Earth’s oxygen cycle and the evolution of animal life

Significance

Earth is currently the only planet known to harbor complex life. Understanding whether terrestrial biotic complexity is a unique phenomenon or can be expected to be widespread in the universe depends on a mechanistic understanding of the factors that led to the emergence of complex life on Earth. Here, we use geochemical constraints and quantitative models to suggest that marine environments may have been unfavorable for the emergence and large-scale proliferation of motile multicellular life for most of Earth’s history. Further, we argue that a holistic evaluation of environmental variability, organismal life history, and spatial ecological dynamics is essential for a full accounting of the factors that have allowed for the emergence of biological complexity on Earth.

Abstract

The emergence and expansion of complex eukaryotic life on Earth is linked at a basic level to the secular evolution of surface oxygen levels. However, the role that planetary redox evolution has played in controlling the timing of metazoan (animal) emergence and diversification, if any, has been intensely debated. Discussion has gravitated toward threshold levels of environmental free oxygen (O₂) necessary for early evolving animals to survive under controlled conditions. However, defining such thresholds in practice is not straightforward, and environmental O₂ levels can potentially constrain animal life in ways distinct from threshold O₂ tolerance. Herein, we quantitatively explore one aspect of the evolutionary coupling between animal life and Earth’s oxygen cycle—the influence of spatial and temporal variability in surface ocean O₂ levels on the ecology of early metazoan organisms. Through the application of a series of quantitative biogeochemical models, we find that large spatiotemporal variations in surface ocean O₂ levels and pervasive benthic anoxia are expected in a world with much lower atmospheric pO₂ than at present, resulting in severe ecological constraints and a challenging evolutionary landscape for early metazoan life. We argue that these effects, when considered in the light of synergistic interactions with other environmental parameters and variable O₂ demand throughout an organism’s life history, would have resulted in long-term evolutionary and ecological inhibition of animal life on Earth for much of Middle Proterozoic time (∼1.8–0.8 billion years ago).

A long-standing and pervasive view is that there have been intimate mechanistic links between the evolution of complex life on Earth—in other words, the emergence and ecological expansion of eukaryotic cells and their aggregation into multicellular organisms—and the secular evolution of ocean−atmosphere oxygen levels (1). Molecular oxygen (O₂) is by far the most energetic of the abundant terminal oxidants used in biological metabolism (e.g., ref. 2). When this energetic capacity is harnessed by mitochondria in eukaryotic cells, the energy flux supported by a given genome size increases by a factor of ∼8,000 (3), potentially paving the way for increased complexity at the cellular level (but see ref. 4). Oxygen is also a crucial component of enzymatic pathways leading to the synthesis of regulatory membrane lipids (5) and structural proteins (6) in eukaryotic organisms and provides a powerful shield against solar UV radiation at Earth’s surface through stratospheric ozone production. Furthermore, O₂ is the only respiratory electron acceptor that can meet the metabolic demands required for attaining the large sizes and active lifestyles characteristic of metazoan life (e.g., ref. 7). There is thus ample support for the view that Earth’s oxygen cycle provided a crucial evolutionary and ecological constraint on the road to increased biotic complexity, both at the cellular level and on the road to motile multicellular life.

However, it is important to distinguish between the O₂ levels required for: (i) the emergence and ecological expansion of complex (multicellular) eukaryotes, including red/brown algae, land plants, fungi, and animals; (ii) the initial origin of the metazoan (animal) last common ancestor (LCA); (iii) the diversification of basal (e.g., Ctenophora, Cnidaria, Porifera) and more derived (e.g., Bilateria) metazoan clades; and (iv) emergence of the larger body sizes and more complex ecological interactions (such as predation and burrowing) likely to leave robust signals in the fossil record (8). The requisite environmental O₂ levels for each of these biotic milestones may vary substantially, and some may depend strongly on environmental variability, whereas others may be more directly linked to the environment achieving a discrete “threshold” or range of oxygen levels. Our focus here is on reconstructing an ecological context for the emergence and expansion of early animal life during the Middle to Late Proterozoic [∼1.8–0.6 billion years ago (Ga)] in the context of evolving Earth surface O₂ levels.

