Abstract
Oceanic oxygen levels are projected to drop in certain areas due to warming climate, but the net effect to the overall ocean redox state is difficult to predict. Here we measured the “stable” uranium isotope composition (238U/235U) in globally representative hydrogenous ferromanganese crusts in order to reconstruct the redox evolution of the global ocean throughout the Cenozoic. Samples averaging ∼3 Myr intervals have analytically indistinguishable 238U/235U throughout the Cenozoic. Combined with a U isotope mass balance model, we suggest that the overall ocean redox state has remained remarkably stable on million year time scales throughout the Cenozoic, despite large surface temperature fluctuations during this time. This suggests that stabilizing feedbacks (for example, nutrient limitation in low oxygen zones) may have prevented dramatic large-scale shifts in oxygen levels in the ocean. However, the Fe-Mn crust record will be unlikely to reflect rapid perturbations in ocean redox state. To investigate these events, sediment archives with faster accumulation rates and redox proxies with faster response time must be explored.
INTRODUCTION
Models predict that warmer surface temperatures will reduce O2 solubility and potentially increase thermal stratification in Earth's oceans (Sarmiento and others, 1998; Shaffer and others, 2009), potentially leading to ocean “deoxygenation” on a broad scale (for examples, Keeling and others, 2010). However, a general lack of well-established long-term and globally representative ocean redox records hinders understanding of potential feedbacks that regulate the spatial extent of low oxygen and/or anoxic marine conditions. Here we apply a uranium isotope approach (238U/235U, expressed as δ238U relative to the standard CRM112a) to hydrogenous ferromanganese (Fe-Mn) crusts in order to reconstruct ocean redox state throughout the Cenozoic [65 million years (Myr) ago to the present], a period during which tectonism, weathering, and surface temperatures fluctuated significantly (Hansen and others, 2008). A similar approach was employed by Goto and others (2014), but this record was derived from a single ocean basin and only extends back to 45 Myr ago. In addition, Goto and others (2014) could not rule out diffusional overprinting of the temporal isotope record. In this work we expand the record both spatially (to three major ocean basins) and temporally (to ∼77 Myr ago) while ruling out an overriding control by diffusional overprinting.
Uranium isotopes have been recently proposed as an oceanographic redox proxy that can be used to track anoxic ocean conditions (Andersen and others, 2014 and references therein). Uranium has two valence states in surface environments: the soluble U(VI) and insoluble U(IV) (Langmuir, 1978). Under reducing conditions U(VI) is reductively removed from seawater. This process results in 238U becoming enriched in the reduced phase, causing the remaining dissolved U(VI) in seawater to be fractionated to lower 238U/235U values. This trend of fractionation—opposite to traditional mass-dependent fractionation—has been well constrained by both empirical observation (Stylo and others, 2015 and references there in) and theoretical prediction based on “nuclear volume effect” (for example, Schauble, 2007). Therefore, at steady state—a condition likely applicable to the modern ocean (Andersen and others, 2016)—the 238U/235U of seawater is expected to shift to lower values when ocean basins characterized by reducing conditions expand.
Strictly speaking, 238U/235U only tracks benthic redox conditions. This is because U(VI) reduction tends to occur below the sediment-water interface rather than within the water column (Anderson and others, 1989; Barnes and Cochran, 1990). However, it is expected that changes in the areal extent of benthic anoxia are difficult to accomplish without significant changes to the volume of Oxygen Minimum Zones (OMZs) in the water column. The reason is that most of anoxic seafloor in the modern open ocean—with the exception of restricted basins such as the Black Sea—is a result of OMZs impinging on the seafloor below coastal upwelling zones (Helly and Levin, 2004; Paulmier and Ruiz-Pino, 2009; Stramma and others, 2010). Because the residence time of U is much longer (∼400 kyr, Ku and others, 1977) than timescales of ocean mixing (∼1 kyr, Broecker and Peng, 1982), it is expected that the U concentration and isotopic composition will be essentially homogeneous in the ocean. Therefore, the U isotope system can potentially track ocean anoxia on a global scale.
