Volume 119, Issue 22 p. 12,660-12,673
Research Article
Free Access

Volcanic sulfate aerosol formation in the troposphere

Erwan Martin

Corresponding Author

Erwan Martin

Sorbonne Universités, UPMC Université Paris 06, UMR 7193, ISTeP, Paris, France

CNRS, UMR 7193, ISTeP, Paris, France

Correspondence to: E. Martin,

[email protected]

Search for more papers by this author
Slimane Bekki

Slimane Bekki

Sorbonne Universités, UPMC Université Paris 06; Université Versailles St.-Quentin, CNRS/INSU, LATMOS-IPSL, Paris, France

Search for more papers by this author
Charlotte Ninin

Charlotte Ninin

Sorbonne Universités, UPMC Université Paris 06, UMR 7193, ISTeP, Paris, France

CNRS, UMR 7193, ISTeP, Paris, France

Search for more papers by this author
Ilya Bindeman

Ilya Bindeman

Department of Geological Sciences, University of Oregon, Eugene, Oregon, USA

Search for more papers by this author
First published: 16 October 2014
Citations: 17

This article was corrected on 6 FEB 2015. See the end of the full text for details.


The isotopic composition of volcanic sulfate provides insights into the atmospheric chemical processing of volcanic plumes. First, mass-independent isotopic anomalies quantified by Δ17O and to a lesser extent Δ33S and Δ36S in sulfate depend on the relative importance of different oxidation mechanisms that generate sulfate aerosols. Second, the isotopic composition of sulfate (δ34S and δ18O) could be an indicator of fractionation (distillation/condensation) processes occurring in volcanic plumes. Here we present analyses of O- and S isotopic compositions of volcanic sulfate absorbed on very fresh volcanic ash from nine moderate historical eruptions in the Northern Hemisphere. Most of our volcanic sulfate samples, which are thought to have been generated in the troposphere or in the tropopause region, do not exhibit any significant mass-independent fractionation (MIF) isotopic anomalies, apart from those from an eruption of a Mexican volcano. Coupled to simple chemistry model calculations representative of the background atmosphere, our data set suggests that although H2O2 (a MIF-carrying oxidant) is thought to be by far the most efficient sulfur oxidant in the background atmosphere, it is probably quickly consumed in large dense tropospheric volcanic plumes. We estimate that in the troposphere, at least, more than 90% of volcanic secondary sulfate is not generated by MIF processes. Volcanic S-bearing gases, mostly SO2, appear to be oxidized through channels that do not generate significant isotopically mass-independent sulfate, possibly via OH in the gas phase and/or transition metal ion catalysis in the aqueous phase. It is also likely that some of the sulfates sampled were not entirely produced by atmospheric oxidation processes but came out directly from volcanoes without any MIF anomalies.

Key Points

  • Topospheric volcanic sulfates are mainly isotopically mass dependent
  • In the troposphere, volcanic SO2 is oxidized by OH and/or TMI-catalyzed O2
  • The study of volcanic sulfate allow to “probe” the atmospheric chemistry

1 Introduction

Volcanic eruptions release large but variable amounts of gas (such as H2O, CO2, SO2, and halogens) into the atmosphere. Sulfur-bearing gases (mainly SO2, H2S, and COS) have a direct impact on the atmosphere and climate, as their oxidation in the atmosphere to SO3 and H2SO4 forms sulfate aerosols that disturb the radiative balance of the atmosphere [Forster and Ramaswamy, 2007; Rap et al., 2013]. Furthermore, sulfate particles also act as cloud condensation nuclei, affecting clouds and climate indirectly [Intergovernmental Panel on Climate Change, 2007; Schmidt et al., 2012a, 2012b]. Therefore, a detailed understanding of the sulfur cycle and the role played by volcanic eruptions is critical for accurate estimation of the impact of sulfur emissions on sulfate aerosol distribution and hence on climate [e.g., Robock, 2004; Schmidt et al., 2012a, 2012b]. Large pyroclastic eruptions having a volcanic explosivity index (VEI) ≥ 4 are the most studied in terms of climate impact because they release huge amounts of sulfur (e.g., 20 Mt for the Mount Pinatubo eruption in 1991) into the stratosphere, where they stay for about a year typically [e.g., Robock, 2000], instead of a week in the troposphere, where the hydrologic cycle washes them down. As a result, they tend to be dispersed on a hemispheric scale or even globally if injected into the tropics. It has also been shown that moderate volcanic eruptions (VEI < 4) can also have an impact on the stratospheric aerosol background [Vernier et al., 2011]. When the eruptive column does not reach the stratosphere or even the tropopause region, volcanic eruptions still affect the atmospheric radiation balance but on more regional scales (e.g., the Laki lava eruption in 1783–1784). Moderate eruptions are by far the most frequent, and this makes them one of the greatest natural sources of atmospheric S-bearing gas [Bates et al., 1992; Andres and Kasgnoc, 1998]. On average, tropospheric eruptions (i.e., the injection of volcanic material mainly into the troposphere) release about 10–20 Mt of SO2 per year, which is 2 orders of magnitude higher than stratospheric eruptions that result in the release of very large amounts of SO2 into the stratosphere but are much less frequent (<0.2 Mt of SO2 per year) [e.g., Bluth et al., 1992; Andres and Kasgnoc, 1998; Wallace, 2001].

To characterize the atmospheric cycle of sulfur, one needs to determine its burden and the different components of its budget: fluxes (that is, source and sink rates). Sources/sinks are associated with processes such as emissions, chemical reactions (resulting in the production or destruction of the compound), transport, and deposition. Conventional methods for inferring chemical reaction terms rely mostly on models that are evaluated and constrained with atmospheric concentration measurements, because there is no direct means of measuring chemical fluxes associated with specific oxidation channels [e.g., Morin et al., 2007].

