Observation and modelling of ozone-destructive halogen chemistry in a passive degassing volcanic plume

. Volcanoes emit halogens into the atmosphere that undergo chemical cycling in plumes and cause destruction of ozone. The impacts of volcanic halogens are inherently difﬁcult to measure at volcanoes, and the complexity of the chemistry, coupled with the mixing and dispersion of the plume, makes the system challenging to model numerically. We present aircraft observations of the Mount Etna plume in the summer of 2012, when the volcano was passively degassing. 5 Measurements of SO 2 — an indicator of plume intensity — and ozone were made in the plume a few 10s of km from the source, revealing a strong negative correlation between ozone and SO 2 levels. From these observations we estimate a mean in-plume ozone loss rate of 1 . 3 × 10 − 5 molecules of O 3 per second per molecule of SO2. This value is similar to observation-derived estimates reported very close to the Mount Etna vents, indicating continual ozone loss in the plume up to at least 10’s km downwind. 10 The chemically reactive plume is simulated using a new numerical 3D model “WRF-Chem Volcano” (WCV), a version of WRF-Chem we have modiﬁed to incorporate volcanic emissions (including HBr and HCl) and multi-phase halogen chemistry. We used nested grids to model the plume close to the volcano at 1 km. The focus is on the early evolution of passively degassing plumes aged less than one hour and up to 10’s km downwind. The model reproduces the so-called ‘bromine explosion’: the daytime bromine activation process by which HBr in the 15 plume is converted to other more reactive forms that continuously cycle in the plume. These forms include the radical BrO, a species whose ratio with SO 2 is commonly measured in volcanic plumes as an indicator of halogen ozone-destroying chemistry. We track the modelled partitioning of bromine between its forms. The model yields in-plume BrO/SO 2 ratios (around 10 − 4 mol/mol) similar to those observed previously in Etna plumes. The modelled BrO/SO 2 is lower in plumes which are more dilute (e.g. at greater windspeed). It is also slightly lower in plumes in the middle of the day compared than in the morning 20 and evening, due to BrO’s reaction with diurnally varying HO 2 . Sensitivity simulations conﬁrm the importance of near-vent products from high temperature chemistry, notably bromine radicals, in initiating the ambient temperature plume halogen cycling. Note also that heterogeneous reactions that activate bromine also activate a small fraction of the emitted chlorine; the resulting production of chlorine radical Cl causes a strong reduction in the methane lifetime and increasing formation of HCHO in the plume. 25 following that of Roberts et al. (2018), with an extrapolation made for the smallest bins. Chlorine and bromine emission ﬂuxes are speciﬁed based on the observed summit 280 chlorine-to-bromine ratio and HCl-to-SO 2 ratio and the SO 2 ﬂux speciﬁed above. We rely on a comprehensive compositional analysis undertaken between June 2010-June 2012 by Wittmer et al. (2014). The chlorine-to-bromine ratio was ﬁxed to 300 by mass (683 by mole), which is an average calculated for the Bocca Nova and North-East craters’ compositions reported by Wittmer et al. (2014). The HCl-to-SO 2 ratio was set to 0.4 mol/mol which is about mid-way in the range of ratios (0.29 to 0.56) for these craters.

Br includes one photolysis reaction and three other reactions. Jourdain et al. (2016) modelled Ambrym plume and found halogens' depletive effect on HO x to further increase the lifetime of SO 2 with respect to oxidation by OH by 36%, and that Cl radicals reduced the in-plume lifetime of CH 4 . Modelling by 95 Roberts et al. (2009) indicates that volcanic halogen chemistry can result in conversion of in-plume NO x to HNO 3 .
Finally, volcanoes are also sources of mercury to the atmosphere (Pyle and Mather, 2003) mainly in the inert form Hg(0) (Witt et al., 2008;Bagnato et al., 2007). A 1D model study by von Glasow (2010) suggested that this mercury could be rapidly oxidised by halogen chemistry in a volcanic plume to more soluble forms, easily removed from the atmosphere (Seigneur and Lohman, 2008). Significant advances in understanding of the kinetics of halogen-mercury chemistry have been made in the 100 last decade (e.g. Saiz-Lopez et al., 2018, 2019 and these are included in the modelling part of this study.

Observation and modelling studies
The current understanding of halogen chemistry within volcanic plumes is based upon a body of observations that have used a variety of techniques, coupled with numerical modelling results, most of which have used zero-or one-dimensional chemical box models.

Observations
There are two main methods for measuring halogens in volcanic plumes, remote sensing and in-situ sampling.
Remote sensing accounts for most observations of reactive halogens in volcanic plumes. Since the first reported detection by Bobrowski et al. (2003), bromine monoxide (BrO) has been observed within the plume of dozens of volcanoes by differential optical absorption spectroscopy (DOAS) (see Gutmann et al. (2018) for a recent catalogue of such observations). A smaller 110 number of measurements of in-plume ClO and OClO have also been reported. These halogen molecules have spectroscopic signatures within the ultraviolet range, meaning they can be identified from the same data that is used to monitor SO 2 , including data collected from long-term DOAS monitoring networks at volcanoes (Dinger et al., 2018;Warnach et al., 2019). As well as ground and airbourne observations, BrO has been observed in the plumes of some larger volcanic eruptions by satellite-based instruments (e.g. Hörmann et al., 2013;Seo et al., 2019), though such large eruptions are the focus of a future study rather than 115 this one.
In-situ sampling of halogens provides the most direct approach to quantify total halogen emissions: time-averaged sampling has for decades been used to quantify total volcanic halogen emission contents for F, Cl, Br and I (e.g. Aiuppa et al. (2004); Wittmer et al. (2014)). Modern techniques now allow for a degree of speciation in the bromine observed through these methods (Rüdiger et al., 2017(Rüdiger et al., , 2020. For most reactive halogen species, these methods required samples to be collected in-situ and then 120 subsequently analysed in-lab. Consequently, there are fewer in-situ observations of reactive halogens than by remote sensing. As well as these direct approaches, ozone measurements can provide indirect evidence for halogen chemistry. Ozone destruction in tropospheric volcanic plumes, caused by volcanic halogen cycling, has been measured in a limited number of cases (Vance et al., 2010;Oppenheimer et al., 2010;Kelly et al., 2013;Surl et al., 2015). In ash-rich explosive eruptions, it is possible that uptake of ozone on ash particles may also contribute to some ozone loss (Maters et al., 2017). Measuring ozone in volcanic 125 plumes in the troposphere region downwind from volcanoes is challenging and typically only achieved using instrumented aircraft. Another difficulty lies in attributing observed ozone losses to halogen chemistry when volcanic bromine emissions and/or bromine radical levels are not well-known. Observations suggest a direct relation between bromine content and ozone depletion (Roberts, 2018), although this is based on only three available volcanic datasets: Mount Etna, Mount Redoubt and Kilauea.
Observations at Mount Redoubt volcano suggest that ozone losses, as a ratio to SO 2 , increase in magnitude with respect to the 130 distance from the source (Kelly et al., 2013).
For the above methods, the observed halogen gas quantities or ozone depletions are often ratioed to simultaneous sulfur or SO 2 measurements. This allows for comparison between plumes of different "strengths" (i.e. density or dilution), and, for example, to trace how halogen chemistry changes as a plume disperses as it travels downwind. This use of SO 2 as a plume tracer presupposes that it has a long atmospheric lifetime relative to the timescale of the given study.