Seafloor Fe-Mn crusts provide a powerful archive for seawater chemistry because they precipitate directly from seawater and preferentially incorporate a wide range of trace elements (see Frank and others, 1999; Koschinsky and Hein, 2003; Siebert and others, 2003; Rehkämper and others, 2004; Hein and Koschinsky, 2014). Goto and others (2014) measured the surface layers of 19 Fe-Mn crusts dredged from the Pacific Ocean and obtained similar δ238U values of -0.65±0.05‰ (2SD, n=19). This value is offset from the modern seawater δ238U (-0.41‰) by -0.24 permil, which agrees well with experimental determination of U isotopic fractionation during adsorption of U(VI) to manganese oxides (Brennecka and others, 2011b). Therefore, both field and experimental studies suggest that Fe-Mn crusts can record contemporaneous seawater δ238U, taking the fractionation attendant to adsorption to Mn-oxide surfaces into account.
SAMPLES AND METHODS
The Fe-Mn crusts examined in this study include: ANTP109D-E and DODO232D-B from the Indian Ocean, D11-1 and CD29-2 from the Pacific Ocean, and ALV539 from the Atlantic Ocean (table 1, fig. 1). Crusts (4–15 cm thick) were embedded in resin and cut perpendicular to the growth layers. The exposed depth profile was sampled by micro drilling (∼4 mm2 drilling area) to obtain ∼30 to 120 mg of material at temporal resolutions ranging from 1.2 to 12.1 Myr (table 2). The resolution is limited by the thickness—and thus accumulation rate—of the crusts. Drilled powders were dissolved in 4 mL 6N HCl for ∼10 hours at room temperature. The amount of detrital grains undissolved by the 6 N HCl accounted for very small portions (<1%) of the total sample materials.
Basic information for the Fe-Mn crusts examined in this study
Map showing the locations of the Fe-Mn crusts examined here.
Results for U isotopic compositions and concentrations
Analytical details are described in the Appendix. Briefly, U was extracted from the sample digests using the UTEVA ion exchange and 233U-236U double-spike method (Chen and Wasserburg, 1981; Horwitz and others, 1993; Stirling and others, 2005; Rademacher and others, 2006; Weyer and others, 2008). Uranium isotopic compositions were measured on a Nu Plasma multicollector ICP-MS, with instrument parameters described in Wang and others (2015). Procedural blanks were 10 to 40 pg, which were negligible (<0.01%) compared to 300 ng sample U. External uncertainties for δ238U, U concentration, and 234U/238U activity ratio (see result section) measurements were ±0.07‰, ±0.1 ìg/g, and 3.7 permil at 95 percent confidence level, respectively, calculated as the root mean square difference (Hyslop and White, 2009) of six duplicate sample preparations and analyses.
AGE MODELS
Age models for the examined Fe-Mn crusts are derived from 187Os/188Os, T1 chemostratigraphy, and cosmogenic 10Be methods. Since the development of a reliable seawater osmium isotope record (see Klemm and others, 2005 for a review), Os chemostratigraphic correlation has been used to date Fe-Mn crusts. Prior to this, age models for Fe-Mn crusts were mainly constructed using cosmogenic 10Be for the past 10 Myr and extrapolated to older ages assuming uniform accumulation rates (O'Nions and others, 1998; Frank and others, 1999; Rehkämper and others, 2004). This extrapolation was supplemented by a Co flux dating method (Puteanus and Halbach, 1988; Frank and others, 1999). In many cases, large discrepancies were recognized between Be/Co and Os age models, especially for older layers (Klemm and others, 2005; Burton, 2006; Klemm and others, 2008; Nielsen and others, 2009; Nielsen and others, 2011). For instance, the Os isotope technique enabled several growth hiatuses to be identified (Klemm and others, 2005; Nielsen and others, 2009). Three of our five examined crusts have published Os-based ages (CD29-2, Klemm and others, 2005; ALV539, Klemm and others, 2008; ANTP109D-E, Nielsen and others, 2011) and therefore the Os age models were used. Ages for crust D11-1 is based on T1 isotope record (Nielsen and others, 2009). Ages for crust CD29-2 is based on Be/Co methods (Rehkämper and others, 2004). The average growth rates based on these age models are summarized in table 1.
RESULTS
The measured δ238U values in all five Fe-Mn crusts fall within a close range, with an average of -0.61±0.09‰ (2SD, n=52, fig. 2 and table 2). The average δ238U value is ∼0.2 permil lower than the modern seawater value of -0.4 permil (Weyer and others, 2008), consistent with previous measurements of the surface layers of Fe-Mn crusts (Weyer and others, 2008; Goto and others, 2014) and previous experimental work on adsorption of U(VI) onto Mn oxides (Brennecka and others, 2011b). A systematic trend in crust DODO232D-B appears to exist but is constant within error. This apparent trend is not seen in any of the other studied crusts, so we will not address it further.
