Research Article

Stepwise Earth oxygenation is an inherent property of global biogeochemical cycling

See allHide authors and affiliations

Science  13 Dec 2019:
Vol. 366, Issue 6471, pp. 1333-1337
DOI: 10.1126/science.aax6459

Stepping to an internal beat

What caused the stepwise nature of the rise of molecular oxygen in Earth's atmosphere since it appeared in large quantities more than 2 billion years ago? Alcott et al. argue that a set of internal feedbacks involving the global phosphorus, carbon, and oxygen cycles, not individual external forces, could be responsible. Their model, which depends only on a gradual shift from reducing to oxidizing surface conditions over time, produces the same three-step pattern observed in the geological record.

Science, this issue p. 1333


Oxygenation of Earth’s atmosphere and oceans occurred across three major steps during the Paleoproterozoic, Neoproterozoic, and Paleozoic eras, with each increase having profound consequences for the biosphere. Biological or tectonic revolutions have been proposed to explain each of these stepwise increases in oxygen, but the principal driver of each event remains unclear. Here we show, using a theoretical model, that the observed oxygenation steps are a simple consequence of internal feedbacks in the long-term biogeochemical cycles of carbon, oxygen, and phosphorus, and that there is no requirement for a specific stepwise external forcing to explain the course of Earth surface oxygenation. We conclude that Earth’s oxygenation events are entirely consistent with gradual oxygenation of the planetary surface after the evolution of oxygenic photosynthesis.

Oxygenation of Earth’s surface environment is thought to have occurred across three broad steps (Fig. 1). The Great Oxidation Event (GOE) occurred around 2.4 to 2.2 billion years ago (Ga) and saw atmospheric O2 rise from trace levels to more than 10−5 of the present atmospheric level (PAL) (1). The following ~1 billion years (Gyr) of the Proterozoic eon likely sustained atmospheric O2 levels of ~10−3 to 10−1 PAL (2). Partial oxygenation of the surface ocean persisted throughout the Proterozoic (3), but deeper waters remained dominantly anoxic (4). The Neoproterozoic Oxygenation Event (NOE) occurred between ~800 and 540 million years ago (Ma) and is generally believed to have resulted in atmospheric O2 levels of 0.1 to 0.5 PAL, as well as the first oxygenation of the deep ocean (5). However, there was considerable variability in the temporal and spatial extent of deep-ocean oxygenation at this time, including the possibility of pulsed oceanic oxic events (6). Evidence for periodic deep-water anoxia remains frequent up until the mid-Paleozoic, when a final major rise in atmospheric O2 concentration occurred around 450 to 400 Ma (7). This Paleozoic Oxygenation Event (POE) appears to have elevated atmospheric O2 to present-day levels and established a dominantly oxygenated deep ocean, which persisted throughout the Mesozoic and Cenozoic eras.

Fig. 1 Redox history of Earth.

Atmospheric O2 based on (47). Crosshatching indicates variable deep-ocean redox between the start of the NOE and the POE. See text for a summary and related references.

These major oxygenation steps are intertwined with the evolution of progressively more complex life-forms. The first eukaryotes evolved either after the GOE or during the run-up to the event when O2 began to rise (8), whereas the NOE was coincident with major eukaryote diversification and the evolution of the first animals (9), followed by the Cambrian explosion, during which animals began to dominate marine ecosystems. The POE was accompanied by a major increase in animal body size, more diverse and specialized predators, and the evolution of vascular land plants (7). However, determining causality between rises in marine and atmospheric O2 levels and the evolution of the biosphere is complex, and there is considerable debate over the role of O2 in driving biological evolution versus the role of life in bioengineering O2 to higher levels (10, 11).

