during the late Cretaceous. The lowest observed Δ¹⁷O (fossil eggshell) value implies CO₂ levels of at least 1200 ppm, and probably closer to 2000 ppm, for at least a short interval of Campanian time. Overall, this triple oxygen isotope approach shows promise for placing constraints on past CO₂ levels. While somewhat limited in precision, it has the benefits of little appreciable loss of sensitivity with increasing pCO₂, no presently-known mechanism for generating 'false positive' estimates of high pCO₂ (except for extremely low atmospheric O₂ levels and low primary productivity), and a basis that is fundamentally different from existing methods, thus allowing for independent new constraints on past CO₂ levels.