Uranium isotope evidence for two episodes of deoxygenation during Oceanic Anoxic Event 2

Significance

Past “Oceanic Anoxic Events” (OAEs) represent important carbon cycle perturbations that offer the opportunity to study Earth’s response to extreme climate warming. A fundamental limitation for understanding OAEs is quantifying the timing and total extent of ocean anoxia. We present a quantitative account of global redox conditions for OAE 2 (∼94 million years ago), using a high-resolution record of uranium isotopes combined with a biogeochemical model. We present new evidence for two discrete intervals of globally extensive anoxia that were coupled to enhanced terrestrial weathering, within the typically defined OAE interval. These anoxic intervals were separated by ocean reoxygenation and the temporary recovery of the carbon cycle.

Abstract

Oceanic Anoxic Event 2 (OAE 2), occurring ∼94 million years ago, was one of the most extreme carbon cycle and climatic perturbations of the Phanerozoic Eon. It was typified by a rapid rise in atmospheric CO₂, global warming, and marine anoxia, leading to the widespread devastation of marine ecosystems. However, the precise timing and extent to which oceanic anoxic conditions expanded during OAE 2 remains unresolved. We present a record of global ocean redox changes during OAE 2 using a combined geochemical and carbon cycle modeling approach. We utilize a continuous, high-resolution record of uranium isotopes in pelagic and platform carbonate sediments to quantify the global extent of seafloor anoxia during OAE 2. This dataset is then compared with a dynamic model of the coupled global carbon, phosphorus, and uranium cycles to test hypotheses for OAE 2 initiation. This unique approach highlights an intra-OAE complexity that has previously been underconstrained, characterized by two expansions of anoxia separated by an episode of globally significant reoxygenation coincident with the “Plenus Cold Event.” Each anoxic expansion event was likely driven by rapid atmospheric CO₂ injections from multiphase Large Igneous Province activity.

Rapid climatic warming events have repeatedly punctuated Earth’s history and have often been associated with prolonged episodes of widespread oceanic anoxia. These intervals represent major disturbances to the global carbon cycle, which contributed to marine faunal turnover and mass extinctions (1). In particular, the Mesozoic Era was characterized by numerous prolonged intervals of globally enhanced anoxia, termed “Oceanic Anoxic Events” (OAEs) (1). The Mesozoic OAEs are considered the model expression of oceanic anoxia in the Phanerozoic geological record, thus informing studies of past hyperthermal events and future environmental change (2, 3). Despite extensive study, the magnitude and timing of oceanic anoxia for OAEs are still poorly constrained, representing a fundamental limitation for defining these events. We address this shortcoming by presenting a quantitative record of oceanic anoxia for OAE 2.

OAE 2 is the most widely recognized and severe OAE, occurring at the Cenomanian−Turonian Boundary (∼94 Ma) (1) and lasting for up to ∼900 thousand years (ky) (4). A leading hypothesis for the initiation of OAE 2 is through the emplacement of one or more of the Caribbean, Madagascar, and High-Arctic Large Igneous Provinces (LIPs) (1, 5–7) and an associated outgassing of CO₂ that caused climate warming and accelerated weathering (8). Such processes likely drove an increase in the supply of nutrients, especially phosphorus (P), to the ocean, and thereby increased productivity and oxygen demand through the aerobic degradation of organic matter (1, 5, 9). Productivity was potentially further amplified by the effective recycling of P from sediments overlain by anoxic waters (10–12). Ultimately, this chain of events led to anoxic water column conditions and the enhanced preservation of organic carbon, which formed distinct black shales in marine sedimentary successions in many ocean basins. The preferential burial of isotopically light carbon (C) also contributed to a broad positive carbon isotope excursion (CIE) across OAE 2 that is utilized as a global chemostratigraphic marker of the event (1). Silicate weathering and organic carbon burial are vital components of the global carbon cycle and represent important negative feedback mechanisms that sequester atmospheric CO₂ and drive climate stabilization during global warming events (8, 12, 13). However, the detailed behavior of these recovery processes is currently limited for numerous carbon cycle perturbations in Earth history, including OAEs.