Results for δ238U in Fe-Mn crusts (B) and previously published Os isotope data (A) for the seawater (shaded band) and the same Fe-Mn crusts. Modern seawater δ238U value is indicated by the thick dashed line (Tissot and Dauphas, 2015 and references therein). The two black thin dashed lines represent the 2σ (±0.07‰) surrounding the mean value (-0.61‰). The fact that Os isotope data in the Fe-Mn crusts align well with the seawater Os data suggests that the Os isotopic compositions in Fe-Mn crusts are not altered by the modern water. Given that Os diffuses ∼30 times faster than U in Fe-Mn crust (Henderson and Burton, 1999), this provides strong evidence that the invariable U isotope trend is a primary signature rather than diffusional artifact. Os data sources: CD29-2, Klemm and others, 2005; ALV539, Klemm and others, 2008; ANTP109D-E, Nielsen and others, 2011; seawater, reviewed in Klemm and others, 2005.
It is crucial to consider open-system behavior of U in Fe-Mn crusts, and this can be assessed through U-series disequilibrium (for example, Henderson and Burton, 1999). Briefly (see Bourdon and others, 2003 for details), in the U-Th decay chain, 238U (half-life=4.46 Gyr, Jaffey and others, 1971) decays through two very short-lived intermediate nuclides (234Th and 234Pa) to form 234U (half-life=245 kyr, Cheng and others, 2000), which further decays to form 230Th. Regardless of the relative amount of 238U, 234U and 230Th that the decay chain starts with, as long as it is not disturbed, it will eventually reach equilibrium (referred to as secular equilibrium). At secular equilibrium, the number of decays per unit time—termed activity—for 238U equals that for 234U, thus their activity ratio—expressed as (234U/238U)—equals 1. The time needed to reach 234U-238U secular equilibrium is ∼1 Myr, determined by the half-life of the relatively shorter-lived 234U. In other words, regardless of the initial (234U/238U) value of a sample, so long as it behaves as a closed system for more than ∼1 Myr, its (234U/238U) value should be ∼1. Modern seawater has a (234U/238U) value of ∼1.14, due to alpha-recoil ejection of 234Th—and thus 234Pa and 234U— into weathering fluids and ultimately seawater (for example, Henderson, 2002; DePaolo and others, 2006). If a sample has (234U/238U) values significantly above the secular equilibrium value of 1, it means there has been open system behavior in the last ∼1 Myr, most likely due to diffusion of modern seawater into the interior of Fe-Mn crust. The (234U/238U) values in most Fe-Mn crust samples explored here are distinct from the modern seawater value and are close to the secular equilibrium value of 1 (fig. 3). Some of the data points are even below the secular equilibrium value—presumably due to the effects of local alpha-recoil. However, Fe-Mn crusts formed in the last 3 Myr yield (234U/238U) values higher than 1, indicating influence by modern seawater. However, such diffusional overprinting seems to have affected only the surface 1 to 2 mm layer of all crusts examined here, except for ALV539 which has elevated (234U/238U) values on the surface and deeper parts, but secular equilibrium values in the middle. The latter pattern is most likely due to bottom-up diffusion of modern seawater through the underlying substrate (Goto and others, 2014).
234U/238U activity ratio, represented as (234U/238U), as a monitor of diffusional overprinting by modern seawater. The upper and lower horizontal dashed lines in (A and B) represent modern seawater value and secular equilibrium value of 1.14 and 1, respectively (Ku and others, 1977). (B) shows magnified details of the rectangular area in (A), with the black curve showing the theoretical (234U/238U) evolution in a closed system (calculated using half-lives from Cheng and others, 2013). Error bars for (234U/238U) are smaller than symbols. Larger age error bars for DODO232D-B are due to lack of Os chemostratigraphy.
U concentrations (fig. 4) reveal the following observations: (1) they vary among crusts from different oceans; (2) they vary temporally within the same crust; (3) the time series do not correlate among crusts. For example, in the past 20 Myr, crusts ANTP109D-E, DODO232D-B and CD29-2 show decreasing trends, while ALV539 and D11-1 show the opposite. These observations suggest that the U concentrations in these Fe-Mn crusts are controlled primarily by growth processes, rather than by temporal changes in seawater U concentrations. In general, lower growth rates seem to be associated with higher U concentrations (fig. 4). There is no correlation among δ238U, (234U/238U), and U concentration (not shown).
