Journal of Earth Science  2019, Vol. 30 Issue (2): 236-243 PDF     0
Seawater Temperature and Dissolved Oxygen over the Past 500 Million Years
Song Haijun 1, Wignall Paul B. 2, Song Huyue 1, Dai Xu 1, Chu Daoliang 1
1. State Key Laboratory of Biogeology and Environmental Geology, School of Earth Sciences, China University of Geosciences, Wuhan 430074, China;
2. School of Earth and Environment, University of Leeds, Leeds LS2 9JT, UK
ABSTRACT: Ocean temperature and dissolved oxygen concentrations are critical factors that control ocean productivity, carbon and nutrient cycles, and marine habitat. However, the evolution of these two factors in the geologic past are still unclear. Here, we use a new oxygen isotope database to establish the sea surface temperature (SST) curve in the past 500 million years. The database is composed of 22 796 oxygen isotope values of phosphatic and calcareous fossils. The result shows two prolonged cooling events happened in the Late Paleozoic and Late Cenozoic, coinciding with two major ice ages indicated by continental glaciation data, and seven global warming events that happened in the Late Cambrian, Silurian- Devonian transition, Late Devonian, Early Triassic, Toarcian, Late Cretaceous, and Paleocene-Eocene transition. The SSTs during these warming periods are about 5-30¦ higher than the present-day level. Oxygen contents of shallow seawater are calculated from temperature, salinity, and atmospheric oxygen. The results show that major dissolved oxygen valleys of surface seawater coincide with global warming events and ocean anoxic events. We propose that the combined effect of temperature and dissolved oxygen account for the long-term evolution of global oceanic redox state during the Phanerozoic.
KEY WORDS: sea surface temperature    global warming    ocean anoxic event    dissolved oxygen    Phanerozoic
0 INTRODUCTION

Global warming is becoming a major environmental problem for humans in the twenty-first century. Global mean surface air temperature rose ~1 ℃ in the past century and it is likely that they will rise by > 1 ℃ toward the end of this century (Brown and Caldeira, 2017; Meinshausen et al., 2009). Rising ocean temperatures have profound impacts on reef ecosystems, triggering mass bleaching of corals (Hughes et al., 2017). Potentially severe consequences of global warming include the declines in dissolved oxygen (DO) in the ocean interior and an expansion in the area and volume of oceanic oxygen minimum zones (Stramma et al., 2008), which may further lead to significant changes in biogeochemical cycles and mass extinctions for marine organisms.

Ocean temperature and DO concentrations have varied greatly in the geologic time. For example, sea surface temperature (SST) rose 4–8 ℃ within a few thousand years at the Paleocene-Eocene transition about 55 Ma, which was accompanied by enhanced seawater anoxia and extinctions of benthic foraminifers (Dickson et al., 2012; Zachos et al., 2003, Kennett and Stott, 1991). The rapid warming of ~10 ℃ at the Permian-Triassic transition about 252 Ma is probably the most conspicuous example (Joachimski et al., 2012; Sun et al., 2012). The temperature rise and associated widespread oceanic anoxia are likely the main killing mechanism for the most severe extinction of marine organisms in the Phanerozoic Eon (Penn et al., 2018; Song et al., 2014).

The SST of ancient oceans is mainly assessed using oxygen isotope values from fossil shells. A critical issue for oxygen isotope paleothermometry is the uncertainty of the composition of seawater δ18O in deep time, whether it remained constant or changed significantly. Previous SST estimates in the Phanerozoic usually used an increased seawater δ18O with ~1‰ per 108 years (Veizer and Prokoph, 2015; Veizer et al., 2000). In this scenario, the resulting SST curve cannot explain the two first-grade cooling events in the Early Paleozoic and Middle Mesozoic because robust glacial evidence is lacking (Royer et al., 2004). Additionally, the cold climates in the Early Paleozoic conflict with extreme high level of atmospheric CO2 derived from the GEOCARBSULF model (Royer et al., 2014; Berner, 2006) and multiple proxies (Foster et al., 2017). Furthermore, recent clumped isotope evidence shows that the composition of seawater δ18O is almost invariant over the past 500 million years (Henkes et al., 2018). Here, we have assumed an invariant oxygen isotope composition of seawater to calculate temperatures in deep time.

Oxygen isotope measurements of carbonate fossils are useful in reconstructing ancient SSTs in the younger Cenozoic and Mesozoic eras, but commonly yield questionable results for the Paleozoic Era (Veizer and Prokoph, 2015; Grossman, 2012). Paleozoic δ18O values from brachiopods show an extremely wide range in composition from ~0‰ to -10‰ and, if the values are primary, then the calculated temperatures are unrealistically high, reaching 70 ℃ in the Early Paleozoic (Grossman, 2012). One scenario is that the older the samples are, the more likely the δ18O values from these samples are diagenetically altered (Grossman et al., 1996). Recent studies have shown that phosphate fossils (e.g., conodont) are more reliable records of original seawater oxygen isotopes than biogenic carbonates (Trotter et al., 2008; Joachimski et al., 2004). Thousands of oxygen isotope measurements on Paleozoic and Triassic conodonts have been published in the past decades. These results, combined with numerous data from the Mesozoic and Cenozoic carbonate fossils, make it possible to establish a reliable SST curve for the past 500 million years.