140
Lagrangian models are either zero-or one-dimensional and simulate the chemical evolution of the cooled plume by calculating the in-plume rates of chemical reactions and include the continual dilution of the plume with background air. Such models are found to better reproduce observations if the initial halogen emissions include a fraction of halogen radicals. This represents the radicals generated by high temperature chemistry in the effective source region.
To our best knowledge, Jourdain et al. (2016) is the only prior 3D Eulerian-type mesoscale chemistry-transport modelling 145 study published to date dealing with halogen chemistry a tropospheric volcanic plume. Volcanic emissions and halogen chemistry were implemented into the CCATT-BRAMS model to simulate the chemistry withing the plume of Ambrym volcano during an intense passive degassing episode in 2005. Their model is similar to the one used in our study. However, their gas emission flux for the Ambrym event is about six times greater than the Mount Etna passive degassing event studied here.
Mechanistically, their results showed the formation of BrO, as well as ozone depletion occurring within the plume's core that 150 impacts bromine speciation The study also simulated in-plume depletion of HO x and NO x , as well as lengthening of SO 2 and methane lifetimes due to halogen chemistry. The model successfully reproduced observed BrO/SO 2 spatial patterns, however the magnitude was somewhat underestimated, and there were no measurements of ozone to provide constraints on the predictions of ozone depletion, a key feature of reactive halogen chemistry. Finally, Jourdain et al. (2016) focuses on the wider-scale impact of volcanic emissions, whereas this study focuses more on the detailed mechanisms of halogen cycling in the early 155 plume with cross-validation against ozone and halogen radical observations.

This study
The present study is devoted to a plume from Etna during July/August 2012, a period when this volcano was passively degassing. We present new airborne ozone and SO 2 measurements which were made during traverses of plumes at distances 7-21 km from the vents. Several other previously published plume measurements were also made around this time. Near-160 simultaneous near-vent (<500 m) ozone measurements and DOAS observations of BrO/SO 2 ratios around 10 km downwind were reported by Surl et al. (2015). Additionally in situ sampling of halogen emissions was undertaken that summer at the crater-rim by Wittmer et al. (2014) and further DOAS measurements were also made of Etna's plume by Gliß et al. (2015).
Consequently, this period of Etna's activity may have the richest overall dataset to date for determining halogen activity in relation to ozone loss in the plume of a volcano.

165
The observational dataset is analysed using a 3D regional chemistry-transport model (WRF-Chem) modified with respect to its handling of volcanic emissions and with halogen chemistry added to a chemical mechanism. The goal of the modelling is to assess the ability of the 3D model to adequately reproduces the key chemistry features (ozone loss, BrO/SO 2 ratios) of the Etna plume given reasonable input parameters, such as the typical halogen emission fluxes for the volcano in a passive degassing state. We then diagnose in-plume chemical processes in the model, exploiting the fact that a model can be interrogated in far 170 greater detail than an observational dataset. The focus is on the chemical processes in the near-downwind plume, up to 10s of km from emissions sources for plume ages of up to 10s min. As well as halogen chemistry and the associated ozone destruction, com/products/oxygen-compound-instruments/t400). Mercury vapour is listed as a potential source of interference; however mercury-detecting instruments were also active during this campaign and the level of gaseous mercury emission from the volcano was determined to be nil or low (Weigelt et al., 2016a). Both the SO 2 and ozone instruments have a temporal resolution 190 of 10s (averaging time) and their response times are 80 seconds and < 30 seconds, respectively.
Three flights were conducted, one each on the mornings of 2012-07-30, 07-31, and 08-01, during daylight hours. These flights started and ended at Reggio Calabria Airport and attempted several transects of the plume. The flight paths are shown in Figure 2.
Since ambient concentrations of ozone vary both spatially and temporally, rather than assessing all of the observation data 195 together, we undertook a systematic approach to identify and isolate separate "major plume encounters" from the dataset, and evaluate separately the ozone variations within these. This approach was designed such that the majority of the variation in ozone within each major plume encounter could be ascribed to plume chemistry rather than variations in the background.
Our approach also fixed a maximum range of distances from the vent that could be considered part of a single major plume encounter so as to minimise any internal variation in ozone losses within a plume encounter due to plume chemistry varying 200 with distance from the source. Ozone varies with altitude so we fixed a maximum range of altitudes that could be considered part of one major plume encounter -this avoids mistaking background ozone variation as the aircraft ascends or descends as a plume signal. Lastly, encounters that are too short or do not reach a sufficiently high plume intensity so as to allow for an identification of signal above background variation are dismissed.
Our algorithm is therefore as follows:  a plume encounter is considered to begin when the SO 2 measurement exceeds 10 ppbv, and ends when SO 2 drops below 10 ppbv -If a datapoint's altitude is more than 300m higher or lower than that of any previous datapoint in the current encounter, the encounter ends and another immediately begins.
-If a datapoint's distance from the source is more than 5 km greater or smaller than that of any previous datapoint in the 210 current encounter, the encounter ends and another immediately begins.
if a plume encounter has maximum SO 2 less than 100 ppbv, or lasts for less than 2 minutes, it is considered a "minor plume encounter" and is discarded This process is presented in full as a flowchart in Figure S1.

