Evidence that low-temperature oceanic hydrothermal systems play an important role in the silicate-carbonate weathering cycle and long-term climate regulation

Authors

  • Laurence A. Coogan,

    Corresponding author
    1. School of Earth and Ocean Sciences, University of Victoria, Victoria, BC, V8P 5C2, Canada
    • Corresponding author: L. A. Coogan, School of Earth and Ocean Sciences, University of Victoria, 3800 Finnerty Road (Ring Road), Victoria, BC, Canada. (lacoogan@uvic.ca)

    Search for more papers by this author
  • Kathryn M. Gillis

    1. School of Earth and Ocean Sciences, University of Victoria, Victoria, BC, V8P 5C2, Canada
    Search for more papers by this author

Abstract

[1] The feedbacks between changes in atmospheric CO2 levels, climate, and CO2 drawdown into rocks are incompletely understood. In particular, the role of the upper oceanic crust in this long-term carbon cycling is debated. Here, a simple model for the precipitation of calcite in the upper oceanic crust is developed with the aim of understanding why Late Mesozoic upper oceanic crust contains several times higher CO2 concentrations (~2.5 wt%) than Cenozoic upper oceanic crust (~0.5 wt%). The modeling shows that neither heating of seawater, nor leaching of Ca from the rock with charge balance maintained by Mg uptake by the rock, can lead to >0.2 wt% CO2 uptake by the oceanic crust. Alkalinity production during fluid-rock reaction in the crust allows substantially more CO2 to be taken up by the crust in calcite, and is consistent with changes in the major element composition of Late Mesozoic upper oceanic crust due to hydrothermal alteration. The higher CO2 content of Late Mesozoic than Cenozoic upper oceanic crust thus requires greater alkalinity production by fluid-rock reactions in the Late Mesozoic. This may have been due to higher bottom water temperature and/or seawater having a different composition leading to different secondary minerals forming in the Late Mesozoic. Irrespective of the mechanism, the negative feedback on atmospheric CO2 levels provided by enhanced hydrothermal CO2 consumption in the Late Mesozoic was of similar magnitude to that from continental weathering.

1 Introduction

1.1 The Silicate-Carbonate Weathering Thermostat

[2] The CO2 content of the atmosphere has played a dominant role in controlling Earth's climate over much, if not all, of Earth's history making understanding the long-term controls on atmospheric CO2 a first-order goal of Earth science. The long-term carbon cycle involves the cycling of C between rocks, and the surface environment (e.g., ocean, atmosphere, biosphere; Berner, [2004]). The inorganic portion of this cycle involves the breakdown of Ca- and Mg-silicates followed by the formation of Ca- and Mg-carbonates, drawing C out of the surface environment into rocks. Metamorphic and diagenetic decarbonation reactions, and volcanism, return C from rocks to the surface environment. It is generally accepted that there must be a negative feedback between C buildup in the surficial environment and C drawdown into rocks that stabilizes the C content of the atmosphere [e.g., Walker et al., 1981; Berner et al., 1983; Berner and Caldeira, 1997; Berner, 2004]. The most widely discussed feedback is that increased atmospheric CO2 levels lead to greenhouse warming that, in-turn, lead (directly and indirectly, for example, through enhanced precipitation) to increased chemical weathering of the continents [e.g., Walker et al., 1981; Berner et al., 1983; Caldeira, 1995; Berner, 2004]. However, the link between temperature and the rate of chemical weathering rate is far from clear [e.g., Willenberg and von Blackenburg, 2010]. Another potential feedback on atmospheric CO2 levels, that has received less investigation, is the role of variations in the amount of CO2 consumed by low-temperature hydrothermal alteration of the oceanic crust [Staudigel et al., 1989; Francois and Walker, 1992; Brady and Gislason, 1997].

1.2 Carbonate Uptake by the Oceanic Crust

[3] Low-temperature hydrothermal circulation through the upper oceanic crust away from the ridge axis, driven by the cooling of the oceanic lithosphere, leads to fluid fluxes into and out of the ocean of similar magnitude to that of rivers [e.g., Mottl, 2003]. Off-axis, hydrothermal circulation is generally thought to be largely restricted to the lava section where permeability is highest (i.e., the upper ~500 m of the oceanic crust commonly referred to as the crustal aquifer). Fluid flows into the crust in regions with little or no sediment cover (outcrops), through the upper crustal aquifer, and exits back into the oceans at other outcrops [e.g., Fisher and Becker, 2000; Anderson et al., 2012]. Fluid-rock reactions within the crust change the composition of the formation fluid and lead to secondary mineral precipitation. Carbonate minerals (largely calcite but also aragonite) are ubiquitous secondary minerals precipitated within these off-axis hydrothermal systems. Carbon isotopic compositions of crustal carbonates are typically between 0 and +4‰ (see compilation in Gillis and Coogan, [2011]) consistent with a source of C largely from dissolved inorganic carbon (DIC) with lighter isotopic compositions where a sedimentary sources are expected on geological grounds (e.g., Juan de Fuca plate and Ocean Drilling Program (ODP) Hole 801C).

[4] The amount of carbonate precipitated within the crustal aquifer, tracked by bulk upper crustal CO2 content, has been shown to be a function of the age of the ocean crust [Gillis and Coogan, 2011]. The abundance of carbonates is very low in typical young (1–3 Ma) crust suggesting that very little carbonate is precipitated within the crustal aquifer at or near the ridge axis [Gillis et al., 2001; Heft et al., 2008]. Older crust shows a bimodal CO2 content with much more carbonate in Late Mesozoic than Cenozoic oceanic crust [Figure 1; Staudigel et al., 1989; Alt and Teagle, 1999; Gillis and Coogan, 2011]. Comparison of the Sr-isotopic composition of secondary carbonates within the upper oceanic crust with the Sr-isotopic composition of seawater allows the age range over which carbonates are formed to be estimated [Staudigel et al., 1981; Staudigel and Hart, 1985; Staudigel and Gillis, 1990; Gillis and Coogan, 2011]. These studies indicate that >80% of the carbonate minerals form within <20 Myr after crust formation. Hence, the different carbonate contents of Late Mesozoic and Cenozoic upper oceanic crust reflect a change in carbonate formation rates (per mass of crust formed) between these time periods [Gillis and Coogan, 2011].

Figure 1.