There is no means to directly measure seawater oxygen content of ancient oceans. The primary source of DO in surface seawater is air-sea exchange with atmospheric O2, leading to approximate saturation. Therefore, the DO of surface seawater relies principally on atmospheric oxygen content, SST and salinity. A reliable O2 curve for Phanerozoic atmosphere has been estimated by using the GEOCARBSULF model (Royer et al., 2014; Berner, 2006). In addition, the salinity changes of the oceans during the Phanerozoic have been constructed by a compiled dataset of volumes and masses of evaporate deposits (Hay et al., 2006). Thus, the evolution of seawater oxygen content in the geologic past can be assessed from estimated atmospheric O2 and ocean salinity, and temperature that we constructed based on oxygen isotopes.

1 OXYGEN ISOTOPE DATASET

A total of 22 796 oxygen isotope measurements of carbonate/phosphate fossils are selected to establish SST curve over the past 500 million years (see Fig. 1 and Daraset S1). The dataset is composed of published δ18O measurements. A total of 8 782 δ18O measurements are from the Phanerozoic database of Veizer and Prokoph (2015), 3 383 measurements are from Phanerozoic database of Grossman (2012), 1 677 measurements are from Cretaceous database of O'Brien et al. (2017), and 161 measurements are from Jurassic database of Dera et al. (2011). The other 8 793 δ18O measurements have been compiled in this study. A total of 3 643 δ18O measurements are from phosphate fossils, including phosphatic brachiopods (46), conodonts (3 299), and fishes (298), while 19 153 δ18O measurements are from carbonate fossils, including belemnites (1 663), bivalves (720), brachiopods (574), benthic foraminifers (129), planktonic foraminifers (15 500), gastropods (179), and coccoliths (388).

 Download: larger image Figure 1. Raw oxygen isotope data used for reconstructing sea surface temperatures over the past 500 million years. A total of 22 796 oxygen isotopic measurements are used (see Dataset S1). The scale of δ18OPhos is used for phosphatic fossils including phosphatic brachiopod, conodont, and fish. The scale of δ18OCarb is used for carbonate fossils including belemnite, bivalve, brachiopod, planktonic foraminifer, and others. The scale of temperature is used for all fossils. Dark red short line: phosphatic brachiopod; purple plus sign: conodont; red diamond: fish; blue circle: belemnite; dark red triangle: bivalve; green star: brachiopod; blue multiple sign: planktonic foraminifer; orange square: other carbonate fossils

Not all data are used to reconstructing the SST curve. The selection of data depends mainly on the following three principles. First, only data from marine fossils that represent organisms living in the shallow water habitats are selected, i.e., shallow planktons (planktonic foraminifers), nektons (conodonts, fishes, and belemnites), and benthos (bivalves, brachiopods, and gastropods), but the bathyal benthic foraminifers are excluded. Second, only lower-paleolatitude data (between 40°N and 40°S) from well-preserved specimens are selected. Not only because δ18O data from lower-paleolatitudes are abundant but also because of the strong temperature gradient between the equator and polar regions. For the regions between 40°N and 40°S, the latitude thermal gradient was much weaker under both present-day cool and Late Cretaceous warm climates (Sinninghe Damsté et al., 2010). Third, only δ18O from phosphate fossils are selected for the Paleozoic Era because of the unreliability of δ18O from carbonate fossils of this age. δ18O from carbonate fossils have a great range in composition and the significant positive correlation with δ13C (n = 4 532, r = 0.64, p < 0.001, Fig. 2a), suggesting these data are probably affected by diagenetic alteration. However, there is no δ13C-δ18O covariance for the Mesozoic and Cenozoic carbonate fossils (n = 2 066, r = 0.03, p = 0.171, Fig. 2b). Therefore, Mesozoic and Cenozoic δ18O are compiled from both carbonate and phosphate fossils.

 Download: larger image Figure 2. Cross plots of carbon and oxygen isotopes of carbonate fossils from Phanerozoic. (a) Cross plot of Paleozoic δ13C versus δ18O showing a clear positive relationship (n = 4 532, r = 0.64, p < 0.001). (b) Mesozoic and Cenozoic δ13C versus δ18O. There is no δ13C-δ18O covariance for the entire data (n = 2 066, r = 0.03, p = 0.171). Data from Grossman (2012)
2 METHODS 2.1 Paleo-Sea Surface Temperature Assessment from δ18O Dataset

Phosphate δ18O values were converted to seawater temperatures by using the equation of Lécuyer et al. (2013), assuming an average δ18O value of the ocean of -1‰

 $t = 117.4 - 4.5 \times ({{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{PO4}}}} - {{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{SW}}}}) \times 1\;000$

where t is the sea surface temperature (℃), δ18OPO4 is the O isotope composition of phosphate fossils (‰, VSMOW); δ18OSW is the O isotope composition of seawater (‰, VSMOW).

