Comment on “What do we know about the evolution of Mg to Ca ratios in seawater?” by Wally Broecker and Jimin Yu


  • This is a commentary on DOI:


[1] A record of the Mg/Ca ratio of ancient seawater is essential to understand variations in the major geological processes that control ocean chemistry and for estimating past ocean temperatures. A recent contribution by Broecker and Yu (2011) regarding past seawater Mg/Ca provides an unbalanced assessment of the uncertainties and key assumptions of the methodologies considered. It misrepresents aspects of a new method to estimate past ocean Mg/Ca proposed by Coggon et al. (2010); here we provide clarification of that approach. We estimate the accuracy and precision of seawater Mg/Ca estimates derived from planktonic foraminifera test calcite, the preferred approach of Broecker and Yu. These are shown to be consistent with other records that suggest major changes in seawater cation chemistry. An explanation of how major changes in the Mg/Ca of seawater might occur is beyond the scope of this comment.

1. Introduction

[2] Broecker and Yu [2011, hereinafter BY11] selectively comment upon the conflicting evidence regarding changes in the seawater Mg/Ca ratio since the Cretaceous. BY11 prefer only a minor (× 1.7) change in seawater Mg/Ca since 55 Ma, based on their interpretation of planktonic foraminiferal calcite Mg/Ca ratios. However, they do note the switch in the dominant mineralogy of marine carbonates, from calcite to aragonite since ∼40 Ma, is consistent with a ∼3-fold increase in the Mg/Ca ratio of seawater [Morse et al., 1997]. Further support for a major increase in seawater Mg/Ca ratios during the late Cenozoic comes from analyses of (1) halite fluid inclusions, (2) marine biogenic carbonates (e.g., echinoderms, molluscs, rugrose corals, and foraminifera), (3) abiogenic marine carbonate cements, and (4) sediment and porefluid geochemical profiles [see Coggon et al., 2010; Ries, 2010; Fantle and DePaolo, 2006, and references therein].

[3] A new approach for estimating past seawater Mg/Ca based on ancient and modern suites of calcium carbonate veins (CCV) formed during low temperature hydrothermal circulation on the mid-ocean ridge flanks also estimates a factor of ∼3 change [Coggon and Teagle, 2011; Coggon et al., 2010]. BY11 reinterpret the Coggon et al. [2010] data without regard to the documented geology of the ridge flank vein suites, and suggest that the CCV data actually indicate a ∼5× change in Mg/Ca. On the basis that BY11 find even a ∼3× change in seawater Mg/Ca difficult to account for, they question the validity of the CCV approach. Here we evaluate the evidence from ridge flank hydrothermal carbonates following BY11's analysis, and then consider the accuracy and precision of their preferred foraminifera-based approach to estimate past seawater Mg/Ca.

2. Estimating Past Seawater Mg/Ca From Suites of Ocean Basalt Calcium Carbonate Veins

[4] Coggon et al. [2010] developed a new method for estimating past seawater cation ratios (e.g., Mg/Ca, Sr/Ca) following earlier observations that related the chemical and isotopic (δ18O, 87Sr/86Sr) compositions of CCVs formed within young basalts (<3.6 Ma) drilled from the hydrothermally active eastern flank of the Juan de Fuca Ridge (JdFR) to basement fluid compositions from those sites [Coggon et al., 2004]. Coggon et al. [2004] estimated fluid Mg/Ca and Sr/Ca ratios from the CCV analyses, using appropriate partition coefficients, and showed that the CCVs record similar linear trends of chemical evolution with temperature (up to ∼65°C), resulting from increasing seawater-basalt exchange, as basement fluids from the same sites. Importantly, the measured and estimated fluid Mg/Ca and Sr/Ca versus temperature trends project back to modern seawater compositions [Coggon et al., 2010]. This observation was then applied to ancient carbonate vein suites from hydrothermally extinct ridge flanks to estimate past Mg/Ca and Sr/Ca seawater compositions, indicating ∼3× increases in these ratios since ∼24 Ma.