Most work on the coevolution of metazoan life and surface oxygen levels can be characterized as either biological (e.g., attempting to constrain threshold environmental O₂ levels for different organisms), or geochemical (e.g., attempting to constrain environmental O₂ levels before, during, and after the emergence of metazoan life). This discussion is typically framed in terms of a relatively sharp dichotomy—e.g., either environmental oxygen levels have had little to do with the timing of the rise of animal life (e.g., refs. 9 and 10) or oxygenation of Earth’s surface has been of primary importance (e.g., ref. 11). Both approaches have emphasized the relationship (or lack thereof) between metazoan evolution and the secular oxygenation of Earth’s ocean−atmosphere system, and both provide a critical backdrop for efforts to understand the importance of environmental factors in early animal evolution.

Experimental work with the modern demosponge Halichondria panacea suggests an empirical O₂ threshold of ∼0.5–4.0% of the present atmospheric level (PAL; corresponding to an equilibrium concentration of ∼1–10 µmol⋅kg⁻¹ for air-saturated water at 25 °C and a salinity of 35‰) for basal metazoan life (12) and perhaps lower for smaller sponge species. Theoretical calculations for the LCA of bilaterian life indicate an even lower threshold range of ∼0.14–0.36% PAL (ref. 13, Fig. 1). In addition, all major eukaryotic lineages contain genes and enzymes allowing for facultatively anaerobic energy metabolism (14), which indicates that either (i) anaerobic pathways of ATP synthesis in primitively aerobic eukaryotes were obtained through interdomain and intradomain lateral gene transfer (15, 16), or (ii) the earliest eukaryotic organisms contained mitochondria that were facultatively anaerobic (17). In any case, it is possible that early animals were able to respire at very low environmental O₂ levels and possibly even survive periodic episodes of local anoxia.

Fig. 1.

Biological thresholds and geochemical constraints on Earth surface O₂ levels. (Left) Estimated minimum metabolic O₂ requirements of sponges (gray) (12) and the LCA of bilaterian metazoans (brown) (13). (Right) Estimated environmental O₂ levels based on the plant fossil and charcoal record during the Phanerozoic (37); fossilized soil profiles during the Proterozoic (blue), recalculated following ref. 38 according to pCO₂ values in ref. 39; chromium (Cr) and manganese (Mn) systematics of marine chemical sediments and ancient soil profiles (red) (11); non-mass-dependent sulfur (S) isotope fractionation in sedimentary rocks (orange) (40); and 0D and 3D Earth system models for Archean oxygen oasis systems (green) (20−22). Downward arrows represent upper bounds on atmospheric O₂ levels, and the range for oxygen oases refers to the estimated atmospheric pO₂ value at gas exchange equilibrium.

Geochemical estimates of atmospheric pO₂ levels during the Proterozoic before the emergence and ecological expansion of metazoan life span orders of magnitude, from less than 0.1% PAL (11) to ∼10% PAL (18, 19). Estimates thus span levels well below what active aerobic metazoans would have been able to tolerate on long timescales to values well above those that should be required for animal respiration (Fig. 1). An additional complication is that, after the evolution of oxygenic photosynthesis, disequilibrium between surface marine and atmospheric oxygen levels would have been common, as it is in the modern oceans. A range of biogeochemical models applied to simulate this disequilibrium indicate that bacterial oxygen production in the surface ocean can sustain local dissolved O₂ concentrations of ∼1–10 µmol⋅kg⁻¹ at steady state (20–22), even at exceptionally low atmospheric O₂ levels (Fig. 1). Thus, local biological O₂ production in some regions of the surface ocean may have allowed for O₂ levels above the respiratory requirements of basal metazoan organisms regardless of background atmospheric pO₂ (10).