Seawater temperatures were calculated from the carbonate δ18O by using equation of Hays and Grossman (1991)

 $\begin{array}{l} \;\;\;\;\;\;\;\;\;\;t = 15.7 - 4.36 \times ({{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{Carb}}}} - {{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{SW}}}}) \times 1\;\;\;\;000 + 0.12 \times [({{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{Carb}}}} - \\ {{\rm{ \mathit{ δ} }}^{18}}{{\rm{O}}_{{\rm{SW}}}}) \times 1\;000{]^2} \end{array}$

where δ18OCarb is the O isotope composition of carbonate fossils (‰, VPDB).

2.2 Dissolved Oxygen of Shallow Oceans

The solubility equation for oxygen is based on equation of Benson and Krause (1984)

 $DO = {{\rm{O}}_2}{\rm{ }}[\frac{{F\left( {1 - {P_{{\rm{wv}}}}} \right)(1 - {\theta _{\rm{o}}})}}{{{k_{{\rm{o,s}}}}{M_{\rm{w}}}}}]$
 $F = 1\;000 - 0.716{\rm{ }}582 \times S$
 ${P_{{\rm{wv}}}} = {10^{8.071\;31 - 1\;730.63/(233.426 + t)}}/760$
 $(1 - {\theta _{\rm{o}}}) = 0.999\;025 + 1.426 \times {10^{ - 5}}t - 6.436 \times {10^{ - 8}}{t^2}$
 ${k_{{\rm{o, s}}}} = {{\rm{e}}^{3.718{\rm{ }}14 + 5{\rm{ }}596.17/T - 1{\rm{ }}049{\rm{ }}668/{T^2} + S(0.022{\rm{ }}503{\rm{ }}4 - 13.608{\rm{ }}3/T + 2{\rm{ }}565.68/{T^2})}}$

where DO is the standard air solubility concentrations of oxygen (mol/kg); O2 is the atmospheric oxygen concentration; F is a salinity factor; S is the salinity of seawater (g/kg); Pwv is vapor pressure of water (atm); θo is a constant that depends on the second virial coefficient of oxygen; ko, s is Henry coefficient for oxygen (atm); Mw is mean molecular mass of sea salt (18.015 3 g/mol); t is temperature in degrees Celsius (℃); T = temperature in kelvins (K).

2.3 Dissolved Oxygen of Deep Oceans

Deep ocean oxygen is calculated by using the equation of Sarmiento et al. (1988)

 $D{O_{\rm{d}}} = D{O_{\rm{h}}} - r(P{O_{{\rm{4d}}}} - P{O_{{\rm{4h}}}})$
 $P{O_{4{\rm{h}}}} = \frac{{P{O_{4{\rm{d}}}}{\rm{ }}{f_{{\rm{hd}}}} - {P_{\rm{h}}}}}{{{f_{{\rm{hd}}}} + T}}$

where DOd is the dissolved oxygen of deep oceans; DOh is the dissolved oxygen of high-latitude surface oceans; r is the Redfield ratio of oxygen consumption to phosphate production accompanying the oxidation of organic matter (r = ~169 for present ocean). PO4d is the phosphate concentration of deep oceans; PO4h is the phosphate concentration of high-latitude surface oceans; fhd is a circulation parameter representing local convective overturning in deep water formation regions (fhd = ~45×106 m3/s for present ocean). Ph is the particulate rain of phosphate from the high-latitude deep water formation regions (Ph = 2.31×106 mol C/s for present ocean). T is the large-scale thermohaline overturning (T = ~19×106 m3/s for present ocean).

3 RESULTS 3.1 Estimating Seawater Temperature

Oxygen isotope data from carbonate and phosphate fossils provide a Late Cambrian to present SST curve (Fig. 3). These temperature estimates exhibit first-order variations characterized by warm climate in the Early Paleozoic–Devonian and Mesozoic–Early Cenozoic, and cool climate in the Late Paleozoic and Late Cenozoic. The Late Cambrian saw the warmest climate of the past 500 million years with SST reaching to ~45 ℃. Although a rapid cooling occurred in the Late Cambrian–Early Ordovician, the SST of the Ordovician, Silurian, and Devonian is maintained at a high level around 30 ℃. Another warm period occurred in the Mesozoic–Early Cenozoic, during which the SST fluctuated around 22 ℃. Two cool intervals are identified in the Late Paleozoic and Late Cenozoic, respectively. Seawater temperature during these two cool intervals fluctuated around 17 ℃ and has a lowest level of 12 ℃ in the Late Carboniferous.

 Download: larger image Figure 3. Low-latitude sea surface temperatures, continental-scale ice sheets, and CO2 concentrations for the past 500 million years. Each temperature dot presents the mean value per one million years (data see Table S1). Vertical error bars show 95% confidence intervals. The smoothed curve is locally weighted linear regressions (LOWESS, red line). Glaciological data for ice sheets are modified from (Crowley and Berner, 2001). Atmospheric CO2 concentrations are from the GEOCARBSULF model (orange dash line) (Royer et al., 2014) and multiple proxies (orange solid line) (Foster et al., 2017). Global warming events (GWEs) are peaks identified from the temperature curve. The peak the Late Carboniferous is not considered as a GWE because the maximum temperature during this interval is only ~20 ℃, which is lower than most of Phanerozoic temperatures. LC. Late Cambrian; SD. Silurian–Devonian; FF. Frasnian–Famennian; ET. Early Triassic; TO. Toarcian; LK. Late Cretaceous; PE. Paleocene–Eocene