[5] Contrary to claims by BY11, the Coggon et al. [2010] approach does not involve the linear extrapolation of ancient calcite Mg content versus formation temperatures. Data handling for ancient and JdFR CCVs is identical; fluid Mg/Ca and Sr/Ca are calculated using appropriate partition coefficients, and their trends with formation temperature extrapolated back to contemporaneous seawater. Modern basement fluid analyses supplement calculated JdFR fluid compositions, but the parent hydrothermal fluids from which ancient CCVs precipitated are unfortunately not available for analysis. The choice of partition coefficients is not arbitrary [see Coggon et al., 2004] but clearly affects calculated fluid compositions (as illustrated for Sr/Ca by Coggon et al. [2010], SOM). Nonetheless, as BY11 demonstrate, uncertainty in partitioning does not alter the result that CCVs record increasing seawater Mg/Ca since the Cretaceous, only the magnitude of this change.

[6] Coggon et al. [2010] considered the geological context of all published ridge flank carbonate analyses (compiled in their SOM) to avoid spurious interpretations (cf., BY11). Some CCV analyses are incomplete, contaminated by Mg-saponite, or have 87Sr/86Sr ratios that clearly indicate non-contemporaneous CCV precipitation at that site. BY11 express concern regarding the treatment of two cool JdFR calcite samples that Coggon et al. [2010] excluded from their analysis because they are geologically and geochemically anomalous. These veins precipitated in a hydrologically isolated diabase sill [Davis and Becker, 2002] that was intruded into sediments half a million years after the formation of the basaltic basement. Despite precipitating from cool (∼3 and 6°C) fluids, these calcites contain a significant basaltic Sr component and fall off the JdFR 87Sr/86Sr-temperature trend recorded by other carbonates and basement fluids [Coggon et al., 2004]. Furthermore, their Mg content (9 to 11% MgCO3) is approximately double that expected for abiotic calcite precipitated from modern seawater at <5°C [Dickson, 2002; Lopez et al., 2009]. Given their anomalous geological location and high Mg contents, which are not solely due to seawater-basalt exchange, these analyses were excluded. This remains our preferred treatment of these data.

[7] There are two possible explanations for the high Mg content of these two sill-hosted calcites: they precipitated from fluids with higher Mg concentrations than JdFR basement fluids of the same temperature, due to reactions with the surrounding sediments and pore fluids; or they precipitated from fluids that do fall on the JdFR ridge flank fluid Mg/Ca-temperature trend, but they partitioned more Mg than expected at this temperature. The latter may indicate that (1) our chosen partitioning relationship is inappropriate for ridge flank fluids at this temperature (as suggested by BY11), (2) the partitioning relationship is inappropriate for fluids with such high Mg/Ca ratio [e.g., Ries, 2010], or (3) some other property of the sill fluids enhanced Mg incorporation.

[8] Additional cool calcites from the JdFR basement are required to determine whether the cold sill calcites are truly anomalous and to improve this portion of the JdFR Mg/Ca-temperature trend. Experiments reveal that carbonate polymorphs occur as a function of both temperature and fluid Mg/Ca [Morse et al., 1997], with calcite only expected to precipitate from modern seawater at less than 6°C. The mineralogy-temperature distribution of JdFR carbonates is broadly consistent with these results, (Figure 1), with aragonite precipitation favored between ∼10 and 30°C. This mineral stability constraint, combined with the observed temperature-dependent evolution of JdFR basement fluids, means that we are unlikely to find calcite samples for the calibration of partition coefficients in this temperature interval.

Figure 1.

The controls of fluid Mg/Ca ratio and temperature on carbonate polymorph occurrence, determined from the mineralogy of experimentally precipitated carbonates by Morse et al. [1997] (white squares = aragonite, black squares = calcite, half filled squares = mixed aragonite and calcite), are compared to the mineralogy-temperature distributions of ridge flank hydrothermal carbonate veins from the JdFR and sites 1256 and 843 [Coggon et al., 2010] (white bar = aragonite, dark gray = calcite, light gray = aragonite and calcite; half height bars indicate mineralogy of JdFR sill-hosted calcites). The solid blue arrow indicates the evolution of JdFR basement fluids (blue crosses) and is broadly consistent with the observed mineralogy-temperature distribution of the carbonates that precipitated from these fluids. The lack of aragonite veins from 129 Ma site 1179 indicates that the Mg/Ca ratio of contemporaneous seawater was below the aragonite field, and is consistent with the Mg/Ca fluid evolution trend determined from analyses of calcites from this site (dashed blue arrow) [Coggon et al., 2010].

