Journal of Geophysical Research: Atmospheres

Aerosol formation in basaltic lava fountaining: Eyjafjallajökull volcano, Iceland

Authors


Corresponding author: E. Ilyinskaya, Icelandic Meteorological Office, Bustadavegi 9, 150 Reykjavik, Iceland. (evgenia@vedur.is)

Abstract

[1] A short-lived episode of basaltic lava fountaining at Eyjafjallajökull volcano (March – April 2010) produced a low-altitude, ash-poor plume. We measured the composition of aerosol particles (sampled using a cascade impactor and filter packs), gases (sampled using filter packs), and volatile species scavenged by scoria and external water in order to investigate the formation and speciation of near-source aerosol (<2 min from emission). Samples were analyzed for volatile species (S, Cl and F) and metals (Na, K, Ca and Mg). The aerosol mass showed two unusual features: the prevalent size mode was finer than typically found in volcanic plumes (∼0.2μm, compared to >0.4 μm), and its composition was dominated by chloride rather than sulfate. We used two thermodynamic equilibrium models (E-AIM and HSC Chemistry v5.1) to show that the formation of particulate Cl by condensation of HCl gas is more responsive to changes in ambient temperature than the oxidation of SO2 to SO42−, so that a low SO42−/Cl ratio in aerosol particles is characteristic of volcanic emissions in cold climates. Field measurements suggested that the efficiency of SO2 to SO42− conversion inside the vent increased with lower explosivity. Volatiles adsorbed on the surface of scoria had significantly higher SO42−/halogen molar ratios than the aerosol samples. Several potential explanations for these differences are discussed.

1. Introduction

[2] Volcanism is an important natural source of aerosol particles; yet relatively little is known about how particles form within volcanic vents [Mather et al., 2003]. This study investigates aerosol formed during an episode of Hawaiian-type lava fountaining, sampled within 1–2 min from emission. Lava fountains are continuous jets of magma propelled up to >100 m, driven by exsolution of volatiles and bubble expansion as magma decompresses on ascent, with comparatively fast magma ascent and closed system degassing [e.g.,Allard et al., 2005]. The majority of the pyroclasts tends to follow near-ballistic trajectories forming the characteristic cone shape around the vent (Figure 1). However, the formation of convective gas-rich plumes above and downwind of lava fountains has been reported [Head and Wilson, 1989] and plumes can rise >10 km in the most energetic events [Thordarson et al., 1996]. Convective plumes result from dynamic mixing and heat exchange between the magmatic and atmospheric gases. It has been shown (using equilibrium calculations) that mixing of magmatic and atmospheric gases at near-magmatic temperatures has important implications for chemical reactions in the near-source plume. This mixing produces elevated concentrations of oxidized species such as SO42− [Gerlach, 2004; Martin et al., 2006], which is one of the most abundant chemical species in volcanic aerosol [Allen et al., 2006, 2000; Ilyinskaya et al., 2010; Mather et al., 2003, 2004a, 2004b]. In Iceland the most common sulfur form in mafic magmas is SO2 [Métrich et al., 1991; Thordarson et al., 1996; Moune et al., 2006], including the eruptive event studied here (T. Thordarson, personal communication, 2012). Sulfur can also be dissolved in magmas as sulfate and sulfite [Métrich et al., 2002], and it is possible that a portion of the magmatic sulfur is exsolved as sulfate. However, the SO42− concentrations typically observed in the field (e.g., SO42−/SO2 ∼ 0.01) [e.g., Allen et al., 2002; Naughton et al., 1975] are too high compared to the predicted sulfate vapor concentrations [Symonds and Reed, 1993]. Thermodynamic modeling [e.g., Martin et al., 2006] indicates that the SO3/SO2 ratio in the eruption plume (and the resulting abundance of SO42−, which is assumed to form by reaction of SO3with water as the plume cools) increases with the efficiency of mixing between magmatic and atmospheric gases in the high-temperature region. SO3has been detected in high-temperature jet aircraft exhaust [e.g.,Curtius et al., 1998; Sorokin et al., 2004] but due to the difficulty with sampling deep parts of volcanic vents, SO3 has not yet detected above magmatic surfaces. Field data that examine this process in volcanic emissions are very limited.

Figure 1.

Hawaiian-type lava fountaining at Eyjafjallajökull in March 2010. The series of photographs allows comparison of eruptive activity between the sampling days and depicts the variability in the eruption vigor. (top) Lava fountaining on March 24, highly vigorous with maximum height of eruptive jet of approximately 150 m above the ground. (middle) The eruption site on March 28. Only one lava fountain was seen. The maximum height of the lava fountain varied from ∼60–100 m above the ground. (bottom) The eruption site on April 1 when the vigor of the eruption had greatly declined. Lava fountaining was very weak and rarely visible above the crater rim. However, a concentrated gas plume was persistently emitted from the crater.

[3] Lava fountaining activity has been observed in all volcano-tectonic settings, i.e., oceanic and continental rifts (e.g., Icelandic volcanoes and Afar depression, Ethiopia), converging plates (e.g., Etna, Italy); and hot spot volcanism (e.g., Hawaiian volcanoes). Gas emissions during episodes of lava fountaining have been investigated in numerous studies, via direct sampling [Crowe et al., 1987; Miller et al., 1990; Naughton et al., 1975; Sigvaldason and Elisson, 1968; Swanson and Fabbi, 1973], spectroscopy [Aiuppa et al., 2007; Allard et al., 2005; Chartier et al., 1988; Edmonds and Gerlach, 2006], the petrological method [e.g., Moune et al., 2007; Thordarson et al., 1996] and modeling [e.g., Parfitt and Wilson, 1995]. Before the eruption of Eyjafjallajökull in 2010, the most recent Icelandic eruption where gas emissions were measured directly was the explosive event of Hekla in 2000 [Moune et al., 2006; Rose et al., 2006].

[4] There are significantly fewer measurements of aerosol emissions during lava fountaining. Aerosol emissions were studied at Kīlauea summit crater [Cadle and Frank, 1968; Naughton et al., 1975] and along its rift zone [Cadle et al., 1971; Crowe et al., 1987; Hinkley, 1991; Miller et al., 1990; Naughton et al., 1974]; as well as at Etna volcano during Strombolian explosions [Allen et al., 2006]. At Kīlauea volcano, the metal/Cl ratio in volcanic ‘fumes’ (gases and aerosol) was found to increase during episodes of lava fountaining compared to episodes when the vent was hot (900°C) but no magma was visible near the surface [Crowe et al., 1987]. The increase in the metal/Cl ratio was attributed to arrival and discharge of more volatile-rich magma [Crowe et al., 1987; Miller et al., 1990]. However, more recent studies [e.g., Edmonds et al., 2009; Mather et al., 2012] indicate that the metal/Cl ratio may be largely dependent on variations of the degassing behavior of Cl with eruptive activity. At Etna, differences in the concentration of metals between the degassing vents have been linked to Strombolian or gas puffing activity [Allen et al., 2006; Gauthier and Le Cloarec, 1998].