The most abundant volcanic sulfur gas is generally SO2. Once emitted into the atmosphere, SO2 can be removed via deposition at the surface level or undergo oxidization through several dominant channels according to the following reactions [e.g., Seinfeld and Pandis, 2006]:
(*: mass-independent O isotopes signature in the troposphere, expressed as Δ17O > 0; TMI: transition metal ion; for Ch4, S(IV) is in aqueous phase and can be present as SO2 · H2O, HSO3, or SO32− and S(VI) corresponds to sulfate SO42−)

The main SO2 oxidants are O3, OH, H2O2, and O2 catalyzed by TMI (TMI: transition metal ion; Figure 1) [e.g., Seinfeld and Pandis, 2006, and references therein]. The oxidation by OH Ch1 takes place in the gas phase only. Sulfuric acid produced via Ch1 can either condense on preexisting particles or form new acid sulfuric particles by homogeneous nucleation. Its oxidation by O3, H2O2, and O2 (Ch2, Ch3, and Ch4, respectively) is heterogeneous, resulting in sulfate production in liquid phases such as cloud droplets. Ultimately, the fate of this sulfate is determined by the fate of the liquid phases themselves, for instance, precipitation or evaporation. It is also worth pointing out that studying the fate of volcanic sulfur can also help to better constrain the atmospheric consequences of anthropogenic SO2 emissions that are 5–10 times higher than volcanic emissions on average (110 Mt per year) [e.g., Smith et al., 2011]. All of these four dominant oxidation channels occur in a background atmosphere and lead to the production of secondary sulfate production. On top of the sulfate generated via the oxidation of sulfur-bearing gases (mostly SO2), there is also some sulfate emitted directly by volcanoes [e.g., Mather et al., 2003, 2006]. These sulfates are called “primary” in opposition to sulfates chemically produced in the atmosphere which are called “secondary.” The sulfate samples analyzed in this study are probably a mix of these two sulfate sources.

Details are in the caption following the image
Simplified diagram showing the 4 dominant oxidation channels through which volcanic SO2 is converted to sulfate aerosols in the atmosphere. Although chemical fluxes 1, 2, 3, and 4 cannot be measured directly, oxygen isotopic composition of sulfate provides constraints in their estimations according to the following mass balance (the same equation is used for Δ17O): urn:x-wiley:2169897X:media:jgrd51791:jgrd51791-math-0005 Percentages correspond to the proportion in sulfate of oxygen originating from the different compounds during each oxidation chain (mainly from Savarino et al. [2000], more details are in the main text). The most likely ranges for δ18O and Δ17O of each compound are provided in Table 2. The resulting O isotope compositions of sulfate aerosols are the following: Channel 1: δ18O = −5 to 1‰ and Δ17O = 0‰ / Channel 2: δ18O = 13 to 25‰ and Δ17O = 1‰ / Channel 3: δ18O = 14 to 27‰ and Δ17O = 8‰ / Channel 4: δ18O = 4 to 15‰ and Δ17O = −0.17 to −0.08‰.

Other oxidation channels have been proposed, some occurring in the vent of the volcanoes. For example, a sulfate such as anhydrite can be formed in the magma itself, in the magma chamber, or possibly during the magma ascent [e.g., Rye et al., 1984]. S-bearing gases can be oxidized very quickly after being exsolved from the magma at high temperatures by the fragmented magma or possibly by magmatic and/or atmospheric oxidant compounds [e.g., Mather et al., 2006]. As these sulfates are produced within volcanic vents, they are seen as primary sulfates. They can also be released during future eruptions by remobilizing the hydrothermal deposits from the conduit's walls [e.g., Rye et al., 1984]. Further on in the volcanic plume, S-bearing gases can be converted to sulfate by chemical and photochemical reactions on mineral and dust surfaces [e.g., Cwiertny et al., 2008]. Among others, processes such as SO2 gas-tephra reactions at relatively high temperatures by Ca2+ diffusion-driven mechanisms [Ayris et al., 2013] or SO2 oxidation by OH produced by UV irradiation of dust particles [Dupart et al., 2012] could be responsible for sulfate formation in dense volcanic plume conditions. Oxidation by halogen species in the aqueous phase has also been invoked in modeling studies [von Glasow et al., 2002]. However, some of the rate constants are still unknown [e.g., Alexander et al., 2012]. Overall, several of the processes are still very poorly characterized or even speculative.

Isotopic ratio measurements can provide insights into the nature and magnitude of fluxes associated with oxidation processes, thus providing unique information regarding processes that are often difficult to quantify from concentration measurements alone. Stable isotope variations in atmospheric compounds can often act as source markers for compounds and provide both qualitative and quantitative constraints on the chemical and physical pathways that determine their fate [Johnson et al., 2002; Brenninkmeijer et al., 2003; Thiemens, 2006; Morin et al., 2008]. Stable isotope analysis has been particularly useful in constraining global trace gas budgets [Quay et al., 1999; Yoshida and Toyoda, 2000; Brenninkmeijer et al., 2003; Harris et al., 2012, 2013]. Kinetic isotope effects associated with sinks can also be used to identify and evaluate chemical reactions occurring in the atmosphere.

Most processes lead to isotopic fractionation according to mass differences in the isotopes. However, some processes, called mass-independent fractionation (MIF), do not follow this general rule of fractionation and generate isotope ratios that do not scale with isotope masses [e.g., Thiemens, 1999]. The most common example is the large and distinctive oxygen-MIF anomaly in the atmospheric ozone that is produced during its formation [Mauersberger et al., 2003; Vicars et al., 2012]. For oxygen, the MIF is quantified from the triple-oxygen isotope ratios with the quantity Δ17O that reflects the difference between measured δ17O and δ18O and those expected from mass-dependent fractionating processes: Δ17O = δ17O − 1000[(1 + δ18O/1000)0.52 − 1]. The O-MIF signature in the ozone is propagated to other oxygen-bearing compounds, in particular oxidants and oxidation products, through chemical reactions [e.g., Lyons, 2001] and can be used to study the atmospheric chemistry.

