Report

Remobilization of crustal carbon may dominate volcanic arc emissions

See allHide authors and affiliations

Science  21 Jul 2017:
Vol. 357, Issue 6348, pp. 290-294
DOI: 10.1126/science.aan5049

Volcanoes find a new carbon platform

The geological carbon cycle assumes that carbon is emitted by volcanic eruptions and removed through various forms of burial. Mason et al. found that not all volcanic eruptions have the same source for carbon in their volcanic gas. Arc volcanic activity appears to harvest carbon from old carbonate platforms, which results in a massive difference in the isotopic signature of the carbon emitted during eruption. This discovery requires revision of the global carbon cycle, decreasing the amount of organic carbon believed to be being buried.

Science, this issue p. 290

Abstract

The flux of carbon into and out of Earth’s surface environment has implications for Earth’s climate and habitability. We compiled a global data set for carbon and helium isotopes from volcanic arcs and demonstrated that the carbon isotope composition of mean global volcanic gas is considerably heavier, at –3.8 to –4.6 per mil (‰), than the canonical mid-ocean ridge basalt value of –6.0‰. The largest volcanic emitters outgas carbon with higher δ13C and are located in mature continental arcs that have accreted carbonate platforms, indicating that reworking of crustal limestone is an important source of volcanic carbon. The fractional burial of organic carbon is lower than traditionally determined from a global carbon isotope mass balance and may have varied over geological time, modulated by supercontinent formation and breakup.

The core, mantle, and crust contain 90% of the carbon on Earth (1), with the remaining 10% partitioned between the ocean, atmosphere, and biosphere. Due to the relatively short residence time of carbon in Earth’s surface reservoirs (~200,000 years), the ocean, atmosphere, and biosphere may be considered a single carbon reservoir on million-year time scales. Carbon is removed from the surface reservoir through formation and deposition of carbonate minerals and organic carbon and added through volcanic and tectonic carbon outgassing. In the preindustrial era, volcanic outgassing at arc, rift, and intraplate areas through vents or diffuse degassing sourced up to 90% of Earth’s surface carbon (2, 3). The remaining carbon came from tectonic regions, through metamorphic decarbonation of carbon-bearing rocks (4), and from the underlying mantle (5). Although we can account for all of the sources of surface carbon, the origin of carbon coming from volcanic arc outgassing is a fundamental yet unanswered question. The amount of carbon derived from the subducting slab compared with the overlying crust is poorly constrained but has implications for the amount of carbon returned to Earth’s deep interior (6) and alters interpretations of the variations in arc volcanic CO2 flux (7).

Determining the cycling of carbon between the surface reservoir and the mantle is important because imbalances greatly influence the amount of total carbon at Earth’s surface, which in turn affects atmospheric partial pressure of CO2 (Pco2) and surface temperature. During subduction, carbon-bearing sediments devolatilize (8) and carbon-bearing minerals may dissolve into fluids (6, 9). Carbon can be transported through primary melts to overlying volcanoes or reprecipitated in the mantle lithosphere beneath the arc (6). Receiving less consideration in the literature is the shallow crustal accretionary carbon cycle, which also can shuffle carbon between different reservoirs (10, 11) (Fig. 1). Mature continental crust contains three orders of magnitude more carbon than the oceans and atmosphere due to the accretion and assimilation of carbonate platforms from past oceans (11). Carbon-rich fluids derived from carbonates can be assimilated into crustal magmatic intrusions, resulting in high-carbon magmas that increase the carbon-outgassing flux from the volcano. In some arcs, this shallow “crustal” carbon cycle dominates over the deeper “subduction” carbon cycle, such that, to balance downgoing carbon with outgassing carbon, most of the carbon contained within sediments on downgoing slabs must be returned to the deep mantle. The return of slab carbon to the deep mantle is predicted by thermodynamic studies of metamorphic decarbonation (12); by observations of isotopically light carbon in diamonds, interpreted to be formed from the recycling of organic carbon associated with subduction (13); and by the relative stability of our atmosphere over geological time, which, owing to its small size relative to Earth’s interior, requires that inputs (via volcanoes) must approximately balance outputs (by subduction).

Fig. 1 Schematic diagram to show the possible sources of carbon in a subduction zone volcanic system and the processes that might fractionate carbon isotopes (22).

Carbon on the downgoing slab is contained within sediments as organic carbon and inorganic carbonate and as inorganic carbonate in the oceanic lithosphere (6). Carbon may be remobilized from the slab by metamorphic decarbonation (12) or by dissolution into ionic supercritical fluids (48) or may be returned to the deep mantle. On ascent through the crust, magmas may interact with crustal carbonate [incorporated into the crust by, e.g., accretion of limestone platforms (11)], assimilating CO2-rich fluids, which then outgas during ascent and eruption at the surface. Isotherms are from (49).