[5] In 2010, a unique opportunity presented itself to investigate emissions during a lava fountaining episode in Eyjafjallajökull volcanic system in Iceland. The eruption was relatively small in terms of erupted volume (∼0.025 km3) [Edwards et al., 2012] and persisted for three weeks. It was possible to follow a change in the activity from energetic lava fountaining (>100 m on March 24) to near-quiescent degassing from the surface of lava ponding in the vent (April 1). We measured the changes in the chemical composition of the gas and aerosol emissions in order to:

[6] 1. Investigate how the relative abundance of sulfate varies with the eruption vigor on the timescale of days (samples collected on March 24, 28 and April 1). Equilibrium models were used to calculate the concentration dependency of sulfate on processes occurring under high temperature (high-T, referring to magmatic or near-magmatic temperatures ≫100°C) conditions (e.g., 2SO2 + O2 → 2SO3) and low temperature (low-T, referring to ambient or near-ambient temperatures ≪100°C) conditions (e.g., SO3 + H2O → H2SO4).

[7] 2. Identify changes in the relative abundance of metals (i.e., Na, K, Mg, and Ca). Previous results, as described in the paragraphs above, have shown a correlation between activity state and metal flux.

[8] 3. Quantify the differences between the abundances in the gas phase, aerosol phase and volatile species scavenged by tephra and external water. Volatiles found on the surface of very young pyroclasts (deposited within ∼1 min after emission) may provide information about the adsorption processes inside the fountain. Volatiles scavenged by external water (surface waters and meteoric precipitation) have been used as proxies for the magmatic volatile composition [Edmonds et al., 2003; Flaathen and Gislason, 2007; Gislason et al., 2002]. By making simultaneous measurements of both in-plume and scavenged species, we attempted to establish whether a significant chemical fractionation occurs during scavenging.

2. Eyjafjallajökull Volcanic System

[9] Eyjafjallajökull (63.63°N 19.65°W) is a glacier-covered stratovolcano rising to ∼1670 m a.s.l. The volcano has erupted four times in the last 1400 years (the penultimate eruption in 1821–23), all of which have been of low volume (≤0.1 km3) [Larsen et al., 1999].

[10] The 2010 eruption started on March 21 with a flank fissure eruption on the ice-free part of Fimmvörðuháls mountain pass at 1100 m a.s.l. (Figure 2). The melt composition was evolved FeTi-basalt (mildly alkalic) with MgO in the range of 4.5–5.1 wt% [Sigmarsson et al., 2011]. It did not interact with the more evolved summit reservoir magma [Sigmundsson et al., 2010]. The eruptive fissure was initially ∼300 m long and shortened throughout the eruption. It produced lava fountains, lava flows and plumes with very low ash content [Gudmundsson et al., 2010; Ilyinskaya et al., 2011]. A second fissure opened on March 31 within approximately 200 m from the March 21 fissure; the two fissures were aligned at approximately 60° angle (Figure 2). A mixed scoria-spatter cone was built up at the first fissure reaching approximately 75–80 m above ground level. A pure spatter cone was built up around the second fissure, reaching approximately 35 m above ground level. Eruptive activity ceased at the first fissure on April 7, and on April 12 at the second fissure. However, post-eruptive gas emissions continued for some time. On April 14 eruptive activity began at the summit crater with energetic phreatomagmatic explosions. Emissions from the summit eruption are not dealt with in this study.

Figure 2.

Location map of the sampling sites during the Eyjafjallajökull flank eruption on Fimmvörðuháls mountain pass (elevations are given in meters a.s.l.). The position of the two eruptive fissures and the outlines of the lava field are based on Edwards et al. [2012]. Samples of gas, aerosol and scoria were collected from the older (easternmost) eruptive fissure. Collection of gas and aerosol on each sampling day took place within the area shown by the white rectangles. The rectangle closest to the eruptive fissure represents the sampling site on March 24. The rectangle further south marks the position of the sampling sites on March 28 and April 1 which were located within a few tens of meters from each other. The zoom area shows in more detail the sampling sites of gas, aerosol and water around a cooling lava flow which had descended into Hrunagil canyon. Temperature of each water sample (W1-W7) is also shown in the zoom area.

3. Methodology

[11] We collected several distinct sample types. Aerosol and gas were sampled in the eruptive plume, and at a high-Tvent in a cooling lava flow. Aerosol was collected both as size-resolved and bulk samples. In addition, we collected tephra fallout from the plume, as well as samples of water which was interacting with a cooling lava flow. Chemical composition of the samples was analyzed using ion chromatography. Thermodynamic equilibrium modeling was then used to investigate the gas-aerosol transformation under high-Tand low-T conditions.

3.1. Sampling Sites

[12] Sampling was carried in two distinct locations: at the eruption site (section 3.1.1), and at a cooling lava flow (section 3.1.2), approximately 2 km northeast of the eruption site (Figure 2). Description of individual samples is in Table 1. The weather conditions were typical for the location and season and remained stable throughout the sampling period. Air temperatures ranged from −15 to −10°C and the relative humidity was 50–60%. The ground was covered by 1–3 m thick snow cover [Edwards et al., 2012]. The weather stayed clear on all of the sampling days, and there was no cloud cover or precipitation. This implies that there was no scavenging of the plume components by meteoric precipitation. The average wind speed was 10–15 m/s (higher in gusts) on all of the sampling days.

Table 1. Samples Collected During Fimmvörðuháls Eruption in 2010a
DateActivity (Maximum Fountain Height)IdentificationSampleNotes
  • a

    FP - filter pack, M – MOUDI cascade impactor, Sc – scoria, W – water sample.

March 24Vigorous lava fountaining (150 m)FP-1a, 1b, 1cGas and suspended PMSampling duration 1 h each
M-1Size-resolved suspended PMSampling duration 5 h
Sc-1ScoriaAge <3 h
Sc-2ScoriaAge <30 min
     
March 28Vigorous lava fountaining (60−100 m)FP-2a, 2b, 2c, 2d, 2eGas and suspended PMSampling duration 1.5 h
M-2Size-resolved suspended PMSampling duration 4 h
Sc-3ScoriaAge <1 h
     
April 1Very weak spattering, lava pond in the vent (10–20 m)FP-3a, 3b, 3cGas and suspended PMDuration 1.5 h
M-3Size-resolved suspended PMDuration 4 h
     
April 13Cooling lava flowW1Surface streamTemp (°C) 2.6 - ambient
W2Water flow through lava22.5
W3“”51.5
W4“”57.5
W5“”80.9
W6“”57.5
W7Surface stream2.6 - ambient
FP-4a, 4bGas and suspended PM above a high-T crackSampling duration 0.75 h

3.1.1. Sampling at the Eruption Site

[13] It was relatively easy to access the eruption site, and sampling was carried out within 100–200 m from the active vent (Figure 2). Emissions from the first eruptive fissure were sampled on March 24, March 28 and April 1. It was not possible to collect samples at the second fissure due to an unfavorable wind direction which precluded safe site access. The results and discussion in the following sections focus therefore solely on the emissions from the first fissure.

[14] The lava fountaining activity was most vigorous on March 24 with maximum fountain height reaching ∼150 m above the ground surface (Figure 1). On March 28, explosive activity was still ongoing but only one lava fountain was active. The fountain height had decreased but was highly variable, reaching maximum of ∼60–100 m (Figure 1). The vigor of the activity had declined significantly by April 1 (Figure 1), and had been reduced to weak spattering or near-quiescent degassing. Fountaining was rarely seen reaching above the vent (∼10–20 m). Lava was ponding in the vent, which was persistently emitting a vapor-rich plume. The rapid decline in eruption vigor between March 31 and April 2 was associated with a shift of activity to the second eruptive fissure which opened on March 31 [Höskuldsson et al., 2011].