Emitted volcanic S-bearing gases and atmospheric oxidants have different oxygen isotopic composition. Therefore, the oxygen isotopic composition of sulfate, the end product of sulfur-bearing compounds oxidation, should reflect not only emission sources but also the relative contributions of the different oxidation channels with their respective fractionations, both mass-dependent and mass-independent (Figure 1).

Measurements of sulfate isotopic composition have already been used successfully in identifying key sources and sinks in the cycle of sulfur in the atmosphere [Alexander et al., 2002, 2005; McCabe et al., 2006; Hill-Falkenthal et al., 2012; Harris et al., 2012, 2013] including volcanic sulfur. For instance, oxygen and sulfur mass-independent signatures have been measured in sulfate from large stratospheric volcanic eruptions in ice cores [e.g., Savarino et al., 2003a, 2003b; Baroni et al., 2007, 2008] or volcanic deposits [e.g., Bao and Reheis, 2003; Bindeman et al., 2007; Martin and Bindeman, 2009]. S-MIFs measured in these volcanic sulfates confirm the existence of MIF channels in stratospheric sulfur oxidation, possibly photooxidation of SO2 while O-MIF measurements tend to indicate that OH radicals contribute very substantially to stratospheric sulfur oxidation. During volcanic eruptions, especially tropospheric eruptions, a large part of the generated sulfate is integrated into the volcanic ash during or after its deposition by different mechanisms (uptake on solid surfaces, condensation, dissolution in liquid layers around particles, etc.). The isotopic composition of this volcanic sulfate should provide constraints on the relative importance of different tropospheric oxidants assuming that the isotopic composition of the volcanic sulfur injected is known.

This paper presents new measurements of the isotopic composition of sulfate found in volcanic ash deposits from worldwide eruptions that are mainly tropospheric. Measured isotopic compositions are then compared to sulfate isotopic compositions derived from oxidant fields provided by a global 3-D atmospheric chemistry model. The consistency between measurements and model calculations is discussed along with the limitations of such a comparison. These comparisons allow us to attempt to test our understanding of volcanic sulfur oxidation in the atmosphere. Previously, high mass-independent signatures in oxygen isotopes have been measured in volcanic sulfate from large eruptions, stimulating the study of such volcanic events on a global scale. Here the focus is on volcanic sulfate produced in the troposphere or the tropopause region. One of the objectives is to explore to what extent isotopic measurements can constrain the fate of the volcanic sulfur in the troposphere.

2 Samples and Methods

2.1 Samples

Ash samples were supplied by several researchers (Table 1) who, along with their colleagues, collected freshly deposited ash during a few days or less after the eruptions, at different distances from the volcanic vents, from 5 km to 400 km and without being washed out by rain (Figure 2). These conditions are optimum for sulfate conservation in the volcanic product. The ash samples considered here were collected during tropospheric eruptions of the Popocatépetl (2008, Mexico), Gjálp (1998, Iceland), Mayon (1968, Philippines), Parícutin (1948, Mexico), Negro Cerro (1947, Nicaragua), and to some degree stratospheric eruptions of Mount Spurr (1992, Alaska), Mount St. Helens (1980, USA), Volcán de Fuego (1974, Guatemala), and Tungurahua (2008, Ecuador); Figure 2. For the latest eruptions, volcanic plumes could have reached almost 20 km (Table 1) but the majority of particles (such as ash and sulfate aerosols) traveled hundreds of kilometers in the upper troposphere or the tropopause region.

Table 1. Oxygen and Sulfur Isotopic Composition of Volcanic Sulfate Extracted From Ash Collected Soon After Tropospheric Eruptions
Volcano and Date of Eruption Sample Namea Distance From Vent (km) Maximum Plume's Height (km) Latitude (°N) Longitude (°W) SO42− (ppm) Δ17O (‰) δ18O (‰) δ34S (‰)
Popocatépetl (15 February 2008) Popo 15/2/2008 b 25 8.4 19.0 98.6 7517 0.35 9.83 9.1
Spurr (18 August 1992) Spurr 42 c 360 16 61.3 152.3 4084 −0.09 2.92 9.4
Spurr (18 August 1992) Spurr 57 c 265 16 61.3 152.3 3914 −0.14 4.26 9.2
Volcán de Fuego (14 October 1974) VF 74–200 c 78 15 14.5 90.9 3015 −0.03 7.32 6.8
Volcán de Fuego (14 October 1974) VF 74–45 c 57 15 14.5 90.9 1215 −0.04 9.15 10.5
Negro Cerro (9 July 1947) 116286-10 d 44 8 17.1 93.0 574 −0.03 7.18 6.6
Negro Cerro (9 July 1947) 116286-11B d 45 8 17.1 93.0 2645 −0.08 7.91 7.4
Negro Cerro (9 July 1947) 116286-9 d 12 8 17.1 93.0 665 −0.06 7.01 7.7
Mayon (1–9 May 1968) 111170 d 10 13.3 236.31 410 −0.04 7.63 7.5
Parícutin (June 1948) 16300-42 d 5 8 19.5 102.3 442 0.13 8.12 6.8
Mt St Helens (18 May 1980) 115315 d 400 19 46.2 122.2 1062 0.02 4.67 9.0
Mt St Helens (18 May 1980) 14 Davis e 400 19 46.2 122.2 - −0.04 0.21 7.6
Tungurahua (February 2008) Tung 04 f <30 14 1.5 78.4 318 −0.02 5.15 3.9
Tungurahua (February 2008) Tung 27 f <30 14 1.5 78.4 351 −0.06 6.61 4.1
Gjálp (December 1998) Gjálp g <30 10 64.4 16.8 - −0.07 13.24 -
  • a The samples from the same eruption have been collected at different location.
  • b Roberge J. (IPN, Mexico).
  • c Rose B. (Michigan Tech, USA).
  • d The rock collection at the Smithsonian Institution of Washington DC.
  • e Sarna-Wojcicki A.M. (USGS).
  • f Hidalgo S. (IGEPN, Ecuador).
  • g Sigmarsson O. (LMV, France/University of Iceland).
Details are in the caption following the image
Location of the studied historical tropospheric eruptions (red dots) and the O isotope composition of annual worldwide precipitation (http://www.waterisotopes.org). The water isotopic composition in the atmosphere close to each volcano is between −5 and −15‰ as reported in Table 2.
Table 2. (Top) Most Likely Isotopic Composition of Volcanic SO2 and Its Oxidant in the Troposphere [Krankowsky et al., 1995; Johnston and Thiemens, 1997; Savarino and Thiemens, 1999; Mather et al., 2006] (www.waterIsotope.org; Figure 2)a
Min Max
Δ17O δ18O δ18O
Volcanic Gas
SO2 0 5 7
Tropospheric Oxidants
OH 0 −15 −5
H2O2 2 45 25
H2O 0 −15 −5
O3 32 60 100
O2-TMI −0.33 23 23
Secondary Volcanic Sulfate Composition
Ch1 (OH) 0 −5 1
Ch2 (H2O2) 1 25 13
Ch3 (O3) 8 14 27
Ch4 (O2-TMI) −0.17 to −0.08 4 15
  • a The Δ17O value for O2 is the one established during our measurements at the University of Oregon and is consistent with what Young et al. [2014] measured. (bottom) Expected isotopic composition of sulfate produced by the different oxidation channels (see the text for further explanations).