[9] The temperature-mineralogy distributions of ridge flank carbonates from other sites of different crustal ages provide further support for considerably lower seawater Mg/Ca ratios in the past. For example, the precipitation of calcite between 8 and 29°C at 129 Ma Site 1179 requires that the Mg/Ca of seawater was at least 2× lower when these veins formed than in the modern oceans (Figure 1). BY11 note that the only other sampled calcite to have formed at near deep-sea temperatures in the last 15 Myr was recovered from 6.9 Ma Hole 504B. That our chosen partition coefficient returns a fluid Mg/Ca ratio of ∼4.7 mol/mol for this sample, only slightly lower than that of modern seawater, is reassuring. However, given that this calcite has 87Sr/86Sr significantly lower than seawater at 6.9 Ma (87Sr/86Sr504BHiMgCC = 0.70889 versus 87Sr/86SrSW since 6.9Ma > 0.70895), this vein precipitated from a fluid significantly exchanged with the host basalt. All carbonate veins from Hole 504B precipitated from fluids with significant basaltic Sr, and there is no 87Sr/86Sr trend with temperature. Hole 504B data cannot be extrapolated to determine past seawater Mg/Ca at the time of CCV formation and were not used by Coggon et al. [2010]. Similarly, Coggon et al. [2010] did not use calcites from 15 Ma Site 1256 to estimate past seawater Mg/Ca because they all precipitated from evolved fluids above 38°C. We concur with BY11 that their estimate of past seawater Mg/Ca from these analyses is suspect.

[10] The dating of carbonate veins by Coggon et al. [2010] was not arbitrary, as suggested by BY11, but was based upon the extrapolated 87Sr/86Sr of the contemporaneous seawater and the marine Sr-isotope curve. For example CCVs from both 110 and 46 Ma crust yield extrapolated 87Sr/86Sr well above seawater at the time of crustal formation, indicating a delay between crustal and CCV formation of 40 and 15 Myr, respectively. Carbonate minerals are commonly the last secondary mineral phase to form during ridge flank hydrothermal circulation, which heat flow measurements indicate must continue for tens of millions of years off axis. The geological context and geochemistry of individual veins and vein suites must be considered.

[11] Observing that there is a trend in the Sr and Mg contents of abiotic shallow water Holocene calcites, BY11 suggest that the same controlling processes might be responsible for both this trend and the range in extrapolated Mg contents of hydrothermal veins from crust of different ages, as these show a trend with extrapolated Sr contents (see their Figure 6). The Sr and Mg contents of JdFR calcites also form a trend, albeit with a different slope to the Holocene calcites [Coggon et al., 2004]. However, this trend reflects the decrease in ridge flank fluid Mg/Ca and Sr/Ca ratios with increasing temperature that is expected from low temperature seawater-basalt reaction and observed from sampled basement fluids. The trend between the extrapolated Mg and Sr contents of calcites from crust of different ages reflects changes in seawater Mg/Ca and Sr/Ca over geological time. Different processes must be invoked to explain the array of Holocene calcites as they precipitated from modern seawater at similar temperatures.

3. Evidence From Foraminiferal Calcite

[12] Mg/Ca ratios of ancient planktonic and benthic foraminiferal tests are important for paleoceanographic reconstructions and potentially provide a constraint on ancient seawater chemistry. However, this approach to reconstructing ancient seawater Mg/Ca ratios, the preferred method of BY11, has its own major uncertainties. BY11 lacks any assessment of these uncertainties and contains a number of errors.

[13] The temperature dependence of the Mg/Ca ratio of foraminiferal calcite, (Mg/Ca)f can be presented in the form

equation image

where T is the “calcification temperature” and A and B are empirically defined constants. The multispecies calibration of Anand et al. [2003] used oxygen isotope thermometry to estimate the foraminifera calcification temperature and determine the coefficients A = 0.090 ± 0.003 and B = 0.38 ± 0.02. Application of this equation to fossil material requires factoring in the effect of changing seawater Mg/Ca ratios through time; the relationship then becomes

equation image

where (Mg/Ca)f′ is the Mg/Ca ratio of ancient foraminiferal calcite, (Mg/Ca)sw and (Mg/Ca)sw′ are the Mg/Ca ratios of modern and ancient seawater respectively [Lear et al., 2000]. Equation (2) can be rearranged to provide an estimate of ancient seawater Mg/Ca.

