Elsevier

Geochimica et Cosmochimica Acta

Volume 68, Issue 17, 1 September 2004, Pages 3521-3530
Geochimica et Cosmochimica Acta

Articles
History of carbonate ion concentration over the last 100 million years1

https://doi.org/10.1016/j.gca.2004.02.018Get rights and content

Abstract

Instead of having been more or less constant, as once assumed, it is now apparent that the major ion chemistry of the oceans has varied substantially over time. For instance, independent lines of evidence suggest that calcium concentration ([Ca2+]) has approximately halved and magnesium concentration ([Mg2+]) approximately doubled over the last 100 million years. On the other hand, the calcite compensation depth, and hence the CaCO3 saturation, has varied little over the last 100 My as documented in deep sea sediments. We combine these pieces of evidence to develop a proxy for seawater carbonate ion concentration ([CO32−]) over this period of time. From the calcite saturation state (which is proportional to the product of [Ca2+] times [CO32−], but also affected by [Mg2+]), we can calculate seawater [CO32−]. Our results show that [CO32−] has nearly quadrupled since the Cretaceous. Furthermore, by combining our [CO32−] proxy with other carbonate system proxies, we provide calculations of the entire seawater carbonate system and atmospheric CO2. Based on this, reconstructed atmospheric CO2 is relatively low in the Miocene but high in the Eocene. Finally, we make a strong case that seawater pH has increased over the last 100 My.

Introduction

The Earth’s carbon cycle is currently being subjected to a severe perturbation in the form of burning of long-buried fossil fuels. Understanding the functioning of the historical carbon cycle may help us understand the implications of our present perturbations to it.

There are still many open questions: for instance, although carbon dioxide is strongly suspected to play a major role in controlling climate, there is still much uncertainty, with evidence of warm climates at times of suspected low atmospheric CO2 (Flower 1999, Pagani et al 1999). The ice core record of atmospheric CO2 concentrations exists only over the last 400,000 yr or so. We do not have any direct evidence to tell us whether the very warm Cretaceous period (135–65 Mya) was caused by high atmospheric CO2. While it is suspected that the slow deterioration in Earth climate since the Cretaceous (the trend towards an icehouse Earth; Zachos et al., 2001) has been caused by declining atmospheric CO2, lack of data prevents a definitive interpretation. More indirect approaches are, therefore, required to reconstruct the long-term history of atmospheric CO2. One possible approach is by reconstructing the history of carbonate chemistry ([CO2(aq)], [HCO32−], [CO32−]) of seawater over time. The atmosphere and the surface ocean reach carbon equilibrium within about a year, and significant imbalances cannot be maintained for longer than this. If the history of surface ocean carbonate chemistry can be calculated, then so too can the history of atmospheric CO2.

In the absence of evidence to the contrary, it was previously assumed that the concentrations of the major ions making up the dissolved salt in seawater (Cl, Na+, SO42−, Ca2+, Mg2+, and K+) were more or less constant over geological timescales Holland 1978, Holland 1984, and more rapid variations are precluded by residence times measured in millions of years (Berner and Berner, 1996). However, recent evidence shows that ocean composition has been far from constant. In this paper, the concern is primarily with calcium and magnesium concentrations. The lines of evidence for slow oscillations in seawater [Ca2+], [Mg2+], and therefore, (Mg/Ca) are as follows:

  • 1.

    The mineralogy of inorganic (nonskeletal) carbonate cements and ooids has varied over time in the geological record, with predominance of aragonite forms at some times and calcite forms at other times (Sandberg, 1983). Laboratory experiments (e.g., Morse et al., 1997) show that either calcite or aragonite precipitates out first from a solution dependent on its temperature and also on its chemistry, particularly its Mg/Ca ratio. The variation in the form of inorganically precipitated calcium carbonate through time led Sandberg to suggest an alternation in seawater chemistry: between “calcite seas” and “aragonite seas.”

  • 2.