Secular changes in the minimum and average temperature of carbonate precipitation (based on oxygen isotope thermometry) and bulk upper crust CO2 content for drill cores from the Pacific and Atlantic ocean basins (modified from Gillis and Coogan, 2011). Late Mesozoic crust contains substantially more CO2 (in secondary carbonate minerals) than Cenozoic crust. This is clearest in the Pacific, where incorporation of sedimentary carbonate is minimal due to faster spreading rates, whereas in the Atlantic, slow spreading rates allow substantial sedimentary carbonate to be incorporated into the crust during crustal accretion (see section 4.2 in Gillis and Coogan, [2011]). The amount of sedimentary carbonate incorporated will likely vary due to local conditions (e.g., topography). The dashed line is sketched to draw the eye to the secular change in CO2 content of the upper oceanic crust and the question mark indicates uncertainty in when the oceanic crust CO2 content changed. Note also that the minimum temperature of carbonate precipitation was ~10°C higher in the Late Mesozoic than the Cenozoic, consistent with higher bottom water temperatures [Huber et al., 2002].

[5] The flux of CO2 into the oceanic crust due to carbonate mineral precipitation is ~0.5–3 × 1012 mol yr−1 [Alt and Teagle, 1999; Gillis and Coogan, 2011; Staudigel et al., 1989], similar to the degassing flux from mid-ocean ridges (~1 × 1012 mol yr−1; Saal et al., [2002]). Hence, the hydrothermal C flux into the oceanic crust is of sufficient magnitude that carbonate precipitation within the oceanic crust could be significant to Earth's long-term C-cycle. To understand the role of low-temperature hydrothermal circulation in the long-term C-cycle, it is critical to understand whether the oceanic crust is acting simply as an environment in which carbonate minerals can precipitate or is actively involved in chemical reactions that induce carbonate precipitation (i.e., is the crust being “weathered”?). For example, heating of seawater within the crust could lead to Ca2+ and HCO3 derived from continental weathering being precipitated as calcite (due to calcite's retrograde solubility)—this would be relatively unimportant to the long-term C-cycle. Likewise, if Mg2+ in seawater is exchanged mole-for-mole for Ca2+ in the basalt driving carbonate precipitation—this would also be relatively unimportant to the long-term C-cycle [e.g., Berner and Kothavala, 2001]. In contrast, breakdown of igneous minerals in the oceanic crust, releasing Ca2+ to the fluid and generating alkalinity, could drive carbonate precipitation—this could make low-temperature alteration of the upper oceanic crust an important part of the long-term C-cycle. This would be especially true if the extent of breakdown of igneous minerals in the oceanic crust was linked, either directly or indirectly, to atmospheric CO2 levels.

[6] The substantially higher rate of uptake of CO2 into the upper oceanic crust in the Late Mesozoic (~2.5 × 1012 mol yr−1; Gillis and Coogan, [2011]) than in the Cenozoic (~0.5 × 1012 mol yr−1; Gillis and Coogan, [2011]) suggests that at least one of the controls on carbonate formation within the crust changed between these time periods. Since the Late Mesozoic was a period of warm climate and high atmospheric CO2, relative to the Cenozoic, the uptake of more CO2 into the oceanic crust during this time may reflect the oceanic crust acting as a negative feedback on atmospheric CO2. Plausible models of how the oceanic crust could act as a negative feedback include through faster reaction rates due to higher deepwater temperature [Brady and Gislason, 1997] or changing seawater composition (e.g., carbonate system chemistry and cation concentrations). Here, we develop a model of carbonate precipitation in the oceanic crust, based on the physics of low-temperature hydrothermal systems and the chemistry of the carbonate system, in order to investigate the potential drivers for the observed change in C-uptake by the oceanic crust. It is shown that the difference in CO2 content of Cenozoic and Late Mesozoic oceanic crust requires a change in alkalinity production via fluid-rock reactions within the crust between these time periods. Fluid alkalinity generated within the crust was consumed by calcite precipitation. Thus, at least over the last ~150 Myr, low-temperature hydrothermal systems have contributed to the stabilization of atmospheric CO2 levels and surface temperatures.

2 A Model for Precipitation of Carbonate in the Upper Oceanic Crust

[7] Previous studies of carbonate in the oceanic crust have concentrated on the abundance and composition of the carbonate with less attention paid to the physico-chemical controls on carbonate precipitation. Here, a model is developed for carbonate precipitation in the upper oceanic crust by treating the crustal aquifer as a single box of average conditions. The model has two components. First, a simple thermal model is used to determine the amount of fluid that passes through the crustal aquifer in a given time interval. Second, the calcite saturation state of the formation fluid is determined from the temperature, pressure, and fluid composition. Calcite is then precipitated until the fluid saturation state decreases to a pre-determined level (≥1). Aragonite is not considered because its saturation state is always less than that of calcite; thus, smaller amounts of carbonate would be precipitated if aragonite was modeled instead. These steps are described below with the details given in the online supporting information.

2.1 Thermal Model

[8] The energy available to heat the formation fluid is determined from the difference between the predicted heat flux from the cooling oceanic lithosphere and the measured heat flow at the seafloor [Stein and Stein, 1994]. The energy flux available is solely dependent on the age interval over which the carbonates form, up to a maximum of ~11 TW if carbonate precipitation occurs over the entire first 65 Myr of the life of a piece of crust (Figure S2 in the supporting information). The increase in fluid temperature within the crustal aquifer prior to calcite saturation can be estimated from the difference between the minimum and average temperature of carbonate precipitation within the crust (Figure 1). This is based on the assumption that the minimum temperature of calcite formation tracks bottom water temperature, which is consistent with thermometry based on benthic foraminifera [Huber et al., 2002]. The average calcite precipitation temperature is ~7°C warmer than bottom water (Figure 1). Regions of anomalously rapid sedimentation, which lead to high upper crustal temperatures, are not considered. An alternative, less direct, approach to determining the average amount the fluid is heated is to calculate modern aquifer temperature from seafloor heat flow measurements and sediment thickness [Johnson and Pruis, 2003]. Using this approach, and assuming a well-mixed aquifer [Fisher and Becker, 2000; Anderson et al., 2012], gives an average temperature of ~10°C above bottom water temperature during the first ~40 Myr of the life of an average piece of modern crust. Based on these constraints, the modeling below explores the effect of varying the temperature increase within the aquifer from 5–10°C.