Global isotopic mass balance models help constrain Earth’s surface carbon cycle over geological time. These models often assume a constant carbon isotope composition reflecting the typical mantle range of around –6 per mil (‰) (1417). The burial of organic carbon removes carbon, along with electrons, from the surface reservoir, allowing progressive oxygenation of the planet (18). The fractional organic carbon burial flux, which is the amount of carbon buried as organic carbon, is determined by measuring the carbon isotope composition of carbonate minerals preserved as limestone over time. However, if the assumption of a constant δ13C of the input is incorrect, this has cascading effects on the calculated fraction of carbon buried as organic carbon and other carbon reservoir estimates over geologic time.

We compiled a global data set for carbon and helium isotopic composition of volcanic gases in arcs and evaluated whether assimilation of carbon from overlying crustal carbonates could dominate the global arc volcanic carbon flux. We identified arcs for which this mechanism may dominate the carbon flux. We found a large effect of crustal carbonates on the carbon budget, which requires reinterpretation of the global carbon isotope mass balance throughout Earth history. This has direct implications for the fractional burial of organic carbon (19, 20) through geological time.

We can use carbon isotopes, combined with other geochemical tracers (helium isotopes), to determine the origin of carbon in volcanic gases because distinct carbon isotopic compositions characterize carbon sources (21, 22) (Fig. 1). Delta notation (δ13C) reports carbon isotopes as the ratio of the heavy 13C isotope relative to the lighter 12C isotope relative to a standard in units of parts per thousand. CO2 released at arc volcanoes derives from one of three sources: (i) mantle (–6.0 ± 2.0‰) (2327), (ii) sedimentary organic carbon (<–20 to –40‰) and (iii) carbonate carbon (~0‰) (21). The 3He/4He ratio is a geochemical tracer of the relative contributions of magmatic and crustal components to volcanic gases. 3He is a conservative primordial stable isotope incorporated into Earth during initial accretion and subsequent accumulation of late veneer material, and the amount on Earth is not increasing. Crustal production of 4He through the decay of uranium and thorium will decrease 3He/4He, providing a strong indicator that the magma has interacted with fluids derived from silicate material in the overlying crust. The isotopic composition of helium is quoted as R/RA, which is the 3He/4He ratio in the sample (R), normalized to that in the atmosphere (RA = 1.38 × 10−6). The canonical mid-ocean ridge basalt (MORB)–He isotope range is 8 ± 1 RA (28), with the mean ratio for volcanic arcs 5.4 ± 1.9 RA (29). R/RA higher than the canonical MORB range is ascribed to an undegassed mantle reservoir and is typical of most ocean island basalts (e.g., Hawaii, Iceland, and Galapagos) (30). CO2/3He ratios are a sensitive tracer of carbon source in volcanoes when combined with the δ13C of the volcanic gas (21). Arcs typically have higher CO2/3He than MORB, due to the addition of either slab- or crustal-derived carbon (21), although it is difficult to distinguish between these sources based on 3He and δ13C alone (22). CO2/3He ratios for volcanic arc gases indicate the mixing of magmatic and sedimentary/crustal sources of CO2 and are up to 100 times as high as the mantle range (21, 31, 32). Here, we suggest that high CO2/3He and δ13C of volcanic gases (21), combined with a low R/RA, provides strong evidence for carbon being derived from the overlying crust as opposed to the downgoing slab.

Our compilation of carbon and helium isotope data (Fig. 2) (22) for volcanic arc gases allows us to discriminate carbon sources and, in doing so, determine the δ13C of modern volcanic arc gas. Our result challenges the fundamental assumption that δ13C of volcanic CO2 has been invariant over geological time. Carbon isotope ratios are available for most arcs worldwide (22) (table S2), but some arcs have little or no data. We did not find a systematic variation of δ13C or R/RA with sample temperature (22). Many arcs emit carbon with a higher δ13C than the typical mantle range (Fig. 2), most notably Italy (3335), the Central American Volcanic Arc, Indonesia (Sangihe and Java-Sunda-Banda), and Papua New Guinea (22). Arcs located in the northern Pacific, such as Japan and Kuril-Kamchatka, release CO2 with a δ13C that lies predominantly within the mantle range. The Cascade and Aleutian arcs outgas carbon that falls within or lower than the typical mantle δ13C range and He with R/RA values consistently within the mantle range. A recent study has linked the δ13C of volcanic gases emitted along the Aleutian arc with the organic sediment flux into the subduction zone (7). A cross-plot of carbon and helium isotope data (Fig. 3) identifies those arcs whose volcanic CO2 output is dominated or influenced strongly by crustal carbon assimilation: Indonesia—East Sunda/Banda; the Italian Campanian Magmatic Province (i.e., Vesuvio and Solfatara); and the Andes—Ecuador, Peru, and Northern Chile.