The second-order variations of temperatures derived from fossil oxygen isotopes reveal seven global warming events (GWEs): in the Late Cambrian, Silurian–Devonian transition, Frasnian–Famennian transition, Early Triassic, Toarcian, Late Cretaceous, and Paleocene–Eocene transition (Fig. 3). The SSTs in the global warming intervals were significantly higher than present-day average temperatures, and witnessed a downward trend to younger intervals (Fig. 3). The SSTs during the Paleocene–Eocene GWE and Mesozoic GWEs are about 5 and 10 ℃ higher than the present-day level, respectively. However, the SSTs in the Late Cambrian and the Early and Late Devonian look extraordinarily high, about 30 and 15 ℃ higher than the present level, respectively.

3.2 Estimating Dissolved Oxygen of Shallow Seawater

Dissolved oxygen of oceans is influenced by many factors such as atmospheric oxygen level, temperature, organic productivity, and ocean currents (Sarmiento et al., 1988). However, for the surface seawater, it mainly depends on the former two because seawater oxygen dominantly comes from atmosphere by air-sea exchanging process. Marine primary producers generate and release a large amount of oxygen in the photic zones through photosynthesis. However, most of the oxygen escapes into the atmosphere within a short time because the surface seawater is near-saturated with oxygen. Therefore, the DO in the surface seawater depends primarily on factors that control the saturation of oxygen in seawaters such as temperature, atmospheric oxygen level, and salinity. The DO of surface seawater has a positive relationship with atmospheric oxygen level, but a negative relationship with temperature and salinity (Figs. 4a, 4b).

 Download: larger image Figure 4. The relationship between seawater oxygen concentration and impact factors in the oceans. (a) Surface seawater oxygen concentration versus temperature (at salinity = 35‰). (b) Surface seawater oxygen concentration versus salinity (at temperature = 25 ℃). (c) Deep-water oxygen concentration versus rate of deep-water formation (represented by fhd). (d) Deep-water oxygen concentration versus phosphorus availability (represented by PO4d–PO4h). The calculations in c and d are done at present-ocean conditions, i.e., Ph = 2.31×106 mol C/s, r = 169, T = 19×106 m3/s, DOh = 325 μM, and PO4d = 2.15 μM (Sarmiento et al., 1988). POL, present oceanic level of dissolved oxygen. POL of high-latitude surface seawater is 325 μM (Sarmiento et al., 1988)

Surface seawater oxygen concentrations in the ancient near-tropic oceans are calculated based on the solubility equation for oxygen by using published salinity (Hay et al., 2006), atmospheric O2 (Royer et al., 2014), and the estimated temperature curve in this study (see Section 2). The result shows four major DO peaks in the Late Ordovician–Early Devonian (~250 μM), Late Paleozoic (~350 μM), Middle–Late Triassic (~250 μM), and Cretaceous–Cenozoic (~270 μM), and four major valleys in the Late Cambrian–Early Ordovician (~110 μM), Late Devonian (~130 μM), Early Triassic (~170 μM), and Early Jurassic (~110 μM) (Fig. 5).

 Download: larger image Figure 5. Dissolved oxygen, sea surface temperature and oceanic anoxic events for the past 500 million years. Dissolved oxygen values are calculated from temperature, atmospheric oxygen, and salinity. The temperature curve is from Fig. 2. The shade shows 90% confidence interval. Oceanic anoxic events (OAEs) are compiled based on Table S2. LC. Late Cambrian; FF. Frasnian–Famennian; ET. Early Triassic; HS. Hettangian–Sinemurian; TO. Toarcian; K. Cretaceous; PE. Paleocene–Eocene
4 DISCUSSION 4.1 The Reliability of Established Temperature Curve

There are still many uncertainties about reconstructing temperatures of ancient oceans. The critical issue is the composition of seawater oxygen isotope that we have discussed above. The second issue is the diverse habitats of different fossil groups. Although marine animals that were used to calculating temperature are shallow dwellers, most of them have a wide habitat in depth from 0 to 100–200 m. Therefore, the estimated temperature is a mixture of values from shelf waters. The third issue is seawater pH. Changes in seawater pH normally result in oxygen isotope incorporation in carbonates and produce bias on δ18O-derived temperatures (Royer et al., 2004). However, oxygen isotopes from carbonate fossils are mainly from the post-Triassic interval when there was much less ${p_{CO}}_{_2}$ variation than in the Paleozoic (Foster et al., 2017; Royer et al., 2014). In addition, other factors such as icehouse climate and diagenesis could also potentially affect the accuracy of oxygen isotope paleothermometry (Grossman, 2012). For the temperature values calculated based on raw oxygen isotope data from different fossil groups, although some data deviate from the trend, the overall change is consistent (Figs. 1, 3), suggesting the established SST curve is reliable.

