Atmospheric Chemistry in Volcanic Plumes

Atmospheric chemistry in volcanic plumes

Roland von Glasow1

School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ, United Kingdom

Edited by Barbara J. Finlayson-Pitts, University of California, Irvine, Irvine, CA, and approved February 22, 2010 (received for review November 15, 2009)

Recent field observations have shown that the atmospheric plumes The model setup is based on observations for Mount Etna, Sicily, of quiescently degassing volcanoes are chemically very active, during nonexplosive stages (3, 7, 8, 9). The crater altitude of the pointing to the role of chemical cycles involving halogen species modeled volcano is 3,300 m; the modeled plume is nonconden- and heterogeneous reactions on aerosol particles that have pre- sing. It is important to mention that most volcanic emissions are viously been unexplored for this type of volcanic plumes. Key into the free troposphere with significantly increased lifetimes features of these measurements can be reproduced by numerical compared to emissions in the boundary layer. A plume that is models such as the one employed in this study. The model shows emitted into the boundary layer from a low-altitude volcano sustained high levels of reactive bromine in the plume, leading to might show significantly different behavior, partly due to the extensive ozone destruction, that, depending on plume dispersal, potential direct contact of the plume with the surface and can be maintained for several days. The very high concentrations of vegetation. sulfur dioxide in the volcanic plume reduces the lifetime of the OH Initial Plume Composition radical drastically, so that it is virtually absent in the volcanic plume. This would imply an increased lifetime of methane in volcanic Where volcanic volatiles and air meet there is a very rapid transi- plumes, unless reactive chlorine chemistry in the plume is strong tion from reducing to oxidizing conditions resulting in a very enough to offset the lack of OH chemistry. A further effect of bro- rapid change of the composition of the gas-particulate mixture. mine chemistry in addition to ozone destruction shown by the For an assessment of the atmospheric effects of volcanic emis- model studies presented here, is the oxidation of mercury. This re- sions, it is crucial to be able to quantify the initial plume compo- lates to mercury that has been coemitted with bromine from the sition. Experimentally it is rather difficult to measure the exact volcano but also to background atmospheric mercury. The rapid composition of volcanic volatiles, and in the literature the oxidation of mercury implies a drastically reduced atmospheric life- majority of the data about volcanic volatiles that are not already time of mercury so that the contribution of volcanic mercury to the influenced by atmospheric processing stem from measurements atmospheric background might be less than previously thought. at fumaroles. Gerlach (10) suggested that the assumption of ther- However, the implications, especially health and environmental ef- modynamic equilibrium might be applicable for high-tempera- fects due to deposition, might be substantial and warrant further ture mixtures of volcanic volatiles and ambient air and studies, especially field measurements to test this hypothesis. proposed, based on results of a thermodynamic equilibrium model [HSC chemistry (11)] that “high-temperature reaction of mag- halogen chemistry ∣ mercury chemistry ∣ oxidation capacity matic gases with air and/or in-plume heterogeneous chemical processes involving aerosols during plume transport” might be raditionally, emissions of volcanoes were regarded to be required to explain observed high mixing ratios of halogen oxides mainly of importance for the atmospheric sulfur cycle, acid in volcanic plumes. Bobrowski et al. (3) and Roberts et al. (12) T based their modeling studies on this idea. In the following, I refer deposition, and stratospheric effects from explosive eruptions. “ ” This view changed recently, mainly because of the observation to an effective source region where high-temperature volcanic of very high concentrations of bromine oxide, BrO, in the tropo- volatiles are assumed to be mixed with a certain volume fraction spheric, nonexplosive plume of Soufrière Hills, Montserrat (1). of air and where the assumption of thermodynamic equilibrium Since then, an increasing number of observations of reactive holds for this mix of hot gases. The addition of oxygen dramati- halogens have been made in noneruptive plumes of various vol- cally changes the composition in the effective source region. canoes (2–4). Many more measurements of the early plumes of The model used here is an improved version of the model used various craters including the chemical composition of aerosol (5) in ref. 3; please see Materials and Methods for more details. The in recent years showed many more facets of the exciting chemical main measurements that are available for a model evaluation with regard to reactive halogen chemistry are, in addition to processes in noneruptive volcanic plumes. For more details and measurements of precursor gases at the crater rim, the remote further references, see, for example, von Glasow et al. (6). It is sensing of bromine oxide, BrO, and sulfur dioxide, SO2, at varying now clear that there is potential for a substantial influence of distances from the crater as performed by, e.g., Oppenheimer nonexplosive volcanic plumes on atmospheric chemistry, which et al. (2) and Bobrowski et al. (3). The variability in SO2 fluxes is why it is necessary to come to a quantitative assessment of these from volcanoes is very large, so that ratios of gases are often used. effects. This paper aims at contributing to this assessment. The main relevance of reactive chemistry in volcanic plumes, In this case, the ratio of the vertical columns (VC) of BrO and and especially halogen chemistry, is changes to the budgets of SO2 could be derived from the field measurements. Fig. 1 shows ozone and other oxidants, the contribution of volcanic bromine the modeled VC of BrO and SO2 and their ratio as well as to the observed free tropospheric background of reactive bro- measurements from Mount Etna; the range of the absolute values and the ratio of the measurements by Bobrowski et al. (3) can be mine, and changes to the atmospheric mercury burden. Ozone reproduced in the model. This is true only for the “base run” is a key driver of atmospheric oxidation and an important green- house gas in the troposphere. Mercury, in its methylated form, is highly toxic. Author contributions: R.v.G. designed research; R.v.G. performed research; R.v.G. analyzed This paper uses a state-of-the-art numerical model to investi- data; and R.v.G. wrote the paper. gate the implications of reactive chemistry in the plumes of quies- The authors declare no conflict of interest. cently (i.e., nonexplosive) degassing volcanoes, with a focus on This article is a PNAS Direct Submission. mercury chemistry. In this paper “volcanic plume” refers to the 1To whom correspondence should be addressed. E-mail: [email protected]. mixture of volcanic volatiles and particulates with ambient air; This article contains supporting information online at www.pnas.org/cgi/content/full/ only the atmospheric evolution of the plume is considered. 0913164107/DCSupplemental.