[15] Sampling equipment was positioned downwind of the active vents, and as close to the source as possible in order to sample the most concentrated emissions. It was necessary to occasionally move the sampling equipment if the wind direction shifted, or if lava began to flow toward the sampling location so that it became unsafe. However, the equipment was never moved more than a few tens of meters. The equipment was positioned in such a way as to sample only the plume emitted from the active vent, and avoid mixing with the plume emitted from the lava flow (verified by visual observations), although this was not always possible where lava flows had been emplaced between the vent and the sampling equipment. However, it has been shown that >70% of the gas carried by the magma toward the surface is released from the eruptive vents, while lava flows generally emit very dilute plumes [Thordarson et al., 1996]. Therefore, it was expected that a small degree of plume mixing would not have a significant effect on the gas and aerosol concentrations collected by the sampling equipment. When extensive mixing of the two types of plumes was deemed inevitable, the sampling was paused and the collection filters for gas and aerosol changed to new ones. The plume from the vent was not grounding constantly, but nevertheless frequently due to sufficiently high wind speeds on all of the sampling days. In addition to the direct sampling, we carried out optical measurements of particle size distributions in the plume using Sun photometry [Ilyinskaya et al., 2011]. Tephra fallout from the plume was sampled on March 24 and 28 (section 3.3).

3.1.2. Sampling at a Cooling Lava Flow

[16] On April 13, samples were collected at a cooling lava flow which had descended into Hrunagil canyon in Thorsmörk valley (Figure 2). An exterior chilled crust had developed that displayed high-Tvents and cracks but no molten material was seen. The temperature in the vents and cracks was at least 150°C, but it was not possible to make a more exact measurement due to the limitations of the available thermometer. It was considered likely that the temperature was significantly higher than 150°C. We collected samples of aerosol and gas above one of the high-T vents.

[17] A glacial water stream was running onto and around the cooling lava flow in Hrunagil (Figure 2). The temperature of the water notably increased upstream (i.e., closer to the lava flow front). Water samples were collected at varying distances from the lava flow front (Figure 2 and section 3.4), in order to assess the degree of chemical interaction between the water and the lava.

3.2. Aerosol Particles and Gas Sampling

[18] Particles can be collected either as bulk samples, or segregated into size-dependent bins. Both approaches have advantages and limitations. While size-segregated particle sampling allows measurement of the particle size distribution, uncertainties are introduced by particle loss between collection stages. Here we employed both types of particle samplers – filter packs for bulk collection (section 3.2.1) and a cascade impactor for size segregated samples (section 3.2.2). These samplers were used simultaneously and in the same location. Filter packs were also used for gas phase sampling. Procedural blanks were collected in the following manner: blank filters were transported into the field and transferred to the sample bags in the same way as the sampling filters. This ensured that there was no unaccounted contamination due to transport and handling of the filters in the field.

3.2.1. Filter Packs

[19] Filter packs were used to collect simultaneous samples of gas and aerosol. We followed the technique described by Allen et al. [2000]. Comparisons of gas ratios from filter pack analyses and simultaneous open-path Fourier transform infrared measurements (OP-FTIR) indicate acceptable differences (30%) in gas ratios [Aiuppa et al., 2004; Martin et al., 2010]. In general, the higher sampling rate of OP-FTIR (∼1 s−1) than filter packs (∼1 h−1) allows for greater confidence in gas ratios (e.g., SO2/HCl) derived from OP-FTIR measurements [Aiuppa et al., 2004; Martin et al., 2010].

[20] The filter packs comprised of one particle filter followed by one base-treated gas filter in an all-Teflon three-stage cartridge. The particle filter used was Millipore, 47 mm, AAWP, pore size 0.8μm. The filters for gas sampling were Whatman 41 mm ashless circles impregnated with 5% K2CO3 and 1% glycerol in distilled deionized (DI) water, following the preparation technique described in Martin et al. [2010]. This base treatment captures acidic gases (e.g., SO2, HF and HCl,) on the filter by conversion to their weakly basic counter-anions (e.g., SO42−, F and Cl). Sample FP-1a was collected using two base-treated gas filters in series. The amount of gas species collected on the second filter was <2% of that on the first filter, and this is believed to be representative of the whole sample set.

[21] Airflow through the filter pack was generated using an external pump (12 V Charles Austen Capex V2 DE pump) running at 22 L min−1 for 60–90 min. Between three and five filter packs were collected each day. Immediately after sampling, the filter pack was closed by a plastic lid to prevent particle loss and contamination. Filters were then transferred into individual polypropylene bags (2 layers) inside a vehicle allowing for easier handling.

3.2.2. Cascade Impactor

[22] A cascade impactor consists of a series of stages with nozzles of decreasing diameter. An external pump (12 V Charles Austen Capex V2 DE pump) creates an airflow through the impactor. Beneath each set of nozzles there is an impaction plate (where a filter substrate is mounted) upon which the air jet impinges and where a fraction of the suspended aerosol particles is collected. Larger particles are deposited at the early separation stages whereas fine particles are not deposited until they have passed through correspondingly narrower nozzles in which the gas flows at higher velocities [e.g., Marple et al., 1991]. The cut-off particle diameter for each impactor stage is conventionally reported asD50which is the diameter of particles retained with 50% efficiency on each stage. In this study, we used a Micro Orifice Uniform Deposition Impactor (MOUDI, model 125NR, MSP corporation). MOUDI cascade impactors have sharp cut-point characteristics and low inter-stage particle loss [Ma and Kim, 2008; Marple et al., 1991], which is of utmost importance for analysis of size-dependent aerosol chemistry. The estimated interstage particle losses for MOUDI instruments, as determined byMarple et al. [1991] using monodisperse aerosol, are greatest for the largest diameters (up to 35% at 15 μm), but <2% between 3.2 and 0.18 μm where the largest proportion of volcanic aerosol tends to be found [e.g., Mather et al., 2003]. The calibration procedures (i.e., determination of D50) have been reported by Marple et al. [1991] and by Marple and Olson [1999].

[23] MOUDI instruments have been successfully used in several studies on volcanic aerosol [e.g., Martin et al., 2008; Mather et al., 2003, 2004a, 2004b]. A new model used here (125NR or nano-MOUDI; previously used byIlyinskaya et al. [2010] and Martin et al. [2011, 2012b]) resolves 14 size bins between 18 and 0.010 μm and operates at 10 L min−1. If the required flow rate is not attained during sampling, the cut-off particle diameter at each stage is increased by a factor of inline image where Qr and Qm are the required and measured flow rates, respectively [Hinds, 1999].

[24] Filter substrates were placed on each impaction stage. Filters used were PTFE, 47 mm, 0.2 μm pore size. The pump used with MOUDI was 12 V Charles Austen Capex V2 DE pump powered by a 12 V car battery. The flow rate obtained in the field was 9 L min−1.

3.3. Scoria

[25] Samples of scoria were collected on March 24 and 28 in the same locations as the gas and aerosol samples (Figure 2). The samples were collected by laying out a polyethylene bag on the ground. After a known time interval (variable for each sample, see Table 1), the scoria which had fallen out on top of the sampling bag was transferred to smaller polypropylene bags (2 layers) and sealed. Caution was taken to keep the samples as dry as possible.

[26] The mean diameter of the scoria was determined by measuring 20 clasts selected at random from the sample. The clasts were all of medium lapilli size (8–32 mm), with a mean of 12 mm. Twenty g of each sample was leached with 18.2 MΩ DI water. The sample/water proportion was 1:10 (weighed to 0.01 g accuracy) and the leaching time was 30 min.