Tectonic evolution has also been considered as a potential driver of the stepwise transitions in Earth surface oxygenation. Changes to plate tectonics have been linked to the GOE through, for example, a change in the fraction of subaerial volcanism (12) or the composition of the crust (13). Some, but not all, supercontinent formation times correspond to oxygenation events (14), as do emplacement times of some large igneous provinces (LIPs), which are proposed to have driven ocean oxygenation through delivery of the limiting nutrient, phosphate (15).

However, the geologically rapid yet ultimately rare nature of Earth’s oxygenation events does not clearly correspond to either tectonic or evolutionary processes. For example, mantle dynamics and the supercontinent cycle are unlikely to produce large-scale changes on time scales of the order of less than ~100 million years (Myr), whereas LIP emplacements are far more common than major rises in O2. Looking to biological innovations, the time scale between the origination of a domain or kingdom of life and its rise to global ecological dominance may also be hundreds of million years (e.g., Eukarya) (16). Furthermore, the oscillations in ocean redox (reduction–oxidation) that are apparent during the NOE are difficult to explain through a sequence of tectonic or biological “switches” acting on the system (17).

It is therefore possible that Earth’s stepwise oxygenation was not the product of individual trigger events and may instead be explained by some inherent property of global biogeochemical feedbacks. This hypothesis has wide implications for the evolution of life on Earth and other planets, and there have therefore been a number of attempts to explain the known stepwise O2 trajectory as a feature of Earth’s internal dynamics: For example, it has been shown that atmospheric feedbacks might have promoted the GOE (18, 19). However, no study has provided a sound theoretical basis that can explain the trajectory and timing of marine and atmospheric oxygenation over Earth’s history without relying on either external trigger events or arbitrary switches in the model itself (such as assuming a transition to greater nutrient availability when O2 crosses a threshold) (20, 21).

Biogeochemical feedbacks

Here we identify a set of feedbacks that exist between the global P, C, and O cycles, which are capable of driving rapid shifts in ocean and atmospheric O2 levels without requiring any stepwise change in either tectonics or the evolution of the biosphere. These feedbacks reproduce the observed three-step oxygenation pattern when driven solely by a gradual shift from reducing to oxidizing surface conditions over time.

Phosphorus (P) is generally considered the ultimate limiting nutrient for marine productivity over geological time scales (22), and P bioavailability exerts a key control on the long-term rate of O2 production through oxygenic photosynthesis and organic carbon (Corg) burial. In the modern ocean, bioavailable P is fixed in the sediments via three primary pathways. Organic-bound P (Porg) is buried with sinking organic matter, iron-bound P (PFe) forms as P is adsorbed or coprecipitated with iron (oxyhydr)oxide minerals, and authigenic P (Pauth) is primarily formed in the sediment by means of “sink switching” of these phases to carbonate fluorapatite and/or vivianite (2325).

The phase partitioning of sedimentary P is largely controlled by redox conditions in the water column and sediments. P may be preferentially released from organic matter during remineralization under anoxic conditions, leading to elevated Corg:Porg ratios in the sediment (26), increased recycling of P back to the water column, and reduced formation of authigenic carbonate fluorapatite (27, 28). The availability of iron (oxyhydr)oxides is also typically considered to diminish in an anoxic system, which limits the burial of PFe (23), although this dynamic becomes more complex when considering the prevalence of low-sulfate, ferruginous (anoxic Fe-rich) oceanic conditions throughout large parts of Earth’s history (4), whereby Fe minerals may trap a proportion of the P delivered to the sediment (29).

Assuming that bottom-water anoxia leads to an overall enhancement of sedimentary P regeneration, two important feedback mechanisms arise that affect global biogeochemistry, with each operating over a different time scale. First, a short-term positive-feedback mechanism (self-promoting) operates, whereby the spread of ocean anoxia results in increased P availability in the water column. This stimulates primary productivity and fuels respiration, thus further increasing P availability (26). These eutrophic conditions rapidly deplete water-column O2 and in turn increase the rate of spread of anoxia. Second, a geologically paced negative feedback mechanism (self-inhibiting) operates on the combined C-O-P cycles, whereby the burial of Corg in marine sediments leads to oxygenation of the atmosphere. This increase in partial pressure of oxygen (Po2) drives higher rates of oxidative weathering of ancient sedimentary Corg (30) and ventilates the ocean, acting to stabilize O2 by both reducing the rate of Corg burial and increasing the consumption of O2 on land.