6594–6599 ∣ PNAS ∣ April 13, 2010 ∣ vol. 107 ∣ no. 15 www.pnas.org/cgi/doi/10.1073/pnas.0913164107 Downloaded by guest on September 29, 2021 scenario (see Materials and Methods); a model run that does not additional bromine has to be “activated” from other bromine re-

employ the idea of a modified initial plume composition in the servoirs in order to explain the continued increase in the BrO to SPECIAL FEATURE effective source region (“pure volcanic volatiles”) shows an in- SO2 ratio. The only relevant gas phase loss of HBr is reaction with crease in BrO∶SO2 ratios that lags behind the measurements. OH; however, the model results show that under all conditions In a model run without aerosol chemistry, neither the BrO VC studied, the core of the plume is OH-free, due to its very rapid nor the BrO∶SO2 ratio can be reproduced, highlighting the loss by reaction with SO2. In this case, HBr either has to react importance of these processes. It has to be stressed that the varia- with OH at the plume edge, where ambient air is in contact with bility in the measurements is very large, partly caused by the the plume or it has to be taken up on aerosol particles, where it variability in emissions and partly by atmospheric variability. reacts with HOBr (stemming from the gas phase) and acidity to The latter includes wind speed, which determines the time that Br2, which is very insoluble and is released to the gas phase where the plume was exposed to atmospheric processing before getting it photolyzes, producing two Br radicals (see, e.g., ref. 14 for a to a certain measurement site, and the presence/absence of general overview of atmospheric halogen chemistry): clouds or a condensed plume, which has a strong influence on heterogeneous chemistry inside the plume. Almost all published BrO þ HO2 → HOBr þ O2; [1] measurements of BrO in volcanic plumes lack such ancillary meteorological observations. HBr → HBraq; [2] Bromine Chemistry in the Plume

Of the gas phase bromine compounds that are expected to be − þ present in volcanic plumes only HBr (e.g., ref. 7) and BrO have HOBraq þ Br þ H → Br2;aq þ H2O; [3] been measured so far. Typically it is assumed that most bromine is released in the form of HBr; however, the concept of an effective → ; [4] source region implies that significant amounts of either Br radi- Br2;aq Br2 cals and/or BrCl and Br2 would be present in the early plume (see, e.g., ref. 10 and Table S2). Roberts et al. (12) and Martin þ hν → 2 ; [5] et al. (13) highlighted the potential of Br radicals and suggested Br2 Br that the initial presence of large amounts of Br radicals would SCIENCES lead to a rapid production of BrO. However, in order to explain 2ð þ Þ → 2ð þ Þ; [6] Br O3 BrO O2 ENVIRONMENTAL an increase of the BrO to SO2 ratio with time, as observed at Mount Etna, the initial presence of Br radicals is not sufficient, þ net: HO2 þ HBr þ 2O3 þ H þ hν → BrO þ 3O2 þ H2O: [7] BrO VC [1014 molec/cm2] HOBr is formed in the gas phase from the reaction of BrO with 1.6 3.7 2.0 4.8 HO2; therefore, the rapid release of bromine from HBr requires 1.2 both the presence of particulate acidity, which is available in co- 0.8 pious amounts in a volcanic plume, and the presence of HO2. Under background conditions, most HO2 is produced in reac- 0.4 tions involving OH. In the model, HO2 concentrations are in- 0.0 creased above background values by a factor of about 1.7× at -5 0 5 15 25 35 45 55 the crater rim (corresponding to 1 min after plume release in 18 2 SO2 VC [10 molec/cm ] the model) but drop to about 8% of the background values within 6.0 10 min after plume release. The initial increase is due to the mod- 5.0 el initialization from HSC, which suggests rather high levels 4.0 of HO2. 3.0 The speciation of bromine is strongly dependent on time. Fig. 2 2.0 shows the speciation of bromine in the core of the plume of the 1.0 base run. In this run, the Br radical accounts for almost a quarter 0.0 of the total bromine at model start. Due to the presence of large -5 0 5 15 25 35 45 55 amounts of NO in the early plume as suggested by the HSC cal- -4 culations, the majority of the Br radicals react with NO2, which BrO/SO2 [10 ] has been rapidly formed from the titration of O3 by NO: 4.0