    Hardie (1996) noted that temporal changes in the mineralogy of potash evaporites in the geological record also track Sandberg’s curve, with potash deposits characterised by MgSO4 salts more common during aragonite seas, and potash deposits characterised by KCl salts more common during calcite seas.

  • 3.

    A wide array of evidence (Stanley and Hardie, 1998) suggests that the variation in the nature of biologically precipitated (skeletal) carbonate rocks obeys a similar variation to that of the inorganic cements and ooids. Among fossilised ‘hypercalcifying’ organisms (corals, sponges, coralline algae, etc.), aragonitic species were more common during Sandberg’s aragonite seas, whereas calcitic species were more common during Sandberg’s calcite seas.

  • 4.

    These indirect suggestions of Mg/Ca oscillations have recently been reinforced by more direct measurements, from fluid inclusions in marine halites (e.g., Lowenstein et al 2001, Horita et al 2002). These fluid inclusions (microscopic globules trapped in salt crystals as they form in evaporating seawater) contain evidence of the ocean chemistry at that time. The partially evaporated nature of the fluid inclusions excludes a completely straightforward reconstruction of past seawater composition, but much information can still be derived. An immediate point of interest is that the chemistry of the fluid inclusions is often very different from that of any point along the evaporation pathway of modern-day seawater, implying very different pre-evaporation chemistries. It is not possible to evaporate modern day seawater to produce a brine resembling many of the Phanerozoic fluid inclusions. Similarities in chemical composition of fluid inclusions in rocks of similar age, but deposited in different parts of the world, argue for control by swings in global seawater composition rather than by local or regional processes Lowenstein et al 2001, Horita et al 2002. The concentrations of ions unlikely to precipitate out until very late in the evaporation sequence, and with very long residence times in seawater (e.g., Br, ∼100 My; (Holland, 1978)), can give an idea of the “degree of evaporation” of each inclusion. Two of the earliest salts to precipitate out as seawater becomes progressively more concentrated are calcium carbonate (CaCO3), then gypsum/anhydrite (CaSO4); the presence of residual [Ca2+] but no [SO42−] in samples from some times, in contrast to residual [SO42−] but no [Ca2+] at other times, points to variations in the initial [Ca2+] and [SO42−]. Through the use of these and other techniques and assumptions, best-guess [Ca2+] and [Mg2+] concentrations Zimmermann 2000, Horita et al 2002 and seawater (Mg/Ca) Lowenstein et al 2001, Horita et al 2002 have been calculated back through time from the fluid inclusions, and they agree well with Sandberg’s calcite and aragonite seas.

  • 5.

    Another recent record for past seawater (Mg/Ca) has been obtained from the (Mg/Ca) of echinoderm skeletons (Dickson, 2002). Echinoderms incorporate Mg and Ca into their shells in a variable ratio linked to that of the seawater they grow in. Their fossilized skeletons have been analysed and the inferred history of Mg/Ca broadly supports that from fluid inclusions (Dickson, 2002).

  • 6.

    Stanley et al. (2002) found in laboratory culture experiments that, like echinoderms, the (Mg/Ca) of the calcite skeletons of coralline algae reflects that of the seawater medium they grow in. The predominance of low-Mg calcite fossils during calcite seas (low seawater Mg/Ca), and of high-Mg calcite fossils during aragonite seas, (high seawater Mg/Ca) (Stanley and Hardie, 1998) therefore, also supports variable seawater (Mg/Ca) through time.

