A cold, hard look at ancient oxygen

The proposal of a “snowball” Earth (1), covered in ice from poles to equator, occupies a special place in the hall of fame of Earth science hypotheses. Spectacular geological evidence for an icy Earth (2), creative solutions for the initiation and escape from a snowball state (1, 3), a multitude of directly testable consequences (1) and potential links to the origin of animals (1, 2) all speak to the fundamentally integrative nature of the proposal. This integrative nature allows for a broad range of scientific specializations—from astrophysics to molecular biology—to contribute to the snowball debate, but has also brought the hypothesis under continuous critical scrutiny. Whether the “original pillars” of the snowball hypothesis have survived this scrutiny is an active debate (see discussion at www.snowballearth.org). However, a crucial buttress for the hypothesis—radiative termination of the glacial interval forced by elevated atmospheric CO₂ levels—is primarily inferred through the intricate veil of oceanic carbonate chemistry (2, 4). Fresh isotopic work (5, 6) promised to lift this veil, offering what seemed to be direct evidence for extremely elevated CO₂ levels following the proposed Marinoan snowball interval 635 million years ago. This conclusion was recently turned on its head, and more moderate CO₂ levels inferred from the same evidence (7). In PNAS, Cao and Bao (8) clarify the reasons for these conflicting interpretations and provide mechanistic support for a high-CO₂ postsnowball atmosphere.

The original isotopic work was motivated by an empirical correlation between CO₂ concentrations and deficiencies in the rarest oxygen stable isotope, ¹⁷O, in O₂ in samples of ancient atmospheres preserved as bubbles in ice cores (9, 10). The relative ¹⁷O deficit in atmospheric O₂ originates primarily from stratospheric photochemical cycles of O₃, O₂, and CO₂ (9). Both CO₂ and O₃ form the ¹⁷O-enriched partners for the ¹⁷O-depleted O₂ and, given a fixed atmospheric lifetime for O₂, isotopic mass balance dictates that increasing CO₂ levels will drive larger relative ¹⁷O deficits in O₂. As a result, the CO₂ level of an ancient atmosphere can be directly estimated if an appropriate record of the isotopic composition of ancient O₂ can be identified. Oxidative weathering of sulfide minerals can trap a partial (∼20%) but robust isotopic signature of atmospheric O₂ in aqueous sulfate species. Sulfate mineral precipitates, therefore, can track the evolution of the relative ¹⁷O deficit in ancient O₂. This exercise requires two assumptions. First, because H₂O is the other major source of O in sulfate minerals, the water from which the sulfate minerals precipitated must not have incurred a ¹⁷O anomaly relative to modern ocean water. Second, the relative proportion of the O in sulfate that originated from atmospheric O₂ must have remained close to modern estimates.

Measured sulfate ¹⁷O anomalies in unique post-Marinoan rocks are larger than at any other time in Earth history and, under the framework laid out above, originally seemed to require CO₂ levels ∼30- to 200-times larger than in the current atmosphere (5, 6). Although the identification of elevated atmospheric CO₂ levels as Earth exited a hypothesized snowball state was a compelling and notable achievement, the relative ¹⁷O anomaly in atmospheric O₂ is not only a function of atmospheric CO₂ levels. Photosynthetic O₂ is characterized by isotopically “normal” oxygen sourced from the global hydrosphere (9). As a result, increased photosynthetic O₂ production dilutes the isotopic anomaly found in atmospheric O₂. Photosynthetic O₂ fluxes directly reflect global primary production (the amount of photosynthetically fixed carbon available for initial heterotrophic consumption in the global ecosystem). Primary production, therefore, was another potential control knob that could have amplified or suppressed postglacial ¹⁷O anomalies.

In fact, diminished postglacial primary production was recently called upon to explain the large ¹⁷O anomalies at 635 million years ago in an atmosphere not much more CO₂-rich than today’s (7). The basic argument is that if primary production were weaker, photosynthetic O₂ fluxes would be lower, leading to a purer stratospheric ¹⁷O signal in tropospheric O₂. A less-dilute signal would compensate for the smaller overall ¹⁷O anomaly that would be produced in an atmosphere with lower CO₂. Under the assumption of a modern atmospheric residence time for O₂, this scenario is consistent with an atmosphere containing 1‰ O₂ and 600 ppm CO₂ (7), very different from the CO₂-rich postglacial atmosphere that had been originally proposed (5, 6). This dilemma has been neatly resolved with a unique mechanistic model for the generation and preservation of ¹⁷O anomalies in O₂ (8).