In principle, one could compare a geochemical estimate of atmospheric pO₂ to an estimate of the dissolved O₂ level required for respiration in a particular metazoan organism and use this to infer whether there would have been any respiratory “metabolic barrier” to animal emergence and ecological expansion during a given period in Earth’s history (e.g., refs. 11 and 23). However, this basic approach has a number of drawbacks. First, it neglects spatial heterogeneity in oceanic O₂ levels—for example, decoupling between atmospheric pO₂ levels and oxygen availability in benthic habitats or the existence of surface ocean regions with relatively high dissolved O₂ as a result of local O₂ production. Second, a system with an average steady-state O₂ level above that required for a given metabolism could nonetheless be inhospitable overall from an evolutionary perspective if O₂ levels oscillate significantly on timescales that are unfavorable for the long-term persistence of robust populations (e.g., refs. 24 and 25). Finally, the comparative threshold approach ignores variable metabolic O₂ demand during an organism’s life history, the ecology of the organism, and the potential synergistic interactions between O₂ and other key environmental variables, such as temperature and pH.

Here we explore a series of quantitative biogeochemical models with the goal of providing a new framework for understanding the constraints posed by a relatively low oxygen Earth system on early animal ecology and evolution. First, we use a 3D Earth system model to assess spatial variability in surface ocean O₂ and surface−benthic decoupling at a range of atmospheric pO₂ values relevant for the Late Proterozoic Earth [∼800–600 million years ago (Ma)]. We then use a regional time-dependent box model of the surface ocean to evaluate the potential for temporal variability in O₂ levels within productive regions of the surface ocean. Finally, we build on these perspectives to propose a simple conceptual model that frames the links between Earth’s oxygen cycle and the early evolution of animal life as a question of habitat viability, highlighting the stability of environmental O₂ levels (e.g., ref. 25) against the backdrop of evolving baseline pO₂ (11, 26), rather than the attainment of threshold O₂ levels for a single life history stage (10, 11, 23). We suggest that this view provides a more realistic and powerful framework for understanding the links between environmental stress and the early evolution of complex metazoan life on Earth. Finally, we argue that spatiotemporal O₂ variability during Middle Proterozoic time would have resulted in a challenging evolutionary and ecological landscape for early animal life, even at relatively high background atmospheric pO_2.

Oceanic Oxygen Landscapes

We evaluate steady-state spatial heterogeneity in surface ocean and benthic O₂ levels by using a 3D Earth system model of intermediate complexity (EMIC) configured for a low pO₂ world (see ref. 22). The model ocean is integrated to steady state at a range of imposed atmospheric pO₂ values, and the resulting distribution of dissolved oxygen concentrations at the surface and ocean interior is evaluated. We explore a range of atmospheric pO₂ levels over two orders of magnitude (0.1–10% PAL), meant to be broadly inclusive of estimates for the Mid- and Late-Proterozoic Earth system (Fig. 1), and we focus our analysis on the spatial arrangement and concentration distributions of both surface and benthic grid cells at steady state.

Although we use an effectively modern continental distribution, we do not mean to suggest that large-scale ocean circulation during the Proterozoic was necessarily similar to that characteristic of modern Earth. However, we consider it likely that the basic range of large-scale “circulatory settings”—low-latitude to midlatitude equatorial upwelling due to Hadley cell convergence, high-latitude deep convection, eastern boundary upwelling—is well represented by the zonal and meridional circulation patterns of modern Earth. This assumption should be particularly true with respect to nutrient recharge to the surface ocean (of primary importance for our simulations here). Although more accurately diagnosing patterns of large-scale ocean circulation under different continental configurations and explicitly linking these to the fossil record of early metazoan life is a crucially important topic for future research, we consider it unlikely that changes in large-scale ocean circulation would dramatically alter our primary conclusions.

Consistent with previous work (20–22), we find that broad regions of the ocean can be characterized by dissolved O₂ levels significantly above ambient gas exchange equilibrium at steady state as a result of localized biological O₂ release (Fig. 2). At atmospheric pO₂ levels below ∼2.5% PAL, the distribution of surface ocean O₂ is controlled principally by the biosphere, with the highest dissolved O₂ levels located primarily in areas of high productivity (and O₂ release) along eastern boundaries and along regions of equatorial divergence (Fig. 2 A and B). However, a critical transition occurs at pO₂ levels above ∼2.5% PAL, at which point surface O₂ distributions are controlled by the atmosphere rather than local O₂ release in the surface ocean. In this regime, high O₂ levels are more commonly observed in colder, high-latitude surface waters where aqueous O₂ solubility is relatively high (Fig. 2D).