215
We use verion 4.1.5 of WRF-Chem (Grell et al., 2005) which is a fully coupled three-dimensional regional model for atmospheric physics, meteorology and chemistry, including cloud and aerosol radiative feedback processes. We have made several modifications to the code, in particular volcanic gas emissions and chlorine/bromine/mercury chemistry. We name this new model "WRF-Chem Volcano" (WCV). Our WCV developments build on the WRF-Chem version developed by the Roland von Glasow group at the University of East Anglia (Surl, 2016) and our WCV developments were made with reference to 220 the model code of Badia et al. (2019), another development on the University of East Anglia version with a focus on marine chemistry.

Mechanism
WCV extends the CBMZ-MOSIAC chemistry scheme with 8 aerosol size bins (Zaveri and Peters, 1999;Zaveri et al., 2008) to include bromine, chlorine, and mercury chemical mechanisms with gas-phase, photolytic, and heterogeneous reactions 225 involving the following species: HBr, Br, BrO, HOBr, BrNO 3 (a.k.a. BrONO 2 ), Br 2 , HCl, Cl, ClO, OClO, HOCl, ClNO 3 , Cl 2 , BrCl, Hg, HgBr, HgCl, HgBr 2 , HgCl 2 , and HgBrCl. We exclude BrNO 2 as previous studies have found it to be a negligible component (Roberts et al., 2014;Rüdiger et al., 2020). These species are also incorporated into the dry-and wet-deposition schemes and the FastJ photolysis scheme (Wild et al., 2000). The rates of heterogeneous reactions involving HOBr, BrNO 3 , and ClNO 3 on volcanic aerosols are calculated on-line accounting for the wet surface area of aerosol and gas-phase diffusion 230 limitations as described by Marelle et al. (in review). The products of HOBr reactive uptake are partitioned between Br 2 and BrCl (i.e. net overall reaction with HBr or HCl) by a parameterization that assumes fast aqueous-phase equilibria between Br 2 , Br 2 Cl -, and BrCl as described by Jourdain et al. (2016). Reactions added to the scheme are listed in the supplement (Tables S1,   S2, S4) along with their rate equations and references for these. Parameters controlling the heterogenous reactions are tabulated in Table S3.

235
Our volcanic emissions pre-processor, a modified version of the the PREP-CHEM-SRC utility (Freitas et al., 2011), provides as inputs to the model fluxes of sulfur, bromine, chlorine, and mercury species as well as an 'at-source' sulfate particle flux, and fluxes of radicals resulting from high-temperature chemistry within the vent (e.g. OH, NO).
We have also introduced artificial tracer species to WCV that do not influence the chemistry but are useful for analysis.
tracer1 is a wholly inert species that is emitted with the same flux rate as SO 2 . tracer1 is used to compute mean (weighted 240 average) in-plume values of parameters -here an in-plume average of a value (e.g. of ozone mixing ratio) refers to the average of this value across all grid boxes where tracer1 exceeds 3 ppbv, weighted by the tracer1 content of the boxes.
tracer2 is similar to tracer1 but undergoes a first-order exponential decay with specified rate. The ratio of tracer1 and tracer2 can therefore be used to derive the mean time-since-emission of any part of the plume. This approach allows us to accurately calculate plume-age in any model grid cell and enables us to monitor how plume parameters (e.g. ozone mixing ratio, BrO/SO 2 245 ratio) vary with the plume time evolution since emission.
We have also added to the model output monitoring of several chemistry diagnostics, such as rates of relevant reactions, in order to carry out species' chemical budgets and therefore facilitate the analysis of the underlying halogen and ozonedestructive chemistry.  Figure 3. The WRF-Chem model area. d02, d03, and d04 are progressively nested domains with two-way nesting.
The total area modelled, at 30 km horizontal resolution, is depicted in Figure 3. Also shown are the extents of the progressively nested domains, each modelled with grid dimension three times smaller than their parent. These domains are two-way nested in exchanging meteorological, chemical, and physical information between them. Here we focus on the near-downwind plume processing (up to about 90 minutes). Therefore all the figures and results presented in this study are from the d04 nest which models an 88 km × 134 km area around the east coast of Sicily with a horizontal spatial resolution of 1.1 km. The model 255 has 50 vertical layers, extending up to 50 hPa.
The model was initialised at 2012-07-29 00:00 and ran for four days, therefore covering all of the days of the aircraft measurements. The first 24 hours are considered spin-up, results are presented from > 24 hours onwards.
The volcano was considered to be a point source of gas and aerosol, emitting at a constant flux rate throughout the simulation period. Although Mount Etna has several active vents (North-east Crater, Voragine, South-East Crater) at the volcano summit, is, and hence the greater this discrepancy is.
Modelled emissions of SO 2 were set to 40 kg s −1 . This flux was estimated by adjusting initial estimates according to comparisons between outputs from preliminary runs and observed SO 2 mixing ratios. A 40 kg s −1 flux results in SO 2 mixing ratios within the centre of the modelled plume being similar to the maximum SO 2 mixing ratios observed from the aircraft at the same distance from the source. We assume these observed maxima correspond to transects which cross or comes close chlorine-to-bromine ratio and HCl-to-SO 2 ratio and the SO 2 flux specified above. We rely on a comprehensive compositional analysis undertaken between June 2010-June 2012 by Wittmer et al. (2014). The chlorine-to-bromine ratio was fixed to 300 by mass (683 by mole), which is an average calculated for the Bocca Nova and North-East craters' compositions reported by Wittmer et al. (2014). The HCl-to-SO 2 ratio was set to 0.4 mol/mol which is about mid-way in the range of ratios (0.29 to 0.56) for these craters.