U concentration for the examined Fe-Mn crusts. No correlation was found among individual crusts, suggesting crust U concentration is not controlled by seawater U concentration but by formation mechanisms. Lower growth rates based on cosmogenic 10Be and Co fluxing methods seem to correlate to higher U concentration, with only one data point falling off the trend. No correlations were found among U concentration, δ238U and (234U/238U) data (plots not shown).
DISCUSSION
Phosphatization
During periods of global climate change throughout the Paleogene, previously formed Fe-Mn crusts were impregnated with carbonate fluorapatite (Hein and others, 1993), potentially altering original U isotope signatures. Phosphatization typically affects crusts formed at water depths shallower than ∼2500 m. Therefore, among the five crusts studied, only the older layers of crusts D11-1 and CD29-2 were phosphatized (Ling and others, 1997). Phosphatization has been known to affect the redistribution of some elements in Mn-Fe crusts, such as Pb and Sr (Koschinsky and Halbach, 1995; Hein and Koschinsky, 2014), but does not seem to change the isotopic ratio of the elements studied thus far (Frank and others, 1999; Rehkämper and others, 2004). Most importantly, the absence of an inflection in the δ238U curve at the phosphatization boundary in D11-1 and CD29-2 indicates that the δ238U signature was not strongly affected by phosphatization.
Diffusional Overprinting
In all five of the Fe-Mn crusts examined in this study, the layers deposited within the last ∼3 Myr have (234U/238U) values slightly above the theoretical decay trajectory (fig. 3). Since the ages of the younger intervals are well constrained by the cosmogenic 10Be and 187Os/188Os methods, such mismatch is not likely explainable by errors on age constraints and is more likely due to addition of 234U from modern seawater, which has a (234U/238U) of ∼1.14 (Ku and others, 1977; Henderson and Burton, 1999). As seawater diffuses inward along the grain boundary, 234U is concurrently consumed by decay. If diffusion is faster than decay, then 234U would accumulate and we would expect to observe (234U/238U) above the theoretical decay curve. We think this is true for the younger layers (<3 Myr) of Fe-Mn crusts (fig. 3). The fact that elevated (234U/238U) was observed in young layers but not older layers can be explained by two possible scenarios: (1) The penetration length scale of diffusive influence is on the order of ∼1 to 2 mm and thus only the young layers are affected; (2) diffusion advances further into the interior of the crust, but the timescale of diffusion beyond ∼1 to 2 mm into the Fe-Mn crust is long enough for secular equilibrium to be achieved, and therefore (234U/238U) appears as if there was no diffusion despite overprinting of δ238U values by recent seawater throughout the crust.
The second model implies that the absence of elevated (234U/238U) in older Fe-Mn crust layers cannot be treated as firm evidence that δ238U values have not been overprinted. However, Fe-Mn crusts are taken to faithfully record the seawater 187Os/188Os trend derived from independent archives (fig. 2). For the U concentrations observed in our Fe-Mn crusts, the effective diffusivity of U in the crust should be on the order of ∼30 times less rapid than that of Os (for example, Henderson and Burton, 1999). As a result, there is no readily available mechanistic explanation for overprinting the δ238U record throughout the crust by recent seawater while simultaneously leaving the 187Os/188Os system unperturbed. Furthermore, the large temporal variations in U content (fig. 4) in all of the crusts examined here also argue against diffusional overprinting, which would tend to systematically smooth out the temporal trends. Therefore, the most plausible explanation for the stability of the δ238U record is that on timescales of ∼3 Myr, the Fe-Mn crusts examined here provide a robust archive for seawater δ238U composition that has remained essentially unchanged throughout the Cenozoic.
Uranium Isotope Mass Balance Modeling
In order to quantitatively relate our seawater U isotope record to the spatial extent of benthic ocean anoxia, we construct a simple isotope mass balance model designed to explore the sensitivity of the U isotope system to shifts in ocean redox state. Our modeling approach differs in design and purpose from similar recent treatments. Goto and others (2014) aimed to investigate the effect of metallic sediment production on seawater δ238U, whereas Tissot and Dauphas (2015) focused on testing the robustness of various ocean U budgets. In contrast, our goal is to explicitly examine the sensitivity of seawater δ238U to changing relative proportions of oxic, suboxic, and anoxic/euxinic seafloor. In essence, our principle aim is to quantify how our current analytical uncertainty on δ238U measurement (±0.07‰) translates into changing ocean redox conditions.