4.4 3.0 NO þ O3 → NO2 þ O2; [8]

2.0 Br þ NO2 → BrNO2; [9] 1.0

0.0 þ hν → þ : [10] -5 0 5 15 25 35 45 55 BrNO2 Br NO2 time since plume release [min] This reaction sequence happens immediately after plume release. Fig. 1. Temporal evolution of the vertical columns of BrO, SO2, and The model output after 1 min represents crater-rim conditions so their ratio. The black, solid line is for the base run, the green, dash-dotted that the initial high levels of Br are not visible in Fig. 2, as Br has line is for pure volcanic volatiles, and the red, dashed line is for a run without already reacted to BrNO2. The only loss of BrNO2 is photolysis aerosol chemistry. The x axis is time in minutes since plume release. The dots are for measurements at Mount Etna, taken from ref. 3. The “time with a lifetime of about 150 sec, rereleasing Br and NO2; BrNO2 since plume release” has been estimated with the assumption of a windspeed therefore acts only as a very short-term reservoir. Because of the of 7 m∕s for all measurements. The three red dots depict concurrent large amounts of NO2 in the plume, it is, however, reformed measurements. continuously. Eventually, the concentration of NO2 is reduced

von Glasow PNAS ∣ April 13, 2010 ∣ vol. 107 ∣ no. 15 ∣ 6595 Downloaded by guest on September 29, 2021 1.0 ∶ BrO mixing ratio. This implies that the ratio of BrO Brtot Br might depend not only on the magma composition but also on 0.8 the wind speed and crater shape, as these parameters influence the initial plume dilution, which has a strong influence on plume tot ∶ 0.6 BrO chemistry. This is shown in Fig. 4, which shows the BrO Brtot Br2 ratio for a number of model runs, employing different magma compositions and initial plume dilution ratios. 0.4 BrNO2 It is important to point out the limitations of these results. The fraction of Br plume near the crater rim will invariably be heterogeneous; how- 0.2 ever, the model simulations presented here assume an idealized, HBr cone-shaped (Gaussian) plume. Heterogeneities will be stronger at high wind speed, when turbulence around the crater increases 0.0 1 5 15 25 35 45 55 as this implies much stronger mixing in the early plume stages time since plume release [min] than considered in this study, which would result in a stronger entrainment of ambient air into the plume. This would increase Fig. 2. Relative speciation of bromine in the core of the plume (base run). the O3 concentration in the plume and thereby change the bro- Shown are the relative contributions of several bromine compounds to the mine speciation, leading to higher BrO mixing ratios. The strong sum of gas phase bromine. Note that the concentration of all bromine com- separation between core center and core edges in the model runs pounds decreases during this time due to mixing with background air. The x axis is time in minutes since plume release, starting with minute 1, which presented here might not occur in reality or only at smaller scales, represent crater-rim conditions that are already significantly different from as turbulent mixing might lead to several smaller plumes that the model initialization. might be more prone to mixing with ambient air. The current model studies should therefore predominantly represent situa- significantly due to horizontal plume dispersal, drastically slowing tions with limited turbulence at the crater rim. As the plume down the production of BrNO2. As the O3 levels have not recov- is typically in the turbulent region around the crater only for a few minutes, turbulent mixing is quickly reduced in the plume, ered at this point in the model, the fraction of Brtot that is present as Br radical is very large; hence, BrNO2 is still the main bromine- which makes comparison of the model results with measurements bearing compound. As can be seen from Fig. 2, BrO accounts for easier. As soon as the plume is outside the region of turbulence at most 15% of total gas phase bromine in the base-run model run influenced by the crater rim, mixing is drastically reduced and the in the first hour after plume release but BrNO2 accounts for more plume maintains its distinct form for a long distance from the than 30%. HOBr never contributes significantly to Brtot. The rea- crater (16). When comparing model results with crater-rim mea- son for this is the very fast uptake on acidic aerosol particles; the surements of, e.g., O3, one should expect a less strong O3 loss in production rate of HOBr is similar to the uptake rate on aerosol the plume but a rather large variability in its measured concen- particles. trations as some parts of the plume will be strongly and others In the model, ozone is quickly destroyed in the volcanic plume, barely diluted by ambient air. mainly due to the self-reaction of BrO, which contributes 84% to Some measurements have shown the presence of large O3 destruction in the first hour and 90% in the first six hours. amounts of ClO and OClO in the near-crater volcanic plume of, Because of the high levels of the Br radical in the plume, O3 re- e.g., Mount Etna (3). Current numerical models simulating halo- covers only at night, when most reactive bromine is present as gen chemistry in volcanic plumes (3, 12) are not capable of re- Br2, so that mixing of O3 from the plume edges into the core producing this quantitatively; the current study is no exception. can actually increase the O3 levels (see Fig. 3). Very little infor- The measurement of chlorine oxides, especially ClO, however, mation about ozone concentrations in volcanic plumes is avail- is technically difficult with passive remote sensing techniques able due to cross-sensitivities of most instruments with SO2. such as MAX-DOAS [Multi Axis Differential Optical Absorption Hobbs et al. (15) have used a chemiluminescence technique and Spectroscopy (3)] because the spectral features of ClO overlap found that ozone was reduced to 10% of its ambient value in the with those of SO2, and a recent study employing active remote plume of Mount Saint Helens. If the total amount of bromine sensing [DOAS (4)] did not detect ClO or OClO above their de- that is released is comparably small or if very strong initial tection limit. See also ref. 12 for a discussion of this topic. The dilution occurs that reduces the concentration of bromine, O3 implications of a potential underestimate of reactive chlorine destruction will be less pronounced and therefore lead to a higher chemistry in the model are the following:

Fig. 3. Contour plots of the RGM to TGM ratio as well as SO2,O3, and BrO for the mode run including reduction of RGM by SO2. The x axis is time since model start; for clarity, only the plume development after the first six hours is shown. Model start is at noon.

6596 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0913164107 von Glasow Downloaded by guest on September 29, 2021 quantified the global importance of bromine chemistry on mer- 0.6 0.5 cury and confirmed the proposition of Holmes et al. (27) that it is SPECIAL FEATURE

tot 0.4 a major oxidant for mercury; however, the inclusion of the pres- 0.3 sure dependence of key reactions in ref. 26 lead to a lower esti- 0.2 mate of the impacts of bromine chemistry than the one in ref. 27, BrO:Br 0.1 namely, that bromine chemistry would reduce the overall lifetime 0.0 of mercury by about 10%. 1 5 15 25 35 45 55 Even though the temperatures in the plumes of quiescently de- time since plume release [min] gassing volcanoes are very different from those encountered in the polar boundary layer, the main “ingredients” for rapid mer- Fig. 4. Ratio of BrO to total gas phase bromine. Black, solid line: base run; × cury destruction are present: reactive halogens. There are still im- red, dashed line: composition from effective source region without the 10 portant uncertainties in our understanding of mercury reaction dilution factor as used in the base run; green, dotted line: dilution factor of 100×; dark blue, long-dashed line: magma composition for “arc mean” com- kinetics, which are obviously present in this study as well. The position (for details, see SI Text); light blue, solid line: pure volcanic volatiles, i. reaction mechanism for mercury is listed in SI Text. The key re- e., no background air present in effective source region. The x axis is time in actions for rapid bromine induced oxidation of GEM and uptake minutes since plume release. to the condensed phase are