[14] The accuracy of the foraminiferal calcite Mg/Ca approach depends on (1) the geochemical fidelity of ancient foraminifera calcite; (2) the conformity of ancient, often extinct, species to the modern empirical temperature calibration; (3) the uncertainty inherent in the modern calibration; and (4) the accuracy of the independent paleothermometry technique used to determine the calcification temperature of ancient foraminiferal calcite.

[15] BY11 base their estimation of early Eocene seawater Mg/Ca ratios on the measurements of test calcite Mg/Ca (∼5.5 mmol/mol) from mixed-layer dwelling planktonic foraminifera in tropical north Pacific sediments that represent conditions of peak warmth during the Paleocene-Eocene Thermal Maximum (PETM) [Zachos et al., 2003]. To estimate ancient seawater Mg/Ca they then use non-PETM, early Eocene, TEX86-derived sea surface temperatures from the continental margin of the Indian Ocean [Pearson et al., 2007]. There are multiple concerns with their approach: (1) the PETM represents a transient, perhaps ≥5°C, global temperature anomaly [Dunkley Jones et al., 2010] and so Mg/Ca data from within the PETM should not be used in combination with background early Eocene temperature estimates; (2) temperature estimates from a different location, indeed a different latitude and different ocean basin, are not appropriate in determining the required estimate of foraminifera calcification temperature; (3) even if TEX86 data were available from the same location, this paleothermometer has its own significant calibration uncertainty (±2.5°C for the GDGT index-2 [Kim et al., 2010]) and the proxy is not well calibrated above 30°C [Kim et al., 2008. 2010; Liu et al., 2009]; and (4) in the modern calibrations TEX86 is a recorder of sea surface temperatures not mixed-layer planktonic foraminifera calcification temperatures. All of these points are concerned with the accuracy of the independent paleothermometer used to estimate the foraminiferal calcification temperature. The accuracy of this temperature estimate is critical in the methodology proposed by BY11; even a 2.5°C error translates into a ∼0.5 to 0.9 mol/mol error in the estimate of seawater Mg/Ca depending on the calibration used (Figure 2). Combining PETM foraminifera Mg/Ca values with non-PETM temperatures could thus cause an overestimate of seawater Mg/Ca on the order of 1 mol/mol.

Figure 2.

Estimated ancient seawater Mg/Ca as a function of the independently estimated fossil foraminifera calcification temperature. Curves represent different choices of the calibration coefficients A and B in the Mg/Ca-temperature relationship. Top curve is the calibration of Anand et al. [2003] bound by envelopes of uncertainty due to the standard error in exponent A (±0.003) (inner dark gray shaded area), and the additional uncertainty due to the standard error in pre-exponent B (±0.02) (light gray shaded area). The middle curve is the equation used by BY11. The bottom curve is the correct form of the Elderfield and Ganssen [2000] calibration, including an envelope of uncertainty given their range in pre-exponent B (0.49 to 0.56) [Elderfield and Ganssen, 2000]. All curves assume the peak-PETM mixed-layer planktonic foraminiferal Mg/Ca shell ratio of 5.5 mmol/mol used by BY11 [Zachos et al., 2003]. The estimates, and uncertainty, of seawater Mg/Ca determined using BY11's assumed PETM temperature of ∼31.5°C at tropical ODP Site 1209 using the BY11 equation (triangle) and calibrations of Anand et al. [2003] and Elderfield and Ganssen [2000] (crosses) are shown. The result is clearly dependent on the choice of calibration and the uncertainty in the independent temperature estimate; for illustration the ±2.5°C calibration uncertainty on the GDGT index-2 [Kim et al., 2010] is shown (the horizontal extent of the shaded uncertainty boxes).