[9] Given the heat flux available to heat the aquifer and the change in fluid temperature, the fluid flux through the upper crustal aquifer is readily determined from an energy balance (equation S1). From this fluid flux, and an estimate of the mass of the aquifer, an average water-to-rock mass ratio can be determined (equation S2). This model assumes that all of the fluids involved in extracting heat from the crust during a given crustal age interval are also available for calcite precipitation; thus, it provides an upper limit on the fluid flux from which DIC can be removed. The maximum water-to-rock ratio determined this way, assuming calcite precipitation from 0 to 65 Myr and an aquifer fluid 7°C above bottom water, is ~3800. A more realistic water-to-rock ratio determined assuming calcite formation occurs between ~1 and ~20 Myr after crustal accretion (see section 1.2) is ~1600. It is informative to compare these water-to-rock ratios with the constraints imposed by the C-budget. For a seawater DIC content of ~2.3 mmol kg−1, formation of calcite at a rate of ~1 × 1012 mol yr−1 requires a minimum fluid flux of ~4 × 1014 kg yr−1 if all the C is precipitated as carbonate in the crust. For more reasonable estimates of 5 to 50% C removal from the fluid, fluid fluxes of 8 to 0.8 × 1015 kg yr−1 are required (~50 to 5% of the present-day river flux). For a 500-m high crustal aquifer through which fluid flow and carbonate precipitation occurs, and a crustal production rate of ~3 km2 yr−1, these fluid flux estimates lead to water-to-rock mass ratios of ~200 (50% DIC removal) to 2000 (5% DIC removal). These are similar to the water-to-rock ratios based on the energy balance and to independent minimum estimates based on B of >400 [Smith et al., 1995]. The water-to-rock ratio explored in the modeling below is controlled by the age interval over which carbonate precipitation occurs and the change in fluid temperature within the crust and ranges from ~550 to ~4300.

2.2 Chemical Model

[10] In the model, the formation fluid composition depends on both the composition of the fluid entering the aquifer (bottom seawater) and the fluid-rock reactions that occur within the aquifer. The composition of bottom seawater is defined using Ca, Mg, B, DIC, and alkalinity (Alk). Fluid-rock reactions within the aquifer are modeled simply as additions of Ca and/or alkalinity to the formation fluid. The saturation state of calcite (Ω), defined by the product of the concentration of the Ca and carbonate ions, divided by the solubility product of calcite (equation S4), is then determined. The solubility product of calcite is dependent on temperature, pressure, salinity, and fluid Mg/Ca [Zeebe and Wolf-Gladrow, 2001]. The fluid temperature within the aquifer is estimated from carbonate O-isotope thermometry as described above. Aquifer pressure is determined from the well-known depth-age relationship in the ocean basins [Stein and Stein, 1994] for the mid-point of the age range over which carbonate is precipitated (equation S3). Changes in aquifer pressure due to changes in sea-level, over the time period considered, have trivial impact on the results. Salinity is assumed to be constant at 35, as is B at 410 µmol kg−1 and Mg/Ca is treated as an input variable in the modeling. While it is known that B is taken up by the crust during low-temperature alteration, which will somewhat influence the role of B in controlling alkalinity, the most B-enriched rock samples only contain ~100 ppm B [Spivack and Edmond, 1987]. For a low water-to-rock ratio of 250 and a fluid B content of 4.5 ppm, this would only lead to a 10% decrease in fluid B content. This is likely a substantial overestimate of the average drop in B in the fluid, and test model runs show that this change in fluid B content makes very little difference to the results, thus changes in fluid B are not included in the modeling.

[11] Calcite precipitation from the formation fluid is modeled until the saturation state of calcite in the fluid reaches a pre-determined level. The maximum amount of calcite is precipitated if the formation fluid saturation state is set to unity, implying no kinetic inhibition of calcite precipitation. In the upper ocean, calcite is substantially supersaturated implying inorganic calcite precipitation can be substantially kinetically inhibited. To model a crustal aquifer in which calcite precipitation can be kinetically inhibited, models are shown in which calcite is precipitated until formation fluid saturation states reach different levels (Ω from 1 to 3).

[12] The changes in fluid Ca content and alkalinity, driven by fluid-rock reactions, are input parameters in the modeling. Bounds on these can be estimated based on the compositional difference between unaltered and altered upper oceanic crust. We consider well-studied drill cores from four locations (Table 1): (i) ODP Hole 504B in the eastern equatorial Pacific that is only 6.8 Myr old and has been shown to be hydrologically active, and thus the compositional changes observed are almost certainly less than those that will occur over the entire lifetime of this portion of crust [e.g., Alt et al., 1986]; (ii) Deep Sea Drilling Project (DSDP) Sites 417 and 418 in the western Atlantic that recovered ~120 Myr old crust [e.g., Staudigel et al., 1996]; (iii) ODP Hole 801C in the western Pacific that recovered ~160 Myr old crust [e.g., Kelley et al., 2003]; and (iv) the Troodos ophiolite, Cyprus, that formed ~90 Myr ago and has preserved its record of off-axis hydrothermal alteration [e.g., Gillis and Robinson, 1990].

Table 1. Summary of Bulk Changes in Lava Composition in Four Drill Cores
LocationAge (Myr)K2O (wt%)Na2O (wt%)MgO (wt%)CaO (wt%)CO2 (wt%)Silicate CaO (wt%)Silicate Ca mol yr-1Alk (mol kg−1 rock)Reference
  1. Of the other major elements Ti, Al, and (probably) Fe are too immobile to impact alkalinity, and Si is lost as aqueous SiO2 and thus does not impact alkalinity. The silicate Ca flux assumes a 500 m thick lava pile, with a density of 2500 kg m−3 and 3 km2 of crust produced per year.

  2. a

    Excludes CO2-richer alkalic cap; value comes from shorter core interval than bulk-rock chemical changes.

  3. b

    Difference between average glass reported by Fisk and Kelley [2002] and supercomposite of Kelley et al. [2003].

  4. c

    CY1 drill core.

ODP Hole 504B6.80.1 0.6−0.70.2−0.9−6.0E + 110.0

Alt et al. [1986];

Alt and Teagle [1999]

DSDP Sites 417/4181200.50.15−0.22.13.3−2.0−1.3E + 120.6Staudigel et al. [1996]
ODP Hole 801Cb1680.5−0.2−0.71.63.1a−2.2−1.5E + 121.1

Alt and Teagle [1999];

Fisk and Kelley [2002];

Kelley et al. [2003]

Troodos ophiolitec912.20.151−1.82.2−4.6−3.1E + 120.6Bednarz and Schminke [1990; all values from their Figure 6]

[13] Changes in the Ca content of the crust depend on how much crustal (silicate) Ca is leached from the crust and how much carbonate is precipitated. Comparison of fresh and altered rock Ca-contents from these four locations show that the upper 500 m of the oceanic crust can be either enriched or depleted in Ca due to fluid-rock reactions (Table 1). However, when the Ca that is stored in carbonate minerals is removed, all locations have lost silicate Ca (between 0.9 and 4.6 wt% CaO). The large water-to-rock ratios in the aquifer mean that such Ca loss from the rock leads to ≤30% increase in the formation fluid Ca content in general. In order to investigate the role of Ca release due to fluid-rock reactions, models are presented for a range from 0 to 4 wt% CaO released from silicate mineral breakdown.