0 þ → ; [11] 1. CH4 reacts very quickly with the chlorine radical. The pre- Hg Br HgBr sence of large amounts of chlorine radicals in volcanic plumes would imply a reduction of the lifetime of CH4, whereas in the HgBr → Hg0 þ Br; [12] current model setup the lifetime of CH4 is increased compared to the background due to the virtual absence of the OH radical in the plume, caused by the high SO2 concentra- HgBr þ Br → HgBr2; [13] tions. Cl atom concentrations would have to be at least 5.8% of those of ambient OH for the whole lifetime of the plume ¼ 287 (calculated for T k), in order to offset the lack of CH4 HgBr þ HgBr → HgBr2 þ Hg; [14] oxidation of OH. Assuming a noontime OH concentration of SCIENCES 107 molecules∕cm3, this would require 5.8 × 105 atoms∕cm3. ↔ ; [15] ENVIRONMENTAL 2. Ozone destruction. High levels of ClO in the plume would lead HgBr2 HgBr2;aq to a further increase in O3 destruction. In the model runs presented here, the effect would be negligible, as O3 is already þ − ↔ −; [16] very efficiently destroyed by bromine chemistry. HgBr2;aq Br HgBr3 Mercury Chemistry − − 2− Volcanoes have for a long time been identified as an important HgBr3 þ Br ↔ HgBr4 : [17] natural source of mercury (see, e.g., refs. 17, 18 for recent inventories). Most of the earlier studies, however, have measured only This sequence is nonlinear as two bromine radicals are involved total mercury and have not made a distinction between gaseous for the production of stable HgBr2, and four bromine atoms are 0 2− elemental (GEM or Hg ) and reactive gaseous mercury [RGM or required for the formation of HgBr4 . The same pathway is active Hg(II)]. Only recently have speciated mercury measurements for chlorine. Reaction [12] is fast as shown by Goodsite et al. (28), been made at volcanic craters (19–22). According to Dedeur- highlighting the need for high levels of the Br radical for the waerder et al. (23), about 62% of the measured mercury resided production of RGM in order to compete with the dissociation in the particulate phase at the rim of a crater of Mount Etna, of HgBr. As shown above, the model suggests that a rather large whereas more recent studies at Masaya, Telica (both Nicaragua), fraction of bromine is present in the early stages of volcanic and Mount Etna (19, 21) showed that most mercury at the crater plumes as a Br radical making a rapid oxidation of GEM feasible. rim resides in the gas phase; in these studies particulate mercury Lohman et al. (29) and Seigneur et al. (30) found strong indica- accounted only for 1–5% by mass of total mercury. During their tions for the reduction of RGM by SO2 in power plant plumes. measurements at Mount Etna (plume ages of a few seconds to a The concentration of SO2 in volcanic plumes is obviously much few minutes), Bagnato et al. (19) determined the speciation of higher than in power plant plumes, so the following net reactions mercury at the crater rim on two occasions and found that were included here: RGM contributed only 1–2% to total gaseous mercury (TGM). þ → þ þ ; [18] This is consistent with the measurements of Witt et al. (21) at HgBr2 SO2 Hg Br2 SO2 Masaya and Telica volcanoes who found a contribution of RGM to TGM of 1%. Bagnato et al. (19) also performed calculations HgCl2 þ SO2 → Hg þ Cl2 þ SO2; [19] with a thermodynamic equilibrium model (HSC) and showed that the speciation of mercury in the effective source region is domi- with a rate coefficient of k ¼ 6.0 × 10−17 molecules∕ðcm3 sÞ [em- 0 nated by Hg , which exceeds the next important mercury com- pirical upper limit (30)]. The inclusion of these reduction reac- pound by more than 3 orders of magnitude. Aiuppa et al. (20) tions, especially the one for HgBr2, leads to a significant shift discussed measurements in the vicinity of a fumarole at La Fossa in mercury speciation; the first minutes after plume release Crater on Vulcano, Italy, and found indications for rapid chemi- are comparable to speciation measurements (19, 21) only if these cal processing of mercury in the plume. This rapid oxidation of reactions are included. Fig. 5 (Left) depicts the mercury specia- mercury was recently confirmed by ref. 22 for several fumarole tion in the plume for the first hour after plume release, showing plumes on Vulcano. that the model can reproduce the measurements that indicate It has been known since the pioneering study of Schroeder that near the crater almost all TGM is present as GEM. Upon et al. (24) that bromine compounds can lead to rapid oxidation dilution, the reduction reaction with SO2 gets less important, of elemental mercury in the polar boundary layer. A large num- leading to a rapid buildup of RGM. ber of follow-on studies has confirmed this as a frequently occur- In order to investigate the atmospheric relevance of the pre- ring phenomenon in polar regions (25). Seigneur et al. (26) dicted rapid transformation of volcanic mercury into RGM and

von Glasow PNAS ∣ April 13, 2010 ∣ vol. 107 ∣ no. 15 ∣ 6597 Downloaded by guest on September 29, 2021 1.0 1.0

0.8 0.8

RGM tot GEM tot GEM 0.6 0.6

0.4 part-Hg 0.4 fraction of Hg

fraction of Hg RGM 0.2 0.2

part-Hg 0.0 0.0 -5 0 5 15 25 35 45 55 0 10 20 30 40 50 60 70 minutes since plume release hours since plume release

Fig. 5. (Left) Evolution of mercury speciation in the first hour after plume release, vertically averaged over the plume. (Right) Evolution of mercury speciation for 72 h after plume release, vertically averaged over 300 m around the plume. The x axis is time since plume release, which is at noon.