Following Goto and others (2014) and references therein, transient global seawater [U] and δ238U can be written as:
where MSW refers to the mass of U in seawater; Fr represents the river input flux and Fi represents anoxic, suboxic, and oxic sink fluxes (mol yr−1); δSW and δr refer to the δ238U values of seawater and the river input; Δi refers to isotopic fractionation factors for anoxic, suboxic, and oxic sinks. At steady state (Andersen and others, 2016), the left side of equation (1) and (2) equals 0, yielding:
where fi is the fraction of individual sink fluxes out of the total input flux: fi = Fi/Fr. If we assume that removal mechanisms are first order with respect to seawater U concentration (Hastings and others, 1996; Partin and others, 2013; Reinhard and others, 2013), we have
where Ai represents the seafloor areas covered by various sinks; [U] denotes seawater U concentration, and ki terms are effective burial rate constants solved for by inverting modern areas and burial fluxes (see table 3).
Parameters used in the mass balance model
Parameters used in our modeling exercises are listed in table 3 and schematically shown in figure 5. Following previous studies (Montoya-Pino and others, 2010; Brennecka and others, 2011a; Goto and others, 2014), the riverine input flux (Fr) is considered to be the dominant source of U to the ocean—aeolian and subsea groundwater discharges are ignored because of their relatively small contributions. Burial sinks for U are divided into anoxic/sulfidic (Fanox), suboxic (Fsubox) and oxic (Fox) sinks (Barnes and Cochran, 1990; Dunk and others, 2002). Note that anoxic/sulfidic, suboxic and oxic are referring to the porewater redox, not the overlying water column. This classification scheme was first proposed by Berner (1981) and used by many subsequent studies (for example, Murray and others, 2007). Specifically, anoxic is defined as [O2] = [H2S] ≈ 0 μmol/L, sulfidic as [O2] = 0 μmol/L and [H2S] ≥ 11 μmol/L, suboxic as 0 < [O2] ≤ 10 μmol/L and [H2S ≤ 10 μmol/L), and finally oxic as [O2] > 10 μmol/L.
A simplified schematic representation of the modern U budget. Detailed discussion about the budget can be found in Tissot and Dauphas (2015). Riverine input is assumed to be the only source of U to the ocean: aeolian and groundwater sources are ignored because of their small contributions. Sinks including carbonates, metallic deposits (for example, Fe-Mn crusts), oceanic crust alteration (high temperature, HT, and low temperature, LT), pelagic clay, and coastal retention are treated as one single oxic sink, with the fractionation factors being the weighted average of these individual sinks (see table 4). All flux data have a unit of 106 mol/yr.
Note that in this modeling exercise several types of sinks are lumped into a single oxic sink, including Fe-Mn crusts, pelagic clays, oceanic crust alteration, and carbonates. Carbonates have been found to be isotopically fractionated from modern seawater by approximately 0.2 permil (Tissot and Dauphas, 2015 and references therein). Fe-Mn crusts are isotopically fractionated from seawater by about -0.24 permil (Brennecka and others, 2011b; Goto and others, 2014). Existing data on pelagic sediments suggest a very small fractionation from seawater (0.04‰, Andersen and others, 2014). Altered oceanic basalts are fractionated from seawater by up to +0.25 permil (Tissot and Dauphas, 2015). Taking the flux data for these individual oxic sinks summarized in Goto and others (2014) (table 4), the overall fractionation factor of the lumped oxic sink from seawater is calculated as the weighted average of these individual sinks, which is 0.005 permil.
Break down of the oxic sink into individual sinks, together with their isotope fractionation factors from seawater (Δi)
Uranium isotope fractionation factors for anoxic and suboxic sinks have been estimated previously (see Tissot and Dauphas, 2015 for a review). The isotope fractionation factors for anoxic sinks are relatively large and current estimates range from +0.4 permil to +0.85 permil (Weyer and others, 2008; Bopp and others, 2010; Romaniello, ms, 2012; Andersen and others, 2014; Basu and others, 2014; Holmden and others, 2015; Noordmann and others, 2015; Stirling and others, 2015; Stylo and others, 2015). Following previous modeling studies (Montoya-Pino and others, 2010; Brennecka and others, 2011a; Goto and others, 2014; Tissot and Dauphas, 2015), +0.6 permil is chosen as the average fractionation factor for anoxic sinks from seawater for our baseline model. The isotope fractionation factor associated with suboxic sinks is less well constrained, and is expected to vary as a function of sediment oxygen penetration depth (Bender, 1990; Clark and Johnson, 2008; Andersen and others, 2014; Reinhard and others, 2014). The +0.1 permil fractionation factor for suboxic sinks used in this study and previous studies (Montoya-Pino and others, 2010; Brennecka and others, 2011a; Goto and others, 2014; Tissot and Dauphas, 2015) is based on a few measurements of Peru margin sediments by Weyer and others (2008).