[14] The carbonate ion concentration of the formation fluid is dependent on its DIC content and alkalinity. The DIC content of the formation fluid is simply dependent on the seawater DIC content and the amount of carbonate precipitated in the crust. The change in the alkalinity of the formation fluid (ΔAlk) due to fluid-silicate mineral reactions (i.e., excluding calcite precipitation) can be estimated from the charge balance associated with the change in the major element composition of the crust [Donnelly et al., 1977; Spivack and Staudigel, 1994]:

display math(1)

where ΔCa, ΔMg, ΔK, and ΔNa are the losses of each element from the silicate rock to the fluid (mol kg-1 of rock) and wr is the water-to-rock mass ratio. Uptake of positively charged ions such as K+ and Mg2+ from the formation fluid into the rock acts to decrease alkalinity. In contrast, leaching of cations such as Ca2+ from the rock produces alkalinity in the formation fluid. Table 1 shows that at the young ODP Site 504B, alkalinity production and consumption due to fluid-silicate mineral reactions cancel out, leaving no net alkalinity change from fluid-rock reactions. The Late Mesozoic drill cores have much larger alkalinity production of 0.6 to 1.1 eq kg−1 of rock. This is largely because they have lost much more silicate Ca than the younger ODP Hole 504B despite these older crustal sites having taken up much more seawater-derived K. In order to investigate the role of alkalinity production by fluid-rock reactions, models are presented for a range from 0 to 1 eq per kg of rock.

3 Results

[15] The model outlined in section 2 can be used to investigate the controls on the amount of calcite precipitated in the ocean crust and hence evaluate the origin of the much higher CO2 content of Late Mesozoic than Cenozoic oceanic crust (Figure 1). As discussed above, the free parameters in the modeling are the average: (i) age range over which calcite formation in the crust occurs; (ii) amount the formation fluid warms within the crust; (iii) alkalinity and Ca gained by the formation fluid via fluid-rock reactions within the crust; (iv) calcite saturation state at which calcite is precipitated from the formation fluid; and (v) composition and temperature of deep seawater. The first four parameters are explored initially using modern deep ocean conditions. This is followed by an exploration of the impact of the secular changes in the composition and temperature of deep seawater in section 3.2.

3.1 Models Using Modern Ocean Conditions

[16] The controls on calcite precipitation within the crust are explored by running a series of models across a broad range of values for the input parameters (i) through (iv) listed above (Table 2) and based on modern bottom seawater entering the crust. In all models, bottom seawater is assumed to be 2°C and contains 2.28 mmol kg−1 DIC and 2.38 meq kg−1 alkalinity (both are global averages from Sarmiento and Gruber, [2006]), 10.2 mmol kg−1 Ca, and 52 mmol kg−1 Mg. These input parameters give a calcite saturation state of unity at 3.8 km depth. Because mid-ocean ridges are shallower than this, and the plate subsides during aging, the fluid entering the crust in the model is calcite saturated if the average crustal age is young enough (~6.5 Myr) that the average crustal depth is less than 3.8 km. If the formation fluid is assumed to have a saturation state of unity, this means that a small amount of carbonate can be precipitated within the crust without any heating or addition of Ca or alkalinity to the fluid from fluid-rock reactions.

Table 2. Range of Parameter Space Explored in Modeling
ParameterMinimumPreferredMaximumUnits
  1. a

    3 if age of initial calcite precipitation is 0.5 Myr and 10 if age of initial calcite precipitation is 1 or 2 Myr. Using a calcite precipitation age window of just 1 or 2 Myr leads to too small a fluid flux, and hence too little DIC supply, to be realistic.

calcite saturation state (Ω)11.53 
age of crust at initial calcite precipitation0.512Myr
age of crust at final calcite precipitation3 or 10a2050Myr
silicate CaO leached from crust024wt%
alkalinity generated per kg of rock0-1Eq kg−1
change in fluid T in crust5710°C

[17] The amount of CO2 taken up in the crust via calcite precipitation is shown in Figure 2 as a function of alkalinity produced by fluid-silicate mineral reactions in the crust using the range of input parameters in Table 2. It should be noted that the alkalinity production is set per kilogram of rock; i.e., for lower water-to-rock ratios, a given value of alkalinity production leads to a greater change in fluid alkalinity. If calcite precipitation is not kinetically inhibited (Ωaquifer = 1), the amount of calcite formed in the crust is strongly correlated with the alkalinity produced by fluid-rock reactions within the aquifer with the other parameters having a second-order control on the amount of carbonate formed (Figure 2). If no alkalinity is produced by fluid-rock reactions between 0 and 0.16 wt%, CO2 is added to the crust across the entire range of parameters considered. This is less than observed in Cenozoic upper oceanic crust, and far less than in Late Mesozoic upper oceanic crust (Figure 1). This suggests that fluid-rock reactions must lead to alkalinity production that in-turn drive calcite precipitation (Figure 2; cf. Spivack and Staudigel, [1994]). The range in the amount of CO2 added to the crust at any given amount of alkalinity produced by fluid-rock reaction reflects the model variables other than alkalinity production (i.e., age range over which calcite forms, Ca added to the fluid via fluid-rock reactions, and increase in fluid temperature within the aquifer). Alkalinity production of at least ~0.5 eq kg−1 of rock must have accompanied fluid-rock reaction within the crust in Late Mesozoic to explain the 2–3 wt% CO2 observed in the crust if bottom seawater composition and temperature were similar to today (see below). The final model fluid compositions typically have alkalinity of 1.5 to 2.3 meq kg−1 with correlated drops in DIC and with pH generally between 7.5 and 7.9. Comparison to measured crustal fluid compositions is complicated because of both: (i) the small number of fluids that have been sampled and, (ii) the fact that the bulk of the fluid sampling that has occurred has been from the highly anomalous Juan de Fuca plate where crustal temperatures are very high due to rapid, early sedimentation.

Figure 2.

Results of model runs showing CO2 uptake by the upper oceanic crust, through calcite precipitation, as a function of alkalinity generated by fluid-rock reactions within the upper crustal aquifer. All models use a modern seawater composition and temperature entering the aquifer. Different symbols reflect different saturation states of the formation fluid required for calcite precipitation. Multiple symbols at a given alkalinity and Ω value reflect models run for calcite precipitation over different crustal age ranges (minimum and maximum values in Table 2), different temperatures in the aquifer (minimum and maximum values in Table 2), and different amounts of Ca leached from the crust (0, 2, and 4 wt% CaO) totaling 126 models. Note that the final alkalinity of the formation fluid is generally less than that of the incoming seawater irrespective of alkalinity production in the crust; this is due to alkalinity consumption during calcite precipitation (fluid pH is generally between 7.5 and 8 after calcite precipitation). The horizontal dashed lines show the CO2 content observed in 0–50 Myr and 90–170 Myr Pacific crust. The inset shows that, for formation fluid saturation states >1, the amount of calcite precipitated in the crust is a strong function of the water-to-rock ratio due to the dilution of alkalinity with increasing water-rock ratio (see text for discussion).