Considering only the last 100 My, the cause of the changes is uncertain, but may involve long-term variations in midocean ridge spreading rates (Hardie (1996); but see also Holland et al. (1996) and Holland and Zimmermann (1998) for counterarguments), or alternatively, a change in the mode of calcium carbonate deposition (Volk, 1989). The timing of the beginning of the most recent seawater calcium decline corresponds approximately with the laying down of the first massive coccolith chalks in the Late Cretaceous (99–65 Mya) and the beginning of significant calcium carbonate flux to deep ocean sediments following the rise to abundance of the main planktonic calcifiers, coccolithophores and foraminifera Volk 1989, Hay 1999. Most shelf sediments are eventually uplifted and the calcium within them then returned by erosion to rivers and then back to the sea; most deep-sea sediments, in contrast, are eventually subducted at continental margins, taking calcium down into the mantle. Increasing Ca2+ loss from the oceans has been accompanied by decreasing Mg2+ loss, probably because dolomitisation (formation of CaMg(CO3)2 rocks) is thought to have only taken place in shallow environments (Holland and Zimmerman, 2000).

Regardless of the cause, the point of interest for this paper is that, taken as a whole, “these studies develop an argument of unprecedented strength for a chemically dynamic ocean over the past half billion years of Earth history” (Montanez, 2002), in particular for [Ca2+] and [Mg2+]. The combined evidence (Fig. 1) suggests that [Ca2+] was more than 100% higher 100 Mya than it is today, whereas [Mg2+] was somewhere near half of today’s value.

The long-term progressive fall in seawater calcium concentration must have affected the ocean carbon system. Calcification and dissolution in the ocean have been shown to be sensitive to the calcite or aragonite saturation state of seawater (Ω), which is defined as Ω=[Ca2+]·[CO32−]/Ksp where Ksp is the stoichiometric solubility product (different for aragonite or calcite), which varies in present-day surface waters primarily as a function of temperature and salinity (Mucci, 1983). The incorporation of some magnesium rather than calcium ions into the crystal lattice affects the solubility of calcite, and we account for this effect of [Mg2+] on Ksp (section 2.2). Expressing the equation in terms of concentrations rather than activities is acceptable for our purposes (section 2.3).

Taking [Ca2+] and [Mg2+] from Figure 1 and assuming all else (including [CO32−]) at present-day values, then Ω at 100 Mya would have been ∼threefold higher than today. This would produce a CCD at ∼10 km depth (Eqn. 4 of Jansen et al. (2002)), that is to say, preventing any dissolution of CaCO3 in the ocean. A 10 km deep CCD is unlikely given the process of carbonate compensation which exerts negative feedback on a timescale of ∼10,000 yr Sundquist 1990, Sigman et al 1998; in any case it is ruled out by the geological data.

To our knowledge there are four sources of information about the history of the calcium carbonate saturation state of the ocean:

  • (A)

    From the history of calcite compensation depth (CCD) in the oceans (Fig. 2). The CCD defines the ‘snow line’ above which calcium carbonate accumulates on the seafloor, below which it dissolves. The deepest ocean sediments recovered by drilling are calcium carbonate-free at all times during the last 100 My. The 0 to 100 Mya CCD record derived from deep ocean cores (Fig. 2) suggests that, although there has been variability (for instance, Hay 1988, Lyle 2003) and a long-term trend to deeper values, nevertheless the ocean average CCD has not varied by more than ∼1.5 kilometres from its current value of ∼4.8 km. The CCD data imply that Ω has been fairly constant since the Cretaceous despite the concomitant decrease in [Ca2+].

  • (B)

    From the abundance of calcifying cyanobacteria (stromatolites) in the fossil record (Arp et al., 2001). These require surface water calcite saturation states ≥10 as a prerequisite for their formation. The geological record contains frequent occurrences of calcifying cyanobacteria throughout most of the Phanerozoic, with the striking exception of the last 100 My, from which time almost no fossilised calcifying cynaobacteria have been found (Arp et al., 2001). One possible interpretation is generally high saturation states through the Phanerozoic, falling to consistently lower values during the last 100 My.

  • (C)

    From analysis of the paleolatitudinal ranges of shallow-water biogenic carbonate (Fig. 7B) of Opdyke and Wilkinson (1993)).