For OAE 2, there is widespread evidence of regionally extensive anoxia and euxinia (anoxic sulfidic water column conditions) in disparate localities, leading to benthic faunal extinctions, globally enhanced organic carbon burial, and the sequestration of redox-sensitive and chalcophilic trace metals (1, 9, 12, 14, 15). Oceanic anoxia is, however, a spatially heterogeneous condition, with a highly variable geochemical and sedimentary expression. A dependence on local redox indicators has significantly limited the integration of this important phenomenon into a global carbon cycle framework. To address this limitation, we present a quantitative record of the global extent of anoxia for OAE 2 using, high-resolution uranium (U) isotope (²³⁸U/²³⁵U, reformulated as δ²³⁸U; see Supporting Information) data, in combination with a biogeochemical model that calculates changes in the coupled global C, P, and U cycles. These geochemical data come from well-preserved and extensively studied carbonate sediments that were deposited under oxygenated conditions in the European shelf sea and Tethyan continental margin during OAE 2. The studied sections (Fig. 1) include deeper shelf pelagic foraminiferal–nannofossil carbonates (chalks) from Eastbourne and South Ferriby (United Kingdom), and shallow-water platform carbonates from Raia del Pedale (Italy). We also present additional lithium isotope data [⁷Li/⁶Li, reformulated as δ⁷Li (8)] for Eastbourne, which supplement previously published results (8) to give a more highly resolved complementary record of changes in the global weathering regime during OAE 2.

Fig. 1.

(A) Paleogeographic reconstruction showing locations of Eastbourne (EB), Raia del Pedale (RDP), South Ferriby (SF), and ODP site 1261, modified with permission from ref. 67. (B) The δ¹³C (15, 34, 68), δ²³⁸U_carb (±2 SE), and U/Ca (ref. 35 and this study) for EB, RDP, and SF in comparison with δ¹³C_org and δ²³⁸U from black shales at ODP site 1261 (14, 55). Correlations (blue shading) are from previous work (8, 15, 35).

Under anoxic conditions, oxidized and soluble U(VI) is drawn down from the water column and deposited in sediments as reduced and immobile U(IV), driving seawater U concentrations ([U]) to lower values (16, 17). The primary isotopic fractionation of U is also strongly redox-dependent, whereby ²³⁸U is preferentially sequestered in anoxic sediments, leaving the residual seawater relatively enriched in ²³⁵U (18–25). This geochemical behavior generates lower δ²³⁸U signatures in seawater that can be recorded in contemporary sediments. Moreover, because of the long residence time of U in the modern ocean [320 ky to 560 ky (16)], sedimentary δ²³⁸U signatures can resolve global changes in oceanic redox conditions (14, 26–29), which cannot be achieved by more traditional proxies.

For OAE 2, application of the U isotope paleoredox tracer is limited to a single study of organic-rich black shales (δ²³⁸U_shale), wherein the data are suggestive of a threefold increase in the extent of anoxia across the event (14). These data (Fig. 1) are of insufficient temporal resolution to provide insights into the rate and magnitude of ocean redox changes across the event or elucidate the temporal relationship of ocean anoxia to other aspects of the Earth system response. Furthermore, the fractionation of U into black shales is strongly influenced by local geochemical controls (30) that complicate attempts to infer the “global” U isotope signature of oceanic redox changes. By contrast, the U isotope signature of primary carbonate precipitates (δ²³⁸U_carb) records the seawater isotopic composition (δ²³⁸U_sw), with minimal additional fractionation under most environmental conditions (21, 25, 31, 32).

The datasets presented here are used to test the current hypotheses for OAE 2 initiation, and the resulting biogeochemical changes, by acting as target outputs for the C–P–U model (see Materials and Methods and Supporting Information). This approach differs from other efforts to model U isotope data in deep time (14, 26–29), in that it calculates dynamic changes to the wider Earth system, including temperature, weathering, nutrient inputs, anoxia, and carbon burial, driven by hypothesized perturbations to the carbon cycle. These changes, in turn, affect the oceanic U cycle. The benefit of this approach is that it gives an internally consistent, quantitative assessment of coupled changes in the carbon cycle.

Results and Discussion

Geochemical Reconstruction of Oceanic Anoxia.