Fig. 2.

Results from the EMIC [Grid Enabled Integrated Earth System Model [GENIE]). Steady-state dissolved oxygen concentrations in the surface layer of the ocean after 20-thousand-year spinup for atmospheric pO₂ values of (A) 0.5%, (B) 1.0%, (C) 2.5%, and (D) 10.0% of the PAL. Note the differing scales for [O₂] for each background pO₂ level.

The precise atmospheric pO₂ threshold at which this transition occurs will ultimately depend on a number of factors, but the presence of this transition and its functional relationship with atmospheric pO₂ is unlikely to be dramatically different from that observed here (Supporting Information and Fig. S1). At pO₂ levels above this transition point, most of the surface ocean will be near gas exchange equilibrium with atmospheric pO₂ (Fig. 2D). Below this transitional atmospheric pO₂, surface ocean dissolved O₂ levels are expected to be extremely variable in time and space and may often have been well below levels that would be expected to challenge aerobic metazoan organisms (Fig. 2 A and B), despite locally elevated values in productive regions of the surface ocean.

Fig. S1.

Sensitivity simulations for steady-state surface ocean O₂ distributions. All panels show oxygen concentration anomalies (∆[O₂], in micromoles per kilogram) for a given experiment relative to the baseline case presented in Fig. 1 (e.g., [O₂]_experiment – [O₂]_baseline, such that negative values represent lower oxygen levels than the baseline case and positive values represent higher levels). (A) Simulations with 0.5× and 0.25× the modern marine nutrient inventory. (B) Simulations with 0.5× and 0.1× the modern organic carbon remineralization length scale.

Perhaps more importantly, we observe strong decoupling between atmospheric pO₂, surface ocean O₂ distributions, and O₂ availability in the benthic realm. On an ocean scale, benthic grid cells are almost entirely devoid of O₂ at steady state at pO₂ levels below ∼2–3% PAL, and, even above this level, the seafloor is dominated by anoxic conditions in our model (Fig. 3). The pervasiveness of benthic anoxia in shallow settings may also have been exacerbated by nutrient limitation of biological oxygen production in the surface ocean during much of Proterozoic time (refs. 27 and 28, Supporting Information). The upper ocean grid size in our model does not fully resolve the shallowest of continental margin settings (29), and it will be important to leverage high-resolution models of the upper ocean (e.g., ref. 30) in future research. Nevertheless, our results imply the potential for a severe O₂-based ecological bottleneck for the large-scale dispersal and expansion of basal metazoan organisms with a benthic life stage (e.g., ref. 31), despite background atmospheric pO₂ above the levels currently considered to be required for respiration (Fig. 2). We emphasize that this decoupling is a robust feature and arises as a natural consequence of the ocean’s biological pump—for example, any changes to the nutrient status of the ocean interior that favor greater local dissolved O₂ accumulation in the surface ocean will tend to exacerbate O₂ deficiency in the benthic realm.

Fig. 3.

(Left) Relative frequency distributions for dissolved oxygen concentration in surface ocean grid cells (f_surface) and (Right) the fraction of benthic grid cells that are anoxic (f_benthic) for a range of atmospheric pO₂ levels. Relative fraction of benthic grid cells that are anoxic are calculated for three different cumulative depth ranges, the upper 1 km, upper 3 km, and upper 5 km of the ocean. Shaded blue boxes denote estimated ranges for basal metazoan organisms derived from laboratory experiments (12), and red boxes show lower limits for bilaterian organisms estimated from theoretical calculations (13).