285
As stated in the introduction, volcanic gases are believed to react at high temperatures immediately following their release in the vent and into the atmosphere, generating radicals, notably HO x and some halogen radicals. These radicals -as well as the primary aerosols --subsequently initiate the onset of the bromine cycling in the cooled plume ( Figure 1). A representation of the high-temperature radicals is therefore needed for the WCV volcanic input. Thermodynamic models have been used previously to represent this high-temperature "effective source region" (Bobrowski et al., 2007) but their assumption of chemical 290 equilibrium is not considered reliable, whereas recently developed kinetic models do not yet include halogens (see Roberts et al., 2019). Here we choose a simpler approach by partitioning the bromine emission flux into hydrogen halide and radicals, with 75% of bromine emitted as HBr and 25% as Br radicals, and by including a HO x emission. This bromine partitioning follows previous thermodynamic modeling estimates (Roberts et al., 2014). Emissions of volcanic HO x are highly uncertain, there exist order-of-magnitude differences between kinetic and thermodynamic model predictions, and in the speciation be- (between reported thermodynamic and kinetic model ranges). The immediate reaction of OH with SO 2 in WCV will generate HO 2 and some additional sulfate. Whilst all volcanic chlorine is emitted as HCl in the model, the reaction with volcanic OH will also quickly generate some Cl radicals.
Although there are open questions regarding the kinetics of high-temperature NO generation in the first few seconds of 300 plume evolution (Martin et al., 2012), we chose to include these emissions to assess its possible effect. We use an NO/SO 2 molar emission ratio of 4.5 × 10 −4 , which is of the order typically produced by high temperature thermodynamic modelling of the early plume-air mix (c.f. 6.6 × 10 −4 used in Roberts et al. (2014)).
Although the aircraft campaign did not find a detectable mercury signal for the plume (Weigelt et al., 2016a), we include a small mercury emission so as to investigate this mechanism. We use a general volcanic emission ratio of 7.8 × 10 −6 mol Hg 305 per mol SO 2 from Bagnato et al. (2014), a quantity too small to significantly interfere with other chemical systems. All of this mercury is emitted as Hg (0).  Emmons et al., 2020) and applied using the MOZBC utility. Since mercury, bromine and chlorine species other than HCl were absent from this CAM-CHEM data, we set their initial and boundary values to zero.
As well as the main model run detailed above, several other model runs were made with varying emissions (Table 2). These include a model run with no volcanic emissions (novolc). Differences between novolc and the other runs are used to quantify volcanic impacts. Sensitivity runs were identical to the main run except for the perturbations to the volcanic emissions listed in 315   Table 2. The nohighT run excludes all species expected to be generated in the high temperature volatile-air mix of the first few seconds after volcanic emission, and therefore includes only H 2 O, SO 2 , HCl, and HBr (all bromine as HBr) as well as primary sulfate.

Aircraft observations quantifying ozone destruction in the plume
The encounter-finding algorithm described in Sect. 2.1 yields 19 major plume encounters, 11 on the 31st July and 8 on the 1st 320 August 2012. Only minor plume encounters occurred on the 30th July. The locations of these major encounters are highlighted in Figure 2. Plots of O 3 vs. SO 2 for these encounters are shown in Figure 4 for the 31st July and Figure 5 for the 1st August.
For the majority of encounters, there is a clear anti-correlation between ozone and SO 2 , with linear fits yielding negative gradients. These observational data show that ozone is depleted within the plume, and this depletion is proportional to the intensity of the plume as quantified by SO 2 measurements.