Theoretical models have previously linked the above feedbacks to the geologically rapid onset of Cretaceous ocean anoxic events and to their delayed termination through increasing atmospheric O2 (28). It has also been shown that under an increased continental weathering input of P, self-sustaining oscillations between oxic and anoxic oceanic states might occur (31). We hypothesize here that the above feedbacks are in fact sufficient to explain the stepwise oxygenation of Earth’s atmosphere and oceans, including apparent cyclic ocean oxygenation and deoxygenation events during the Neoproterozoic and early Phanerozoic (6), which are followed by the transition to a sustained oxic deep ocean.

First, the GOE can occur when the weathering of Corg becomes the principal long-term O2 sink. This can be achieved once the rate of photosynthetic O2 production outpaces the consumption of O2 by way of reaction with reduced gases and reduced seawater species. Second, cyclic oxygenation events in the NOE would be a likely consequence of the combined positive and negative feedbacks between ocean oxygenation and sedimentary P recycling, whereby a small shift toward a more oxygenated planetary surface results in gradual oxygenation of oceanic bottom waters. This would limit P regeneration from sediments, reducing short-term productivity and O2 demand and thus further increasing dissolved O2 concentrations. This positive feedback could oxygenate ocean basins, but only temporarily, because reduced P availability for primary productivity then leads to less O2 production over geologic time scales, thus reducing atmospheric O2, which eventually leads to a return to marine anoxia. Finally, a combination of the two mechanisms outlined above can result in sustained oceanic oxygenation. In this case, a sufficiently large increase in surface redox potential, coupled with a greater contribution of oxidative weathering to overall O2 regulation, allows deep-ocean oxygenation to be maintained.

Modeling results

We test our hypothesis by building on a well-established conceptual model of marine biogeochemistry (28, 32, 33). The model tracks the global cycles of P, C, and O2 in a four-box ocean system representing shelf, open-ocean, and deep-water environments (Fig. 2). We add to the model an atmospheric O2 reservoir, a global geological O2 cycle, oxidative weathering of Corg, and an open-ocean scavenging flux of P by upwelled Fe, basing these on other previous models (29, 30).

Fig. 2 Ocean and atmosphere box model.

Boxes with background color show hydrospheric reservoirs, and gray arrows denote mixing between them. White boxes show chemical reservoirs, and black arrows denote biogeochemical fluxes. (A) Carbon cycle: C exists as dissolved inorganic carbon (DIC) or Corg. Prod., primary productivity; Remin., remineralization. (B) Phosphorus cycle: P exists as soluble reactive phosphorus (Preac) and Porg. W, weathering. (C) Oxygen cycle. Single O2 reservoir encompasses all ocean boxes that exchange with the atmosphere. Resp., respiration. See text for full description; see methods and other supplementary materials for equations.

Phosphorus-dependent primary productivity occurs in all surface ocean boxes, and redox-dependent burial of P is included in all boxes that are in contact with the sediments. As in previous versions of the model, sedimentary inventories are not calculated explicitly, and regeneration of P from sediments is addressed through the net P burial terms, which follow previous model derivations (28, 32). Deep-ocean dissolved O2 concentration is calculated explicitly, as is the O2 content of the atmosphere. Following previous model versions, the O2 content of the ocean boxes in contact with the atmosphere is represented by a “degree of anoxia” parameter, termed fanoxic, which represents the balance between O2 diffusion from the atmosphere and O2 use (32). An optional scavenging flux of P sorption to upwelling iron particles (29) is implemented to test the effects of additional P drawdown in a ferruginous (iron-rich) ocean, a state that may have persisted throughout much of Earth’s history (4). The scavenging flux directly follows previous models (29), operating at O2 concentrations below 1 μM and removing 25% of the P that is upwelled into the open ocean. Full model equations are shown in the supplementary materials. As well as testing different limits on the scavenging flux, we also ran sensitivity tests to vary the degree of redox dependency of the Porg and Pauth burial terms (following previous versions of this model).