The stratigraphically expanded Eastbourne section has been extensively studied for a multitude of paleoenvironmental proxies, owing to its lack of significant diagenetic alteration (1, 8, 15, 33, 34) (Supporting Information). In particular, the carbonates from Eastbourne have very low values of total organic carbon (TOC) (34) and were not subject to appreciable diagenetic sulfate reduction in pore waters (33), which should have minimized alteration of the primary δ²³⁸U_carb signatures. As such, the δ²³⁸U_carb record for Eastbourne should closely reflect variability in δ²³⁸U_sw throughout the OAE 2 interval and is hence used as the primary reference curve in the present study. The δ²³⁸U_carb record for Eastbourne has the highest resolution of the three sites, and systematic temporal trends are statistically identified using a locally weighted polynomial smoothing function (LOESS) to prevent undue emphasis being placed on single data points. The South Ferriby section, although also showing a low degree of diagenetic alteration (Supporting Information), is more lithified than Eastbourne, raising the possibility of greater secondary influences on δ²³⁸U_carb. It also contains a significant unconformity (Fig. 1), which is overlain by an organic-rich black shale. Mobilization of U from this horizon could potentially explain the comparatively high δ²³⁸U_carb values above the unconformity that are more typical of anoxic sediments (18, 19, 23, 30). Deposits at Raia del Pedale show evidence for partial recrystallization and dolomitization of originally mixed aragonite and high-magnesium-calcite producers (Supporting Information). In particular, the lowest δ²³⁸U_carb of the section are from dolomitized samples with anomalously high U/Ca (Fig. 1 and Fig. S3), consistent with the negative δ²³⁸U_carb offset and U enrichment observed in modern dolomites (21, 31).

Multiple lines of evidence support the fidelity of the Eastbourne δ²³⁸U_carb record. First, despite potential diagenetic influences in the Raia del Pedale and South Ferriby sections, stability is seen in the δ²³⁸U_carb records postdating OAE 2 at all three sites (Fig. 1). It is important to note that, during this stable post-OAE 2 interval, δ²³⁸U_shale values from the organic-rich black shales of Ocean Drilling Program (ODP) site 1261 (14) are ∼0.5‰ higher than contemporaneous δ²³⁸U_carb (Fig. 1). This offset is consistent with the observed magnitude and direction of U isotope fractionation during U reduction in modern organic-rich sediments under an anoxic water column (18, 19, 23, 30) and therefore suggests that Eastbourne δ²³⁸U_carb closely reflects absolute values in δ²³⁸U_sw. Furthermore, during this stable period, δ²³⁸U_carb from Raia del Pedale and South Ferriby are ∼0.2 to 0.3‰ higher than that from Eastbourne, which is consistent with observations from some modern bulk carbonate sediments that show higher δ²³⁸U_carb than seawater due to early diagenesis (31). These offsets, compared with Eastbourne, therefore likely reflect the diagenetic gradient between the carbonate sites and support a near-primary δ²³⁸U_sw signature at Eastbourne. In addition, there are two systematic decreases in δ²³⁸U_carb at Eastbourne [Anoxia Expansion (AE)1 and AE2, Fig. 2] which are associated with the progressive lowering of U/Ca and other redox-sensitive or chalcophilic trace metals (TM_redox) at multiple localities, consistent with the removal of these elements into anoxic sediments. Finally, there is a broad coupling between two decreases in the δ²³⁸U_carb and δ⁷Li records (Fig. 2), which would not be preserved during diagenesis, and substantiates the veracity of both datasets.

Fig. 2.

Complementary geochemical datasets (δ¹³C, δ¹⁸O, and δ⁷Li) (8, 34), together with the δ²³⁸U_carb and TM_redox datasets for Eastbourne (35). AE1 and AE2 are identified from coupled negative δ²³⁸U_carb excursions and TM_redox decreases. Unusual boreal faunas (blue bar) and less negative δ¹⁸O are indicative of cooler waters and mark the PCE (35, 38, 56). Black dashed arrows indicate periods of broad coupling between δ⁷Li and δ²³⁸U_carb. The positive δ¹³C excursion (34) is traditionally used to define OAE 2 (gray shading). Timescale (right axis) is based on estimates for OAE 2 of up to ∼900 ky (4) and a linear age model, with 0 ky marking the start of the CIE.