[16] It is also regrettable that BY11 erroneously use, or misattribute, the coefficients used in their thermometric equation, with pre-exponent (B) of 0.38 following Anand et al. [2003] and exponent (A) of 0.1 following Elderfield and Ganssen [2000], although they only cite Elderfield and Ganssen [2000] who determine pre-exponent (B) as 0.52. The choice of calibration, and error in the calibration equation used, have an effect of up to a factor of 2 on the final estimate of ancient seawater Mg/Ca (Figure 2). The significant offsets between the calibrations (Figure 2) partly reflect the uncertainty in pre-exponent B. This uncertainty does not affect Mg/Ca paleothermometry applications in which relative temperature changes are calculated because the pre-exponent cancels out, but has a significant impact on the estimation of ancient seawater Mg/Ca following the method of BY11.

[17] Given these potentially large uncertainties in the foraminiferal-based estimate of ancient seawater Mg/Ca caused by (1) uncertainty in the Mg/Ca-temperature calibration and (2) uncertainties in the independent estimate of calcification temperature, we argue that early Paleogene (∼55 Ma) seawater Mg/Ca ratios cannot be reliably constrained using the data presented by BY11 from the PETM (Figure 2). Indeed, the envelope of uncertainty in these estimates overlaps with seawater Mg/Ca estimates from other approaches and does not rule out a significant change in the Mg/Ca ratio of seawater since the PETM. Given the ongoing difficulty of obtaining accurate, to the degree or less, sea surface temperature estimates from late Mesozoic and Paleogene successions, efforts to constrain ancient seawater Mg/Ca ratios might be better focused on comparisons between bottom water temperatures and benthic foraminiferal Mg/Ca ratios [cf. Lear et al., 2000].

4. Accounting for Past Changes in Seawater Mg/Ca

[18] Various lines of evidence indicate that the Mg/Ca ratio of seawater has increased since the Cretaceous (Table 1). While an explanation of how such changes in seawater Mg/Ca might occur is beyond the scope of this comment, we do not agree with BY11 that our current knowledge of processes, and estimates of present and past cation fluxes are sufficiently quantitative to a priori exclude some of these estimates of ancient seawater Mg/Ca ratios. BY11 claim that there are only four mechanisms that control the Mg/Ca ratio of seawater: (1) the Mg/Ca ratio of rivers, (2) the rate of seafloor spreading, (3) the rate of dolomitization, and (4) the residence time of Ca in seawater. They dismissively claim that only (3) has changed significantly through the Cenozoic but do not quantify changes in (1) or (2) and make no discussion of (4). There are, however, published arguments that all four of these mechanisms, as well as others, may have varied significantly through the Cenozoic.

Table 1. Estimates of Seawater Mg/Ca Since 180 Maa
Evidence UsedAge (Ma)Seawater Mg/Ca (mol/mol)
  • a

    Errors are quoted in parentheses.

  • b

    Data from selected ages along the seawater Mg/Ca curve of Fantle and DePaolo [2006].

  • c

    Note that there is an offset between the value quoted by BY11 (3.02) and that given by the equation of BY11 in Figure 2 (triangle). This appears to come from two sources: first, BY11 use a calcification temperature of 31.2°C, a temperature derived from Mg/Ca thermometry using the BY11 equation, a PETM planktonic foraminifera Mg/Ca value (5.5 mmol/mol), and modern seawater Mg/Ca (5.1 mol/mol), giving 26.7°C, but with the addition of 4.5°C to account for warmer tropical temperatures in the Eocene based on Tanzanian TEX86 data [Pearson et al., 2007]. Second, BY11 use this temperature estimate (31.2°C) to calculate ancient seawater Mg/Ca but have erroneously used 5.1 as both the value of modern seawater Mg/Ca and as the value of ancient foraminiferal Mg/Ca, which should, instead, be the 5.5 mmol/mol quoted in their text. This appears to be the only way to arrive at a value of 3.02 mol/mol.