325
Weighting by the duration of each encounter and their R 2 values, the average O 3 vs. SO 2 gradient for the plume encounters is -0.018 molec molec −1 , and the average distance from the source is 14 km. Assuming that ozone destruction is a continuous process, and that, at distance zero, ozone destruction is zero, these value can be used to quantify the rate of ozone destruction as a ratio of SO 2 per km traveled; the resulting value is 0.0015 molec molec −1 km −1 . This could be converted to a destructionper-second value by dividing by wind speed. No wind speed data was collected during the flights, so to do this we inspect the 330 meteorological output from the model, which yields for both days a wind speed for the plume of approximately 9 m s −1 at the time of the flights. Using this value yields a rate with respect to time of −1.3 × 10 −5 molec molec −1 s −1 . Interestingly, this rate is very close to a value of −1.0 × 10 −5 molec molec −1 s −1 derived from in-situ measurements made within 500 m of the vents on 27-30 July 2012 (Surl et al., 2015), supporting the theory of a continuous process beginning from no ozone depletion at source. 335 We note that an analysis to evaluate the trend in ozone depletion with respect to distance within the dataset yielded a null result: ozone depletion to SO 2 ratios were calculated for each in-encounter datapoint by using the y-axis intersect of each plume encounter as an estimate for background ozone for all datapoints within that encounter. The output from this analysis across the whole dataset was too noisy to discern an overall trend.   In several cases we use plume age as a variable. This is determined from the tracers, as discussed in the Methods. Before investigating the halogen chemistry of the volcano plume, we first look at SO 2 . The volcano emits 40 kg of SO 2 355 per second throughout the simulation. This produces a plume that travels downwind, dispersing (i.e. diluting) during transport -SO 2 mixing ratios decrease with time and distance from the source (Figures 6, 14). The model SO 2 chemistry includes gas-phase oxidation by OH, generating secondary sulfate aerosol.
For plumes aged less than an hour, the modelled mixing ratios of SO 2 and tracer1 in the plume are nearly identical ( Figure   6), indicating that SO 2 losses in this period are negligible. This gives confidence to the use of SO 2 as a plume tracer, though we 360 should note that the model does not contain other potentially significant SO 2 loss mechanisms which occur in the liquid phase (Galeazzo et al., 2018).
Notably, the average mixing ratio of SO 2 in the plume of the 30th July is significantly less than on the other two days, although the trend is similar. This difference is due to the fact that the wind speed on the 30th is much higher than on the other two days: the average wind speeds in the < 60 min old plume at 0800 UTC are 19, 9, and 9 m s −1 for the 30th July, 31st July, 365 and 1st August respectively. Therefore, volcanic emissions are released into a greater volume of air on the 30th, yielding lower mixing ratios. The volcano is a direct source of aerosols, with a flux of 1.2 kg of sulfate per second in the model. This is important for the halogen chemistry as it provides a surface for HOBr uptake, enabling heterogeneous reactions. Figure 7a shows that, shortly after emission, the in-plume ratio of sulfate to SO 2 is slightly above the emission ratio of 0.03 because of the early oxidation produces additional sulfate. This ratio continues to increase with plume age due to ambient temperature oxidation by OH mixing in from background air.
There are similar trends for the aerosol surface area-to-SO 2 ratio (Figure 7b), the increase in this ratio shows that the secondary aerosol formation notably increases the surface area available for HOBr uptake. When considered in absolute terms, 375 this secondary aerosol formation partly offsets the decline in aerosol surface area density caused by plume dispersion. This can be seen on Figure 6 where the aerosol surface area density declines at a slower rate than SO 2 mixing ratio. For the 30th the aerosol surface area density is approximately constant after 10 minutes, indicating that the secondary formation compensates for the dispersion in this regard.
Although the oxidation of SO 2 is not significant over these time scales with regard to SO 2 mixing ratios, the oxidation that 380 does occur is significant for in-plume aerosol and HO x levels. There is a substantial depletion of OH within the plume, and a moderate depletion of HO 2 (Figure 8). This result is consistent with model findings for the Ambrym plume (Jourdain et al., 2016), and occurs despite the modelled volcano being a source of OH -this emitted OH is consumed very quickly.
Volcanic halogens and SO 2 compete for reaction with the available OH. The abundance of SO 2 in the plume results in substantial conversion of OH to HO 2 via the SO 2 + OH reaction. This starts a chain of reactions with short-lived intermediate species that simplified to a single SO 2 + OH HO 2 + H 2 SO 4 reaction in the model (Bekki, 1995;Galeazzo et al., 2018).
Volcanic HCl is also abundant and removes OH by the HCl + OH H 2 O + Cl reaction. A similar reaction of HBr also occurs, but HBr is much less abundant. HO 2 is consumed as part of the bromine cycle in the BrO + HO 2 HOBr reaction. it can be seen that the non-halogen components of the plume are sufficient to cause substantial OH depletion, whilst halogens are the cause of HO 2 depletion -the hal00 plume actually has greater HO 2 compared to the novolc case.
Although the instantaneous lifetime of SO 2 (with respect to oxidation by OH) is substantially increased in the halogen-free model plume, we note that the addition of halogen emissions to the model further suppresses OH, increasing the SO 2 lifetime and having a reductive effect on secondary aerosol production both in terms of mass and surface area (Figure 7). This result 395 for a tropospheric volcanic plume mirrors findings from a recent study of a stratospheric volcanic cloud .
Whilst a detailed analysis of the aerosol microphysics and climate impacts of volcanic aerosols lies beyond the scope of this study, our simulations show substantial differences in the plume sulfate particle surface area density for WCV model simulations with and without volcanic halogen emissions. As plume halogen chemistry exerts an important influence on the oxidation rate of volcanic SO 2 and associated formation of secondary aerosol, our results suggest that models simulating 400 chemistry-climate impacts of volcanic sulfur should not ignore the chemistry of volcanic halogens.

Bromine speciation and BrO/SO 2 ratio
In the model output, the bromine is emitted from the volcano as HBr and Br in a 3:1 ratio. During daylight this is rapidly converted to other forms, including BrO. Figure 9 shows how the forms which this bromine takes vary between plume of different ages at 08:00 UTC on the 30th, 31st and 1st. In this model output, HOBr becomes the dominant form of bromine 405 within the plume, followed by BrO. The fraction as BrO increases over approximately the first 20 minutes before reaching an approximately stable fraction. A significant amount of Br 2 is formed shortly after emission but this fraction declines, with BrCl being the larger fraction of the two halogen dimers. A significant reservoir of BrNO 3 that forms shortly after emission declines slowly over time.
There are significant quantitative differences between the bromine evolution on these three days, although the trends are  HBr and BrNO 3 , and HOBr persist in the plume for much longer on the 30th July. The balance between HOBr and BrO is more greatly tilted towards the former on that day because the in-plume aerosol surface area density is lower on the 30th, reducing the rate of heterogeneous reactions that consume HBr, BrNO 3 , and HOBr. Additionally, the reaction of BrO with background HO 2 to form HOBr is suppressed under more concentrated plume conditions due to the depletion of HO 2 discussed in Sect. Although the bromine speciation appears roughly stable after approximately 30 minutes of evolution, this does not indicate that no further chemistry is occurring, bromine is continually cycled between forms. This is shown in Figure 10 which depicts the rates of transfers between bromine species. "Stability" indicates a state where the chemical formation and loss of each species is approximately balanced, i.e. a steady state.

420
The ratio of total bromine and SO 2 is mostly invariant in the plume, therefore the variations in bromine speciation with plume age yield variations of in-plume BrO/SO 2 ratios. There is an initial rise in BrO/SO 2 , followed by a small decrease in some cases, and then a plateau. This pattern varies with time of day, as shown in Figure 11. Sunrise is shortly after 04 UTC and sunset shortly after 18 UTC. Negligible BrO is formed in the model plume at night. We find on all three days that BrO/SO 2 ratios are generally greater in the morning and evening than during the middle of the day. This occurs because BrO is more rapidly 425 converted to HOBr in the middle of the day when atmospheric HO 2 is at a maximum. Although moderate in magnitude, this phenomenon may be significant when comparing spectroscopic columns at different times, including datasets from low-earth orbit satellites with overpasses at different local times.    this level. This rapid rise, and higher ratio, has better agreement with the observations of the 31st and 1st.
More generally, the shape of the modelled BrO/SO 2 versus distance/time trend seen in Figures 11, 12, and 13a -of an initial rise followed by a steady value -is in agreement with the general trend observed for Etna and other volcanoes (Gutmann et al., 2018).