Model results are presented in figure 6. We first vary the areal extent of anoxic seafloor area while keeping the areal extent of suboxic seafloor the same as the modern value (solid black curve in figs. 6A and 6B), with oxic seafloor maintaining overall balance. The modeling result shows that with our current understanding of global U mass balance and U isotope fractionation in different marine settings, seawater U concentration and δ238U both decrease as anoxic seafloor area increases. However, with current analytical uncertainty on δ238U measurement (±0.07‰), the anoxic seafloor fraction has varied by no more than ∼3 times the modern value (0.21%–0.35% of total seafloor area, Veeh, 1968; Bertine and Turekian, 1973; Tissot and Dauphas, 2015). In reality, however, suboxic seafloor area is likely to covary with anoxic seafloor area. The effect of this scenario is also shown figures 6A and 6B again with the areal extent of oxic seafloor area being balanced accordingly in each model scenario. The results show that suboxic seafloor area variation has a very small effect on our basic conclusion.
Results from isotope mass balance model discussed in the text. The shaded boxes denote modern conditions. In B, D, and E, the horizontal dashed line represents the modern seawater minus the analytical uncertainty (0.07‰), with the vertical dashed lines denoting the uncertainties on our reconstructed anoxic seafloor area. With an analytical uncertainty of ±0.07‰ on seawater δ238U reconstruction, anoxic seafloor area has remained less than ∼1% of the total seafloor area (∼3 times the modern value). This basic conclusion is not changed after considering potential concurrent variation of suboxic seafloor area, and considering uncertainties on the fractionation factors of anoxic (C and D) and suboxic (E and F) sinks. For C to F, different fractionation factors were input into the same model exercises as in A and B, with the suboxic seafloor area fixed at the modern value (6% of total seafloor area), and oxic seafloor area being balanced. Note that fractionation factors have effects on U concentrations (C and E) because modifying fractionation factors requires rebalancing different sink fluxes to satisfy the modern U isotope mass balance in the ocean as described by equation (4). For interpretation of colors the reader is referred to the web version of this article.
The crossover between red and blue lines in figures 6B, 6D, and 6F highlights an important point. For instance, in figure 6B an increase in anoxic seafloor area by a factor of 3 brings the seawater δ238U from modern value of -0.40 permil down to -0.49 permil, keeping suboxic seafloor unchanged. However, if the increase in anoxic seafloor area is accompanied by a concurrent increase in suboxic seafloor area from a modern value of 6 percent to 20 percent, seawater δ238U is reduced by a slightly smaller offset. In reality, however, it is also likely that anoxic seafloor area expands at the expense of suboxic seafloor. In this case, the drawdown in seawater δ238U is slightly enhanced (down to -0.51‰). Although the difference is small under the scenarios examined here, it highlights the necessity to explicitly consider variation of the areal extent of suboxic seafloor when using ancient δ238U data to interpret anoxic seafloor areal extent variations.
We note that there are substantial uncertainties on current estimates of the U isotope fractionation factors (Δi) associated with suboxic and anoxic sinks. However, after incorporating these uncertainties (figs. 6C-6F) into our model exercises, our basic conclusion remains the same – the anoxic seafloor areal extent has varied by less than a factor of 3 from the modern value. This basic conclusion is broadly consistent with conclusions based on the Mo isotope system (Siebert and others, 2003).
Implications of Ocean Deoxygenation
The invariant Cenozoic δ238U values recorded in Fe-Mn crusts suggest that benthic anoxia has remained remarkably stable on a broad scale. This observation is somewhat surprising given the expectation of low oxygen levels in the warmer climate of the early to middle Cenozoic. A ∼10 °C surface temperature change over the last 60 Myr should have caused >65 μmol L−1 decreases in dissolved O2 in the surface ocean based purely on gas exchange. With all other factors affecting marine oxygen levels being held constant, this should have caused well over an order of magnitude increase in ocean anoxia compared to today, given the spatial distribution of modern low oxygen surface water (Weiss, 1970; Jaccard and Galbraith, 2012). Specifically, approximately 15 percent of the upper 1.5 km of today's ocean, the most biogeochemically active part of the marine realm, is characterized by <65 μmol L−1 dissolved oxygen levels (Jaccard and Galbraith, 2012). In addition, approximately 8 percent of the shallow ocean is occupied by OMZs with <20 μmol L−1 O2 (Paulmier and Ruiz-Pino, 2009). Further, the most recent efforts to model atmospheric oxygen levels suggest slightly lower atmospheric pO2 during the early and middle Cenozoic (for example, Berner, 2006). All these factors discussed above should have driven the shallow oceans toward lower oxygen conditions in the early to middle Cenozoic. In response to decreased surface seawater oxygen levels, and potentially to increased ocean stratification due to warming, the deep ocean should also have experienced decreased oxygen levels.