Fig. 2 Carbon and helium isotope compositions of volcanic and fumarolic gases measured at volcanic arcs globally (22).

Each arc is picked out by different color symbols, labeled with arc name and, in some cases, with country. The mean value is plotted as a darker symbol, and the standard deviation is lighter in shade on each side. The red-shaded columns show the canonical upper mantle values of δ13C (–6.0 ± 2.5‰) and helium isotope ratio 3He/4He normalized to the atmospheric value (R/RA) (8 ± 1). At the top are carbon and helium data for a number of nonarc volcanoes.

Fig. 3 Plot of helium isotope composition, 3He/4He, normalized with the same ratio for air (R/RA) against carbon isotopic composition, in δ13C, ‰, of volcanic arc gases, where both are measured in the same sample (22).

Mantle ranges are shown as orange panels. The data points are each from an individual arc volcano (22) and are colored by volcanic arc; colors as in Fig. 1. Assimilation of crustal material (containing radiogenic He from the decay of U and Th in crustal rocks) and crustal carbonate (supplying isotopically heavy carbon) would push values to the bottom right corner of the plot, shaded gray (i.e., high δ13C and low R/RA). Plots of CO2/3He are given in (22).

To estimate a global average δ13C for gases released at volcanic arcs, we must weight δ13C by CO2 flux. Estimating global volcanic CO2 fluxes is a nontrivial problem (36). We cannot measure CO2 fluxes directly due to the high concentration of CO2 in the background atmosphere and the absorption interference in the ultraviolet or infrared range. Estimates instead are usually derived by combining the flux measurement of SO2 with the ratio CO2/SO2 in volcanic gas emission compilations (7, 37, 38). Arcs currently represent between 33 and 63% of the total volcanic (arc + mid-ocean ridge + intraoceanic hot spot) outgassing flux (39). We do not show intracontinental hot spots—e.g., Yellowstone (40)—and rifts because of poorly constrained fluxes. Intracontinental rifting—e.g., East Africa—may generate a relatively large CO2 flux, although the absolute number depends on scaling up from a relatively narrow temporal window and a limited number of sampling locations (41). Diffuse degassing may equate to 50% of the passive volcanic CO2 flux from arcs (2), thereby enhancing the contribution of arcs to the global volcanic CO2 budget. We do not take account of tectonic fluxes of CO2, which are not well known. We used these fluxes, combined with median δ13C, to calculate the average δ13C of volcanic gases released today (table S4) (22). The nonweighted global arc mean δ13C is –4.3 ± 2.6‰ (with a median value of –3.8‰). After weighting the median arc values for CO2 flux at each arc, the global arc average δ13C is –2.8 to –3.3 ± 0.5‰ (with missing arcs assigned an average δ13C of –3.0‰ for the lower-end estimate and –5.0‰ for the higher end). These estimates yield mean global volcanic outgassing carbon compositions of –3.8 to –4.6‰, assuming that arcs represent 33 to 63% (39) of the total volcanic input to the atmosphere-ocean system (table S4) (22), with mid-ocean ridges accounting for much of the remainder. The volcanic arcs with the highest CO2 fluxes also have a higher δ13C, suggesting that assimilation and outgassing of crustal carbon may dominate global volcanic CO2 fluxes (Fig. 4).

Fig. 4 Plot of CO2 flux per kilometer of arc length against median carbon isotope composition δ13C for each arc, in ‰.

The horizontal bars represent the standard deviation around a median value. Estimated uncertainties on CO2 flux are not shown (22). The solid gray line is a linear regression, and the dashed gray lines are the 95% confidence limits.

The carbon isotope signature of volcanic gases reflects their source (organic carbon, limestone, crust, and mantle) (Fig. 1) but may also be modified by carbon isotope fractionation during magma degassing, assimilation of near-surface organic carbon, and precipitation of calcite in the subsurface geothermal or hydrothermal systems beneath volcanoes (22). The high δ13C of volcanic gases; the high proportions of radiogenic helium; and the mature, continental nature of the overlying crust (10, 11) all point to outgassed carbon being sourced dominantly from crustal limestones for a subset of arcs (Central America, Aegean, Papua New Guinea, Indonesia, and parts of the Andes). The limestones may be remnants of accreted carbonate platforms. Magma geochemistry provides additional evidence supporting substantial interaction of these magmas with the crust (22). Although we recognize that the carbon isotopic composition of volcanic gases may be influenced by a range of factors (22), we suggest that crustal carbonate assimilation is a key parameter controlling both the magnitude of the CO2 flux and its carbon isotope composition in the arcs that dominate global carbon outgassing. If a large fraction of the outgassed carbon is sourced from the overlying crust in some arcs, the implication is that a larger proportion of subducted carbon may return to the deep mantle, as predicted by models of metamorphic decarbonation (8, 12, 42).