particulate mercury, the model run was continued for 72 h after mosphere by precipitation. This leads to a much reduced atmo- plume release. The mercury speciation for this time period is spheric lifetime of volcanic mercury but, as background mercury shown in Fig. 5 (Right), however, not averaged over the vertical is affected by the plume chemistry as well, also of background extent of the plume (as in the left panel of that figure) but over GEM. Based on these results, the contribution of volcanic mer- 300 m with the plume located in the upper third in order to in- cury to the atmospheric GEM pool should be reassessed. The clude sedimentation effects. The mercury speciation shows a contribution to the atmospheric mercury loading will be much diurnal variation, highlighting that key atmospheric reaction se- reduced, in some areas, where the volcanic plume is condensing quences involving mercury are actually reversible and dependent and/or exposed to precipitation, volcanic plumes might actually on photochemistry. This graph shows that the extent of halogen act as a net sink for atmospheric mercury, if not only volcanic induced oxidation of mercury is not restricted to the mercury but also background mercury (that has been oxidized by volcanic emitted by the volcano itself but that also GEM in the back- halogens) is removed from the atmosphere. ground air is being oxidized. This is also visible in Fig. 3, showing This has repercussions not only for the atmospheric budget of a contour plot of the RGM to TGM ratio, which increases with mercury but also for the environment and human health, due to time also outside of the core plume area due to slow vertical mix- the toxicity of the heavy metal mercury once it has been methy- ing of halogens out of the core of the plume. The vertical mixing lated. Deposition of oxidized mercury downwind of a volcano that in the model is, however, not enough to provide a flux of O3 that emits appreciable amounts of halogens and mercury might be is strong enough to counter the rapid destruction by halogen substantial and might pose a risk to human health for populations chemistry. living in close proximity of such volcanoes. The plumes of some The previous discussion was based on a model start at local low-altitude volcanoes can ground for several 10 km downwind of noon, when actinic fluxes peak. In order to show that the effects the plume, so that RGM might deposit directly on vegetation or discussed here are not dependent on photochemistry in the first food crops. The exact effects are hard to assess at this point as few hours after plume release, the model scenarios with the re- deposited oxidized mercury can be reduced and rereleased to the duction of RGM by SO2 has been run 12 times, with the start atmosphere. See, e.g., ref. 31 for a recent compilation of review delayed by 2 h for each run. Fig. 6 shows 24-h averages of the papers about the global cycle of mercury and its environmental first 24 h of plume evolution. The 24-h averaging shows a consis- and health effects. tent and strong transformation of GEM into RGM. The ratio of Experimental verification of the hypothesis of rapid mercury RGM∶TGM in each of these 12 runs has a very consistent day- oxidation in volcanic plumes is possible by making airborne time maximum of about 0.83 and a nighttime minimum of 0.03. mercury speciation measurements in volcanic plumes as well as ground-based measurements or analysis of vegetation samples Implications of Rapid Mercury Oxidation exposed to grounding volcanic plumes. These measurements GEM has an atmospheric lifetime of about a year, and previous would allow a quantification of the hypothesized mechanism, studies about volcanic mercury emissions mainly considered the which is very important, given the large uncertainties in our contribution of volcanic mercury emissions to the atmospheric understanding of chemical processes in volcanic plumes and in GEM pool. Based on the results shown, I feel that this view has mercury chemistry. to be revised. The atmospheric lifetime of RGM and particulate In order to further our understanding of the global mercury mercury is much shorter than that of GEM and is mainly deter- cycle, it seems imperative to improve our kinetic understanding mined by deposition. As shown in Fig. 5, about 2–10% (three-day of the key chemical reactions. Furthermore, if the hypothesis put average of 7.8%) of the total mercury in a 300-m vertical layer forward here is correct, one would have to consider the relative around the volcanic plume is present as particulate mercury source strength of mercury, halogens, and SO2 in order to yield and 7–46% (three-day average of 23%) as RGM. Both par- improved estimates of the net contribution of volcanic emissions ticulate mercury and RGM can easily be scavenged from the at- to the atmospheric mercury pool.

Fig. 6. Twenty-four-hour average of RGM to TGM ratio as well as SO2; see the text for more details. The x axis is time since model start.