Importance of high-temperature volcanic products 450
The modelled conversion of HBr into BrO and other forms is much slower in the noHighT run ( Figure S2). This indicates the importance of initial volcanogenic radicals from the "effective source region" in the autocatalytic processes of the bromine explosion. In their absence this process is nonetheless initiated, but it is only partially completed even for 60 minutes plume age. Note however that the model does not have any background bromine, which in the absence of volcanogenic radicals could potentially contribute to initiating this process.

Chlorine species
The uptake to particles and subsequent reaction of HOBr with hydrogen halides (HBr, HCl) has the effect of transferring halogen from halides to reactive forms. When HBr and HCl are both present in the plume the product of this is almost exclusively Br 2 . However after a short time HBr is almost totally depleted in the plume and is consumed as fast as it is produced, whereas HCl remains abundant. The subsequent photolysis of BrCl produces reactive chlorine. This reactive chlorine forms 460 the spectroscopically detectable species ClO and OClO within the plume -the reaction of Cl with ozone produces ClO, and the reaction of ClO with BrO produces OClO. Figure 13 shows the vertical columns of these species within the plume as ratios  Spatially, lower ClO/SO 2 column ratios are found in the more concentrated parts of the plume (and higher ClO/SO 2 at the 465 plume edges). Conversely, OClO/SO 2 shows the opposite spatial pattern, and is highest shortly after emission, because of the initial burst in OClO production from ClO and BrO in the concentrated early plume.
OClO and ClO are rarely observed above instrumental detection limits in volcanic plumes. Our model results are broadly consistent in magnitude with the few reported OClO observations in in Etna plume reaching 1 × 10 −5 (Gliß et al., 2015) whilst General et al. (2014) report OClO/SO 2 up to 10 −4 mol/mol. Some observational studies (Bobrowski et al., 2007;General et al., 470 2014) report greater BrO/SO2 and OClO/SO2 at the plume edges compared to the centre, which is not seen in our model. This might be due to the horizontal spatial resolution of WCV, or it might reflect a real difference in the chemistry of the plume.
Note that box-modeling findings indicate that the magnitude of such 'edge effects' depends on volcanic conditions such as emitted the HBr/SO 2 .
High levels of CH 4 -oxidising chlorine radicals (Cl) in the plume reduce the instantaneous lifetime of CH 4 in the plume, 475 which, in the early plume considered here, more than compensates for the decrease in CH 4 oxidation from the reduced levels of OH. However, at the edges of the plume, the lifetime-extending effect is greater, leading to the spatial pattern seen in Figure   S4. Oxidation of CH 4 produces HCHO, and therefore the plume has elevated mixing ratios of this species ( Figure S5).

Ozone depletion
During the day, ozone levels are lower in the plume than in the surrounding air ( Figure 14). On the 30th, the in-plume ozone 480 loss is only a few ppbv, whereas it reaches of the order of a few 10s of ppbv on the 31st and 1st.
Comparing output from the main and novolc model runs allows for a precise calculation of ∆O 3 , the change in ozone due to the volcano. Following the approach used in Kelly et al. (2013) and Surl et al. (2015), we compute ∆O 3 /SO 2 to isolate the chemical signal from physical dispersion effects. Figure 15 shows the variation of this ratio with plume age at different times of day. Because this ozone destruction is slower on the 30th compared to the other two days, and because the plume is 485 travelling faster due to the greater wind speed, the smaller ozone loss on 30th compared to the other days is even more apparent if ∆O 3 /SO 2 is plotted against distance ( Figure 16). The absence of halogen chemistry means that, at night, ∆O 3 /SO 2 is close to zero. During the day the ratio is negative, and increases in magnitude with plume age. This indicates that chemical ozone destruction in the plume is a continuous, ongoing, process. Although the data depicted in Figure 15 are Eulerian snapshots rather than Lagrangian traces of the plume, these lines' gradients are an indication of the rate of the ozone loss process. For 490 2012-08-01 08:00, the gradient is about −7.5 × 10 −6 molecules of O 3 per molecule SO 2 per second.
Secondly, by inspecting the rates of reaction for the main model run as shown in Figure 10, and computing the differences between the rates of ozone-destructive and ozone-forming reactions, we find that these halogen reactions result in an average net ozone loss rate of 3.1 × 10 7 molec cm −3 s −1 within this part of plume. Dividing this by the weighted average plume SO 2 mixing ratio yields an instantaneous loss rate of of 7.

Attribution of ozone loss to halogen reactions
A detailed analysis of the model outputs allows us to attribute ozone loss to specific bromine reaction cycles. Ozone is destroyed by its reaction with Br to form BrO, but the net ozone loss depends on the subsequent fates of BrO and HOBr (the product of BrO and HO 2 ). If BrO undergoes a reduction chemistry that reforms ozone there is no net impact, while if a reduction path 500 does not reform ozone there is a net ozone loss. Table 4 tabulates the relative rates of the BrO and HOBr reduction reactions which yield Br, Br 2 , or BrCl without reforming ozone in the plume and therefore can be "credited" with ozone destruction. We find that for this young plume, the most important of these bromine reduction reactions are the reactions of BrO with itself and of BrO with ClO, which together account for about three-quarters of the bromine recycling.
The relative importance of the BrO self-reaction decreases slightly as the plume dissipates and evolves, whilst the two 505 reactions of BrO with ClO maintain the approximately same level of importance. The importance of HOBr photolysis increases over time but remains minor. Although the reactions of HOBr are responsible for only a minor fraction of the bromine reduction in this case, the heterogenous reactions of HOBr are important for transferring bromine from HBr and HOBr to the more potent ozone destructive forms, and for generating the reactive chlorine involved in the BrO + ClO reactions.
The overall rate of ozone destruction within the plume is dependent upon the quantity of bromine cycling. As shown in 510 Figure 9, compared to the 30th, bromine is transferred faster out of HBr and BrNO 3 on the 31st and 1st as these are denser (more concentrated) plumes with higher surface area density. Additionally, because several of the reactions listed in Table 4 are between halogen species originating from the volcanic emissions, they are faster in denser plumes. These factors result in a slower ozone destruction for 30th, as shown in Figure 15a. Because this ozone destruction is slower, but the plume travels faster, this difference is magnified if ∆O 3 /SO 2 is plotted against distance (as is done in Figure 16).