4.2 The Reliability of Established Dissolved Oxygen Curve

A critical limitation when reconstructing the oxygen content of surface seawater is the uncertain of paleo-O2 and salinity. Atmospheric O2 in geologic time is usually reconstructed through geochemical models, i.e., GEOCARB-style models (Krause et al., 2018; Royer et al., 2014; Berner, 2006; Falkowski et al., 2005; Bergman, 2004). However, these constructions are only moderately constrained by proxies. Charcoal is a product of wildfire that occurs when oxygen content is over 15% (Belcher and McElwain, 2008). Sedimentary records show that charcoal is present in the sedimentary rocks of the past 400 Ma (Glasspool and Scott, 2010), suggesting oxygen content over 15% for much of the Phanerozoic. This observation is in keeping with geochemical models except in the Jurassic interval (Belcher and McElwain, 2008). Another limitation of reconstruction of seawater oxygen content is that paleo-salinity records are uncertain for the limited knowledge of the history of water on Earth. However, salinity of the past 500 million years was likely within a narrow range of between 30‰ and 50‰ (Hay et al., 2006). In this range, salinity only has a small contribution to the uncertainty of reconstruction of DO content (Fig. 4b).

4.3 The Possible Drivers of Phanerozoic Temperature Change

The first-order temperature variations are supported by robust and independent glaciological data for continental-scale ice sheets (Crowley and Berner, 2001) and carbonate clumped isotope data of fossil brachiopod and mollusk shells (Henkes et al., 2018; Came et al., 2007). For instance, the two major intervals of continental glaciation from the Late Paleozoic (ca. 338–256 Ma) and Late Cenozoic (ca. 40–0 Ma) (Crowley and Berner, 2001) coincide closely with the two lowest-temperature intervals in the past 500 Ma (Figs. 1, 3). The transition from Early Paleozoic warm climate to Late Paleozoic cool climate can be also reflected in the carbonate-clumped isotopes (Henkes et al., 2018). The estimated SST curve shows that the Late Paleozoic ice age witnessed the coldest climate over the past 500 Ma. The lowest temperature of shallow seawater during the Late Paleozoic is about 5 ℃ lower than the present-day level, which is in accord with the glaciological evidence that shows the longest and largest glacial period of the Phanerozoic at this time (Fielding et al., 2008).

Independent evidence is available for many of the second-order variations of temperatures. For example, the Paleocene-Eocene thermal maximum, PETM has been identified using multiple proxies in addition to oxygen isotope including foraminiferal magnesium/calcium ratios (Tripati and Elderfield, 2005) and TEX86 derived from the membrane lipids of marine picoplankton (Sluijs et al., 2006; Zachos et al., 2006). Toarcian GWE is independently supported by the abrupt increase of atmospheric CO2 obtained from fossil leaf stomatal frequency (McElwain et al., 2005). Early Triassic GWE reflected by marine conodont O isotopes coincided with the elevated air temperatures recorded by O isotopes from continental tetrapods (Rey et al., 2016).

Atmospheric CO2 concentrations from the GEOCARBSULF model show a strong positive correlation between SST and CO2 (r = 0.853, n = 51, p < < 0.001, see Fig. 6 and Table S2). Warm periods are characterized by higher CO2 horizons, i.e., Early Paleozoic, Mesozoic, and Early Cenozoic. Cool periods are marked by lower CO2 horizons such as Late Paleozoic and Late Cenozoic. However, the latest Ordovician ice age is an exception. Although temperature in this interval is the lowest of the Ordovician (Figs. 1, 3), CO2 is at a high level of ~2 500 ppm (Royer et al., 2014). It might be because the duration of the latest Ordovician glaciation was very short (~1 million years) (Finnegan et al., 2011), but the GEOCARBSULF model generally has a time-step of 10 million years (Royer et al., 2014). The overall striking correspondence between CO2 concentrations and surface temperatures supports the proposal that high CO2 concentrations drive or amplify high global temperatures (Berner, 2001; Crowley and Berner, 2001).

 Download: larger image Figure 6. Cross plots of sea surface temperature and atmospheric carbon dioxide (r = 0.853, n = 51, p < 0.001; Table S3). The shade shows 95% confidence interval. Atmospheric CO2 concentrations are from the GEOCARBSULF model (Royer et al., 2014)

In addition, temperature evolution calculated by oxygen isotope data seems to be associated with galactic cosmic ray flux (Shaviv and Veizer, 2003). In principle, the variations of cosmic ray flux likely contribute to the temperature evolution via controlling the formation of clouds at low latitudes (Shaviv and Veizer, 2003). Cold ages in the Late Paleozoic and Late Cenozoic coincide with high cosmic ray fluxes, whereas most warm climate intervals are characterized by low cosmic ray fluxes, e.g., Cambrian, Devonian, Triassic and Cretaceous. Only the Mid-Jurassic warm period is an exception because it occurs during a time of low cosmic ray level (Shaviv and Veizer, 2003).

4.4 The Drivers of Phanerozoic Ocean Redox State

The DO curve shows similar features of atmospheric O2 derived from Geocarbsulf model (Royer et al., 2014; Berner, 2006), suggesting atmospheric O2 level is a major factor controlling long-term evolution of oxygen content of surface seawater. The second-order variations are associated intimately with temperature, e.g., valleys of DO closely coincide with global warming events (see Fig. 5).