[18] In model runs in which calcite is assumed to only precipitate at formation fluid saturation states >1, the amount of CO2 added to the crust for a given alkalinity gain is lower, and varies more as a function of the other parameters, than for the case of Ω = 1 (Figure 2). For Ω = 2 (or higher), no calcite is formed without alkalinity production via fluid-rock reaction. The increased variability in the amount of calcite formed is almost entirely due to the variation in fluid alkalinity induced by changing the water-to-rock ratio between different models at constant alkalinity generation (Figure 2, inset). Because alkalinity production from fluid-rock reactions is set in the model per mass of rock, increased water-to-rock ratios decrease the fluid alkalinity if all other parameters are held constant. Water-to-rock ratio impacts the amount of calcite precipitated in the crust much more for models in which calcite precipitation only occurs at Ω > 1 than in models in which calcite precipitation occurs for Ω = 1. The reason for this is that for Ω = 1 the fluid is initially calcite saturated or very close to calcite saturation. Thus, although increased water-to-rock ratios dilute alkalinity addition from fluid-rock reactions, the supply of greater amounts of (near) calcite-saturated fluid offsets this effect in the total amount of carbonate precipitated. In contrast, if a calcite saturation state of 2 or higher is required prior to calcite precipitation, no calcite is formed without alkalinity generation from fluid-rock reactions. In this case, increased water-to-rock ratios dilute the alkalinity of the fluid and hence decrease the amount of calcite precipitated (inset, Figure 2).

[19] In summary, precipitation of sufficient calcite to add >0.2 wt% CO2 to the upper oceanic crust requires that fluid-rock reactions act as a source of alkalinity to the formation fluid (Figure 2). Using a modern ocean composition feeding the upper crustal aquifer, the addition of ~0.5 wt% CO2 to the crust, as observed in Cenozoic crust, requires ~0.1 eq of alkalinity to be produced by hydrothermal alteration of each kilogram of upper crustal rock (assuming Ω = 1; more if Ω > 1; Figure 2). In contrast, the ~2 to 3 wt% CO2 added to Late Mesozoic age upper oceanic crust requires ≥0.5 eq of alkalinity to have been produced by alteration of each kilogram of upper crustal rock at this time (Figure 2); i.e., this modeling suggests a fivefold decrease in alkalinity production by fluid-rock reactions within the upper oceanic crust between the Late Mesozoic and Cenozoic. For a Late Mesozoic crust production rate of ~4 km2 yr−1 [Seton et al., 2009], this amounts to ≥2.5 × 1012 eq yr−1. For context, this is ~20% of the estimated present-day alkalinity flux of rivers derived from continental weathering of silicate minerals [e.g., Meybeck, 1987].

3.2 Models Using a Simple Ocean-Atmosphere Carbon Model to Constrain the Impact of Changing Seawater Composition and Temperature

[20] The modeling presented in section 3.1 shows that for modern deep-ocean conditions (temperature and composition), large amounts of CO2 can only be added to the upper crust by hydrothermal circulation if fluid-rock reactions generate significant alkalinity. However, Late Mesozoic seawater is thought to have been more Ca-rich and more Mg-poor than modern seawater, based on fluid inclusions in evaporates, and biogenic carbonate compositions [e.g., Lowenstein et al., 2001; Horita et al., 2002; Timofeeff et al., 2006]. Late Mesozoic deep seawater is also thought to have been warmer by 10–20°C than modern deep seawater [Huber et al., 2002; Figure 1]. Likewise, atmospheric CO2 concentrations have changed over this time and, in combination, these factors will have changed the carbon chemistry of the ocean [e.g., Tyrrell and Zeebe, 2004]. Could these differences in seawater chemistry and temperature explain the higher CO2 content of Late Mesozoic age upper oceanic crust compared to that formed in the Cenozoic?

[21] In this section, the range of plausible compositions of deep seawater entering the crustal aquifer is determined using a simple model of the ocean-atmosphere carbon system that ties changes in ocean Ca and Mg contents and temperature to changes in carbonate chemistry. The basic premise is that deep marine carbonate sediments buffer the calcite saturation state of the deep ocean. Since the depth of carbonate compensation has not changed dramatically (always between ~3 and 5 km) since the Jurassic [Van Andel, 1975], changes in oceanic Ca content must have been accompanied by changes in the oceanic carbon system (specifically, deep ocean inline image concentration). If ocean Ca and the calcite saturation depth are defined, the deep water inline image content is also fixed (equation S4). Surface water inline image content can then be determined based on an estimate of the gradient in the inline image ion content between deep and shallow water. This defines one of the six parameters in the carbonate system for the surface ocean. Given atmospheric and hence upper ocean CO2 content, the entire carbon system of the upper ocean is then defined allowing, for example, upper ocean DIC and alkalinity to be determined (e.g., from the intersection of the two sets of lines in Figure 3; Sundquist, [1986]; Broecker and Sanyal, [1998]; Zeebe, [2001]). Finally, if the difference in DIC (or some other parameter) between the shallow and deep ocean can be estimated, the entire deep ocean C-system is also defined (Figure 3).

Figure 3.

Possible past ocean DIC content and alkalinity used in modeling the impact of changing ocean composition on calcite precipitation in the crust (Figure 5). Diagonal blue lines show the covariation of DIC and alkalinity for different atmospheric CO2 concentrations. Diagonal red lines show the covariation of DIC and alkalinity for different saturation depths (Ω = 1) of 3 and 5 km for an oceanic Ca content of 10 mmol kg−1 (modern). Diagonal green lines are the same but for an oceanic Ca content of 25 mmol kg−1 (possible Late Mesozoic ocean). Given the Ca content of the ocean, the saturation depth, atmospheric CO2 content, the upper ocean DIC, and alkalinity are defined by the intersection of the two sets of lines [Sundquist, 1986; Broecker and Sanyal, 1998; Zeebe, 2001]. Deep ocean DIC and alkalinity can then be determined by assuming a given difference between upper ocean and deep ocean DIC (here assumed to be 0.3 mmol kg−1 following Sundquist, [1986]). The black dots are the range of surface ocean conditions considered in the modeling shown in Figure 5 and the arrows connect these to the corresponding deep water conditions (see text for details). Constructed for seawater Mg content of 52 mmol kg−1, deep water temperature of 2°C, and surface water temperature of 20°C.

[22] Tyrrell and Zeebe [2004] discuss the gradient in carbonate ion concentration in the ocean and their approach of assuming that the surface water carbonate ion concentration has been roughly double that of the deep ocean since the Late Mesozoic is followed here. Figure 3 shows that for a modern ocean Ca content of 10.2 mmol kg−1, a calcite saturation depth of 4 km (i.e., depth at which Ω = 1), and an atmospheric CO2 of 500 ppm, upper ocean DIC and alkalinity are ~2.15 mmol kg−1 and ~2.4 meq kg−1, respectively. Increasing atmospheric CO2 to 1000 ppm while maintaining the same calcite saturation depth and ocean Ca leads to an increase in DIC to ~2.9 mmol kg−1 and an increase in alkalinity to ~3.1 meq kg−1. These changes reflect the need for higher DIC and alkalinity to maintain the same shallow water carbonate ion concentration at higher CO2 levels.