2.2 Sulfate Extraction, Oxygen, and Sulfur Isotopes Measurements

The nature of sulfates in our samples has not been characterized. Sulfates might consist of various alkali-bearing sulfates mainly precipitated from aerosol liquid droplets (B. Grobéty, personal communication, 2014). All soluble sulfates were extracted from each sample by leaching in diluted HCl (<0.1 M) solution. All sulfates in solution were integrally precipitated into barite and purified by double dissolution-reprecipitation in a chelating diethylene triamine penta acetic acid (DTPA) agent [Bao, 2006; Martin and Bindeman, 2009]. For practical reasons, more than 95% of the precipitated barite was collected, leading to a slight underestimation of the SO42− concentration in the ash samples. The O2 was extracted and purified from 4 to 6 mg of purified barite on a laser fluorination line, and its isotopic composition was measured on a MAT 253 at the Stable Isotope Laboratory at the University of Oregon. Because BaSO4 fluorination has a 35–45% yield, we used NBS127 and IBO4-13 standards [Bindeman et al., 2007] to correct the measured raw δ18O values (8.6‰ correction was required). This correction has no effect on the Δ17O value as fractionation processes during measurements are mass dependent. The reader is referred to Martin and Bindeman [2009] for further details on analytical method. Based on duplicates, analytical error (in 2σ) on oxygen isotopes is 0.05‰, 0.16‰, and 0.3‰ for Δ17O, δ17O, and δ18O, respectively. The error (in 2σ) of this whole method (sulfate extraction + analysis) is about 0.1‰, 0.5‰, and 1.0‰ for Δ17O, δ17O, and δ18O, respectively. Note that in this study all duplicates have been analyzed in only five very reproducible sessions in a short period of time (2 weeks).

Sulfur isotopes were analyzed via an elemental analyzer at the New Mexico Tech Stable Isotope Lab. Multiple analyses of international standard NBS127 give δ34S = 15.94‰ ± 0.09 (in 2σ, n = 4), δ34S = 12.28‰ ± 0.02 (in 2σ, n = 2) for NBS123, and δ34S = −8.98‰ ± 0.06 (in 2σ, n = 2) for an in-house standard Cpy-4.6. Considering the certified values of 21‰, 17‰, and −4.6‰ for NBS127, NBS123, and Cyp-4.6, respectively, the samples measured during the analytical session were corrected using a factor that is between 4.3 and 4.9‰.

2.3 Modeling

As illustrated in Figure 1, the O isotopic composition of the collected sulfate depends on the relative contributions of different oxidants to sulfur oxidation as well as the isotopic composition of the SO2 source. In order to see whether our measurements of sulfate O isotopic composition are consistent with expected levels of oxidant, we extracted concentration fields of oxidants from a present-day multiyear global 3-D chemistry transport model simulation. The model is the Model for Ozone and Related Chemical Tracers (MOZART) [Horowitz et al., 2003; Emmons et al., 2010a], run at 1.9° latitude and 2.5° longitude and forced with NASA GEOS 5 analyses from 1 January 2007 to 31 December 2011. The simulation uses anthropogenic emissions based on David Streets' inventory for Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (http://www.cgrer.uiowa.edu/arctas/emission.html) and fire emissions from FINN-v1 [Wiedinmyer et al., 2011]. MOZART model calculations have been extensively evaluated against a range of atmospheric chemical measurements [Horowitz et al., 2003; Pfister et al., 2006, 2008; Emmons et al., 2010b]. Like almost all global models, MOZART does not account for the oxidation of sulfur catalyzed by TMI in condensed phases. Therefore, TMI concentrations have to be specified. It has to be stressed that our model calculations are only valid for very diluted and mixed volcanic plumes. Some of the sulfates are produced in dense volcanic plumes where complex multiphase processes are important. This type of model calculation is difficult to constrain because some kinetic data and rate constants are poorly characterized or even missing.