The global carbon isotope mass balance describes the burial of organic carbon as a fraction of total carbon burial (f) through the balance of an exospheric carbon input (δ13Cin) and organic carbon (δ13Corg) and carbonate carbon (δ13Ccarb) outputs.

Embedded Image(1)

Ultimately, the goal is to solve this equation for forg, which allows us to resolve changes in organic carbon burial and thus the redox balance of Earth’s surface. In practice, only the δ13Ccarb is measured in limestone across a stratigraphic interval. The value for the organic carbon endmember δ13C composition is assumed to be 25 to 27‰ lower than the δ13Ccarb due to carbon isotope fractionation during photosynthesis. The volcanic input value is usually set at the assumed “bulk Earth” value, between –5 and –6‰; changing this value (δ13Cin) affects the calculated fractional organic carbon burial. For example, an increase of δ13Cin from ~–5‰ to ~–4.1‰ (table S4) (22) decreases the modern forg from ~20 to ~15% if volcanic arcs supply about half of the CO2 flux to the surface environment (fig. S4) (22).

We suggest that fractional organic carbon burial may be a smaller part of the total carbon removed from the surface of the planet today than previously assumed, based on our data compilation. A present-day forg lower than 20% (i.e., closer to 15%; see above) can occur by the incomplete oxidation of old carbon (43, 44) and a return of a substantial fraction of carbon to the ocean-atmosphere system as methane (45). This mechanism may explain the discrepancy noted between forg from isotope mass balance (0.19 to 0.34) and from inventory mass balance (how much sedimentary carbonate is present, 0.10 to 0.17) (44). Our forg (0.15) falls within the range quoted for the inventory mass balance and may be a more reasonable estimate of fractional organic carbon (forg) burial today.

It is likely that the δ13C of the volcanic input, and thus the overall calculation of forg , has varied over Earth history. Broadly, it has been proposed that transitions between continental-arc dominated and island-arc dominated states, related to the amalgamation and dispersal of continents through Earth history, have the potential to influence atmospheric Pco2 (46). During supercontinent breakup, continental arcs dominate over island arcs. Present-day subduction zones are dominated by island arcs compared with other periods in Earth’s past [e.g., during the Cretaceous, where continental arcs have been shown to be as much as 200% longer than today (10, 11)]. With continental arcs showing particularly high δ13C (due to contamination by crustal carbonates), the volcanic δ13C input during continental-arc–dominated periods, often associated with the closing of ocean basins and the breakup of supercontinents, has the potential to be substantially higher than today. We see a prime example of this in the Cretaceous, when overall high δ13C of marine carbonates has been linked to increased organic carbon burial, and by proxy increased atmospheric oxygen; this has been linked to the evolutionary radiation of mammals (47). A substantially higher δ13C of arc volcanoes associated with the closing of the Tethys Ocean at this time would require a reinterpretation of this record. Furthermore, over Earth history with the breakup of two supercontinents (Kenorland at 2.1 Ga and Rodinia at 800 Ma) there is sustained high δ13C measured in marine limestones; both of these high δ13C are associated with increases in atmospheric oxygen. It is unlikely that a high δ13C of volcanic CO2 alone can account for excursions to sustained δ13Ccarb as high as +10‰ because the δ13C input required to produce such values is unreasonable in the context of the highest δ13C of volcanic CO2 released today (22). However, at higher forg (>0.15), the δ13C of the volcanic input required for excursions to δ13Ccarb = +5‰ is more reasonable, at 0 ± 1‰. We suggest that some of these excursions in δ13C of carbonate minerals may result from increased δ13C of the outgassed carbon accompanying continental volcanism during supercontinent breakup due to crustal carbonate assimilation.

Supplementary Materials

www.sciencemag.org/content/357/6348/290/suppl/DC1

Materials and Methods

Figs. S1 to S4

Tables S1 to S5

References (50106)

References and Notes

  1. Materials and methods are available as supplementary materials online.
Acknowledgments: The data reported in this paper are available in the supplementary materials. This study was supported by the Alfred P. Sloan Foundation and the Deep Carbon Observatory. This work was supported by a European Research Council Starting Investigator Grant (307582) to A.V.T.
View Abstract

Subjects

Navigate This Article