Figure 3 shows steady-state responses of the baseline model to variable fluxes of reduced gas to the surface system, representing overall changes in net surface redox over Earth’s history. All other parameters remain at their present-day values, and processes vary only by means of internal feedbacks. The redox dependence of the Porg and Pauth burial fluxes were varied and are shown as different lines. The model runs with stronger redox dependencies (40% for Pauth, 25% for Porg) respond first to oxygenation (leftmost lines in Fig. 3), with weaker dependencies (35% for Pauth, 25% for Porg, and 10% for both Pauth and Porg) plotted to the right of these. Under a very high reductant flux (left side of x axis in Fig. 3), as proposed for the Archean (≫1 × 1013 mol O2 equivalent reductant input) (34, 35), O2 production is overwhelmed and atmospheric O2 is stable at ~10−5 PAL, consistent with pre-GOE conditions (36). The O2 balance is primarily maintained by reaction of O2 with reduced gases (30). All surface and deep-ocean boxes are anoxic because of the low O2 supply. A GOE occurs in the model when consumption of atmospheric O2 by means of reduced gases is reduced to a value lower than the total O2 source from Corg burial (at ~1.5 × 1013 mol O2 equivalent). Atmospheric O2 rises by several orders of magnitude to ~0.06 to 0.25 PAL, which is broadly consistent with several estimates for the mid-Proterozoic (0.01 to 0.1 PAL) (37), and the O2 balance is primarily controlled by oxidative weathering (30). Deep-ocean O2 concentration also rises to ~0.1% of the modern value, whereas the proximal and distal shelf environments remain anoxic because O2 diffusion from the atmosphere is not sufficient to outweigh O2 demand.

Fig. 3 Model stable solutions with respect to overall surface redox state.

The model is run to steady state for changes to the reduced gas input rate. (A) Atmospheric O2 reservoir. (B) Deep-ocean O2 reservoir (mol) and shelf oxic fraction of seafloor (foxic). (C) Molar C and P burial ratio in sediments. Three lines for each zone represent different choices of redox dependence for P burial fluxes (see text). Breaks in solid lines indicate periods when no stable solution exists; in these parameter spaces, the model produces an oscillating solution.

Under a further decrease in reductant input, shelf environments undergo rapid oxygenation events. Gradually rising atmospheric O2 concentrations cause a step change in water-column redox due to the positive feedback between bottom-water O2 concentration and net P burial. Gradual ventilation of shelf bottom waters results in greater net removal of P, reducing overall O2 demand and promoting further oxygenation. As the input of reduced gas declines, this transition happens first in the proximal shelf and then in the distal shelf environment because the latter is more strongly buffered against oxygenation, owing to the upwelling of P from anoxic deeper waters. Finally, as reductant input declines further, the deep ocean becomes fully oxygenated (reductant input of <0.5 × 1013 mol O2 equivalent). The positive feedback between dissolved O2 concentration and net P burial again causes a rapid transition. In our model, the deep ocean is oxygenated when atmospheric O2 reaches 0.7 to 0.8 PAL, which is consistent with values reported in other studies (38).

Figure 3C demonstrates the degree of sedimentary P recycling in the model (shown as Corg:Porg). These changing ratios reflect the positive feedbacks between bottom-water redox and P recycling. This P recycling is dependent only on the extent of bottom-water anoxia, and thus the model predicts a substantial degree of recycling before the GOE. However, under global ferruginous conditions in the early Archean, P may have been more effectively trapped in the sediment (29) (a test of this can be found in the supplementary materials). Nevertheless, this P trap would ultimately help to stabilize O2 at low levels by reducing surface ocean productivity.