At Eastbourne, the onset of the CIE, which typically defines the beginning of OAE 2, is closely associated with a systematic decrease in δ²³⁸U_carb from a maximum of −0.24 ± 0.09‰ (2 SE) to a minimum of −0.71 ± 0.06‰ (AE1). This drop in values is accompanied by a decrease in TM_redox in multiple localities (35–37), including Eastbourne (Figs. 1 and 2) and Raia del Pedale (Fig. 1), suggesting the global expansion of oceanic anoxic conditions. After AE1, δ²³⁸U_carb rapidly increases to −0.41 ± 0.07‰, suggesting a decrease in the extent of oceanic anoxia. This dramatic change directly corresponds to the onset of the Plenus Cold Event (PCE), an interval characterized by cooler temperatures at a range of latitudes (35, 38–45). Higher δ²³⁸U_carb values averaging −0.38 ± 0.12‰ (1 SD) were maintained throughout the PCE and correspond to widely observed TM_redox enrichments (35–37) (Figs. 1 and 2), indicating the decreased sequestration and oxidative liberation of these metals under more oxygenated conditions (35). The rise in δ²³⁸U_carb is also consistent with observations of local benthic faunal repopulation and local reoxygenation in some basins, during the PCE (35, 36, 44–47). Importantly, the δ²³⁸U_carb dataset indicates that the PCE reoxygenation was globally significant, returning the ocean to a redox state similar to preperturbation conditions.

The end of the PCE corresponds to a second systematic decrease in δ²³⁸U_carb (AE2). Again, the concentrations of TM_redox decrease at both Eastbourne (Fig. 2) and Raia del Pedale (35), indicating a return to more expanded oceanic anoxia. Interestingly, the recovery of δ²³⁸U_carb from this decrease is poorly defined and appears more gradual, unlike the rapid changes in δ²³⁸U_carb observed for the PCE. The lack of a distinct post-OAE 2 recovery may indicate a lingering degree of anoxia after the CIE, or that insufficient time is recorded in these sections to observe the complete restabilization of the U and C cycles.

Both AE1 and AE2 are coupled to increased global temperatures, as indicated by δ¹⁸O in Eastbourne (Fig. 2) and organic geochemical proxies in North Atlantic sites (34, 42, 43, 45). Moreover, there is a coupled relationship between decreasing δ²³⁸U_carb and δ⁷Li during both AE1 and AE2 (Fig. 2), which is suggestive of a shift to a more congruent weathering regime [greater primary mineral dissolution than secondary mineral formation (8, 48)] during these two anoxic intervals. These two intervals are also accompanied by independent evidence for an enhanced hydrological cycle (42), together suggesting two discrete periods of enhanced global weathering flux. In contrast to AE1 and AE2, the increasing δ²³⁸U_carb and reoxygenation during the PCE corresponds to cooler temperatures and drier conditions (34, 42, 45), as well as the recovery of δ⁷Li to more positive values. The coupling of datasets therefore empirically supports the hypothesized link between enhanced weathering activity under warmer climates, increased nutrient fluxes, heightened productivity, and expanded oceanic anoxia (1, 8, 9, 42). Furthermore, the first interval of decreasing δ⁷Li and δ²³⁸U_carb (AE1) also temporally corresponds with a prominent decrease in osmium (Os) isotopes to more unradiogenic, basalt-like signatures that marks the main phase of LIP emplacement associated with the CIE (5–8). The return to more radiogenic Os isotope signatures is not seen until after the PCE (5–7), reflecting a degree of continued volcanism across the cooling event, although two separate peaks in sedimentary Os concentrations may indicate multiple phases of heightened volcanic activity (5). LIP activity is therefore likely to have driven the two intervals of enhanced weathering activity and resulted in expanded oceanic anoxia.

A Biogeochemical Model for OAE 2.