3.4. Water Samples

[27] The lava flow produced during the eruption partially filled a steep-walled canyon, Hrunagil in Thorsmörk valley, which has a glacial water stream running in it (Figure 2). The water stream was sampled at five locations at varying distances from the lava front (Figure 2). The water temperature in Hrunagil ranged from 22°C (∼2000 m from the lava front) to 80°C (at the lava front). The stream increased in flow away from the lava front, as smaller streams from the canyon sides converge. Two ‘uncontaminated’ streams from nearby canyons which had not come in contact with the 2010 lava were sampled for comparison (water temperature ∼3°C). The water samples were collected into 50 ml polypropylene vials with metal-free caps. The vials were cleaned with DI water prior to fieldwork and then rinsed twice with the sample water before collection.

3.5. Analytical Procedures

[28] The filter substrates from the impactor and filter packs were handled in a regular laboratory using ceramic scissors and tweezers to minimize metal contamination. The outer rim of the filter which was in contact with the filter holder was cut off. The filters were then transferred into metal-free polypropylene vials. All vials and vial caps used in this study were pre-cleaned with ultra-pure DI water (18.2 MΩ).

[29] The acid gas filters were extracted with 10 ml of the ultra-pure DI water. A few drops of analytical grade 30% H2O2 were added to the extract to oxidize SO32− to SO42−. The extracts were sealed with metal-free caps and left for 24 h.

[30] Particle filters (hydrophobic PTFE) were wetted with a few drops of propan-2-ol, and the ultra-pure DI water was added to make a solution of 5 ml. The solutions were placed on a mechanical shaker for 30 min to promote extraction. This procedure is intended to extract only the water-soluble fraction of the collected particle mass (including that which is adsorbed on the surface of silicate particles, if any are present). Glass dissolution of silicate particles (ash) was considered negligible during this mild extraction procedure.

[31] All samples (extracts of gas and particle filters, leachates of scoria, and water samples) were analyzed using ion chromatography. Gas filter extracts were diluted 100× to avoid saturation of the ion chromatograph column, as concentrations in undiluted volcanic gas extracts can be in excess of 1000 ppm. We used a Dionex ICS-3000 ion chromatography system (Department of Earth Sciences, University of Cambridge). Anion calibrations were achieved using 10, 2, 1, 0.1, 0.05 and 0.01 ppm solutions made by dilution of Cl and SO42− standards (Fischer Scientific J/4546/05, J/4564/05), and using 5, 1, 0.5, 0.25, 0.125, 0.025, 0.0125, 0.0025 ppm solutions made by dilution of F, Br and NO3 standards (Fischer Scientific J/4548/05, J/4544/05, J/4556/05). Cation calibration (Na+, K+, Mg2+, Ca2+ and NH4+) was achieved using 4, 2, 1, 0.2, 0.1, and 0.05 ppm dilutions of a mixed standard solution (Fisher Scientific J/4554/05).

[32] The analytical method detection limit (MDL) and precision (MP) are listed in Table 2. The MDL was determined by

display math

where cbl is the mean concentration of blank measures, σ is the standard deviation of repeated standard measures (1 ppm standard solution measured repeatedly between every 14 samples) and k is a t-distribution value chosen at 95% confidence level (df =N − 1 at P = 0.05, where N is the number of repeated standard measures). Measurement precision (MP) was defined as the percent error of repeated analysis of the 1 ppm standard.

Table 2. Method Detection Limits (MDL; μmol/Sample Extract) and Method Precision (MP; ±%) for Each Sample Seta
 Ionic Species
SO42−FClBrNO3Na+K+Ca2+Mg2+
  • a

    Measurement precision (MP) was defined as the % error of repeated analysis of 1 ppm standard (5–8 repetitions in each analysis run).

MDL0.000180.00140.00180.000350.000120.000340.000320.00200.0059
MP132.69.54.01.25.95.97.18.1

3.6. Thermodynamic Equilibrium Modeling

[33] Thermodynamic equilibrium modeling requires the assumption that the components of interest (in the high-Tcase, magmatic and atmospheric gases; and in the low-T case, gases and particles within the plume) have reached equilibrium. Equilibrium models are appropriate because the uptake of gases into particles may occur over timescales as short as a few seconds [Wexler and Clegg, 2002]. Even if equilibrium has not been attained at the point of sampling, the modeling results provide insights on the effect of different variables (including temperature, relative humidity and species concentrations) on the proportions of the plume components. Here we calculated the high-T equilibrium composition using HSC Chemistry model v5.1 (section 3.6.1), and the low-T equilibrium composition was calculated using the Extended Aerosol Inorganics Model (section 3.6.2).

3.6.1. HSC Chemistry

[34] High-temperature speciation in mixtures of magmatic and atmospheric gases have been investigated using equilibrium models, such as Gasmix [Kress et al., 2004]; Solvgas [Symonds and Reed, 1993]; and HSC Chemistry [Gerlach, 2004; Martin et al., 2006]. Here we used the HSC Chemistry v5.1 to calculate the equilibrium constant for the reaction SO2 + 1/2 O2 = SO3 between 800 K and 1400 K. The equilibrium constant and the SO3/SO2 molar ratio (assumed to equal SO42−/SO2 from measurements) constrain pairs of values for the temperature (T*) and the relative volume of atmospheric and magmatic gases (VA/VM) of the gas mixture. However, since T* and VA/VM are related through the specific heat capacity of the gas mixture, both can be determined by measurement of SO42−/SO2. We selected the initial magmatic gas temperature of 1243 K (1170°C as determined by Keiding and Sigmarsson [2012]) with a specific heat capacity equal to that of H2O(g), and an atmospheric gas temperature of 248 K or 298 K with a specific heat capacity equal to that of N2(g). The main limitation of our calculations was that we did not consider the presence of H2 and H2S in the gas mixture. If present, these gases may inhibit the oxidation of SO2 to SO3 [Martin et al., 2009]. However, petrological analyses of tephra and lava from the same eruption [Gunnlaugsson et al., 2011; Keiding and Sigmarsson, 2012], indicated that dominant sulfur form in the magma (and the magmatic gas) was SO2 so the amounts of H2 and H2S present are expected to have been small.

3.6.2. Extended Aerosol Thermodynamics Model

[35] The Extended Aerosol Inorganics Model (E-AIM) [Wexler and Clegg, 2002] has been applied in multiple studies of volcanic aerosol [Martin et al., 2010, 2012a; Mather et al., 2004b; Oppenheimer et al., 2006; Roberts et al., 2009]. The model considers the equilibrium speciation of H+-NH4+-Na+- SO42−-NO3-Cl over a range of temperatures and relative humidity.

[36] Here, we used E-AIM version IV to investigate gas-aerosol partitioning of HCl(g) to Cl(aq,s) with variations in the ambient temperature (between −9°C and +9°C). The relative humidity (RH) was set at 60%. Ice formation was excluded from the model scenario. Water in atmospheric clouds can be cooled much below the equilibrium freezing temperature without forming ice particles. Homogeneous ice particle formation requires supercooling below −35°C; although heterogeneous nucleation may begin at temperatures around −10°C [Pruppacher and Klett, 1997]. The rate of heterogeneous nucleation is dependent on the abundance of particles capable of acting as ice nuclei; this process is considered to be important in plumes rich in fine (<1000 μm) ash [Durant et al., 2008; Seifert et al., 2011]. Ice nucleation in young and essentially ash-free plumes is generally thought to be insignificant [Durant et al., 2008; Seifert et al., 2011].