The oxygenation of the distal shelf environment has the potential for oscillatory behavior (limit cycles), denoted by the break in steady-state lines in Fig. 3. The oxygenation of the entire shelf environment results in a large reduction in overall Corg burial rates (because the shelves are the major locus of Corg burial). Over geological time scales, this reduction in C burial sufficiently reduces the O2 content of the atmosphere to return the shelves to anoxia. Figure 4 shows this cyclic response of the model under reductant inputs of 1 and 2.5 × 1013 mol O2 equivalent per year. The cyclic regime includes temporary oxygenation of both the distal shelf and the deep ocean, as the rapid oxygenation event results in increased supply of oxic water by means of down-welling. These ocean oxygenation events (OOEs) last between 2 and 5 Myr for the range of redox dependencies tested in the model and occur on approximate 5- to 20-Myr time scales.

Fig. 4 Oscillating redox solutions.

Transient model responses demonstrating limit cycles of frequency of ~5 to 20 Myr. (A) Conservative redox dependency for deep-ocean P burial terms (50% Pauth, 25% Porg) with 1 × 1013 moles of O2 consumption. (B) Stronger redox dependencies (90% Pauth, 50% Porg) with 2.5 × 1013 moles of O2 consumption (28).

Our model demonstrates that gradual oxygenation of Earth’s surface over time results in distinct oxygenation events. This occurs because, in our model, the atmosphere, continental shelves, and deep ocean act as distinct compartments of the Earth system (39, 40), which are controlled by local rather than global feedbacks (41). First, a “great oxidation” of the atmosphere occurs, followed by oxygenation of near-shore shelf environments, and then distal shelf environments. This oxygenation of the whole shelf is manifest as an oscillating solution, which would likely lead to a series of OOEs. Finally, the deep ocean becomes resiliently oxygenated. This sequence of events tracks the apparent oxygenation history of Earth as recorded by multiple redox proxies, including cycling between oxic and anoxic deep-ocean states during the Neoproterozoic and early Paleozoic (6). Our model neglects some potentially stabilizing negative feedbacks, such as the climate-silicate weathering link, although models that do include these processes still exhibit rapid shifts in marine anoxia (42).

In Fig. 5, we examine the potential for our model to recreate Earth’s oxygenation history by performing transient model simulations under a continuous decline in reductant input constrained by mantle thermal evolution (43). In light of the potential for high Archean reductant availability and outgassing, and considering the relatively high rate of O2 production in our model, we test starting fluxes of 4.5 × 1013 mol O2 equivalent per year at 4 Ga (35, 43). With or without the inclusion of open-water P scavenging, the model is able to recreate the broad observed pattern of atmospheric and oceanic oxygenation over Earth’s history. This includes a “great oxidation” of the atmosphere at around 2.5 to 2 Ga and unstable oxygenation of the deeper ocean starting at around 1 Ga, which continues until permanent oxygenation of the deep ocean is established at around 0.7 to 0.4 Ga.

Fig. 5 Possible O2 evolution over Earth’s history.

The model is run from 4 Ga to present, subject to a decrease in reductant input, illustrative of a gradual shift in net redox and demonstrating possible evolution of surface O2 levels. Dotted boxes represent the boundaries of limit cycles in the solution. (A and C) Atmosphere and deep-ocean O2 abundances. (B and D) Total shelf P burial rates (red; left axis) compared with P abundance in marine shales (right axis) (29). The gray line shows the 50-Myr binned moving average through compilation data. wt %, weight %. (A and B) No iron-bound scavenging and redox dependencies of 50% (Pauth) and 25% (Porg) (32), with a starting O2 consumption via reductants of 45 × 1012 mol/year and a linear decrease. (C and D) Iron-bound scavenging included with a maximum rate of 5 × 1010 mol P/year. Plots show stronger redox dependencies (90% Pauth, 50% Porg) (28). Starting O2 consumption via reductants was 45 × 1012 mol/year, with an exponential decrease.