We use a C–P–U model (Supporting Information) to examine whether plausible fluxes of CO₂ from LIP activity can indeed drive the magnitude of changes observed in δ²³⁸U_carb, U/Ca, and δ¹³C at Eastbourne. The model is informed by previously established biogeochemical models (11, 13, 49–52) and calculates the coupled dynamics of C, P, and U cycling associated with changes in temperature, weathering, and oceanic anoxia, in response to hypothesized CO₂ perturbations. The C, P, and U cycles are coupled together with the following processes (Fig. S2): (i) LIP CO₂ emission, which acts as a source of C; (ii) seafloor spreading, which acts a source of C and a sink of U; (iii) silicate weathering, which is a sink of C and a source of P and U; (iv) primary production of organic matter, which is controlled by P availability and whose burial is a sink of C; and (v) oxygen demand in the ocean, which is controlled by primary productivity and organic matter remineralization rates, and determines the extent of seafloor anoxia and the burial of U(IV) in anoxic sinks. The burial of organic C also results in an input of O₂ to the atmosphere. The U isotope mass balance is included in the model, allowing calculated δ²³⁸U_sw to trace the predicted extent of seafloor anoxia, while accounting for changes in U inputs linked to silicate weathering. Similarly, δ¹³C traces relative changes in the sources and sinks of inorganic and organic C.

The model is set up for mid-Cretaceous boundary conditions, and the sensitivity to uncertainty in these parameters is discussed in Supporting Information. Background atmospheric O₂ is a particularly important boundary condition. Varying O₂ in the model serves as a convenient way to account for uncertainty in the preexisting degree of anoxia during the pre-OAE 2 interval, and hence the size of the ocean U reservoir (Fig. 3) and the sensitivity of the U cycle to CO₂ perturbations. Estimates of mid-Cretaceous atmospheric O₂ suggest greater than modern levels (49, 51, 53). Here, three baseline settings [O₂ = 1.0, 1.1, and 1.2 times present atmospheric levels (PAL)] are used to provide an envelope of uncertainty that reasonably captures pre-OAE 2 conditions, satisfies U cycle constraints, and covers the range of uncertainty in δ²³⁸U_carb (Fig. 2). The highest O₂ concentration setting (1.2 PAL) is the least sensitive to CO₂ perturbations and predicts modern levels of pre-OAE seafloor anoxia (0.3%) with the largest oceanic U reservoir [120% of today’s size (16)], and a higher baseline δ²³⁸U_sw of −0.30‰ [compared with modern seawater of −0.39 ± 0.01‰ (18, 21, 25, 30, 54)].

Fig. 3.

C–P–U model outputs in relation to events of OAE 2 identified in the δ²³⁸U_carb dataset of this study. Time (vertical axis) is given relative to the approximate start of the CIE, as in Fig. 2.

AE1.

Two CO₂ emission pulses were modeled, corresponding to AE1 and AE2, where the rate of CO₂ injection was kept constant for two discrete intervals of 150 ky. According to the C–P–U model, the δ²³⁸U_carb excursion during AE1 can be explained by a CO₂ degassing event of 0.9 to 2.25 × 10¹⁸ mol C (i.e., 0.6 to 1.5 × 10¹³ mol C⋅y⁻¹), which represents ∼7 to 19% of current total degassing estimates for the combined Caribbean and Madagascar LIPs (Supporting Information). This perturbation results in an increase of atmospheric CO₂ to 3.4 to 4.8 PAL (Fig. 3) and warming of 1.1 °C to 2.3 °C, similar to independent proxy reconstructions (39, 41, 45). The resultant increase in silicate weathering drives a ∼16 to 39% increase in P input (F_Pw; Fig. S4), thereby increasing seawater P concentrations ([P]), promoting the expansion of anoxia and the sequestration of reduced U(IV), and preferentially ²³⁸U, in anoxic sediments. The responses of ocean nutrient levels, anoxia, and δ²³⁸U_sw are delayed compared with the CO₂ emission, with the maximum extent of seafloor anoxia occurring up to ∼100 ky after the end of the CO₂ injection, and covering between 8% and 15% of the total seafloor area (Fig. 3 and Fig. S4). This temporal lag is a function of the positive P feedback mechanism (Fig. S4). When overlain by anoxic waters, sedimentary P burial efficiency is decreased (10, 11, 50, 51) due to (i) the decreased flux of P absorbed to Fe oxides (F_FeP) and (ii) an increased C/P of buried organic matter [(C/P)_organic], which suppresses productivity-driven increases in the burial of organic bound P (F_OrgP). These processes inhibit seawater P depletion in the model and fuel further productivity even when external nutrient inputs are declining, a result which is consistent with published sedimentary P data for OAE 2 (12).

PCE.