[23] Using Figure 3 to constrain possible surface-ocean compositions and assuming a deep ocean DIC 0.3 mmol kg−1 higher than the upper ocean [Sundquist, 1986; Key et al., 2004], the deep ocean alkalinity can be determined for different ocean Ca contents, calcite saturation depths, and atmospheric CO2. A series of deep ocean compositions were calculated for a wide range of atmospheric CO2 (200–2000 ppm), different calcite saturation depths (3 to 5 km), and ocean Ca contents (10 and 25 mmol kg−1; Figure 3). Using these deep ocean compositions and the preferred estimates from Table 2 for all other parameters, CO2 uptake by the crust is almost entirely controlled by the extent of alkalinity production in the crust (Figure 4) as is the case for modern ocean chemistry (Figure 2). This is despite the wide range of deep ocean alkalinity (1.8 to 3.8 mmol kg−1), DIC (1.8 to 3.8 mmol kg−1), and Ca (10 to 25 mmol kg−1) considered. Increasing the bottom water temperature from 2°C to 15°C while maintaining the same saturation depth, seawater Ca content, and atmospheric CO2 leads to a small decrease in the CO2 uptake by the crust (0.2–0.4 wt%). Changing the Mg content of the ocean changes the solubility product of calcite and the speciation of C in seawater. Models were run using the empirical model of Tyrrell and Zeebe [2004] to account for the former and of Ben-Yaakov and Goldhaber [1973] to account for the latter; these models show that changing Mg/Ca has a trivial effect on the amount of C taken up by the crust (<0.1 wt%). While this empirical approach to quantifying the effect of Mg on calcite saturation has been questioned (see Figure 4 in Roberts and Tripati, [2009]), whatever the effect of Mg on the crustal system is, buffering of the deep ocean chemistry by carbonate sediments will mean that this has little effect on calcite precipitation within the crust.

Figure 4.

Results of models showing carbon uptake by the upper oceanic crust, through calcite precipitation, as a function of alkalinity generated by fluid-rock reactions within the upper crust for realistic paleo-ocean compositions based on a simple ocean carbon model (Figure 3). Symbols indicate different seawater Ca contents, and the different models at a given seawater Ca content are for different atmospheric CO2 and saturation depth (which define seawater DIC and alkalinity as discussed in the text and shown in Figure 3). Each model corresponds to the head of an arrow in Figure 3. The dashed lines indicate the approximate CO2 content of Cenozoic and Late Mesozoic upper oceanic crust.

[24] The sensitivity of these results to the assumptions made in the modeling shown in Figure 4 has been investigated using a series of numerical experiments. These show that the gradient in carbonate ion concentration in the ocean assumed in this modeling is not critical to the results obtained. For example, using a constant carbonate ion concentration with depth or a surface carbonate ion concentration three times that of the deep ocean (for atmospheric CO2 = 500 ppm; ocean Ca = 10.2 mmol kg−1 and all other inputs as above), leads to negligible changes in CO2 uptake by the crust (in the second decimal place). This is despite these changes in carbonate ion gradient leading to large changes in deep water DIC and alkalinity (3.0 to 1.8 mmol kg−1 and 3.1 to 1.9 meq kg−1 at surface carbonate ion concentrations of 3× and 1× deep ocean concentrations, respectively). Again, the insensitivity of the amount of calcite formed in the crust to these changes is because the saturation state of the deep ocean water entering the crust is not changing. The only model runs in which the carbonate ion gradient was found to be important in controlling calcite precipitation in the crust were for very small differences between deep and shallow water carbonate ion concentration and low water-to-rock ratios (i.e., short durations of carbonate precipitation). The small difference between deep and shallow water carbonate ion concentration in these model runs leads to low surface water carbonate ion contents, especially for high ocean Ca contents and hence low seawater DIC. At small water-to-rock ratios and low DIC, large amounts of calcite cannot be precipitated within the crust because this requires too high a fraction of the fluid DIC to be converted to inline image. Numerical experiments also show that varying the difference in DIC between the surface and deep ocean from 0 to 0.6 mmol kg−1 makes no significant difference to the results (again, due to buffering of the deep ocean calcite saturation state).

[25] In summary, in an ocean in which deep-water carbonate sediments act as a buffer on the carbon system, realistic changes in seawater Ca, DIC, alkalinity, and atmospheric CO2 cannot directly drive substantial changes in carbonate formation within the upper oceanic crust. In this buffered situation, the main controlling factor is the compensation depth. That said, changing the compensation depth from 3 to 5 km (i.e., increasing the calcite saturation state in the deep ocean) only increases the amount of CO2 taken up by the crust by ~0.3 to 0.4 wt%. Since, the CCD in the Late Mesozoic was similar to, or shallower than, the CCD in the Cenozoic [Van Andel, 1975], changes in ocean chemistry do not appear to be likely causes of changes in CO2 uptake by the crust. Thus, an increase in alkalinity production in the upper crustal aquifer (Figures 2 and 4) is the most realistic explanation for the higher carbonate content of Late Mesozoic than Cenozoic oceanic crust. At earlier times in Earth history, before deep marine carbonates became common, the system could have behaved differently [Ridgwell, 2005].

4 What Causes Secular Variation in Alkalinity Production in the Oceanic Crust?

[26] Modeling the precipitation of calcite in the upper oceanic crust indicates that changes in alkalinity generation via fluid-rock reactions within the crust are required to explain the observed secular variation in the amount of carbonate within the oceanic crust [Gillis and Coogan, 2011]. Two models that could potentially explain this secular change in alkalinity generation, either on their own or in combination, are considered: (i) changes in bottom water temperature, and hence the kinetics and thermodynamics of fluid-rock reactions, and (ii) increased K uptake by the crust in secondary feldspar rather than clay minerals. These processes would act as negative feedbacks on atmospheric CO2 contents.

4.1 Change in Bottom Water Temperature

[27] Brady and Gislason [1997] suggest that changes in bottom water temperature could drive increased hydrothermal alteration of the oceanic crust. They show that plagioclase is increasingly undersaturated in seawater with increasing temperature. This means that for time periods in which bottom seawater was warmer, such as in the Late Mesozoic (Figure 1; Huber et al., [2002]), and hence crustal formation fluids were warmer, there was a greater thermodynamic drive to dissolve (igneous) plagioclase. Additionally, the rates of fluid-rock reaction increase with increasing temperature. These factors are both expected to lead to an increase in the extent of fluid-rock reaction during periods of high bottom water and hence global surface temperature. Deep seawater was ~10–15°C warmer in the Late Mesozoic than Cenozoic. Brady and Gislason [1997] suggest that a change in temperature of this magnitude would roughly double the rate of seafloor basalt dissolution potentially having a substantial impact on alkalinity production and hence carbonate formation within the crust.