The stark contrast between our finding and O2-solubility-based expectation suggests that benthic marine redox, and by inference water column oxygenation on a broad scale, has been somehow stabilized against a backdrop of large changes in tectonic activity, Earth surface temperatures, and ocean circulation. Marine dissolved oxygen concentrations reflect the balance between oxygen supply via gas-water equilibration at the surface and oxygen consumption throughout the water column. Given the expected markedly lower saturation oxygen concentrations on the surface during the early and middle Cenozoic, the redox stability suggested by our U isotope data implies broad decreases in oxygen consumption during the same period.
We suggest that decreased oxygen consumption was probably due to decreased primary productivity. Intense local primary productivity has been suggested to be the prerequisite of ocean anoxia (Pedersen and Calvert, 1990). Anoxic conditions in the modern open ocean are rare because of inhibited N fixation and thus primary productivity. This is due to Fe-P co-limitation (for example, Mills and others, 2004) and intense nutrient-N loss from the marine system due to denitrification and anaerobic ammonium oxidation (Kuypers and others, 2003; Kuypers and others, 2005; Ward and others, 2009). In fact, enhanced denitrification has been suggested to accompany oceanic anoxic events in a greenhouse climate (Jenkyns and others, 2001; Junium and Arthur, 2007). This negative feedback between ocean oxygenation and the nitrogen cycle provides a potential stabilizing mechanism—at least in localized low oxygen zones—that may have prevented the large-scale expansion of anoxic seafloor area (Canfield, 2006; Boyle and others, 2013).
The effect of warming-induced ocean stratification on ocean oxygen levels is not as straightforward. It is generally agreed that warmer climate (for example, the early to middle Cenozoic) tends to cause ocean stratification due to freshening of high-latitude surface water (Sarmiento and others, 1998; Keeling and Garcia, 2002; Zachos and others, 2008). The effects of stratification on ocean oxygen levels are two-fold: On the one hand, more intense stratification will in general reduce oxygen supply to the deep ocean. On the other hand, stratification also reduces nutrient fluxes to the photic zone (Behrenfeld and others, 2006). When there is less organic matter produced, there is less oxygen demand in the water column. Although the net effect of thermal stratification seems to decrease ocean oxygen levels on short time scales (for example, Sarmiento and others, 1998; Keeling and Garcia, 2002), the long time scale effect (a few million years) is still unclear (Zachos and others, 2008).
Although potential stabilizing feedbacks may have prevented large-scale shifts in ocean redox during the Cenozoic, sporadic ocean anoxic events (OAEs) apparently occurred during the Paleozoic and Mesozoic. Moving forward, it will be important to establish why certain internal stabilization mechanisms work effectively for some time periods but not others. For example, Cretaceous OAEs have been attributed to multiple phenomena acting synergistically, including climate optima, realignment of major ocean current systems, and release of volcanic CO2 or methane gas hydrate (Jenkyns and others, 2001; Jenkyns, 2010). Certain combinations of forcing may thus conspire to overwhelm the stabilizing influence of, for example, redox-nutrient feedbacks in oxygen-deficient zones of the oceans.
It is important to emphasize that the effects of transient events [for example, the Paleocene-Eocene Thermal Maximum; see Dickson and Cohen (2012) and Dickson and others (2012)] on ocean oxygenation would be unlikely to register at the temporal precision of the records obtained from Fe-Mn crusts. The reason is that these events lasted for only short time periods (for example, the PETM only lasted for ∼0.17 Myr, Röhl and others, 2007) which is much shorter than the ∼3 Myr diffusional smoothing we observe in Fe-Mn crusts. Shorter transient perturbations in benthic marine redox are important to explore, but will require archives with higher sedimentation rates and proxies with faster response time.