Vertical profiles of oxidant concentrations (from the surface to 10 km) at the location (latitude, longitude) of each volcanic eruption were extracted by simple linear interpolation of the global 3-D MOZART fields. We only consider the 3 days of profiles following the date of each eruption. As the considered volcanic eruptions mostly occurred prior to the period covered by the simulation (from 1 January 2007 to 31 December 2011), model-calculated oxidant vertical profiles are averaged temporally over the 5 years of simulation in order to produce climatological atmospheric oxidant levels corresponding to the present-day conditions over each volcano. For the aqueous oxidation of SO2 by O3 and H2O2, aqueous concentrations of SO2, O3, and H2O2 are calculated from Henry's law constants exploring a range of pH values (from 1 to 9) and liquid water content (from 0.005 to 3 g of water · m−3 of air) [Seinfeld and Pandis, 2006]. The calculation of the SO2 aqueous concentrations also takes into account the fact that some of the dissolved SO2 dissociates twice to form HSO3 and SO32−. This dissociation is a reversible reaction that reaches equilibrium rapidly, resulting in an effective solubility of SO2 in aqueous phases strongly dependent on the pH [Seinfeld and Pandis, 2006]. The rate constant for the gas phase reaction between SO2 and OH, the rate constants for the aqueous reactions of SO2 with O3 and H2O2, the Henry's law coefficients for SO2 by O3 and H2O2, and the equilibrium constants between SO2 and HSO3 and between HSO3 and SO32− are taken from recommended chemical kinetic and photochemical data [Sander et al., 2006]. The TMI concentrations in the liquid phase that are specified in the model are based on recent field measurements in bulk cloud water where TMIs are derived from mineral dust [Harris et al., 2013]. We focus on Fe3+ and Mn2+, the most widely studied system. Both ions oxidize sulfur, S(IV), but they also have a very strong positive synergistic effect on the oxidation rate [Seinfeld and Pandis, 2006]. As done in Alexander et al. [2009], reaction rates for sulfur oxidation by Fe3+ and Mn2+ are taken from Martin and Good [1991].

Finally, the overall oxidation rate of volcanic SO2 for each oxidant is calculated by vertically integrating the profiles of instantaneous oxidation rates assuming that volcanic SO2 plumes are distributed homogeneously from the surface to 10 km. We tested spreading volcanic plumes over different altitude ranges, a couple of kilometers from the surface. However, in terms of the relative contributions of the different oxidation channels (the determining factor for the oxygen isotopic composition of the sulfate), the results remain unchanged.

There are important caveats that have to be kept in mind when analyzing the results of the comparisons. First, as the atmospheric composition has changed substantially over the last century, the comparison is not entirely fair for some isotopic measurements that correspond to eruptions occurring half a century earlier (for instance, Negro Cerro in 1947). Second, the comparison assumes that oxidant levels are not affected by the eruptions themselves; this is only valid for very diluted volcanic plumes. As a result, model-calculated profiles can only be viewed as rough indicators of oxidant levels prevailing in the background atmosphere during the volcanic eruptions.

3 Results

3.1 Oxygen and Sulfur Isotopes

Overall, the SO42− concentration in ash samples varies from 350 to 7500 ppm (Table 1 and Figures 3c–3e). Note that these units of mass mixing ratio (ppm) refer to the mass of extracted BaSO4 compared to the mass of ash leached. The δ18O of the extracted sulfate covers a wide range from 0.2‰ to 13.2‰ as well as the δ34S that varies from 3.9 to 10.5‰. However, the Δ17O ranges from −0.14‰ to 0.35‰ with 80% of the samples between −0.1 and 0.1‰, which is relatively homogeneous. In the Δ17O versus δ18O diagram (Figure 3a), with the exception of Mexican samples, the data set describes a horizontal trend that indicates no correlation between Δ17O and δ18O. The same observation holds between δ34S and δ18O as the data set looks scattered in the δ34S versus δ18O diagram (Figure 3b). In detail, as shown in Figure 3, samples from the same eruption but collected at different locations have a significantly similar composition. Only the Mount St. Helens samples show a significant difference in δ18O but it is still in the same range, among the lowest values. However, differences in the sulfate concentration (up to 2100 ppm) can possibly be observed in samples from the same eruption. It is noteworthy that in our data set there is no apparent correlation between the sulfate concentration and the isotopic composition (Figure 3). This confirms that at least in our samples, processes affecting the sulfate concentration (e.g., volcanic SO2 emission and natural ash leaching by rain before sampling) are clearly distinct from and do not interfere with processes determining the isotopic composition (e.g., oxidant composition and fractionation processes). Finally, an apparent correlation between δ18O and the distance from the vents is observed (Figure 3f). The same correlation is observed between δ18O and the maximum height of the volcanic plumes, but it might not be relevant as the larger the eruption is, the farther from the vent volcanic material can be deposited and hence collected.

Details are in the caption following the image
Oxygen and sulfur isotope composition and sulfate concentration of the samples. The grey area is the initial magmatic SO2 composition which is very close to the magmatic value for O isotopes [e.g., Eiler, 2001] and potentially up to 2‰ higher than mantle value for S isotopes [e.g., Maarten de Moor et al., 2010]. However, the S isotope composition of magmatic SO2 is highly variable because it strongly depends on (a) the mantle source composition that can differ from the mantle value (δ34S = −0.91 ± 0.50‰) [Labidi et al., 2012, 2013] by interactions with subduction fluids, (b) the magma degassing processes, and (c) the oxidation state of magmatic sulfur [e.g., Rye et al., 1984; Alt et al., 1993; Mather et al., 2006; Maarten de Moor et al., 2010].

The main result of this geochemical study is the compositional difference between the initial volcanic SO2 and the sulfates extracted from volcanic ash (Figure 3). Indeed, differences of up to 6‰ in δ18O and up to 0.25‰ in Δ17O are observed, which is significantly higher than the analytical uncertainties (error bars in Figure 3). However, the variations measured in sulfate δ34S are comparable to what can be expected for the volcanic SO2 worldwide (Figure 3).

3.2 Atmospheric Modeling

Model-calculated respective percentage contributions of O3, H2O2 OH, and O2-TMI to the total volcanic SO2 oxidation (expressed as percent of the total) are presented in Figure 4 as a function of the pH of the liquid phase. The colored areas in Figure 4 represent the range of values obtained for all the different volcanic eruptions for varying liquid water content (from 0.005 to 3 g of water · m−3 of air). Our model results are consistent with previous modeling results [Seinfeld and Pandis, 2006; Harris et al., 2012, 2013]. As expected, at pH <6, H2O2 is by far the most effective oxidant. At higher pH (> 6–7), O3 becomes the most important oxidant. However, because of the SO2 oxidation, volcanic liquid phases are expected to be rather acidic. Therefore, according to the equilibrium calculations, H2O2 should be the overwhelmingly dominant oxidant for volcanic SO2. In addition, as long as there are significant amounts of liquid phases in the volcanic plumes (amounts equivalent to cloudy conditions in the atmosphere), the oxidation of SO2 in the gas phase (by OH) is almost negligible. The results vary little with the location and time of the year of the volcanic eruptions considered here.