The reconstructed DO curve is supported by independent data on oceanic anoxic events (OAEs). Sedimentary and geochemical data show at least nine OAE intervals in the past 500 million years, i.e., Late Cambrian, Frasnian–Famennian transition, Early Triassic, Hettangian–Sinemurian, Toarcian, Aptian–Albian, Cenomanian–Turonian, Coniacian–Santonian, and Paleocene–Eocene transition (Song et al., 2017; Meyer and Kump, 2008, and Table S3). Most of those OAEs coincide with low levels of DO, especially the very low levels (<150 μM) seen in the Late Cambrian, Late Devonian, and Early Jurassic (Fig. 5). Additionally, OAEs show good correspondence with global warming events. Under a warm climate, surface seawaters absorb less oxygen, but probably also important is the likelihood of less efficient ocean circulation at these times (Sarmiento et al., 1988). Temperature is a major contributing factor on the rate of formation of deep water (fhd) by controlling the strength of ocean thermohaline circulation. In modern oceans, fhd is positively correlated with deep seawater oxygen (Fig. 4c). Therefore, deep ocean anoxia usually appeared under low fhd associated with global warming events.

Oxygen content of surface seawater is a significant factor affecting the redox state of deep oceans. When surface DO is up to 1.5 times of present-level, it is difficult to cause deep-water anoxia even under low fhd (Fig. 4c). Conversely, when surface DO fall to half, it can easily lead to anoxia in deep oceans. Another essential factor controlling the redox state of deep ocean is nutrient availability (represented by PO4dPO4h) (Sarmiento et al., 1988), which has a negative relationship with deep seawater oxygen content (Fig. 4d). Some OAEs occurred in the periods of high surface DO levels, such as Cretaceous OAE1, 2, and 3, and Paleocene–Eocene OAE. High temperatures might be the most important factor for these OAEs by reducing the the solubility of oxygen and producing sluggish ocean circulations. In addition, global warming was usually accompanied by enhanced continental weathering and increased nutrient input to oceans (Jenkyns, 2010), which would further increase the rate of oxygen consumption and lead to a global OAE faster.

ACKNOWLEDGMENTS

We thank Zhipu Qiu for collecting data, Ján Veizer for comments on earlier drafts, and Dana L. Royer for providing atmospheric oxygen and carbon dioxide data. This study is supported by the National Natural Science Foundation of China (Nos. 41821001, 41622207, 41530104, 41661134047), the State Key R & D Project of China (No. 2016YFA0601100), and the Strategic Priority Research Program of Chinese Academy of Sciences (No. XDB26000000), a Marie Curie Fellowship (No. H2020-MSCA-IF-2015-701652), and the Natural Environment Research Council's Eco-PT Project (No. NE/P01377224/1), which is a part of the Biosphere Evolution, Transitions and Resilience Program (BETR). The final publication is available at Springer via https://doi.org/10.1007/s12583-018-1002-2.

Electronic Supplementary Materials: Supplementary materi-als (Dataset S1, Tables S1, S2, S3) are available in the online version of this article at https://doi.org/10.1007/s12583-018-1002-2.