Relying on a clever calibration of photochemical experiments to release the constraints of geochemical uniformitarianism, the new study lays bare how CO₂ levels and photosynthetic O₂ fluxes each influence ¹⁷O anomalies in O₂ (8). Although there are low CO₂ and high CO₂ solutions to the model at steady state, a low CO₂ atmosphere can produce ¹⁷O anomalies like those seen at 635 million years ago only if photosynthetic O₂ fluxes are ∼5- to 30-times lower than modern levels. Calling on a variety of evidence for nonnegligible primary production at the end of the Marinoan glacial interval—including a recent bold extrapolation of modern zinc isotope biogeochemistry to the Marinoan marine arena (11)—Cao and Bao (8) propose that the low CO₂, low production solution is not applicable. Therefore, it seems that elevated atmospheric CO₂ levels remain a robust inference as Earth exited the Marinoan snowball.

Importantly, the Cao and Bao model is dynamic, which enabled a pair of unique constraints on the Marinoan earth system. First, large postglacial atmospheric ¹⁷O anomalies last for only 10⁵ to 10⁶ years (8). This timescale agrees well with recent empirical estimates of time represented by the ¹⁷O anomalies in sulfate-rich Marinoan-aged rocks from south China (12). Second, the model discriminates between two styles of snowball glaciations: an “isolated” snowball in which ocean–atmosphere gas exchange is severely limited, and a “crevassed” snowball with continuous ocean–atmosphere gas exchange throughout the glacial interval. In the isolated case, atmospheric O₂ needs to be built up to levels that can support the O₃-O₂-CO₂ photochemistry that produces the ¹⁷O anomalies. The result is the delayed development of a single anomalous ¹⁷O pulse that disappears when CO₂ is drawn down below the levels necessary to develop large ¹⁷O anomalies (8). In contrast, the presence of O₂ in the crevassed atmosphere, coupled with different timescale for O₂ and CO₂ consumption in the postglacial world, leads to a pair of distinct anomalous ¹⁷O pulses (8). Preliminary stratigraphic variation of ¹⁷O anomalies in South China (12) suggests that the crevassed snowball scenario is more likely, as might be expected from the recent demonstration of a vigorously circulating snowball ocean (13). There are provocative hints, however, that these achievements may eventually be overshadowed by another potential application of the new study.

Viewed in light of the balance between a stratospheric source and a biospheric sink for the ¹⁷O anomaly, CO₂ abundances and ¹⁷O anomalies in O₂ in gas samples extracted from ice cores can constrain relative changes in the globally integrated photosynthetic O₂ flux (normalized gross biospheric production, ) and, by extension, global primary paleo-production back to the Upper Pleistocene (Fig. 1) (9, 10). The work of Cao and Bao (8) shows that this approach should be applicable even on geological timescales. Applying their steady-state analysis to high-quality datasets of ¹⁷O anomalies in sedimentary sulfate-bearing minerals (5) and proxies for atmospheric CO₂ (14) reveals what has long been a geobiological holy grail: quantitative global biospheric productivity estimates for the geological realm (Fig. 1).

Fig. 1.

Estimates of globally integrated photosynthetic O₂ flux relative to recent preanthropogenic values. Red-filled circles are exploratory estimates made from CO₂ proxies and measured ¹⁷O anomalies in sulfate minerals. Blue-filled rectangles estimated from vegetation–biogeochemistry model and general circulation model simulations (19). Black-filled rectangle estimated from CO₂ abundances and ¹⁷O anomalies in O₂ in gas samples extracted from ice cores (9, 10). Gray band is an envelope of variability for the exploratory estimates, green solid line is the recent preanthropogenic value, and black dashed line is the Permian–Triassic boundary.

Much of the exploratory record is consistent with the few previous quantitative estimates of global primary paleo-production (Fig. 1). For example, the late Carboniferous biosphere produced photosynthetic oxygen and fixed carbon, at nearly the same rate as the modern global biosphere (Fig. 1). The vast coal deposits that give the Carboniferous its name may reflect a less pervasive process than the geologic record suggests, or perhaps Carboniferous marine primary production was relatively feeble. The most striking feature of the exploratory record is that the end-Permian biosphere appears to have had a broader photosynthetic foundation than at any other time in the past ∼300 million years. Elevated primary production presaged the end-Permian mass extinction, a unique event that fundamentally restructured Earth’s biosphere over a geologically brief period (15). Modeling of the Permian Earth system provides a quantitative narrative in which overproduction sustained a lethal oceanic storehouse for dissolved CO₂ and H₂S (16); eruption of the Siberian Trap lavas initiated this process through feedbacks associated with CO₂- and CH₄-induced global warming (17). The exploratory record underscores the viability of this narrative, and highlights global eutrophy as a fundamental component of the end-Permian biotic catastrophe.

Despite its clear importance to extreme Earth-system events like the end-Permian and the Marinoan snowball Earth, examination of the initial arbiter of geologic trends in atmospheric CO₂ and O₂—global biospheric productivity—has remained mostly qualitative, inspired inferences from the fossil record notwithstanding (18). The paper by Cao and Bao (8) places these examinations on quantitative footing and opens the door for new understanding of the coevolution of life and Earth.