Details are in the caption following the image
Model-calculated contributions (in %) of OH, O3, H2O2, and O2-TMI to SO2 oxidation as a function of pH of the liquid phases (note the logarithmic scale). The colored areas represent the range of values obtained for all the different volcanic eruptions for varying liquid water content (from 0.005 to 3 g of water · m−3 of air). Gas phase concentrations of OH, O3, and H2O2 are taken from 3-D chemistry transport model calculations (see text for more details). The only TMIs considered here are Fe3+ and Mn2+. Based on recent field measurements in bulk cloud water, [Fe3+] = 1 μM and [Mn2+] = 0.1 μM.

4 Discussion

4.1 Oxygen and Sulfur Isotopes

The range of oxygen isotope compositions of the initial volcanic SO2 is relatively small and strongly depends on the magmatic isotope composition which commonly varies between 5 and 7‰ [e.g., Eiler, 2001] (Figure 3). Sulfur isotope compositions of magmatic SO2 show a much wider range because of its strong dependence on the mantle source composition that can differ from the “mantle value” [Labidi et al., 2012, 2013] by interactions with high δ34S subduction fluids but also by magma degassing processes and by different oxidation states of magmatic sulfur (δ34SSO2 ≈ 0–9‰; Figure 3) [e.g., Rye et al., 1984; Alt et al., 1993; Mather et al., 2006; Maarten de Moor et al., 2010].

After the oxidation and transport of volcanic SO2 within the atmosphere, the resulting products (sulfate aerosols) are expected to differ in terms of δ34S and δ18O from the initial magmatic SO2. Our data set clearly confirms this for oxygen isotopes (Figure 3). However, it is less clear for sulfur isotopes because the variations in isotopic composition among initial SO2 are as large as the variations from the resulting sulfates (Figure 3). In addition, all the samples whose S isotopic compositions have been determined are from subduction zones where the mantle source can be more or less metasomatized by relatively high δ34S fluids.

The high δ34S variability of initial magmatic SO2 more or less covers the composition range observed in sulfate compositions. Therefore, it is difficult to quantitatively discuss processes that can generate this type of isotopic fractionation during SO2 oxidation. Indeed, isotopic fractionation can occur in the volcanic plume itself by distillation/condensation processes as described by the Rayleigh distillation equation [e.g., Castleman et al., 1974] and also during homogeneous and heterogeneous oxidation reactions [Harris et al., 2012]. The only way to discuss such processes in detail would be to know exactly what the initial magmatic SO2 isotopic composition is for each of the studied eruptions, which is not the case here. It is worth pointing out that even for a specific volcano, the sulfates extracted from ash deposits have isotopic compositions (δ18O and δ34S) that can vary from one eruption to another as observed, for instance, at the Popocatépetl [Armienta et al., 2010].

The 13‰ variation observed in δ18O can also be at least partially explained by processes such as distillation/condensation processes in the volcanic plume as described by the Rayleigh distillation equation. However, the contributions of oxidants with contrasted δ18O compositions (Table 2) can significantly affect the sulfate composition too. Through the oxidation channel Ch1, 50% of the sulfate oxygen atoms come from SO2, 25% from OH, and 25% from H2O [Savarino et al., 2000] (Figure 1), leading to sulfate having δ18O between −5 and 1‰. Considering now the oxidation channel Ch2, Savarino et al. [2000] experimentally demonstrated that 50% of the sulfate oxygen comes from H2O2, but the other 50% can be derived from SO2 only or 25% SO2 and 25% H2O (Figure 1). In both cases, the resulting δ18O of the sulfate is similar, between 25 and 13‰ (Table 2). Considering now the SO2 oxidation via O3 Ch3, 50%, 25%, and 25% of the sulfate oxygen comes from SO2, O3, and H2O respectively [Savarino et al., 2000] (Figure 1), leading to sulfate δ18O of about 14–27‰ (Table 2). Finally, the O2-TMI channel Ch4, is expected to generate sulfate having δ18O between 4 and 15‰ (Table 2) as 25–50% of the sulfate's oxygen atoms are thought to come from the initial SO2, 0–25% from H2O, and 25–50% from O2 (Figure 1).

As a consequence, the O isotope compositions of secondary sulfate are the result of the mixing between the isotopic signatures of the initial volcanic SO2 and the atmospheric oxidants. Therefore, if we consider a low primary sulfate contribution, the lowest δ18O measured in our volcanic sulfate could be explained mainly by OH oxidation Ch1 and most of our samples δ18O could be derived from the O2-TMI channel Ch4. Furthermore, there are isotopic fractionation processes (e.g., distillation/condensation) operating in the initial volcanic plume and during the atmospheric transport of S-bearing volcanic gas and aerosols that can contribute to δ18O variations in the analyzed volcanic sulfate. Finally, one should keep in mind that the contribution of potential primary sulfate as well as their isotopic composition is very difficult to assess.