4.2 Increased K Uptake in K-Feldspar

[28] Low-temperature alteration of the upper oceanic crust is known to act as a substantial sink for K added to the oceans by rivers and high-temperature hydrothermal fluids. Potassium is added to the oceanic crust largely in K-rich clay minerals, K-rich zeolites, and K-feldspar. Potassium-rich clays and zeolites are found in all ages of oceanic crust but K-feldspar is only a significant constituent of Late Mesozoic oceanic crust. Potassium-feldspar is a common alteration product, pseudomorphing igneous plagioclase, in samples from Late Mesozoic crust recovered from the: (i) ~110–155 Ma sites drilled on DSDP Leg 17 [Bass et al., 1973]; (ii) ~120 Ma DSDP Sites 417 and 418 [Alt and Honnorez, 1984]; (iii) ~91 Ma Troodos ophiolite [Gillis and Robinson, 1990]; and (iv) ~80 Ma DSDP Site 543A [Biju-Duval and Moore, 1984]. In contrast, K-feldspar is extremely rare or absent in Cenozoic crust sampled at the: (i) ~48 Ma Site 1224 [Paul et al., 2006]; (ii) ~15 Ma DSDP Hole 319 [Bass, 1976]; (iii) ~15 Ma IODP Hole 1256D [Alt et al., 2010]; (iv) ~13 Ma DSDP Hole 396B [Bohlke et al., 1980]; and (v) ~6 Ma ODP Hole 504B [Alt et al., 1986]. The higher K-feldspar content in Late Mesozoic upper oceanic crust is accompanied by a higher average K-content (Jarrard et al., [2003]; Table 1). Another difference in alteration mineralogy between Late Mesozoic and Cenozoic upper oceanic crust appears to be in the abundance of Al-rich clays (beidellite) which are more commonly reported for the older cores as well as in association with the small amounts of K-feldspar in the younger cores [Alt and Honnorez, 1984; Alt et al., 2010]. However, the paucity of data for these Al-rich clay minerals prevents any firm conclusion about this being drawn.

[29] The occurrence of significant amounts of K-feldspar in Late Mesozoic, but not Cenozoic, upper oceanic crust can be interpreted as an indication either that K-feldspar only forms in old crust or that alteration conditions have changed through time. Coogan et al. [2011] present preliminary radiometric ages for K-feldspar in DSDP Holes 417A and 543A that indicate formation within ~20 Myr of crustal accretion leading us to prefer the second model. Changing alteration conditions could include higher alteration temperature (section 4.1) and/or a change in seawater composition. The simplest change in fluid composition that could lead to the common replacement of plagioclase by K-feldspar would be an increase in the K content of the fluid leading to increased K-uptake by the crust. While this is not the only way to explain increased saturation state of K-feldspar, there is independent evidence that this may have been the case. Potassium is added to seawater from rivers and high-temperature hydrothermal circulation and removed largely through low temperature alteration of the oceanic crust and sediment diagenesis. Fluid inclusions in evaporites indicate that the K content of seawater has changed little over the Phanerozoic [Timofeeff et al., 2006; Horita et al., 2002]. In contrast, several lines of evidence suggest that the abundances of other major cations in seawater that have similar sources to K, such as Ca, have changed several folds. This suggests that there is an efficient buffer mechanism that leads to increased K delivery to the ocean being match by increased K sinks. Demicco et al. [2005] suggest that off-axis hydrothermal systems are the most likely sink for extra K added to the ocean during times of enhanced continental weathering and/or high-temperature hydrothermal circulation. Higher rates of oceanic crustal production have been inferred for the Late Mesozoic than Cenozoic [Seton et al., 2009] and the lower 87Sr/86Sr of seawater in the Late Mesozoic than Cenozoic is consistent with large hydrothermal inputs into the ocean at this time.

[30] We hypothesize that during the Late Mesozoic, an enhanced K flux into the ocean (from increased river and/or increased hydrothermal fluxes) was accompanied by enhanced K uptake by off-axis hydrothermal systems with K-feldspar formation being an important K-sink. In this scenario, alkalinity would be produced through the reaction of igneous plagioclase (here written as a 7:2 solid solution of anorthite and albite), aqueous K, SiO2, and H+ to form K-feldspar and beidellite and releasing Ca into the formation fluid:

display math(2)

[31] Alternatively, K-feldspar formation can be written in terms of the carbon system as consumption of CO2 and bicarbonate ions and precipitation of calcite:

display math(3)

[32] This type of reaction can explain the observed occurrence of Al-rich clay minerals associated with K-feldspar and calcite pseudomorphing plagioclase. These reactions contrast with those that form K-rich clays, which consume alkalinity.

[33] Addition of 2.5 wt% CO2 to the crust via these reactions requires the addition of ~0.75 wt% K2O and leaching of ~3.2 wt% CaO from silicate minerals; these are similar to the observed changes in Late Mesozoic crustal composition (Table 1). This comparison is not meant to suggest that these are the only reactions that impacted the Ca–K–C-alkalinity budgets of the Late Mesozoic upper crust—this is clearly not the case. Instead, this is simply noted as a demonstration that the order-of-magnitude effect of having K-feldspar serve as an important sink for K in Late Mesozoic oceanic crust is significant to the system.

5 The Role of Low-Temperature, Off-Axis, Hydrothermal Circulation in the Long-Term C-Cycle

[34] The modeling presented in section 3 indicates that formation of substantial amounts of carbonate within the oceanic crust is only possible if fluid-rock reactions generate alkalinity. Neither leaching of Ca from the rock (e.g., in exchange for Mg) nor increases in fluid temperature, without alkalinity generation, can lead to more than ~0.2 wt% CO2 uptake in calcite. The ~0.5 wt% CO2 in Cenozoic upper oceanic crust requires production of at least ~0.1 eq of alkalinity for each kilogram of rock reacted. In contrast, the ~2.5 wt% CO2 in Late Mesozoic upper oceanic crust requires production of at least ~0.5 eq of alkalinity for each kilogram of rock reacted. The higher alkalinity production due to fluid-rock reactions in the upper oceanic crust in the Late Mesozoic than Cenozoic means that hydrothermal alteration of the oceanic crust acted to remove C from the surficial environment into rocks more efficiently in the Late Mesozoic. If this resulted from higher bottom water temperatures, due to higher polar temperatures (Brady and Gislason, [1997]; section 4.1), alteration of the upper oceanic crust acted as an important direct negative feedback on atmospheric CO2 levels in the long-term climate-carbon cycle. Alternatively, if the higher alkalinity production in the Late Mesozoic was due to differences in ocean composition, such as higher K content, then this would also be an important, albeit less direct, negative feedback on atmospheric CO2 levels in the long-term carbon cycle.