[37] It must be noted that the RH calculations require the assumption that the ambient RH equals the in-plume RH. This assumption is supported by the fact that plumes generated by quiescent degassing, or mildly explosive activity, are at ambient temperature at the point of measurement.

[38] The input data were an average of all near-vent filter pack samples inμmol m−3. Na+ is the only metal cation considered by the model; so we added the concentrations of all measured cations according to (1):

display math

where [Na+]model was the total concentration (i.e., charge weighted) of cations used for the model input. This approach is in agreement with previous studies [Martin et al., 2010; Mather et al., 2004b]. H+ was calculated by assuming a charge balance between the cations (i.e., [Na+] + [K+] + 2[Mg2+] + 2[Ca2+] + [H+]) and the anions (i.e., [F] + [Cl] + 2[SO42−]). The input values (in μmol m−3) were: Cl = 606 (combined gas and particle phase); H+ = 605 (combined gas and particle phase); Na+ = 0.77; SO42− = 0.15; NH4+ = 0; NO3 = 0.

4. Results

[39] Samples of gases and aerosol were collected in the eruptive plume at a lava fountaining fissure (March 24, 28 and April 1) and at a partially cooled lava flow (April 13), and subsequently analyzed for composition. In addition, we sampled and analyzed the composition of (i) volatile species adsorbed on surface of scoria clasts (March 24 and 28), and (ii) volatiles dissolved in river waters that came into contact with the lava flow (April 13). The analytical results are summarized in Table 3 and the results for each individual cascade impactor stage can be found in the auxiliary material. It must be noted that the different sample collection methods may cause some difference in analytical results. Specifically, chemical species of low abundance in the aerosol phase are more likely to be above detection limits in bulk samples (filter packs and scoria surface) rather than in size-segregated bins (cascade impactor).

Table 3. Concentrations Measured in Each Samplea
Date, ActivityIdentificationHFHClSO2FClSO42−Na+K+Mg2+Ca2+Aerosol Charge Balance
  • a

    Abbrevation n/a – not analyzed. Units used: filter packs (FP) and cascade impactor (M) - (μmol m−3); scoria leachates – mg kg−1 scoria; water – mg L−1. Concentrations on individual stages of the cascade impactor are included in the auxiliary material.

March 24, syn-eruptiveFP-1a53721200.911.40.170.860.430.0450.211.5
FP-1b95891901.61.70.261.60.530.070.261.4
FP-1c15026037001.20.100.340.2100.241.4
M-1n/an/an/a00.140.0270.190.1800.0860.4
Sc-1n/an/an/a2.24.919100.360.926.31.8
Sc-2n/an/an/a4.62.320100.390.796.71.8
             
March 28, syn-eruptiveFP-2a82530250000.340.0320.120.05800.0651.3
FP-2b79300110000.210.0490.0890.05100.0351.5
FP-2c3301900780000.770.150.360.1900.131.3
FP-2d59029001100000.400.170.300.2500.120.9
FP-2e67400150000.200.060.080.0400.031.8
M-2n/an/an/a000.00620.0780.030000.1
Sc-3n/an/an/a8.34.725160.561.29.41.7
             
April 1, syn-eruptiveFP-3a31468.000.0740.0400.0420.04500.0211.2
FP-3b39764800.270.100.100.05300.0352.1
FP-3c44832500.300.160.110.05800.0382.5
M-3n/an/an/a000.00370.0840.10000.04
             
April 13, post-eruptiveFP-4a35477.900.2500.050.04300.0251.7
FP-4b2253161.03.11.05.040.4100.151.1
W1–2.6n/an/an/a0.554.911211.27.3160.4
W2–22.5n/an/an/a2.02533484.89.9210.8
W3–51.5n/an/an/a4.04562899.116350.9
W4–57.5n/an/an/a4.24662919.417380.8
W5–80.9n/an/an/a6.04574102111.59.81.5
W6–57.5n/an/an/a4.14560919.418380.8
W7–2.6n/an/an/a0.184.72.08.80.852.86.30.3

[40] The best evaluation of results can be made by comparing ratios (Table 4) rather than absolute concentrations. This is due to the fact that the absolute concentrations of different species depend on the sampling distance from the source, and meteorological factors (i.e., the extent of dilution by ambient air, and wind direction relative to the sampling site).

Table 4. Molar Ratios in the Gas and Aerosola
DateSample TypeGasAerosol/Scavenged SpeciesAerosol/Gas
SO2/HClSO2/HFHCl/HFSO42−/ClSO42−/FCl/FNa+/K+SO42−/SO2Cl/HClF/HF
  • a

    Aerosols: suspended, water dissolved and absorbed on scoria. FP – filter pack; n/a – not applicable.

March 24FP (n = 3)1.7 ± 0.352.2 ± 0.231.3 ± 0.400.13 ± 0.0260.17 ± 0.0171.3 ± 0.342.2 ± 0.681.4 × 10−310.2 × 10−311.3 × 10−3
Impactorn/an/an/a0.20--1.0n/an/an/a
Scorian/an/an/a1.5 ± 0.11.7 ± 0.10.11 ± 0.147 ± 1.3n/an/an/a
            
March 28FP (n = 5)4.0 ± 0.4022 ± 6.25.4 ± 1.10.16 ± 0.12--1.8 ± 0.310.022 × 10−30.5 × 10−30
Impactorn/an/an/a0.18--2.6n/an/an/a
Scorian/an/an/a1.11.00.548.5n/an/an/a
            
April 1FP (n = 3)0.37 ± 0.240.68 ± 0.491.8 ± 0.250.48 ± 0.094--1.6 ± 0.584.4 × 10−33.1 × 10−30
Impactorn/an/an/a0.12--2.7n/an/an/a
            
April 13FP (n = 2)0.23 ± 0.0910.47 ± 0.341.9 ± 0.730.33 ± 0.20.96 ± 0.403.0 ± 1.56.8 ± 1.30.06333.5 × 10−322.7 × 10−3
Water (n = 5)n/an/an/a0.52 ± 0.0502.9 ± 0.295.7 ± 0.9816 ± 0.50n/an/an/a

4.1. Gas Composition

[41] The gaseous SO2/HCl and SO2/HF molar ratios increased between March 24–28 (by a factor of 2 and 10, respectively) and then decreased by 1–2 orders of magnitude between March 28 and April 1 (Table 4). The drop in SO2 gas emissions on April 1 is correlated with the decreased eruption vigor at the first eruptive fissure. This is supported by spectroscopic gas flux measurements of Burton et al. [2010], which showed that on April 1 only a third of the total SO2 flux was emitted by the first fissure.

[42] The HCl/HF molar gas ratio is higher in our in-plume measurements (1.8–5.4) than the Cl/F ratio in melt inclusions and groundmass of the erupted tephra (0.6–0.7) [Moune et al., 2012]. However, our results are broadly in agreement with those obtained by spectroscopical measurements in the plume [Burton et al., 2010]. Burton et al. [2010]observed that the HCl emissions were highly variable between the sampling days, and report molar HCl/HF ratios between 2 and 6. The discrepancies between the petrological and in-plume measurements suggest that the degassing behavior and/or near-vent processing of the emissions at Fimmvörðuháls eruption are not straightforward.