In summary, our 3D model results suggest a complex interplay between atmospheric pO₂, surface ocean dissolved O₂ distributions, and O₂ availability in the benthic marine realm over the range of atmospheric pO₂ values envisioned for the Late Proterozoic Earth system. Surface ocean dissolved O₂ levels would have been spatially variable and, in many cases, well above those predicted based on ambient atmospheric pO₂. At the same time, O₂ availability in benthic habitats would have been largely insulated from surface ocean−atmosphere O₂ dynamics, making dispersal and long-term survival of metazoans with a benthic life stage extremely difficult—even at background atmospheric pO₂ levels well above those conventionally assumed to be challenging for basal metazoan life.

Temporal Variability in Surface O₂ Levels

Although the Earth system model results provide insight into the long-term mean state of the ocean’s oxygen landscape, we can also diagnose temporal variability in surface O₂ levels using a simpler regional model (21) that tracks dissolved O₂ concentrations in a generalized coastal surface ocean environment as governed by physical transport, gas exchange, and biogeochemical processes. This approach has a number of advantages for our purposes. First, it allows us to apply dynamic wind stress on diurnal and seasonal timescales such that air−sea gas exchange can be periodically forced. Second, it allows us to explicitly explore the role of externally sourced deep ocean reductants (e.g., Fe²⁺) on regionally large O₂ production fluxes.

Briefly, our model assumes that a fraction of primary production in surface waters escapes respiration and becomes exported beneath the photic zone. This exported carbon is offset by a stoichiometric production of O₂. Oxygen that accumulates in the surface ocean is transported by upwelling circulation and equilibrates with the atmosphere through air−sea gas exchange on a timescale governed by temperature, salinity, and surface wind speed (21). The upwelling circulation also injects ferrous iron (Fe²⁺) into surface waters, which can react with O₂ produced by photosynthesis and thus represents a potential sink term for O₂. Ferrous iron concentrations were kept constant in the simulations explored here. It is likely that pulsed benthic and/or hydrothermal sources of Fe²⁺ would have resulted in temporally variable bottom water ferrous Fe concentrations, leading to periodic anoxia. Given poor empirical constraints on local and global variability in Fe²⁺ fluxes, we do not explicitly model these processes here, rendering our conclusions conservative.

Seasonal (monthly) oscillations in biological carbon fluxes and air−sea gas exchange result in oscillatory changes in dissolved [O₂] that are, in some cases, very large (Supporting Information and Fig. S2). To diagnose the potential biological consequences of this temporal variability, we evaluate the minimum surface O₂ level attained during a seasonal cycle at a range of atmospheric pO₂ values similar to those explored above and assuming a range of deep water dissolved [Fe²⁺] levels (Fig. 4). Relatively modest [Fe²⁺] levels well below some recent predictions for ferruginous Proterozoic oceans (32) can significantly depress seasonal minima and maxima in dissolved [O₂], even at very high rates of biological productivity (Fig. 4B). Indeed, under most conditions, the estimated tolerance range for basal metazoan organisms (12) is only safely exceeded throughout the year at pO₂ on the order of ∼5–10% PAL.

Fig. S2.

Representative results for seasonal [O₂] oscillation in the time-dependent model. (A−C) Dissolved [O₂] in the surface ocean box is shown as a function of time according to the time-dependent forcing in organic carbon export (green) and surface wind speed (blue) shown in D. Results are shown for atmospheric pO₂ values of (A) 2.5%, (B) 1.0%, and (C) 0.5% of the PAL. In each panel, curves correspond to assumed deep ocean [Fe²⁺] values of 50 µmol⋅kg⁻¹ (solid), 150 µmol⋅kg⁻¹ (dashed), and 250 µmol⋅kg⁻¹ (dotted).

Fig. 4.

Results of the time-dependent biogeochemical model. (A) A schematic depiction of the main processes represented by the model. (B) Minimum seasonal dissolved oxygen concentrations in the surface ocean as a function of background atmospheric pO₂ value for two different baseline carbon export rates (Left and Right), and at deep [Fe²⁺] levels of 50 µmol⋅kg⁻¹ (blue), 150 µmol⋅kg⁻¹ (red), and 250 µmol⋅kg⁻¹ (black). Gray shaded box encompasses the biological limits denoted in Fig. 1.