CONCLUDING REMARKS
A new U isotope record from hydrogenous ferromanganese crusts sampled from three major oceans at different ocean depths yields invariant δ238U values. Combined with a simple isotope mass balance model, we suggest that the spatial extent of anoxic seafloor throughout the Cenozoic has remained with a factor of 3 of the modern value, well below ∼1 percent of total seafloor area. This implies a system of long-term feedbacks that has maintained a remarkably stable state of internal ocean oxygen despite large changes in tectonic activity, weathering, nutrient delivery, Earth surface temperature, and ocean circulation. However, this conclusion is only applicable on ∼3 Myr time scales due to a diffusional smoothing effect. Sediments with higher accumulation rates and proxy systems with more rapid response times will be required to explore shorter transient perturbations to ocean oxygenation.
ACKNOWLEDGMENTS
This work was supported by the Agouron Institute Geobiology Postdoctoral Fellowship Program, the NASA Exobiology Program, and the Afred P. Slogan Foundation. We thank Associate Editor Freidhelm von Blanckenburg, reviewer Stephen Romaniello, and two anonymous reviewers for providing comments that helped improve the quality of the original manuscript.
Appendix
Methods
A 233U-236U double spike method was used to correct for instrumental mass bias and potential isotopic fractionation during sample preparation (Weyer and others, 2008; Shiel and others, 2013). The double spike method also allowed isotope dilution calculations for U concentrations, based on measured 238U/236U and the amount of double spike added. Sample aliquots with 300 ng U were spiked with appropriate amount of double spike to yield a 238U/236U ratio of ∼30. The sample-spike mixtures were then slowly evaporated to dryness and redissolved in 3 N HNO3.
Samples were purified with the UTEVA ion exchange resin (based on ref. Weyer and others, 2008). Briefly, U samples in 3 N HNO3 were loaded onto 0.2 mL UTEVA resin columns (inner diameter of 0.33 mm) pre-cleaned with 4 mL 0.05 N HCl and preconditioned with 1 mL 3 N HNO3. Matrix elements were eluted with 2 ml 3 N HNO3. Thorium was eluted with 2 mL 3 N HCl (Rožmarić and others, 2009). U was then released from the resin with 2.5 mL 0.05 N HCl. Collected U samples were evaporated to dryness and treated with a few drops of concentrated distilled HNO3 to destroy potential residual organics leached from the resin.
Purified U samples were dissolved in 2 mL 0.3 N HNO3 and introduced into the MC-ICP-MS with a DSN 100 desolvating system (Nu Instrument, UK) housed at Department of Geology, University of Illinois at Urbana-Champaign. Detailed information on instrument parameters and analysis routines can be found in Shiel and others (2013) and Wang and others (2015). Briefly, isotopes 232Th, 233U, 234U, 235U, 236U and 238U were measured simultaneously on Faraday collectors connected to 1011 Ω amplifiers, except that 238U was measured with a Faraday connected to 1010 Ω amplifier yielding 30–40 volt 238U signal.
The CRM 112a (New Brunswick Laboratory, U.S. Department of Energy) standard was analyzed every 3 samples to monitor drift in instrumental mass bias. The δ238U and (234U/238U) values of samples were normalized to the average values of the bracketing CRM 112a, assuming δ238UCRM 112a = 0‰ and (234U/238U)CRM 112a = 0.9619. (234U/238U) values are calculated using decay constants for 238U and 234U of 1.55×10−10 and 2.83×10−6, respectively (Cheng and others, 2013; Villa and others, 2016). Secondary standards IRMM REIMP 18a (JRC, Brussels, Belgium) and CRM 129 (New Brunswick Laboratory, U.S. Department of Energy) measured in each session before samples yielded δ238U values of -0.17±0.09‰ (2SD, n = 5) and −1.77±0.07 (2 SD, n = 5), respectively, consistent with other studies (Weyer and others, 2008; Shiel and others, 2013). External uncertainties for δ238U, U concentration, and (234U/238U) measurements were ±0.07‰, ±0.1 μg g−1, and 3.7‰ at 95% confidence level, respectively, calculated as the root mean square difference of six duplicate sample preparations and analyses (Hyslop and White, 2009). Procedural blank was ∼ 10–40 pg, which was negligible (<0.01%) compared to 300 ng sample U.
REFERENCES
In this issue
Jump to section
Related Articles
Cited By...
- Extensive marine anoxia during the terminal Ediacaran Period
- Multiple episodes of extensive marine anoxia linked to global warming and continental weathering following the latest Permian mass extinction
- Uranium Isotope Fractionation
- Integrated geochemical-petrographic insights from component-selective {delta}238U of Cryogenian marine carbonates