[43] Variations in S/Cl ratios between syn- and post-eruptive gas and aerosol phases are shown inFigure 3. The composition of gas emitted from the cooling lava flow (collected on April 13) had a similarly low SO2/halogen molar ratio as the gas emitted from the eruptive fissure on April 1 (Table 4), although it must be noted that the number of samples collected at the cooling lava flow is limited. SO2 is generally less soluble than HCl and HF in silicate melt and therefore tends to exsolve at greater pressure [Carroll and Webster, 1994]. For this reason, plumes emitted from more degassed magma bodies generally have lower SO2/halogen ratios [Allard et al., 2005; Burton et al., 2003; Martin et al., 2008; Naughton et al., 1975].

Figure 3.

Molar ratios of sulfur and chlorine species in emissions at Fimmvörðuháls. ‘Water’ refers to the composition of a stream which comes in contact with cooling lava. Units are: post-eruptive gas -μmol m−3; syn-eruptive gas -μmol m−3; aerosol -μmol m−3; water – mmol l−1. Trendlines are shown for data sets which have >2 points, i.e., syn-eruptive gas (blue), syn-eruptive aerosol (green) and post-eruptive water (red).

4.2. Aerosol Composition

4.2.1. Composition and Size Distribution in the Plume

[44] The aerosol phase contained Cl, Na+, SO42−, K+ and Ca2+, in decreasing order of abundance (mol%, Table 3). Mg2+ and Fwere detected on March 24 only in bulk aerosol collected by filter packs and not in the size-segregated samples; it is likely that the concentrations in each size bin were too low to be detectable individually. The charge balance (Table 3) showed slight anion excess (ratio of ∼1.5), likely due to the presence of H+ in the aerosol. A contribution from other metal cations is not excluded, but is unlikely to account for a significant proportion of the soluble aerosol. Concentration of H+ was therefore calculated based on the charge balance (Table 3).

[45] The size-resolved composition of aerosol is shown inFigure 4. Sulfate was concentrated in fine particles sized 0.2–0.3 μm on all sampling days. Chloride was found predominantly in the 0.2–0.3 μm size bin on March 24, and in a coarse mode only (>18 μm) on March 28 and April 1. The size distribution of Na+ and K+ correlated with both SO42− and Cl. On March 24 the concentration peak was ∼0.2–0.3 μm, and on March 28 and April 1 the size distribution was bimodal (∼0.2–0.3 μm and >18 μm).

Figure 4.

Size-resolved aerosol composition. Abundance of each ion has been normalized to the total molar concentration across all size bins.

4.2.2. Sulfate

[46] Normalization of concentrations removes the effect of plume dilution and allows a more straightforward comparison between samples (although it requires an assumption that the measured species contribute ∼100% of the total composition). Figure 5 compares the normalized compositions in aerosol from March 24, 28 and April 1. The abundance of sulfate increased throughout the three days of sampling when normalized to the bulk composition (4, 10 and 20 mol% of bulk, respectively, Figure 5). Sulfate abundance did not have a linear correlation with the abundance of SO2 gas. The daily average SO42−/SO2 molar ratios for March 24, 28 and April 1 are 1.4 × 10−3, 2 × 10−5 and 4.4 × 10−3 respectively (Table 4).

Figure 5.

Normalized composition of chemical species in the aerosol. Concentration of H+ was not measured directly but calculated from the charge balance of other ions.

4.2.3. Halogens

[47] The abundance of Clin the syn-eruptive aerosol had a very strong correlation with SO42− (R2 = 0.81). The SO42−/Cl molar ratio (Table 4) was, however, one of the lowest reported for a volcanic plume (Table 5). Possible reasons for this are explored further in section 5.2.

Table 5. The SO42−/SO2 and SO42−/Cl Molar Ratio in Plumes From Different Degassing Volcanoes, as Reported in Several Recent Studiesa
VolcanoSO42−/SO2SO42−/Cl
Fimmvörðuháls0.00005–0.0010.3
Kīlauea0.0016–0.006 (1)24 (1)
Lascar0.012(2)43(2)
Masaya0.006–0.012 (3,4)5.0 (4)
Erebus-0.3 (5)
Etna0.01–0.05 (6)16 (6)
Villarica0.023(2)170(2)

[48] The F/HF ratio was similar to the Cl/HCl ratio (Table 4), but it is worth noting that Fwas only detected in two syn-eruptive, and one post-eruptive aerosol sample. The reason for the low abundance of F in the suspended aerosol phase is not clear. One possible explanation might be that F preferentially adsorbs onto silicate particles [e.g., Oskarsson, 1980], and is therefore efficiently removed from the suspended aerosol mass.

4.2.4. Metals

[49] The coarse particle size mode (>18 μm) was composed of Cl, Na+ and K+ in the inferred form of NaCl and KCl, as these were the most abundant ionic species with similar size distributions. The fine particle size mode (∼0.2–0.3 μm) was composed of SO42−, Na+ and K+ on March 28 and April 1 (inferred mode Na2SO4 and K2SO4), as well as Cl (March 24 only). This size mode was somewhat finer than the most abundant mode (∼0.5–1.0 μm) at Masaya, Etna and Kīlauea [Martin et al., 2008, 2011; Mather et al., 2003], but similar to Erebus volcano [Ilyinskaya et al., 2010].

4.2.5. Scoria Leachates

[50] Leachates of scoria had higher SO42−/halogen molar ratios than the non-adsorbed aerosol by almost an order of magnitude (mean values of 1.3 and 0.15, respectively;Table 4). This large difference suggested a real chemical fractionation between the two types of samples. It is unlikely that this was an artifact of analytical calibration as the scoria leachates and filter extracts were analyzed in the same run. The Cl/Fmolar ratio was approximately an order of magnitude lower in the scoria leachates than in the non-adsorbed aerosol (0.1 and 1.3, respectively), as well as in the gas phase (HCl/HF molar ratio, mean value of 2.8). As stated in 4.2.3, the reason for this might be preferential scavenging of F by tephra.

4.3. Surface Water Chemistry

[51] The stream running through the cooling lava showed elevated concentrations of all the measured ions compared to the water in the nearby ‘uncontaminated’ streams (Table 3). The most abundant anion by mass was SO42− followed by Cl and F. Molar ratios of SO42−/Cl and SO42−/F(0.52 and 2.9 respectively) were similar to, or higher than in the gas phase (0.23 and 0.47, respectively) and aerosol phase (0.33 and 0.96, respectively), which were measured above a high-T vent in the cooling lava. The ionic concentrations in the water were plotted against water temperature (Figure 6) and temperature was used here as a direct proxy for distance from the lava front (Figure 2). The concentrations of SO42−, Cl, F, Na+ and K+ had a linear relationship with distance from the lava front, increasing closer to the lava (Figure 6). This suggests that the temperature and chemical trends were controlled by the same mechanism - dilution of hot, alkali-rich (Na+ and K+) stream water with colder, alkali-poor water. The concentration of Ca2+ and Mg2+ showed no such correlation with low values in the sample collected closest to the lava flow front (although still elevated above the ambient concentrations in the ‘uncontaminated’ streams).

Figure 6.

The relationship between the concentrations of (left) anions (mmol l−1; SO42− Cl, F) and (right) cations (mmol l−1; Na+, K+, Ca2+, Mg2+) in stream water, and the water temperature (°C). The temperature of the steam water decreased with distance from the lava. Symbols which are not filled with color represent background samples (collected in streams not in contact with lava).