We compare our model results with a compilation of P concentrations from marine shales (29) (Fig. 5). The data show an approximate fourfold increase in P weight percent between the Precambrian and Phanerozoic baselines, and this corresponds to an increase in P burial rate in the model shelf environment between the Proterozoic and oxygenated ocean states. Our model does not calculate sediment P weight percentages but does produce an upward baseline shift in the shelf sedimentary phosphorus burial rate, which is qualitatively consistent with the data. This increase in P burial is much greater when open-ocean scavenging is included, as the shutdown of scavenging on deep-ocean oxygenation results in substantially more P remaining in the system, in exactly the manner described by Reinhard et al. (29). We also note a two- to fourfold increase in shelf P burial rates when the model does not include open-ocean scavenging. This occurs simply because more P is trapped in the sediments when the shelves become oxygenated.

The reductant-driven oxygenation of the surface system that we analyze here, although plausible, is only one way in which to drive gradual net surface redox changes over time. Others include the long-term build-up of Corg in the crust (13, 43), the escape of hydrogen to space (44), or a gradually increasing supply of P to the ocean. Net redox changes driven by continual removal of hydrogen or Corg from Earth’s surface should operate in a similar way to the addition of reduced gases that we explore here. Notably, in these model runs, under a fixed present-day rate of riverine phosphate delivery and no scavenging flux, the model predicts a declining inventory of ocean P and declining rates of productivity and C burial over Earth’s history as sedimentary P recycling is curtailed (full details can be found in fig. S4). When scavenging is considered, the P inventory and overall Corg burial rates are broadly static through early Earth history and increase slightly when the deep ocean is oxygenated. Both model outputs reproduce the increasing P concentrations in shales through time (29), and although we cannot produce a δ13C record with this model (because it lacks an inorganic C cycle), it would likely be consistent with the geological record, first because the model can produce either an increase or decrease in Corg burial and second because ocean δ13C is buffered by the adjustment of oxidative weathering rates at low O2 (30) and by higher rates of inorganic C degassing and deposition on the early Earth (45). Nevertheless, although P input over Earth’s history is highly uncertain (46), it is commonly assumed that the P inventory, along with productivity and C burial rates, has increased substantially over time. Thus, in the supplementary materials, we rerun our model under a varying riverine P input and show that the stepwise oxygenation events are indeed reproducible when Earth’s gradual redox shift is driven by a steady increase in P supply to the oceans.

We demonstrate here that relationships between the global P, C, and O cycles are fundamental to understanding the oxygenation history of Earth. Our model confirms that observed oxygenation events throughout Earth’s history may be driven by well-defined internal system feedbacks between these cycles, without the requirement for extensive external forcing. The results of this analysis are far-reaching. It appears that oxygenation of Earth’s surface did not require any biological advances beyond simple photosynthetic cyanobacteria and was simply a matter of time, which substantially increases the possibility of high-O2 worlds existing elsewhere.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Tables S1 to S3

References (4852)

References and Notes

Acknowledgments: We thank C. Slomp and K. Wallmann for sending computer code; we also thank the reviewers of this work for constructive and useful comments. Funding: L.J.A. is funded by a Leeds Anniversary Research Scholarship. B.J.W.M. acknowledges support from a University of Leeds Academic Fellowship. S.W.P. acknowledges support from a Leverhulme Research Fellowship and a Royal Society Wolfson Research Merit Award. B.J.W.M. and S.W.P. are funded by the UK Natural Environment Research Council (NE/R010129/1 and NE/S009663/1). Author contributions: L.J.A. and B.J.W.M. designed the research and developed the model. L.J.A. performed model runs. All authors wrote the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials. Model code and output data are available from the corresponding author on request.

Stay Connected to Science


Navigate This Article