In summary, seasonal variability in biological O₂ production, carbon export, and surface winds would have resulted in significant temporal variability in surface O₂ within productive regions of the surface ocean and in shallow continental margin settings. In addition, even modest amounts of Fe²⁺ relative to some predictions for the Proterozoic ocean interior (18, 32) would have been sufficient to depress seasonal minima in dissolved O₂ to levels that would be expected to have inhibitory metabolic effects on emerging metazoan organisms for much of Proterozoic time. Such temporal variability in redox structure, particularly in light of the spatial complexity of the long-term mean state oxygen landscape discussed above, would have provided an additional environmental constraint on the early evolution of animal life.

Conclusions and Future Directions

Taken together, our results imply a long-term mean surface ocean oxygen landscape during the Proterozoic that would have been characterized by significant spatial variability (Fig. 2). Superimposed on this long-term mean state were surface ocean dissolved O₂ distributions that would have varied significantly on seasonal timescales (Fig. 4), all against a backdrop of consistently low benthic O₂ levels (Fig. 3). Although it should be stressed that many key biological innovations must have preceded the Late Proterozoic emergence of basal metazoan clades (33), we suggest that early multicellular organisms would have had to contend with a “patchy” and evolutionarily restrictive oxygen landscape for much of Proterozoic time, despite background local O₂ levels that may, in some cases, have been sufficient to fuel their metabolic needs. Although our discussion here focuses on metazoans, if periodic anoxia was common throughout shallow oceans during the Neoproterozoic, all eukaryotic life may have been oxygen-stressed on evolutionary timescales. This relationship may, in part, explain the early rise but protracted diversification of eukaryotes (34).

More broadly, we stress that geochemical reconstructions of background atmospheric pO₂ (11, 26) and experimental/theoretical reconstructions of resting O₂ demand in basal animal organisms (12, 13), although critical, only provide part of the picture. Proterozoic marine environments would have been characterized by O₂ levels that departed strongly from background mean pO₂, and the oxygen demand and marine habitat of different life history stages could present physiological bottlenecks that are currently poorly known for basal metazoan organisms. Indeed, microfossil evidence for metazoan life stages adapted for large-scale dispersal across inhospitable environments may represent a paleontological signal for such ecosystem oxygen stress during the Late Proterozoic (24). In this light, our results strongly suggest that the distribution of O₂ within the ocean−atmosphere system played a significant role in structuring evolutionary innovation and ecological success among early animals.

Finally, we hypothesize that the interplay between long-term changes in base-level pO₂ and some degree of spatiotemporal oxygen variability may promote rather than inhibit biological and ecological innovation. For example, theory predicts that environmental fluctuation can often reinforce evolutionary transitions in individuality (35), such as multicellularity. A critical combination of baseline pO₂ and spatiotemporal variability may thus have ratcheted the evolution of more advanced forms of multicellularity in some eukaryotic lineages. In addition, variability in the marine oxygen landscape against the backdrop of increasing baseline O₂ during the Late Proterozoic (36) may have acted as a creative evolutionary force, ultimately favoring rapid phylogenetic diversification and ecological expansion among early animal lineages. Future work pinpointing when and how this transition occurred will be essential for fully understanding how changes in surface O₂ levels have affected the tempo and mode of early metazoan evolution.

Spatial Heterogeneity in Steady-State O₂ Levels

To explore spatial variations in surface ocean dissolved oxygen levels, we perform simulations using the GENIE framework, an EMIC (e.g., ref. 41). GENIE considers a frictional geostrophic 3D ocean circulation coupled to a marine biosphere that specifies nutrient-limited export production, aerobic respiration, denitrification, sulfate reduction, and methanogenesis. The marine biogeochemistry module also includes a range of oxidative processes, including aerobic methanotrophy, nitrification, and sulfide oxidation (see ref. 29). The ocean is represented as a 36 × 36 equal-area horizontal grid with 16 depth layers, and is coupled to a 2D energy/moisture balance model, a sea ice model, and a 2D atmospheric model for calculating sea−air exchange fluxes and atmospheric chemistry (42). The model is well calibrated for simulations of the Cenozoic Earth system (29, 43) and has also been used successfully to interrogate older intervals of Earth’s history, including the Cretaceous (41), the end-Permian (45), and the Archean/Proterozoic (22).