4.4. HSC and E-AIM Modeling Results

[52] The HSC Chemistry model was used to explore the relative abundance of SO42− with varying temperature and mixing ratio of atmospheric and magmatic gases. Figure 7shows the modeled temperature of a magmatic-atmospheric gas mixture (which decreases with increasingVA/VM), and the SO3/SO2 molar ratio which was used here as a proxy for the SO42−/SO2ratio in the plume. If we assume that the high-T chemical equilibration is quenched at the time of sampling, these calculations show that a plume with SO42−/SO2 = 0.001 was quenched at a lower VA/VM ratio than a plume where SO42−/SO2 = 0.01.

Figure 7.

Modeled temperature of the magmatic-atmospheric gas mixture (green lines) and the quenching conditions constrained by three fixed SO3/SO2 molar ratios (blue lines). The initial temperature of the magmatic gas was selected as 1243 K (1170°C) [Keiding and Sigmarsson, 2012]. The SO3/SO2 molar ratio is used as a proxy for the SO42−/SO2 molar ratio in the volcanic plume. The figure shows that a plume which contains 0.1% SO42−/SO2 was quenched at a lower VA/VM ratio than a plume with 1% SO42−/SO2. The figure also shows that the considered variation in the atmospheric temperature (Tatm) has an insignificant effect on the temperature of the magmatic-atmospheric gas mixture at lowVA/VMratios. If the near-source sulfate is primarily formed in the high-T region of the volcanic plume, it is unlikely to be affected by the ambient atmospheric temperature.

[53] The E-AIM model [Wexler and Clegg, 2002; Martin et al., 2012a] was used to investigate the gas-to-aerosol transformation as a function of ambient atmospheric temperature. The model results predict that condensation of HCl from gas to aerosol in volcanic plumes will be enhanced by lower ambient temperatures (Figure 8). The Cl in the aerosol increases at low-T conditions for two reasons: (1) decreasing T causes more water to condense, so to maintain the Henry's law constant (kH = [HCl]/pHCl) more HCl has to enter the aerosol to balance water uptake; and (2) the value of kHalso increases at low-T due to the decreasing vapor pressure of HCl [Wexler and Clegg, 2002].

Figure 8.

Thermodynamic partitioning of Cl between gas and aerosol (solid and liquid) phases as a function of ambient temperature (calculations made by E-AIM IV model, relative humidity was set at 60%). Note that the gas and aerosol phases are plotted on separate axes. The proportion of Cl decreases in the gas phase and increases in the aerosol phase at lower ambient temperatures.

5. Discussion

5.1. Effect of the Eruptive State

[54] One of the primary objectives of our study was to investigate how the chemical speciation of aerosol varies with differences in eruptive activity. For this purpose, we sampled the eruptive plume during stages of different eruptive activity: explosive activity with highly vigorous lava fountaining (March 24); ongoing explosive activity but a decreased lava fountain height (March 28); weakly spattering, or quiescently degassing lava pond (April 1). Equilibrium models were then used to investigate the chemical speciation further in relationship to the eruption vigor. Our interpretations and conclusions are discussed below.

5.1.1. Sulfate Formation

[55] An important source of sulfuric acid (H2SO4) in near-source plumes may be high-T oxidation of SO2 to SO3 by atmospheric O2 (R1), followed by reaction with H2O (R2) [Allen et al., 2002; Cadle et al., 1971; Mather et al., 2006]. Sulfuric acid is highly hygroscopic and is rapidly converted to aqueous aerosol (R3).

display math
display math
display math

[56] The oxidation reaction depends on the mixing temperature of the atmospheric and magmatic gases as the equilibrium constant for SO3 formation increases with temperature (because (R1) is an endothermic process (NIST Standard Reference Database Number 69, http://webbook.nist.gov/chemistry). Assuming that most of the near-source sulfate forms through high-temperature SO2oxidation, our field data suggests that the efficiency of high-temperature SO2 oxidation appears to decrease with increased eruption vigor. This is in agreement with the only other study which has measured SO42−/SO2 ratios through different eruptive stages [Naughton et al., 1975]. Naughton et al. [1975] compared episodes of lava fountaining and quiescent degassing at Kīlauea volcano and found that the SO42−/SO2molar ratio decreased by an order of magnitude (from 0.016 to 0.0013) during periods of increased activity. Based on our field and modeling results, we propose the following mechanism: during a quiescent state, a certain amount of atmospheric gases is able to penetrate into the high-T region of the vent. At very low VA/VM(≪1), the atmospheric-magmatic gas mixture remains at high, near-magmatic temperatures, which efficiently oxidizes some of the SO2 to SO3. However, during explosive eruptions, the exit velocity increases and atmospheric air is excluded from the vent region. Upon exiting the vent, the plume is very rapidly diluted and cooled. The mixture is thus quenched (i.e., high-T chemical equilibration ceases) and formation of SO3 is inhibited. The SO42−/SO2 ratio at Fimmvörðuháls was generally low compared to that typically measured in plumes of quiescently degassing volcanoes (Table 5), which further supports our interpretation.

5.1.2. Metal Speciation

[57] Metals are believed to be released from the magma as gaseous halogen-compounds (e.g., NaCl(g)) and condense into aerosol as metal halide salts immediately below magmatic temperatures [Oskarsson, 1980; Symonds and Reed, 1993]. If this aerosol subsequently reacts with liquid H2SO4 at lower temperature, the increased acidity may cause revolatilization of halides (e.g., HCl) and instead form metal sulfate aerosol (which exist either as a solution or salt, depending on composition and relative humidity [Martin et al., 2011; Naughton et al., 1974; Toutain et al., 1995]). Coarse-sized (>1μm) metal-sulfate particles have been found in plumes of e.g., Masaya [Martin et al., 2009] and Kīlauea (E. Ilyinskaya et al., Aerosol in emissions from Pu‘u ‘Ō‘ō and Halema‘uma‘u craters at Kīlauea volcano, Hawai‘i, 2007–2009, submitted to Chemical Geology, 2012), however, in the Fimmvörðuháls plume, there was no evidence for metal-sulfate particles aerosol larger than 1μm (Figure 4). The low abundance of SO42− may limit the extent to which it may displace and revolatilize HCl from particles. Aerosol collected on March 24 has Cl primarily in the 0.2–0.3 μm size bin (Figure 4). It is likely that the particularly low SO42−/Cl ratio on March 24 was not sufficient to displace Cl from the particles. On March 28 and April 1 ionic balance calculation shows no evidence for H+ in the fine aerosol mode as the [SO42−] was balanced by [Na+] and [K+]. This suggests that the displacement reaction (R4) went to completion in the fine aerosol mode as the SO42−/Cl ratio increased. Other possible explanations include primary emission of Na2SO4 and K2SO4from the magma, although magmatic emissions of sulfate-species are unlikely to be in high concentrations, since sulfate is largely unstable at magmatic temperatures according to e.g., thermodynamic modeling bySymonds and Reed [1993] and field observations by Naughton et al. [1975].

display math

[58] The ratio of Na+/K+ in the aerosol appeared to be higher in the emissions from the cooling lava flow than at the active fissure (7 and 2, respectively, Table 4). In comparison, aerosol and gas measurements at Kīlauea volcano [Naughton et al., 1975] revealed an order of magnitude increase in the Na/K ratio with decrease in explosivity (from 0.57 during lava fountaining to 4.7 during passive degassing of a lava lake). On the other hand, aerosol samples collected at Etna [Gauthier and Le Cloarec, 1998] did not show a significant change in the Na/K ratio with varying volcanic activity (comparing lava flows, Strombolian explosions and quiescent degassing). There is evidently much scope to investigate the reasons for these apparent discrepancies in metal abundances with changes in eruptive activity. It is possible that melt-vapor fractionation of the two alkali metals differs with varying conditions at source; experimental and/or modeling may elucidate this process.