REFERENCES CITED
 Belcher, C. M., McElwain, J. C., 2008. Limits for Combustion in Low O2 Redefine Paleoatmospheric Predictions for the Mesozoic. Science, 321(5893): 1197-1200. DOI:10.1126/science.1160978 Benson, B. B., Krause, D. Jr., 1984. The Concentration and Isotopic Frac-tionation of Oxygen Dissolved in Freshwater and Seawater in Equilib-rium with the Atmosphere1. Limnology and Oceanography, 29(3): 620-632. DOI:10.4319/lo.1984.29.3.0620 Bergman, N. M., 2004. COPSE:A New Model of Biogeochemical Cycling over Phanerozoic Time. American Journal of Science, 304(5): 397-437. DOI:10.2475/ajs.304.5.397 Berner, R. A., 2001. GEOCARB Ⅲ:A Revised Model of Atmospheric CO2 over Phanerozoic Time. American Journal of Science, 301(2): 182-204. DOI:10.2475/ajs.301.2.182 Berner, R. A., 2006. GEOCARBSULF:A Combined Model for Phanerozoic Atmospheric O2 and CO2. Geochimica et Cosmochimica Acta, 70(23): 5653-5664. DOI:10.1016/j.gca.2005.11.032 Brown, P. T., Caldeira, K., 2017. Greater Future Global Warming Inferred from Earth's Recent Energy Budget. Nature, 552(7683): 45-50. DOI:10.1038/nature24672 Came, R. E., Eiler, J. M., Veizer, J., et al., 2007. Coupling of Surface Temperatures and Atmospheric CO2 Concentrations during the Palaeozoic Era. Nature, 449(7159): 198-201. DOI:10.1038/nature06085 Crowley, J. K., Berner, R. A., 2001. CO2 and Climate Change. Science, 292: 870-872. DOI:10.1126/science.1061664 Dera, G., Brigaud, B., Monna, F., et al., 2011. Climatic Ups and Downs in a Disturbed Jurassic World. Geology, 39(3): 215-218. DOI:10.1130/g31579.1 Dickson, A. J., Cohen, A. S., Coe, A. L., 2012. Seawater Oxygenation during the Paleocene-Eocene Thermal Maximum. Geology, 40(7): 639-642. DOI:10.1130/g32977.1 Falkowski, P. G., Katz, M. E., Milligan, A. J., et al., 2005. The Rise of Oxygen over the Past 205 Million Years and the Evolution of Large Placental Mammals. Science, 309(5744): 2202-2204. DOI:10.1126/science.1116047 Fielding, C. R., Frank, T. D., Isbell, J. L., 2008. The Late Paleozoic Ice Age-A Review of Current Understanding and Synthesis of Global Climate Patterns. In: Fielding, C. R., Frank, T. D., Isbell, J. L., eds., Resolving the Late Paleozoic Ice Age in Time and Space, 441: 343-354 Finnegan, S., Bergmann, K., Eiler, J. M., et al., 2011. The Magnitude and Duration of Late Ordovician-Early Silurian Glaciation. Science, 331(6019): 903-906. DOI:10.1126/science.1200803 Foster, G. L., Royer, D. L., Lunt, D. J., 2017. Future Climate Forcing Potentially without Precedent in the Last 420 Million Years. Nature Communications, 8: 14845. DOI:10.1038/ncomms14845 Glasspool, I. J., Scott, A. C., 2010. Phanerozoic Concentrations of Atmos-pheric Oxygen Reconstructed from Sedimentary Charcoal. Nature Ge-oscience, 3(9): 627-630. DOI:10.1038/ngeo923 Grossman, E. L., Mii, H. S., Zhang, C. L., et al., 1996. Chemical Variation in Pennsylvanian Brachiopod Shells-Diagenetic, Taxonomic, Micro-structural, and Seasonal Effects. SEPM Journal of Sedimentary Research, 66(5): 1011-1022. DOI:10.1306/d4268469-2b26-11d7-8648000102c1865d Grossman, E. L., 2012. Oxygen Isotope Stratigraphy. In: Gradstein, F. M., Ogg, J. G., Schmitz, M. D., et al., eds., The Geologic Time Scale 2012. Elsevier. 195-220 Hay, W. W., Migdisov, A., Balukhovsky, A. N., et al., 2006. Evaporites and the Salinity of the Ocean during the Phanerozoic:Implications for Climate, Ocean Circulation and Life. Palaeogeography, Palaeoclimatology, Palaeoecology, 240(1/2): 3-46. DOI:10.1016/j.palaeo.2006.03.044 Hays, P. D., Grossman, E. L., 1991. Oxygen Isotopes in Meteoric Calcite Cements as Indicators of Continental Paleoclimate. Geology, 19(5): 441. DOI:10.1130/0091-7613(1991)019<0441:oiimcc>2.3.co;2 Henkes, G. A., Passey, B. H., Grossman, E. L., et al., 2018. Temperature Evolution and the Oxygen Isotope Composition of Phanerozoic Oceans from Carbonate Clumped Isotope Thermometry. Earth and Planetary Science Letters, 490: 40-50. DOI:10.1016/j.epsl.2018.02.001 Hughes, T. P., Kerry, J. T., Álvarez-Noriega, M., et al., 2017. Global Warming and Recurrent Mass Bleaching of Corals. Nature, 543(7645): 373-377. DOI:10.1038/nature21707 Jenkyns, H. C., 2010. Geochemistry of Oceanic Anoxic Events. Geochemis-try, Geophysics, Geosystems, 11(3): Q03004. DOI:10.1029/2009gc002788 Joachimski, M. M., van Geldern, R., Breisig, S., et al., 2004. Oxygen Isotope Evolution of Biogenic Calcite and Apatite during the Middle and Late Devonian. International Journal of Earth Sciences, 93(4): 542-553. DOI:10.1007/s00531-004-0405-8 Joachimski, M. M., Lai, X., Shen, S., et al., 2012. Climate Warming in the Latest Permian and the Permian-Triassic Mass Extinction. Geology, 40(3): 195-198. DOI:10.1130/g32707.1 Kennett, J. P., Stott, L. D., 1991. Abrupt Deep-Sea Warming, Palaeoceano-graphic Changes and Benthic Extinctions at the End of the Palaeocene. Nature, 353(6341): 