Our default simulations assume many effectively modern boundary conditions, including climate (e.g., pCO₂ and pCH₄), continental configuration (e.g., large-scale ocean circulation), marine nutrient inventory, and organic carbon remineralization length scale in the water column (e.g., the “penetration depth” of organic carbon exported from the photic zone). However, it is clear that all of these parameters have changed dramatically throughout Earth’s history. It is thus important to establish broadly the sensitivity of any model result to plausible changes in these parameters through time. We focus below on the sensitivity of our results to ocean circulation, nutrient inventory, and biological carbon fluxes through the water column, while noting that our assumption of roughly modern preindustrial greenhouse gas inventories is conservative for our purposes in that the relatively cold preindustrial climate state should on its own act to increase O₂ solubility in the ocean (e.g., ref. 46).

Quantitative reconstructions of marine nutrient inventory that are currently available suggest that long-term oceanic fertility was, if anything, less than that of modern Earth for most of Earth’s history (47−50). The effects of this are illustrated in a series of sensitivity simulations in which the initial oceanic phosphorus inventory is reduced to 0.5× and 0.25× the modern value (Fig. S1A). Intuitively, decreasing the marine nutrient inventory leads to decreased local production and accumulation of O₂ in surface marine environments. For reasonable decreases in nutrient inventory as envisioned for much of the Precambrian (50), surface ocean oxygen levels for a given set of identical boundary conditions are lower by up to ∼2 µmol⋅kg⁻¹. Thus, as is argued here, all other things being equal, the marine nutrient inventory would need to be significantly elevated relative to that of modern Earth to effectively counter the magnitude of variability in oxygen concentrations implied by our time-dependent simulations. In addition, it is possible (if not likely) that nutrient availability would have exacerbated the spatial heterogeneity in surface oxygen levels discussed here.

It has long been suggested that reduced sinking speeds of marine aggregates during much of the Precambrian would have led to shallower organic carbon penetration into the ocean interior (see ref. 9 for a recent review). However, this conceptual model was originally based on the notion of enhanced vertical transport due to zooplankton fecal pellets (51), which have been shown to be a relatively small component of vertical carbon transport, even in modern surface ocean ecosystems (52). In addition, zooplankton activity is very effective at promoting particle deaggregation as particles sink through the water column (e.g., ref. 53). In other words, zooplankton may, in fact, reduce overall sinking speeds of marine aggregates while contributing only minor direct fluxes due to fecal pellet production. Perhaps more importantly, this framework has yet to rigorously incorporate the clear importance of picoplankton in vertical carbon fluxes within modern marine ecosystems (54, 55). Our overall understanding of evolving biological pump strength is thus rather poorly developed at present, and it is far from clear that Precambrian surface ecosystems would result in markedly reduced sinking speeds of organic aggregates through the water column.

Nevertheless, if shallower remineralization of sinking organic matter were pervasive during Precambrian time, this could result in more efficient nutrient recycling and increased input fluxes of nutrients through the main thermocline. Such a mechanism could, to some degree, counteract the effects of a smaller overall oceanic nutrient inventory. Indeed, simulations with shallower organic carbon remineralization result in relative increases in local surface oxygen levels for a given set of parameters (Fig. S1B). However, rather extreme decreases in organic carbon penetration would be required to counteract the scale of seasonal variability observed in our time-dependent simulations—for example, simulations in which the remineralization length scale for organic carbon in the model is decreased by an order of magnitude result in rather modest local increases in O₂, generally less than ∼1 µmol⋅kg⁻¹. This is well below the temporal variability we observe in our regional model and could easily be counteracted by modest reductions in nutrient inventory.