Comparison of model and aircraft data on ozone loss
Here we compare the aircraft observations to model outputs. Because the model plume does not precisely trace the same path as the observed plume, using the exact coordinates of each in-plume observation to identify the plume in the model domain would certainly often result in missing the modelled plume. Instead, for each of the plume encounters discussed in Sect. 3, we identify the equivalent of the observations in the model by the following method. Model data are considered "equivalent" for a 520 plume encounter if they satisfy the following criteria: -Time is closest to the median time of the observed encounter.
-Grid box is wholly or in part within the altitude range of the observed plume encounter.
-Grid box centre is within the range of distances from Etna for the observed plume encounter.
-SO 2 mixing ratio is in excess of 10 ppbv.

525
This effectively delimits a 3D space within the model. This 3D space is more likely to include the most concentrated part of the plume than the 1D transect made by the aircraft, and as such, the model data tends to include grid boxes with SO 2 greater than the maximum values from the observed plume encounter. For this reason our observation-model comparisons are based on O 3 vs. SO 2 trends rather than absolute values.
The model output is hourly, and the looping flight path (Figure 2) means that many of the plume encounters were made 530 minutes apart at nearly the same points in space. As a result, several of these model-equivalent sets share many model data points and are nearly identical. Model equivalents of Figures 4 and 5 are shown, respectively, in Figures S7 and S8. The coefficients from simple linear regression of these data are tabulated in Table 5.
The gradients of ∆O 3 /SO 2 versus plume age from applying a linear regression to the model and observation data are very similar. Weighting by the quantity and R 2 of the fit of the observational data, the root mean square difference of the two 535 gradients is 0.005, with a weighted mean bias of less than 0.001. Figure 16 plots modelled ∆O 3 /SO 2 and these plume encounter gradients against distance from the volcano. The range of distances of the plume encounters is insufficient to determine any trend from the observations beyond those made in Sect. 3, however there is a reasonable match between observed and modelled ∆O 3 /SO 2 values for distances of around 15 km from the volcano. This gives confidence that the model has skill with regards to the ozone chemistry of the plume.

540
The y-intercept of the lines of best fit in the model are consistently 15-20 ppbv lower than those of the observations, reflecting a somewhat higher background ozone. The cause of this is most likely bias in the initial and boundary conditions used in the model. We do not expect this offset to have a significant impact on the main results of this study, which are based on changes in ozone, (∆O 3 ), rather than its absolute magnitude.

545
As tabulated in Table 2, we ran the model with different volcanic fluxes in order to assess how volcanic impacts on ozone could vary for different passive degassing emission compositions. In Table 6 we report, for the main, mag, hal, and oth runs  Table 5. Data relating to the major plume encounters, and the equivalent model data, depicted in Figures 4 and 5