225-229. DOI:10.1038/353225a0 Krause, A. J., Mills, B. J. W., Zhang, S., et al., 2018. Stepwise Oxygenation of the Paleozoic Atmosphere. Nature Communications, 9(1): 4081. DOI:10.1038/s41467-018-06383-y Lécuyer, C., Amiot, R., Touzeau, A., et al., 2013. Calibration of the Phos-phate Ǆ18O Thermometer with Carbonate-Water Oxygen Isotope Frac-tionation Equations. Chemical Geology, 347: 217-226. DOI:10.13039/501100004794 McElwain, J. C., Wade-Murphy, J., Hesselbo, S. P., 2005. Changes in Carbon Dioxide during an Oceanic Anoxic Event Linked to Intrusion into Gondwana Coals. Nature, 435(7041): 479-482. DOI:10.1038/nature03618 Meinshausen, M., Meinshausen, N., Hare, W., et al., 2009. Greenhouse-Gas Emission Targets for Limiting Global Warming to 2¦. Nature, 458(7242): 1158-1162. DOI:10.1038/nature08017 Meyer, K. M., Kump, L. R., 2008. Oceanic Euxinia in Earth History:Causes and Consequences. Annual Review of Earth and Planetary Sciences, 36(1): 251-288. DOI:10.1146/annurev.earth.36.031207.124256 O'Brien, C. L., Robinson, S. A., Pancost, R. D., et al., 2017. Cretaceous Sea-Surface Temperature Evolution:Constraints from TEX 86 and Planktonic Foraminiferal Oxygen Isotopes. Earth-Science Reviews, 172: 224-247. DOI:10.1016/j.earscirev.2017.07.012 Penn, J. L., Deutsch, C., Payne, J. L., et al., 2018. Temperature-Dependent Hypoxia Explains Biogeography and Severity of End-Permian Marine Mass Extinction. Science, 362(6419): eaat1327. DOI:10.1126/science.aat1327 Rey, K., Amiot, R., Fourel, F., et al., 2016. Global Climate Perturbations during the Permo-Triassic Mass Extinctions Recorded by Continental Tetrapods from South Africa. Gondwana Research, 37: 384-396. DOI:10.1016/j.gr.2015.09.008 Royer, D. L., Berner, R. A., Montañez, I. P., et al., 2004. CO2 as a Primary Driver of Phanerozoic Climate. GSA Today, 14(3): 3-7. DOI:10.1130/1052-5173(2004)014<4:caapdo>2.0.co;2 Royer, D. L., Donnadieu, Y., Park, J., et al., 2014. Error Analysis of CO2 and O2 Estimates from the Long-Term Geochemical Model GEOCARB-SULF. American Journal of Science, 314(9): 1259-1283. DOI:10.2475/09.2014.01 Sarmiento, J. L., Herbert, T. D., Toggweiler, J. R., 1988. Causes of Anoxia in the World Ocean. Global Biogeochemical Cycles, 2(2): 115-128. DOI:10.1029/gb002i002p00115 Shaviv, N. J., Veizer, J., 2003. Celestial Driver of Phanerozoic Climate?. GSA Today, 13(7): 4-10. DOI:10.1130/1052-5173(2003)013<0004:cdopc>2.0.co;2 Sinninghe Damsté, J. S., van Bentum, E. C., Reichart, G. J., et al., 2010. A CO2 Decrease-Driven Cooling and Increased Latitudinal Temperature Gradient during the Mid-Cretaceous Oceanic Anoxic Event 2. Earth and Planetary Science Letters, 293(1/2): 97-103. DOI:10.1016/j.epsl.2010.02.027 Sluijs, A., Schouten, S., Pagani, M., et al., 2006. Subtropical Arctic Ocean Temperatures during the Palaeocene/Eocene Thermal Maximum. Nature, 441(7093): 610-613. DOI:10.1038/nature04668 Song, H. J., Wignall, P. B., Chu, D. L., et al., 2014. Anoxia/High Temperature Double Whammy during the Permian-Triassic Marine Crisis and Its Aftermath. Scientific Reports, 4(1): 4132. DOI:10.1038/srep04132 Song, H. J., Jiang, G. Q., Poulton, S. W., et al., 2017. The Onset of Wide-spread Marine Red Beds and the Evolution of Ferruginous Oceans. Na-ture Communications, 8(1): 399. DOI:10.1038/s41467-017-00502-x Stramma, L., Johnson, G. C., Sprintall, J., et al., 2008. Expanding Oxy-gen-Minimum Zones in the Tropical Oceans. Science, 320(5876): 655-658. DOI:10.1126/science.1153847 Sun, Y., Joachimski, M. M., Wignall, P. B., et al., 2012. Lethally Hot Temperatures during the Early Triassic Greenhouse. Science, 338(6105): 366-370. DOI:10.1126/science.1224126 Tripati, A., Elderfield, H., 2005. Deep-Sea Temperature and Circulation Changes at the Paleocene-Eocene Thermal Maximum. Science, 308(5730): 1894-1898. DOI:10.1126/science.1109202 Trotter, J. A., Williams, I. S., Barnes, C. R., et al., 2008. Did Cooling Oceans Trigger Ordovician Biodiversification? Evidence from Conodont Thermometry. Science, 321(5888): 550-554. DOI:10.1126/science.1155814 Veizer, J., Godderis, Y., François, L. M., 2000. Evidence for Decoupling of Atmospheric CO2 and Global Climate during the Phanerozoic Eon. Nature, 408(6813): 698-701. DOI:10.1038/35047044 Veizer, J., Prokoph, A., 2015. Temperatures and Oxygen Isotopic Composi-tion of Phanerozoic Oceans. Earth-Science Reviews, 146: 92-104. DOI:10.1016/j.earscirev.2015.03.008 Zachos, J. C., Wara, M. W., Bohaty, S., et al., 2003. A Transient Rise in Tropical Sea Surface Temperature during the Paleocene-Eocene Thermal Maximum. Science, 302(5650): 1551-1554. DOI:10.1126/science.1090110 Zachos, J. C., Schouten, S., Bohaty, S., et al., 2006. Extreme Warming of Mid-Latitude Coastal Ocean during the Paleocene-Eocene Thermal Maximum:Inferences from TEX86 and Isotope Data. Geology, 34(9): 737-740. DOI:10.1130/g22522.1