With the exception of one sample from Mexico that shows significant oxygen-MIF signature (Δ17O > 0.1‰), all the other samples are mass-dependent (MD) within 2σ error of our method (Δ17O = 0 ± 0.1‰). As the initially emitted volcanic SO2 is not expected to contain any significant MIF isotopic signature and that fractionation processes such as the distillation/condensation occurring in the volcanic plume are MD, the formation of MD sulfates with a significant isotopic composition range has two possible sources: (1) mostly primary sulfate, which is unlikely at least for samples collected hundreds of kilometers from the volcanic vents or (2) secondary sulfates formed by SO2 oxidation via atmospheric oxidants that are overwhelmingly MD. In the troposphere, H2O2 and O3 exhibit consistent oxygen-MIF signatures with Δ17O ~2‰ and ~32‰, respectively (Table 2). Regarding the oxidation channels Ch2 (H2O2) and Ch3 (O3), 50% and 25% respectively of the sulfate oxygen atoms are from the MIF oxidant (H2O2 and O3). Consequently, we should expect volcanic sulfate generated through Ch2 and Ch3 to have Δ17O of ~1‰ and ~8‰, respectively (Table 2 and Figure 1). OH radicals that derived from the photodissociation of O3 should inherit some MIF signature. This is the case in the stratosphere where the water content is very low (few ppmv), several orders of magnitude lower than in the troposphere. However, in the humid troposphere, OH very quickly loses its MIF signature through isotopic exchanges with MD water vapor. Therefore, tropospheric OH radicals can be assumed to have a Δ17O = 0‰ resulting in secondary volcanic sulfates Ch1 that are MD (Figure 1 and Table 2). Tropospheric O2 has a slightly negative Δ17O (−0.33‰) resulting in secondary volcanic sulfates Ch4 that can be very slightly mass independent with a negative Δ17O between −0.17 and −0.08‰, which is barely detectable in our measurements. From our data set, it can be inferred that in the troposphere, volcanic sulfates that are MD must be overwhelmingly generated by the oxidation of S-bearing gas via OH radicals and/or O2-TMI (Ch1 or Ch4; Table 2) and certainly not H2O2 Ch2 and O3 Ch3.

4.2 Atmospheric Modeling

The modeling results illustrated in Figure 4 show that in the expected acidic volcanic plume conditions [e.g., Symonds et al., 1990; Giggenbach et al., 2001, and references therein], H2O2 is by far (>>90%) the main SO2 oxidant in the troposphere Ch2 and the oxygen isotopic composition of volcanic sulfate should reflect that. However, as shown in 9, this is not consistent with the isotopic measurements which indicate that non-MIF channels (gaseous OH channel Ch1 and/or aqueous O2-TMI channel Ch4) are the dominant pathways in the formation of volcanic sulfate in the troposphere. This contradiction can be explained by the fact that we assume in our model that oxidant levels are not affected by the oxidation of volcanic SO2. This hypothesis necessarily implies that H2O2 concentrations are well in excess of the SO2 available for oxidation [Kreidenweis et al., 2003]. However, SO2 levels in volcanic plumes are extremely high compared to the tropospheric background, which is typically < 0.2 Dobson unit (DU; where 1 DU = 2.69 × 1016 molecules/cm−2). Satellite SO2 vertical column measurements during minor volcanic eruptions [Yang et al., 2007; Lee et al., 2009; Yang et al., 2009] (see also http://so2.gsfc.nasa.gov/) can exceed tens of Dobson unit with peaks at 1000 DU, even during effusive eruptions [Yang et al., 2009]. Once converted into local concentrations, such amount of SO2 is equivalent to approximately tens of ppbv of SO2 throughout the tropospheric column. In situ measurements of SO2 in volcanic plumes from moderate eruptions indicate concentrations ranging from a few tens of ppbv [e.g., Carn et al., 2011; Rix et al., 2012] to about a ppmv [e.g., Hunton et al., 2005; Rose et al., 2006]. In contrast, atmospheric H2O2 concentrations do not typically exceed the ppbv level [e.g., Bey et al., 2001]. Overall, SO2 levels are expected to exceed H2O2 levels by several orders of magnitude in plumes from large volcanic eruptions. Therefore, our equilibrium calculations that assume H2O2 is constant (well in excess of SO2) do not appear to be applicable to SO2 oxidation even after moderate volcanic eruptions. H2O2 is probably consumed very early in volcanic plumes, and hence, only a small fraction of volcanic SO2 can be oxidized by H2O2.

As the volcanic liquid phases are expected to be rather acidic [e.g., Symonds et al., 1990; Giggenbach et al., 2001, and references therein], the SO2 oxidation by O3 cannot be significant (Figure 4). Therefore, most of the volcanic sulfate should be produced via gas phase oxidation by OH and/or via O2-TMI aqueous oxidation. This is consistent with the lack of significant isotopic MIF anomalies in our measurements. The most likely fate of volcanic SO2 during a large tropospheric eruption would be first to react quickly in liquid phases with H2O2, the most effective oxidant (Figure 4), leading to rapid H2O2 depletion. Since volcanic SO2 concentrations would vastly exceed H2O2 concentrations, only a very small fraction of SO2 could be oxidized by H2O2. Once H2O2 is depleted within the volcanic plume, the second most effective oxidant, the OH radical, can take over the S-bearing gas oxidation (Figure 4). Furthermore, OH concentrations are not expected to be as affected by SO2 oxidation as H2O2, because OH radicals are regenerated on a time scale of the order of a second [Brasseur and Solomon, 1986], whereas H2O2 is typically regenerated on a time scale of a day [von Kuhlmann et al., 2003]. It is also possible that O2-TMI oxidation could play a significant role in this phase of volcanic SO2 oxidation but only if the pH does not drop below 3. No data about the amount of TMI present in volcanic plumes are considered here. However, recent field measurements demonstrated that O2-TMI oxidation could be one of the dominant pathways for in-cloud sulfur oxidation [e.g., Harris et al., 2013]. In terms of isotopic signature, it would lower the sulfate Δ17O to slightly negative values (Table 2), which could partly explain some sample compositions with small negative Δ17O‰ (down to −0.14‰; Table 1 and Figure 3).