Carbon sequestration from silicate weathering and the burial of marine-derived organic matter increase during AE1 by 16 to 30% and 20 to 33%, respectively, compared with pre-OAE 2 conditions (Fig. S4). These processes drive the decrease of atmospheric CO₂ after the LIP emission ended, resulting in a temperature and weathering decrease, and hence also a decrease in F_Pw (Fig. S4). The combination of decreased F_Pw, continued high Ca-bound P burial (F_CaP), and high F_OrgP burial leads to a [P] decrease (Fig. S4), and hence biological productivity also decreases. The burial of organic matter under anoxic conditions further drives a minor (∼0.02 PAL) net increase in atmospheric O₂ during AE1 (Fig. S4). As a result of these controls, ocean oxygenation progresses and is positively reinforced by the enhanced P removal from increased F_FeP burial and decreased (C/P)_organic (Fig. S4). In response, δ²³⁸U_sw recovers relatively quickly as oceanic anoxia declines, due to the depleted global U reservoir, while [U] recovers more slowly, reaching maximum values up to 100 ky later (Fig. 3). This process could explain the stratigraphic lag between increasing δ²³⁸U_carb and U/Ca during the PCE (Figs. 1 and 2; see the discussion below). During the onset of oxygenation and increase in δ²³⁸U_sw, δ¹³C also continues to rise owing to the slower dynamics of the larger C reservoir, again consistent with the trends shown at Eastbourne for the start of the PCE (Fig. 2). Eventually, the decreased burial of organic matter, and thus isotopically light carbon, during the reoxygenation episode begins to generate the carbon isotope “trough” that is used for the chemostratigraphic correlation of the PCE (35, 45, 55, 56).

Despite the δ²³⁸U_carb increase during the PCE being consistent with the observations of cooler conditions and local benthic reoxygenation (35, 36, 44–47, 57), the magnitude and abrupt rise of the δ²³⁸U_carb at Eastbourne is difficult to simulate with the C–P–U model, assuming that this feature is not an artifact of a condensed sedimentary interval. Indeed, the modeled δ²³⁸U_sw recovery better corresponds to the later phase of the PCE, when TM_redox also recover. The rapidity of the δ²³⁸U_carb increase, and delay of the TM_redox enrichments, implies a greater sensitivity of δ²³⁸U_carb to changing redox conditions, compared with trace metal concentrations. Such sensitivity was likely driven by an additional decrease in the dynamic residence time of U, already greatly reduced to a minimum of ∼45 ky (O₂ = 1 scenario) during AE1, due to the burial of U(IV) in anoxic sinks. This reduction might have been accomplished by further U drawdown in the “hypoxic” U sink [i.e., sediments that are reducing at depth, beneath a fully or poorly oxygenated water column (19)], which is not accounted for in the model, and would have had little impact on δ²³⁸U_carb. Under such conditions, δ²³⁸U_carb would more precisely capture the onset of reoxygenation than TM_redox. Alternatively, this model–data discrepancy may be explained by the oxidation and loss of U from shales previously deposited under anoxic conditions, as has been observed in modern continental margin sediments (58), and is similar to the mechanism proposed to explain transiently light δ³⁴S_CAS values during the PCE (15, 35). This additional supply of isotopically heavy U to the ocean would act to amplify the δ²³⁸U_sw increase due to the decrease in anoxic sinks, and could further explain the faster response compared with TM_redox. It is also plausible that reoxygenation could have been accelerated by mechanisms not considered in our model, such as sea level or temperature changes (9, 59), or the additional supply of O₂ caused by pyrite burial (49, 53).

AE2.

An assumed second CO₂ injection, of lesser magnitude (0.4 to 1.05 × 10¹³ mol C⋅y⁻¹) but of the same duration as the first event, generates a return to warmer conditions after the PCE. This increase in temperature results in a return of widespread anoxia and a smaller negative shift in δ²³⁸U_sw by −0.2‰, equivalent to an expansion of oceanic anoxia of 3 to 8% of the total seafloor area. This modeled CO₂ injection also results in the addition of isotopically light carbon to the ocean−atmosphere system that augments the decrease in ¹²C-enriched carbon burial during the PCE reoxygenation episode, thereby accelerating the development of the δ¹³C trough of the PCE. Despite this minor negative excursion, relatively elevated δ¹³C values occur throughout the entire OAE 2 interval, owing to the slower dynamics of the relatively large carbon reservoir.