5.2. Effect of the Ambient Atmosphere

[59] The SO42−/Cl molar ratio was one of the lowest reported for a volcanic plume (see Table 5 for comparison with other volcanoes). The only notable exception is Erebus volcano in Antarctica [Ilyinskaya et al., 2010]. The dominance of Clin the aerosol phase at Erebus could be attributed to the unusually halogen-rich composition of the magmatic gases compared to most other volcanoes (SO2/HCl molar ratio <1; see compilation in Pyle and Mather [2009]). However, the gas composition at Fimmvörðuháls was more SO2 rich than that at Erebus (molar SO2/HCl ratio ∼3, which is of the same order of magnitude as at e.g., Masaya and Etna) and therefore could not easily explain the observed Cl/SO42− enrichment in the aerosol phase.

[60] Our thermodynamic calculations predicted that condensation of HCl from gas to aerosol in volcanic plumes will be enhanced by lower ambient temperatures (Figure 8). The abundance of the near-vent SO42− aerosol, on the other hand, is shown to be largely unaffected by the fluctuations in the ambient temperature (Figure 7). We conclude that the condensation of HCl gas to Cl aerosol is more responsive to a change in ambient temperature than the transformation of SO2 to SO42−. This causes a low SO42−/Cl ratio in volcanic aerosol emitted into the cold atmosphere (<0.5 at Fimmvörðuháls and Erebus).

[61] It has been previously suggested that the ambient conditions are an important factor in particle growth [Mather et al., 2003], and that they control the degree of volatile absorption on ash [Oskarsson, 1980]. The conditions during the Fimmvörðuháls eruption were similar to Erebus (approximately −15°C and 50–60% RH). It is noteworthy that the particle size modes were noticeably finer at Erebus and Fimmvörðuháls than elsewhere. Low ambient temperatures may inhibit particle growth, as well as a relatively high metal/H+ ratio since H2SO4 is more hygroscopic than Na2SO4 [Martin et al., 2011], although further work is needed to establish the exact mechanisms.

5.3. Aerosol Scavenging

5.3.1. Adsorption on Tephra Surface

[62] The chemical composition of species adsorbed on scoria surfaces had a significantly higher sulfate/halogen ratio than the aerosol sampled in-plume (Table 4). Erupted tephra clasts lose heat by radiation and forced convection, and grain size is the most important control on the cooling rate [Thomas and Sparks, 1992]. Lapilli-sized scoria, with a low surface area to volume ratio, therefore retains heat for significantly longer than aerosol, ash, or gases. The temperature of scoria clasts can remain >500°C for several 10s of minutes [Thomas and Sparks, 1992]. Tephra deposition during lava fountaining is very rapid compared with Plinian eruptions, where clasts are lifted to ≫1 km above ground level. Experimental work has shown that while the adsorption of halogens onto the surface of tephra is negligible above ∼600°C, the adsorption rate increases linearly with falling temperatures until ∼200°C [Oskarsson, 1980]. The high-T adsorption behavior of SO2gas is largely unknown but preliminary experimental results show that it may adsorb at significantly higher temperatures than halogen-bearing gases (P. Delmelle, personal communication, 2010). If this is the case, SO2will be preferentially adsorbed in the high-T region of the plume. Another possibility is that hot scoria might provide a reaction surface for heterogeneous oxidation of SO2 to SO42−, resulting in a higher SO42−/halogen ratio than the suspended aerosol mass. More experimental and/or modeling work is needed to explain these processes. The low Cl/Fmolar ratio in the scoria leachates (compared to the non-adsorbed aerosol, and the gas phase) is not fully explained. It is possible that it is caused by preferential adsorption of F onto silicate particles [e.g.,Oskarsson, 1980]. It may also partially account for the significant fluctuations seen in the gas phase ratios.

5.3.2. Scavenging by Surface Waters

[63] The water stream interacting with the cooling lava did not appear to preferentially scavenge Cl over S compared to the post-eruptive gas and aerosol composition (Table 4 and Figure 3). Proportions of S and F species were also fairly similar between the stream water and the in-plume components, although the number of data points was very limited (Figure 3). Scavenging by liquid water may be compared to that studied by Edmonds et al. [2003]who compared in-plume gas measurements with the composition of rain falling through the plume. The results ofEdmonds et al. [2003]agree with ours, as they noted no significant difference between S/Cl ratios in the gas phase and in the rainwater. Studies which directly compare the composition of the gases and aerosol in-plume with that of scavenged volatiles are sparse, and therefore only tentative conclusions can be made at this point. It appears that the S/Cl ratio in volcanic emissions is not affected when scavenged by condensing water. Samples of snow and rainfall are therefore likely to provide a good proxy for the in-plume S/Cl volatile ratios.

6. Conclusions

[64] Syn-eruptive emissions were measured at an active lava fountaining fissure, and during residual degassing of a cooling lava flow (Eyjafjallajökull volcano flank eruption, March – April 2010). Syn-eruptive gas emissions had the highest S/Cl molar ratio (average of ∼2) in the gas phase. The S/Cl ratio was found to be significantly lower in the post-eruptive gas emissions (∼0.2), although the number of samples was limited. The water flowing through cooling lava did not appear to preferentially scavenge Cl over S compared to the post-eruptive gas and aerosol composition. Volatiles adsorbed on the surface of scoria had significantly higher SO42−/halogen molar ratios than the aerosol samples.

[65] Sulfate is one of the most abundant and environmentally important species in volcanic aerosol, and this study advances our understanding of its formation. Our measurements showed that the SO42−/SO2 ratio appears to decrease with increasing eruption vigor. We suggest that the increasing eruption vigor causes less efficient SO2 to SO42− oxidation as ambient air is excluded from the vent region.

[66] It was also demonstrated that although the known majority of volcanic plumes generated by open-vent magmatic degassing are dominated by sulfate, chloride can be the most abundant species under certain conditions (Eyjafjallajökull and the previously reported Erebus volcano). Similarly, the particle size modes were noticeably finer at these two volcanoes than elsewhere. Using a thermodynamic model we showed that low SO42−/Cl ratio (<1) in volcanic aerosol may be a characteristic feature of volcanic plumes in cold atmospheres. These results highlight that the ambient atmospheric conditions are an important factor in the aerosol formation within a minute after magmatic emission.

Acknowledgments

[67] The authors are deeply grateful to the Institute of Earth Sciences at the University of Iceland for generous assistance with field- and laboratory work. In particular, we would like to thank Magnús Tumi Guðmundsson, Andri Stefánsson and Hanna Kaasalainen. Emmanuel Pagneux is thanked forFigure 2. Talfan Barnie is kindly acknowledged for providing 4 photographs in Figure 1. Lucas Laursen is thanked for assistance in the field and Guðrún Nína Petersen for proofreading. E.I. thanks the Cambridge Gates Trust for fieldwork and research funding. R.S.M. thanks Christ's College, University of Cambridge for a research fellowship. C.O. acknowledges generous support via NERC grant NE/F001487/1 and the National Centre for Earth Observation “Dynamic Earth and Geohazards” project (http://comet.nerc.ac.uk/). The authors are grateful to T. Thordarson and three anonymous reviewers for their comments which greatly improved the manuscript.