6598 ∣ www.pnas.org/cgi/doi/10.1073/pnas.0913164107 von Glasow Downloaded by guest on September 29, 2021 Materials and Methods cycling of mercury, especially the interactions with halogens, is still limited

In this study an established one-dimensional model of multiphase chemistry so that the uncertainties in the kinetics for mercury are higher than for SPECIAL FEATURE (32) was used originally developed for studies of the marine boundary layer the other elements. A listing of the mercury reactions can be found in but also successfully used for the investigation of chemical processes in vol- SI Text. The background GEM mixing ratio was assumed to be canic plumes (3, 33). A major focus is on the interaction of gas and particulate 0.18 pmol∕mol, which, at Mount Etna’s summit height, corresponds to about phase (aerosol and cloud droplet) chemistry. Radiative fluxes as well as 1.1 ng∕m3. The base run assumes a mixture of 85% volcanic volatiles and photolysis frequencies are calculated online so that effects of particles on 15% air in the effective source region for which the equilibrium composition the radiative transfer can be explicitly modeled. The chemical mechanism is calculated with HSC [(11); for Mount Etna conditions, see SI Text). In order contains the most important reactions of O, H, C, N, S, Cl, Br, I, and Hg species to be able to reproduce measured crater-rim SO2 mixing ratios, this gas mix is both in the gas and particulate phases. The current model version is signifi- diluted 10× with ambient air. The sensitivity studies presented in Figs. 1 and 4 cantly improved compared to the one used in Bobrowski et al. (3). Most are for runs without the assumption of entrainment of ambient air into the important are an improved treatment of horizontal dispersion and the initial effective source region (“pure volcanic volatiles”), for a run without aerosol composition of the volcanic plume [gaseous, see below, and aerosol after chemistry, for a 100× dilution with ambient air, and for a different magmatic (8, 9)], an update of chemical reactions and the inclusion of mercury composition (“arc mean”;seeSI Text). chemistry. The mercury mechanism is based on previous modeling studies but somewhat extended based on a number of laboratory studies and com- ACKNOWLEDGMENTS. I thank Sandro Aiuppa, Alex Baker, Nicole Bobrowski, pilations of mercury reaction rate coefficients with the goal of having a Christoph Kern, Tamsin Mather, Tjarda Roberts, Leif Vogel, and an anon- rather comprehensive gas and aqueous phase reaction mechanism and ymous reviewer for helpful discussions and/or comments on this manuscript. ensuring that the mechanism is “closed” in the sense of having sinks for This work was partly funded by the Deutsche Forschungsgemeinschaft all intermediate species, which was not always the case in previous model (DFG): Emmy-Noether Junior Research Group MarHal, GL 353-1/1,2, and is studies (28, 34–36). Our quantitative understanding of the atmospheric a contribution to the IGAC/SOLAS task “Halogens in the Troposphere.”

1. Bobrowski N, Hönninger G, Galle B, Platt U (2003) Detection of bromine monoxide in a 20. Aiuppa A, et al. (2007) Real-time simultaneous detection of volcanic Hg and SO2 at La volcanic plume. Nature 423:273–276. Fossa Crater Vulcano (Aeolian Islands, Sicily). Geophys Res Lett 34:L21307 doi: 10.1029/ 2. Oppenheimer C, et al. (2006) BrO formation in volcanic plumes. Geochim Cosmochim 2007GL030762. Acta 70:2935–2941. 21. Witt MLI, et al. (2008) Mercury and halogen emissions from Masaya and Telica 3. Bobrowski N, et al. (2007) Reactive halogen chemistry in volcanic plumes. J Geophys volcanoes, Nicaragua. J Geophys Res 113:B06203 doi: 10.1029/2007JB005401. Res 112:D06311 doi: 10.1029/2006JD007206. 22. Zambardi T, Sonke J, Toutain J, Sortino F, Shinohara H (2009) Mercury emissions and 4. Kern C, et al. (2008) Halogen oxide measurements at Masaya volcano, Nicaragua using stable isotopic compositions at Vulcano Island (Italy). Earth Planet Sci Lett active long path differential optical absorption spectroscopy. Bull Volc doi: 10.1007/ 277:236–243.