[35] The approximate fivefold differences in CO2 content of, and alkalinity production by, Late Mesozoic and Cenozoic upper oceanic crust indicate large changes in the role played by the oceanic crust in the long-term carbon cycle. It is informative to compare this change in CO2 consumption by alteration of the oceanic crust to that required to buffer atmospheric CO2 levels. This is because it is changes in CO2 consumption, not total amount of CO2 consumed, that provide feedbacks on the C-system. Unfortunately this comparison is not straightforward because modern CO2 degassing rates (and hence, assuming a pre-Anthropocene steady-state system, consumption rates) are not well known. As a first-order approach, a range of plausible CO2 degassing rates from 4 and 10 × 1012 mol yr−1 for the modern system and a 1.5 to 2× higher degassing rate in the Late Mesozoic are taken from Berner [2004]. It should be noted that the higher modern estimate makes CO2 degassing at mid-ocean ridge (~1 × 1012 mol yr−1; Saal et al., [2002]), a small fraction of the global degassing flux. Cenozoic and Late Mesozoic CO2 consumption rates by the oceanic crust were ~0.5 × 1012 and 2.5 × 1012 mol yr−1, respectively; i.e., Late Mesozoic CO2 consumption rates by the oceanic crust were ~2 × 1012 mol yr−1 higher than in the Cenozoic (Gillis and Coogan, [2011]; section 3). Given these values and the assumption of steady-state, the increase in continental CO2 consumption (inorganic and organic) required to maintain a steady-state system can be readily determined from mass balance; i.e., whatever amount of the degassing is not taken up by oceanic crust alteration must be taken up by continental processes. For the lower estimate of modern CO2 degassing rates (4 × 1012 mol yr−1), Late Mesozoic continental CO2 consumption would have been 0 or 2 × 1012 mol yr−1 higher than Cenozoic rates for 1.5× and 2× higher degassing rates, respectively; i.e., the change in CO2 consumption by hydrothermal alteration of the oceanic crust between these time periods was as large or as larger than that due to continental processes in this scenario (Figure 5). For the upper end estimate of modern CO2 degassing rates (10 × 1012 mol yr−1), Late Mesozoic continental CO2 consumption would have been 3 or 8 × 1012 mol yr−1 higher than Cenozoic rates for 1.5× and 2× higher degassing rates, respectively. In this scenario, changes in consumption of CO2 by the oceanic crust were less important than changes in continental CO2 consumption but were still non-trivial (Figure 5).

Figure 5.

Histograms showing the proportion of the change in CO2 consumption between the Cenozoic and Late Mesozoic that is due to changes in continental processes (weathering and organic C burial) and seafloor hydrothermal processes. The calculations assume a steady state in which CO2 degassing and uptake match, and use the change in oceanic crust CO2 consumption from Gillis and Coogan [2011] and leave the remainder for continental process. The different bars reflect different estimates of modern CO2 degassing (=consumption) and different estimates of how much greater this was in the Late Mesozoic. The values in the boxes are fluxes in 1012 mol yr−1.

[36] The significant role played by hydrothermal alteration of the oceanic crust allowing CO2 consumption to change over time in response to variation in CO2 degassing (Figure 5) is not incorporated in most models of the long-term carbon cycle. These generally either exclude uptake of C by the oceanic crust or treat it as a sink that correlates in size linearly with the rate of oceanic crust production [e.g., Berner et al., 1983; Berner and Kothavala, 2001; Berner, 2004; Wallmann, 2001]. These approaches assume that the CO2 concentration in the upper oceanic crust is constant over time, something that is inconsistent with the data [Gillis and Coogan, 2011; Figure 1]. Francois and Walker [1992] suggested that hydrothermal alteration of the oceanic crust is important in the long-term carbon cycle; however, Caldeira [1995] showed that the ocean pH-based feedback on seafloor CO2 consumption rate proposed by Francois and Walker [1992] was untenable. Alternative feedback mechanisms, such as through bottom water temperature [Brady and Gislason, 1997; section 4.1] and changes in ocean cation (excluding Ca) abundances (section 4.2), provide plausible mechanisms for off-axis hydrothermal CO2 consumption to play an important role in the long-term C-cycle and thus long-term climate. Considering that on a global scale, most basalt erupts at mid-ocean ridges, and this basalt is in contact with warm circulating seawater for much of its life, it is perhaps not surprising that seafloor basalt alteration is important in the long-term carbon cycle.

6 Summary and Conclusions

[37] A simple model for the precipitation of calcite in the upper oceanic crust during low-temperature, off-axis, hydrothermal circulation based on a well constrained thermal model for these systems, and well-understood carbonate chemistry has been developed. The model was used to explore the controls on carbonate precipitation in the oceanic crust and, in particular, what factors could have led to the several fold higher amounts of carbonate precipitated in the crust in the Late Mesozoic than the Cenozoic [Gillis and Coogan, 2011]. Model results show that only ~0.2 wt% CO2 can be added to the crust if fluid-rock reactions do not generate alkalinity. The larger amounts (~0.5 wt%) of carbonate found in Cenozoic upper oceanic crust require ~0.1 eq of alkalinity to be produced per kilogram of rock. The much larger (~2.5 wt%) amounts of CO2 observed in Late Mesozoic than Cenozoic crust require that fluid-rock reactions within the crust generated more alkalinity (consumed more CO2) in the Late Mesozoic than in the Cenozoic. Differences in both the secondary mineral abundances and bulk composition of Late Mesozoic and Cenozoic crust are consistent with this model. Greater CO2 drawdown into the oceanic crust in the Late Mesozoic, when atmospheric CO2 levels and surface temperatures were higher, suggests that hydrothermal alteration of the oceanic crust acts as a negative feedback on atmospheric CO2 levels in a manner that is not included in current models of the long-term C-cycle. Comparison of the CO2 consumption fluxes into the oceanic crust with those required to maintain a steady-state atmospheric CO2 content indicates that low-temperature alteration of the oceanic crust provides a feedback of similar importance to continental weathering (Figure 5) in stabilizing atmospheric CO2 levels on long-timescales.

Acknowledgments

[38] Journal reviews by John Higgins and Sidonie Révillion along with the comments of editor Louis Derry, are gratefully acknowledged. Informal reviews by and/or discussion with Paul Hoffman, Colin Goldblatt, and, in particular, Jay Cullen are gratefully acknowledged; their reputations should not, however, be sullied by association to the ideas put forward.

Ancillary