  • (D)

    From analysis of the paleolatitudinal ranges of inorganically precipitated ooids and cements (Fig. 3, Fig. 4 of Opdyke and Wilkinson (1990)). Neither of these two ranges show large contractions or expansions in the past, such as might be expected to accompany any large shift in average surface ocean Ω.

We take our lead in this paper from the CCD record because of the large number of ocean cores that have been drilled and because of the unmistakable appearance of a CCD shallowing or deepening through a core location (colour change of the core). We use a smoothed fit to the long-term trends in CCD (Fig. 2), which averages out short-term CCD variations such as during the last 20 My (e.g., Lyle, 2003), during temporary episodes such as the Paleocene-Eocene Thermal Maximum (PETM) (Thomas, 1998), and during glacial-interglacial cycles (Barker and Elderfield, 2002). The CCD records deep ocean saturation state. We assume that surface saturation state tracks deep saturation state, but also explore sensitivity to this assumption in the Appendix.

The partitioning of the CaCO3 flux between shallow and deep seas can be an important control on ocean carbonate chemistry, and therefore atmospheric pCO2 Opdyke and Wilkinson 1989, Kump and Arthur 1997. In the GEOCARB model (Berner, 1994; Berner and Kothavala, 2001), atmospheric CO2 during the last 100 My was found to be quite sensitive to this partitioning (Fig. 11 of Berner (1994)). Large-scale CaCO3 deposition in shallow seas during the Late Cretaceous has been succeeded by increasing importance of deep-sea CaCO3 deposition through the Cenozoic (Hay, 1999). This study, however, is a reconstruction of [CO32−] and atmospheric pCO2 from data. It is not a mechanistic model. The location and processes of CaCO3 burial are, therefore, irrelevant to our purpose except as possible explanations of the reconstructions obtained.

Respiration of organic carbon in sediments can lead to a partial decoupling between deep ocean chemistry and the CCD (Archer and Maier-Reimer, 1994). However, this effect is likely to be of minor importance to this study (see discussion in section 4.5 of Zeebe and Westbroek (2003)). The near-constancy of the differential between planktic and benthic δ13C (Broecker and Peng, 1998) suggests that the organic carbon fluxes of today are similar to those of the past. The respiration effect is not included in our calculations.

Section snippets

Methods

Given the evidence against a large decrease in Ω over time, we reconstruct a best estimate of the evolution of [CO32−] over the last 100 My, by assuming nearly constant Ω (Fig. 2) in the face of the changes to [Ca2+] (22 down to 10.6 mMol kg−1) and [Mg2+] (30 up to 55 mMol kg−1) shown in Figure 1.

Carbonate ion concentration ([CO32−])

Combining the ‘best-fit’ scenarios (Horita et al., 2002) for [Ca2+] (22 down to 10.6 mMol kg−1) and [Mg2+] (30 up to 55 mMol kg−1) with near-constant calcite saturation state derived from Figure 2 and Eqn. 4 of Jansen et al. (2002), we calculate that surface ocean [CO32−] rose by approximately fourfold, from ∼55 μMol kg−1 at 100 Mya to its present-day value of ∼200 μMol kg−1 (Fig. 3a). The near-constant CCD documented in marine sediments records deep-water saturation state, but our

First multimillion year reconstruction of carbonate ion concentration

Although foraminifera shell thickness has been developed as a proxy for carbonate ion over relatively short timescales (0–0.05 Mya; Barker and Elderfield, 2002), there are, however, no previous reconstructions over timescales longer than a million years. The Berner, Lasaga and Garrels (BLAG) model contained variable [Ca2+], [Mg2+], and [HCO3] (Figs. 9 and 10 of Berner et al. (1983)); but it was not possible at that time to constrain their time histories with data, and they do not resemble

Conclusions

The first 100 million year-long reconstruction of carbonate ion concentration is presented here, derived from fluid inclusion evidence of variable major ion concentrations and from ocean drilling evidence of saturation state. It gives new insight into the evolution of the oceanic carbonate system over the last 100 My. It provides a strong constraint on the global carbon cycle and atmospheric CO2 against which future data and models can be tested.