Considering the quick tropospheric H2O2 depletion in volcanic plumes, it can be estimated that the formation of MD volcanic secondary sulfates (−0.1‰ < Δ17O < 0.1‰) requires the participation of H2O2 for less than 10% and consequently OH radicals and/or O2-TMI for more than 90%. This estimation is in agreement with our isotopic data set indicating that volcanic sulfates generated in the troposphere are massively mass dependent. The only exception would be one Mexican sample for which the H2O2 oxidation channel could have been responsible for up to 35% of the total sulfate formation. Without further data, notably on the isotopic composition of volcanic SO2 and possibly of the fraction of primary versus secondary sulfates in the sampled volcanic ash, it is not possible to conclude unambiguously on the importance of the H2O2 channel for this specific volcanic eruption.

In the troposphere, volcanic sulfates produced by the oxidation channels Ch1 (OH) and Ch4 (O2-TMI) are expected to have δ18O between −5‰ and 1‰ and between 4 and 15‰, respectively (Table 2). Sulfate produced by channel Ch2 (H2O2) should have δ18O between 13‰ and 25‰. From our data sets, we have inferred that >90% of the volcanic sulfates in the troposphere are generated via Ch1 and/or Ch4 and <10% via Ch2. Consequently, the mass balance calculation based on Table 2 indicates that volcanic sulfate should have δ18O between −5‰ and 14‰, which is similar to the composition range observed in our data set (Figure 3). However, as discussed before, fractionation processes such as distillation/condensation could be also responsible for at least some of the δ18O variations observed in our samples. Indeed, during the release of S-bearing volcanic gases, the first sulfate aerosols formed quickly fall to the ground and relatively close to the volcanic vent. The not yet oxidized S-bearing gases are oxidized later on after traveling, depending on the wind, in the troposphere or in the tropopause region. The sulfate aerosols generated from these volcanic gases fall to the ground farther away from the volcanic vent than the first sulfate generation. According to the Rayleigh distillation equation, the first and latest generation of aerosol sulfate should be isotopically the heaviest and the lightest, respectively. Therefore, if distillation/condensation processes play an important role, we should expect sulfate having the highest δ18O to be found in the volcanic deposits closest to the vent.

As discussed above, the δ34S values from this study cannot be interpreted in terms of distillation/condensation processes as that would require a precise knowledge of the initial magmatic SO2 isotopic composition for each eruption. In order to better constrain the formation of sulfate and its isotopic composition, it is also necessary to have more precise fractionation factors than those experimentally determined by Harris et al. [2012] during SO2 oxidation in the atmosphere. Then the discrimination between distillation/condensation processes and fractionation during SO2 oxidation reactions would allow us to link the sulfate δ34S with the distillation/condensation processes which should be somewhat a function of the distance to the studied volcano. As a consequence, the fractionation factors of the O and S isotopes during such processes that occur in complex environments such as volcanic plumes could be explored, as initiated by Harris et al. [2013]. This could certainly increase our knowledge of volcanic plume conditions and behavior in the troposphere. For instance, it could lead to the estimation of the distance from the source and potentially the altitude that a volcanic plume reached during a specific eruption based on the isotopic composition of the sulfate found in a volcanic deposit. This could obviously have major implications in tephra stratigraphy where the location and magnitude of eruptions are investigated from the study of different volcanic ash deposits.

5 Summary and Conclusions

In this paper, we illustrate how sulfate extracted from volcanic ash deposits could be useful to understand the oxidation of S-bearing gases in the atmosphere. Indeed, combining the S and O isotope compositions of these volcanic sulfates and model calculations may help to constrain the relative importance of some of the oxidation channels through which sulfate aerosols are generated in the atmosphere. While volcanic sulfates generated in the stratosphere during large eruptions show that most of the time, MIF oxidation channels dominate, volcanic sulfate formation in the troposphere is poorly constrained. In this study, we extracted sulfate from ash deposits of moderate eruptions, in order to assess volcanic sulfates that had been most likely generated in the troposphere and the tropopause region. Our isotopic measurements and modeling indicate that if H2O2—a MIF anomaly carrier—is by far the most important SO2 oxidant in the background atmosphere, it must be quickly consumed by the volcanic SO2 in the initial phase of the eruption. Oxidation of volcanic sulfur in the troposphere appears to be dominated by significantly non-MIF oxidation channels, notably oxidation via tropospheric OH radicals in the gas phase and by O2 catalyzed by transition metal ions in the aqueous phase. Other more speculative non-MIF channels could be involved.

It has to be emphasized that the volcanic sulfate sampled was certainly a mix of primary sulfate (i.e., directly emitted by the volcanoes) and secondary sulfate (i.e., produced in the atmosphere). The MIF anomaly of primary sulfate is expected to be negligible. Nonetheless, in order to conclude unambiguously on the isotopic anomaly resulting from atmospheric processes, the isotopic composition of primary sulfate should be characterized for each eruption. This study shows that the individual variations in δ34S and δ18O for specific eruptions could also help to better constrain volcanic sulfur oxidation processes. However, in order to do so, the isotopic composition of initial volcanic SO2 for specific eruptions needs to be known. Better constraints on volcanic sulfur oxidation processes would have important implications for our understanding of volcanic sulfate aerosol formation and hence the impact of volcanic eruptions on climate. Finally, the study of the isotopic composition of sulfate deposits from volcanoes that are found everywhere and have been present on Earth from its early stages could provide a window onto the atmospheric chemistry of the past.


We would like to thank S. Hidalgo, J. Roberge, B. Rose, A.M. Sarna-Wojcicki, O. Sigmarsson, and the rock collection at the Smithsonian Institution of Washington DC for providing volcanic ash samples as well as V. Atudorei for sulfur isotope analyses at the University of New Mexico. Constructive reviews from T. Mather and anonymous reviewers improved the manuscript. Finally, we are grateful to L. Whiteley for proofreading this manuscript. The city of Paris and the Agence Nationale de la Recherche (ANR) via contract 14-CE33-0009-02-FOFAMIFS are acknowledged for the financial support. Data can be requested from Erwan Martin ([email protected]).


  1. In the originally published version of this article, typos were present in the Introduction and in the caption for Figure 1. These typos have since been corrected and this version may be considered the authoritative version of record.