It is difficult to accurately diagnose the effects of changing oceanic circulation for the vast majority of Earth’s history, simply because accurate continental reconstructions are not available until the latest Proterozoic (56). We do not mean to suggest that large-scale ocean circulation during the Proterozoic was necessarily similar to that characteristic of modern Earth. However, it is likely that the basic range of large-scale “circulatory settings”—low-latitude to midlatitude equatorial upwelling due to Hadley cell convergence, high-latitude deep convection, eastern boundary upwelling—is well represented by the zonal and meridional circulation patterns of modern Earth. This should be particularly true with respect to nutrient recharge to the surface ocean (of primary importance for our simulations here), although large-scale patterns of ocean ventilation should change dramatically with varying continental configuration. In short, we consider it difficult to imagine an open ocean system unusually favorable for local O₂ accumulation that is not represented in some region of the modern Earth system. Although more accurately diagnosing patterns of large-scale ocean circulation and how they have responded to shifts in continental configuration through time is a crucially important topic, and should, in general, be a major focus of future quantitative modeling efforts, we consider it unlikely that changes in continental configuration would dramatically alter our primary conclusions here.

Temporal Heterogeneity in Surface Ocean O₂ Levels

We attempt to diagnose temporal heterogeneity in surface oxygen levels by using a simple regional model of a coastal upwelling system (Fig. 2A), modified from ref. 21, that tracks dissolved oxygen in the surface ocean as governed locally by physical transport, biological productivity, gas exchange, and biogeochemical processes. Briefly, concentrations of dissolved O₂ in the surface ocean are governed by sources minus sinks due to physical transport (∆_mix), gas exchange (J_gas), and biogeochemical processes (J_bio),

$$h_p\frac{d}{dt}{\left[{\text{O}}_2\right]}_{surface}={\mathrm{\Delta }}_{mix}+J_{gas}+J_{bio},$$

where

$$\begin{array}{c}{\mathrm{\Delta }}_{mix}=w_u\left({\left[{\text{O}}_2\right]}_{deep}-{\left[{\text{O}}_2\right]}_{surface}\right),\\ J_{gas}=k_{O2}\left(\left[K_H(T,S)p{\text{O}}_2\right]-{\left[{\text{O}}_2\right]}_{surface}\right),\\ J_{bio}=r_{O:C}{J}_{\text{exp}}-r_{O:Fe}{k}_{ox}{\left[{\text{Fe}}^{2+}\right]}_{surface}{\left[{\text{O}}_2\right]}_{surface}{\left[{\text{OH}}^{-}\right]}_{surface}^2.\end{array}$$

In the above expressions, w_u represents the upwelling rate (meters per day), brackets denote concentrations in either surface or deep waters (the latter of which are specified as boundary conditions for the active surface layer), k_O2 is the gas exchange coefficient for O₂ (governed by surface wind speed and corrected Schmidt number at a given temperature), K_H is the Henry’s Law constant for O₂ at a given seawater temperature and salinity, and r_i values represent stoichiometric factors for a given reaction. Physical transport and gas exchange are driven by surface winds as described in ref. 21. The model has been modified to allow for time-dependent forcing of physical boundary conditions and deepwater chemistry, most notably by including an explicit expression for the kinetics of Fe²⁺ oxidation. The rate constant of this process (k_ox) depends on seawater temperature and ionic strength according to ref. 57, with [OH⁻] derived based on an assumed pH (taken to be constant at 8.3). We neglect CH₄ in the present analysis, for simplicity, but note that this is conservative for our purposes here.

We assume a given deepwater [Fe²⁺] value and allow the system to equilibrate from an initial assumed surface [O₂] value (taken here to be 5 µmol⋅kg⁻¹, but the exact value is irrelevant) while subjecting the model to periodic forcing in surface wind speed (and thus upwelling rate) and export production (net rates of O₂ production by photosynthesis). We explore two timescales of periodic forcing. In the first case, oscillations have a period of 24 h, meant to represent diurnal fluctuations in, for example, photosynthetic activity. In the second case, oscillations have a period of ∼120 d, chosen to represent roughly seasonal variability in upwelling, nutrient recharge, and biological productivity (ignoring subtle differences in the length of the Precambrian day and year). The latter timescale is examined in detail here.