Acknowledgements

We are grateful to Eric Sundquist, Paul Wilson, Martin Palmer, Dieter Wolf-Gladrow, Howard Spero, and Bradley Opdyke for comments and stimulating discussions. We also thank Robert Berner, Juske Horita, and Klaus Wallmann for sending model results and preprints, and Tony Dickson for data. T.T. has benefitted from UK Natural Environment Research Council (GT5/98/15/MSTB) and SOC Research Fellowships.

References (67)

  • D. Archer et al.

    Effect of deep-sea sedimentary calcite preservation on atmospheric CO2 concentrations

    Nature

    (1994)
  • S. Barker et al.

    Foraminiferal calcification response to glacial-interglacial changes in atmospheric CO2

    Science

    (2002)
  • R.A. Berner et al.

    The carbonate-silicate geochemical cycle and its effect on atmospheric carbon dioxide over the past 100 million years

    Am. J. Sci.

    (1983)
  • R.A. Berner

    3GEOCARB IIa revised model of atmospheric CO2 over phanerozoic time

    Am. J. Sci.

    (1994)
  • E.K. Berner et al.

    Global EnvironmentWater, Air, and Geochemical Cycles

    (1996)
  • R.A. Berner et al.

    GEOCARB IIIA revised model of atmospheric CO2 over phanerozoic time

    American J. of Science

    (2001)
  • W.S. Broecker et al.

    Tracers in the Sea

    (1982)
  • W.S. Broecker et al.

    Greenhouse Puzzles (III)Walker’s World

    (1998)
  • K. Caldeira et al.

    Seawater pH and atmospheric carbon dioxide

    Science

    (1999)
  • R.V. Demicco et al.

    Atmospheric pCO2 since 60 Ma from records of seawater pH, calcium, and primary carbonate mineralogy

    Geol.

    (2003)
  • J.A.D. Dickson

    Fossil echinoderms as monitor of the Mg/Ca ratio of phanerozoic oceans

    Science

    (2002)
  • B.P. Flower

    PalaeoclimatologyWarming without high CO2?

    Nature

    (1999)
  • L.A. Hardie

    Secular variation in seawater chemistryAn explanation for the coupled secular variation in the mineralogies of marine limestones and potash evaporites over the past 600 my

    Geol.

    (1996)
  • L.D.D. Harvey

    A quasi-one-dimensional coupled climate-carbon cycle model 2. The carbon cycle component

    J. of Geophysical Research-Oceans

    (2001)
  • W.W. Hay

    Paleoceanographya review for the GSA centennial

    Geol. Soc. Amer. Bull.

    (1988)
  • W.W. Hay

    Carbonate sedimentation through the late Precambrian and Phanerozoic

    Zbl. Geol. Palaont. Teil I

    (1999)
  • W.W. Hay et al.

    Evolution of sediment fluxes and ocean salinity

  • H.D. Holland

    The Chemistry of the Atmosphere and the Oceans

    (1978)
  • H.D. Holland

    The Chemical Evolution of the Atmosphere and Ocean

    (1984)
  • H.D. Holland et al.

    On the secular variations in the composition of Phanerozoic marine potash evaporitesReply

    Geol.

    (1998)
  • H.D. Holland et al.

    The dolomite problem revisited

    International Geology Review

    (2000)
  • H.D. Holland et al.

    On the secular variations in the composition of Phanerozoic marine potash evaporites

    Geol.

    (1996)
  • H. Jansen et al.

    Modeling the dissolution of settling CaCO3 in the ocean

    Global Biogeochemical Cycles

    (2002)
  • Cited by (0)

    1

    Associate editor: L. R. Kump

    University of Hawaii at Manoa, SOEST, 1000 Pope Road, MSB 504, Honolulu, Hawaii 96822 USA

    View full text