Wider Implications for OAEs.

Characterizing the timing, duration, and extent of oceanic anoxia during OAE 2 is fundamental to understanding the driving mechanisms of these extreme perturbations to the carbon cycle and climate system. In an attempt to replicate the distribution of anoxia indicated by local redox proxies, Earth system modeling suggests 50% of the global ocean volume was dysoxic or anoxic (9), while the shale-derived δ²³⁸U estimate suggests only 1 to 2% of the seafloor was overlain by anoxic waters (14). By contrast, the carbonate-derived δ²³⁸U dataset presented here provides a better systematic resolution of the timing and magnitude of ocean deoxygenation (Fig. 1), suggesting that between 8% and 15% of the seafloor was overlain by an anoxic water column. This reconstruction is in close agreement with recent estimates from thallium isotopes (60) and is compatible with 2 to 5% of seafloor being euxinic, based on modeling of the sulfur isotope and molybdenum isotope systems (15, 46, 61), as this extreme chemical state would only make up a fraction of the total anoxic ocean, and δ²³⁸U does not discriminate between anoxic and euxinic redox states. The importance of differentiating these two redox states is increasingly being recognized for periods of expanded anoxia (e.g., ref. 62). Crucially, the high resolution of the δ²³⁸U_carb data uniquely resolves at least two fluctuations between episodes of globally expanded and contracted anoxia during the traditionally defined OAE 2 interval. Such findings are likely applicable for other OAEs where there are suggestions of local redox and temperature variability, such as the Aptian OAE 1a (1).

The C–P–U model successfully generates the magnitude of U isotope excursions through fairly conservative CO₂ degassing levels and subsequent changes in global temperature. These amounts of CO₂ degassing probably represent upper estimates, since the model does not consider many contributory deoxygenation mechanisms, including the inverse relationship between temperature and O₂ solubility, or changes in O₂ supply through variable ocean ventilation rates (9). We suggest that these additional mechanisms are of second-order importance to changes in terrestrial weathering for driving OAE 2, although they could help refine our understanding of the precise temporal and geographical distribution of oceanic redox changes.

In summary, the coupled geochemical and modeling approach adopted here for OAE 2 allows for an internally consistent examination of global-scale deoxygenation during carbon cycle perturbation events. The U isotope dataset highlights globally significant oscillations in oceanic redox conditions within the traditionally defined OAE 2 interval that are closely coupled to changes in global temperature and hydrological regime. The C–P–U model successfully reproduces the trends in δ²³⁸U_carb through discrete atmospheric CO₂ injections, suggesting that the exceptional longevity of OAE 2 was the result of multiphase LIP activity.

Materials and Methods

U Isotope Analysis.

Carbonate samples were mechanically cleaned for weathered surfaces. Sample powders underwent oxidative−reductive cleaning (63, 64), before selective digestion using 1 M buffered sodium acetate. U isotope composition was determined using a ²³⁶U−²³³U double spike to correct for instrumental mass fractionation, where U was preconcentrated by coprecipitation and purified by ion exchange chromatographic procedures using a single column packed with UTEVA resin (Eichrom Technologies) (20, 21, 23, 65, 66). U isotope measurements were performed by multiple collector inductively coupled plasma mass spectrometer (MC-ICP-MS) at the Centre for Trace Element Analysis, University of Otago (19, 21–23).

Li-Isotope Methods.

An additional eight samples were analyzed from the Eastbourne section for Li isotopes, to increase the resolution of the already existing dataset, following the same original protocols (8). Carbonate samples were leached using 0.1 M HCl for 1 h and sample solutions were purified using a two-stage cation-exchange procedure. Analyses were performed by MC-ICP-MS at Oxford University (8).

Coupled C–P–U Model.

The box model is a simplified version of elements of the Geochemical Carbon Cycle model (GEOCARB) and the Carbon, Oxygen, Phosphorus, Sulfur, and Evolution (COPSE) Earth system model (11, 13, 49–52), focusing exclusively on the shorter-term processes of the carbon and phosphorus cycles. The model incorporates U reservoirs and the isotopic mass balance equations for C and U isotopes as tracers of the predicted outcomes to assumed CO₂ emission perturbations.