s00445-008-0252-8. 23. Dedeurwaerder H, Decadt G, Baeyens W (1982) Estimations of mercury fluxes emitted SCIENCES 5. Mather TA, Allen AG, Oppenheimer C, Pyle DM, McGonigle AJS (2003) Size-resolved by Mount Etna volcano. Bull Volc 45:191–196. characterisation of soluble ions in the particles in the tropospheric plume of Masaya 24. Schroeder WH, et al. (1998) Arctic springtime depletion of mercury. Nature ENVIRONMENTAL volcano, Nicaragua: Origins and plume processing. J Atmos Chem 46:207–237. 394:331–332. 6. von Glasow R, Bobrowski N, Kern C (2009) The effects of volcanic eruptions on atmo- 25. Steffen A, et al. (2008) A synthesis of atmospheric mercury depletion event chemistry spheric chemistry. Chem Geol 263:131–142. in the atmosphere and snow. Atmos Chem Phys 8:1445–1482. 7. Aiuppa A, et al. (2005) Emission of bromine and iodine from Mount Etna volcano. 26. Seigneur C, Lohman K (2008) Effect of bromine chemistry on the atmospheric mercury Geochem Geophy Geosy 6:Q08008 doi: 10.1029/2005GC000965. cycle. J Geophys Res 113:D23309 doi: 10.1029/2008JD010262. 8. Watson IM, Oppenheimer C (2001) Photometric observations of Mt. Etna’s different 27. Holmes CD, Jacob DJ, Yang X (2006) Global lifetime of elemental mercury against aerosol plumes. Atmos Environ 35:3561–3572. oxidation by atomic bromine in the free troposphere. Geophys Res Lett 33:L20808 9. Allen AG, et al. (2006) Sources, size distribution, and downwind grounding of aerosols doi: 10.1029/2006GL027176. from Mt. Etna. J Geophys Res 111:D10302 doi: 10.1029/2005JD006015. 28. Goodsite ME, Plane JMC, Skov H (2004) A theoretical study of the oxidation of Hg0 to 10. Gerlach TM (2004) Volcanic sources of tropospheric ozone-depleting trace gases. HgBr2 in the troposphere. Environ Sci Technol 38:1772–1776. Geochem Geophy Geosy 5:Q09007 doi: 10.1029/2004GC000747. 29. Lohman K, Seigneur C, Edgerton E, Jansen J (2006) Modeling mercury in power plant 11. Roine A (2007) HSC chemistry 6.1. Tech. rep. Outotec Research Oy. plumes. Environ Sci Technol 40:3848–3854. 12. Roberts TJ, et al. (2009) Modelling reactive halogen formation and ozone depletion in 30. Seigneur C, Vijayaraghavan K, Lohman K (2006) Atmospheric mercury chemistry: volcanic plumes. Chem Geol 263:151–163. Sensitivity of global model simulations to chemical reactions. J Geophys Res 13. Martin RS, Roberts TJ, Mather TA, Pyle DM (2009) The implications of H2S and H2 111:D22306 doi: 10.1029/2005JD006780. stability in high-T mixtures of magmatic and atmospheric gases for the production 31. Pirrone N, Mason Robert, eds. (2009) Mercury Fate and Transport in the Global Atmo- – of oxidized trace species (e.g., BrO and NOx). Chem Geol 263:143 150. sphere (Springer, Dordrecht, The Netherlands). 14. von Glasow R, Crutzen PJ (2007) Tropospheric halogen chemistry. Treatise on 32. von Glasow R, Sander R, Bott A, Crutzen PJ (2002) Modeling halogen chemistry in the Geochemistry Update 1, eds HD Holland and KK Turekian (Elsevier-Pergamon, Oxford), marine boundary layer. 1. Cloud-free mbl. J Geophys Res 107:4341 doi: 10.1029/ vol.4.02, pp 1–67. 2001JD000942. 15. Hobbs PV, Tuell JP, Hegg DA, Radke LF, Eltgroth MW (1982) Particles and gases in the 33. Aiuppa A, et al. (2007) The tropospheric processing of acidic gases and hydrogen emissions from the 1980–1981 volcanic eruptions of Mt. St. Helens. J Geophys Res sulphide in volcanic gas plumes as inferred from field and model investigations. Atmos 87:11062–11086. Chem Phys 7:1441–1450. 16. Tiesi A, et al. (2006) Estimation of dispersion coefficient in the troposphere from 34. Ryaboshapko A, et al. (2002) Comparison of mercury chemistry models. Atmos Environ satellite images of volcanic plumes: Application to Mt. Etna, Italy. Atmos Environ 36:3881–3898. 40:628–638. 35. Hedgecock IM, Trunfio GA, Pirrone N, Sprovieri F (2005) Mercury chemistry in the MBL: 17. Nriagu J, Becker C (2003) Volcanic emissions of mercury to the atmosphere: Global and Mediterranean case and sensitivity studies using the amcots (atmospheric mercury regional inventories. Sci Tot Env 304:3–12. chemistry over the sea) model. Atmos Environ 39:7217–7230. 18. Pyle DM, Mather TA (2003) The importance of volcanic emissions for the global 36. Ariya PA, Peterson K, Snider G, Amyot M (2009) Mercury chemical transformation in atmospheric mercury cycle. Atmos Environ 37:5115–5124. the gas, aqueous and heterogeneous phases: State-of-the-art science and uncertain- 19. Bagnato E, et al. (2007) Degassing of gaseous (elemental and reactive) and particulate ties. Mercury Fate and Transport in the Global Atmosphere, eds N Pirrone and R Mason mercury from Mount Etna volcano (southern Italy). Atmos Environ 41:7377–7388. (Springer, Dordrecht, The Netherlands), pp 459–501.

von Glasow PNAS ∣ April 13, 2010 ∣ vol. 107 ∣ no. 15 ∣ 6599 Downloaded by guest on September 29, 2021