Halite-hosted fluid inclusions [Lowenstein et al., 2001]54.05 (0.75)
 154.3 (0.7)
 403.7 (0.6)
 1002.3 (0.45)
 1201.6 (0.25)
Halite-hosted fluid inclusions [Horita et al., 2002]53.6
Halite-hosted fluid inclusions [Timofeeff et al., 2006]112 – 941.2 – 1.7
 120 – 1121.1 – 1.3
Echinoderm ossicles [Dickson, 2002, 2004]531.3 – 2
 1000.6 – 1.2
 1301.6 – 2.3
 1521.1 – 1.9
 1561.1 – 1.6
 1640.6 – 1.3
 1660.8 – 1.7
 1841.1 – 1.8
Benthic foraminifera [Lear et al., 2002]493.95 (0.65)
Porefluid chemical profile modelingb [Fantle and DePaolo, 2006]52.8 – 3.5
 102.6 – 2.8
 153.5 – 5
Ridge flank hydrothermal carbonate veins [Coggon et al., 2010]1.65 (1.65)5.5 (0.5)
 24 (0.1)2.3
 30 (0.2)2.1
 32 (0.1)2.0
 33 (0.1)2.4
 36 (0.7)2.4
 40 (0.1)1.8
 79 (4)1.8 (0.3)
 109 (1)1.3 (0.2)
 130 (2)1.8 (0.2)
 168 (2)1.5 (0.3)
Planktonic foraminifera [Broecker and Yu, 2011]553.02c

[19] Quantifying the variability in river chemistry and discharge over geological time is challenging. The imbalance of many oceanic geochemical budgets (e.g., 87Sr/86Sr; [Davis et al., 2003]) indicates that modern river chemistries are not representative of the long-term average for elements with long residence times in the oceans and that weathering rates are strongly influenced by glacial/interglacial cycles [Vance et al., 2009]. Consequently our estimates of ancient riverine inputs could be significantly in error.

[20] The hydrothermal contribution to the ocean includes axial and ridge flank exchanges and is not a simple function of global crustal production rate. Global fluxes resulting from seawater-basalt exchange are influenced by other parameters including spreading rates, the ratio of fast versus slow spreading ridges, the proportion of arc versus mid-ocean ridge ocean crust, and the age-area distribution of the ocean crust. All of these parameters have changed since the Cretaceous [Seton et al., 2009]. The relationship between axial hydrothermal fluxes and spreading rate remains poorly calibrated. The critical sinks for Mg are low temperature (<100°C) seawater-basalt reactions that precipitate Mg-saponite [e.g., Alt and Teagle, 1999]. These reactions occur during black smoker-type hydrothermal recharge at the ridge axis but are much more volumetrically significant on the vast ridge flanks where reactions occur for tens of millions of years. The decrease in Mg/Ca ratio of ridge flank fluids [Elderfield et al., 1999] and carbonates [Coggon et al., 2004, 2010] indicate that virtually all Mg is removed from ridge flank fluids before they are heated beyond 40°C, with significant depletion at lower temperatures. The magnitude of this Mg sink depends on the temperature-area (and hence age-area) distribution of the seafloor [Coggon et al., 2010; Mottl and Wheat, 1994]. Additional, as yet only poorly quantified, hydrothermal processes occur on slow spreading ridges, where tectonic exposure of reactive fresh gabbros and upper mantle peridotites [Bach et al., 2001] could greatly increase Mg uptake. Evaluation of the potential processes that might change seawater Mg/Ca and Sr/Ca indicates that only a decrease in ridge-flank seawater-basalt exchange provides a mechanism for the observed increases in both these ratios, which resulted in seawater Mg/Sr remaining approximately constant (∼0.6) since the Cretaceous [see Coggon et al., 2010].

5. Conclusions

[21] To make progress with the reconstruction of past seawater Mg/Ca it is essential that the uncertainties associated with all methods be prudently evaluated. Uncertainties associated with estimates based on ridge flank hydrothermal carbonate vein suites would be reduced with an improved understanding of trace element partitioning behavior. This could be achieved, in part, with an expanded database of both fluid and mineral trace metal analyses from the JdFR and other well-characterized ridge flanks, which provide important natural laboratories. Improving knowledge of trace element exchange in abiotic and biotic fluid-mineral systems is a major ongoing challenge to environmental and material sciences.


[22] We thank Harry Elderfield and Rainer Zahn for their insightful and constructive reviews, which improved this manuscript.