(observations) and Figures
S7 and S8 (model). dur = duration of plume encounter, alt = range of altitudes of plume encounter, dist = distance from Etna, grad = gradient of line-of-best fit of O 3 vs. SO 2 slope, int = y-intercept of this line the average ∆O 3 values (compared to the novolc run) in ppbv for modelled plume aged 60 ± 5 minutes at 2012-08-01 08:00 as a metric for the volcanic impact on ozone. Collectively, these runs explore a two-dimensional parameter space of variations in halogen and other emissions. In all runs the chlorine emission is scaled to the bromine emission (Cl/Br = 300 by mass), 550 the bromine is emitted in the same 3:1 fixed proportions of HBr and Br, and all other species (H 2 O, Hg, OH, NO, at-source aerosol) are scaled to the SO 2 emission. In the absence of both SO 2 and bromine emissions the ozone loss relative to the novolc run is by definition zero. In the absence of halogens (only), the plume is slightly ozone productive, a phenomenon we ascribe to the impact of the NO emissions. Table 6 confirms that it is the halogen emission that causes the ozone loss.
Increasing the modelled flux of all species other than the halogens above that of the main case does not significantly change 555 the depletion amounts. However, decreasing this flux by two-thirds limits ozone depletion by around 20%. As was the case for the 30th in the main model run, the surface area density is insufficient to quickly move bromine from HBr to the ozonedestructive cycle -though in this case this is due to weaker aerosol emissions rather than faster wind speeds.   Table 6. ∆O 3 values (compared to novolc run) in ppbv for plume aged 60 ± 5 minutes at 2012-08-01 08:00 for various model runs with varying emissions. For all runs the Cl/Br, Br/HBr, OH/SO 2 , NO/SO 2 , H 2 O/SO 2 , Hg/SO 2 , and aerosol/SO 2 emission ratios remain the same as in main. The 40 kg s −1 row and 30 g s −1 column indicate the emission fluxes used in main.
The four data points for different bromine fluxes at 40 kg SO 2 s −1 can be fitted to a second order polynomial, (∆O 3 ) 60min = −3.5 × 10 −3 x 2 − 0.15x + 0.6 ppbv, where x is the volcanic bromine flux in g s −1 . We interpret this combination of a second-560 and first-order terms to be the product of the complexities of the chemistry. The reactions that recycle BrO though reaction with other halogen oxide species would be expected to have rates approximately proportional to the square of the amount of halogens in the plume, while the rates of those that recycle HOBr, including those that generate reactive halogens from hydrogen halides, would vary approximately linearly with this quantity.
We caution against scaling these results to the plumes of large eruptions. Such plumes could have near-total depletion of 565 ozone and/or HO x and produce non-linear effects beyond the scope of this study. The halogen chemistry of large eruptions is being investigated using the WCV model and will be the focus of a future study. Although the volcano degassing is a source of NO in the model, the plume is nearly totally depleted in NO x (Figure 17a).
The reason for plume NO x being below background levels is the reaction sequence BrO + NO 2 BrNO 3 followed by the heterogeneous reaction of BrNO 3 with hydrogen halide that has the net effect of converting NO 2 into HNO 3 , a phenomenon discussed by Roberts et al. (2014). As a consequence, the plume is elevated in HNO 3 compared to the background (17b).
As well as the conversion of the volcanogenic and background NO x , displacement of nitrate from background aerosol can also contribute to the in-plume HNO 3 . The acidic plume, rich in sulfuric and hydrochloric acid, displaces nitrate from background aerosol into the gas phase as HNO 3 . As shown in Figure 17c, the aerosol-phase nitrate content within the plume is much lower than the background in the model. The contributions of background NO x (via BrNO 3 ) and background nitrate (via 580 acid displacement) to the plume HNO 3 enhancement are of similar magnitude. In addition, conversion of volcanic NO x (via BrNO 3 ) into HNO 3 contribute to further enhance the plume HNO 3 .
As discussed by Martin et al. (2012) it is unclear from reaction kinetics if volcanoes are sources of reactive nitrogen. The levels of background NO x and nitrate aerosol in the free-tropospheric environment modelled in this study (July-August 2012) would be too low to yield, by themselves, HNO 3 concentrations of the order measured by (Mather et al., 2004a) (Voigt et al., 2014). However background NO x and nitrate may be significant contributors to volcanic HNO 3 in more nitrogen-polluted environments. Voigt et al. (2014) states that typical conversion times of atmospheric NO x to HNO 3 is days in summer midlatitudes, and so cannot explain formation of HNO 3 in volcanic plumes. Our modelling results show in-plume gas-phase HNO 3 being generated quickly by the mechanisms of acid displacement of background nitrate aerosol and volcanic plume halogen 590 chemistry that converts background NO x as well as volcanic NO x (if present) into HNO 3 via BrNO 3 . These results suggest that analyses of HNO 3 measurements within plumes used to assess volcanogenic NO x or NO y need to account for background reactive nitrogen in both the gas and particulate phases.
Due to the formation of the BrNO 3 reservoir, a volcanic NO x emission (or, potentially, background NO x of similar magnitude to the plume bromine) could impact the bromine chemistry and speciation and limit the amount of reactive bromine available 595 to form BrO. The quantity of BrNO 3 that accumulates depends upon the abundance of NO x available to react with BrO and the rate at which BrNO 3 decays via heterogeneous chemistry. Of the three days modelled, only on the 30th, where the in-plume heterogeneous rates are low (due to a higher windspeed causing greater along-plume dilution), is a substantial fraction of plume bromine held in the BrNO 3 reservoir for several minutes.

Halogen impacts on mercury 600
The volcano is modelled to emit, per mol of SO 2 , 7.8 × 10 −6 mol of Hg in an unoxidised state. A simplified Hg chemistry scheme has been implemented in WCV to evaluate the extent to which this Hg can become oxidised by interaction with the volcanic halogen halogen chemistry. Importantly, our mechanism includes the recently identified photo-reduction pathways  Overall, there is a small net oxidation of Hg occurring in the plume, but it is slow -for daytime plumes the average in-plume 610 lifetime of Hg(0) is of the order of a few hours and this oxidation is mostly offset by photo-reduction. Despite the slow rate of oxidation in the plumes aged 10s of minutes, we find that in the early plume, modelled levels of mercury oxidation can be several % ( Figure S6) from the first few minutes of evolution during both night and day. We attribute this oxidised mercury to oxidation occurring in the very early plume (first few seconds) where volcanogenic radicals are in high concentrations. This near instantaneous oxidation accounts for the vast majority of oxidised mercury in the modelled plume further downwind.

615
Supporting this interpretation, we find negligible Hg oxidation in the first hour of plume evolution in output the noHighT run.
Therefore variations in the Hg oxidation with plume age and time of day are due mostly the conditions at the point of emission, rather than any processes occurring within the downwind plume.  We conclude that, for our case study of a Mount Etna passive degassing plume that focused on mercury-halogen interactions only, the net in-plume oxidation rate of mercury by halogen chemistry is near-negligible in the dispersed plume but could be 620 significant very close to the source. Our findings contrast with the model study of von Glasow (2010) that predicted substantial oxidation of mercury to Hg(II) occurring in the dispersed volcanic plume of Mount Etna. However, that study did not include the very fast HgCl and HgBr photo-reduction pathways that critically impact the overall Hg oxidation, although it did include a SO 2 -mediated reduction of Hg(II) hypothesised by (Seigneur et al., 2006). We note that this slow net oxidation rate in the modelled evolved plume occurs despite the absence of this reduction pathway in the mechanism. 625 We invite caution regarding the interpretation of these results due to the very simple mercury chemical scheme used, and the apparent importance of the first few seconds of plume evolution which would be, spatially, poorly represented even at 1 km grid resolution. Further investigation of mercury chemistry in volcanic plumes is needed, across a range of volcanic and meteorological conditions. This requires modeling at higher resolution to capture very near source processes both in the young cooled plume 630 as well as investigations of hot plume chemistry just after emission, and the mercury chemistry in much larger eruption plumes that may differ from passive degassing case. The mercury oxidation-reduction scheme should also be extended to investigate possible roles of other gases (e.g. NO x ). More comprehensive observation studies of speciated mercury at Mt. Etna and other volcanoes are also needed.

635
Volcanoes emit halogens that are converted into active chemical radicals in plumes and whose chemistry, notably bromine, causes ozone destruction. However, the plume processes driving this halogen conversion, the so-called halogen activation, Measurements of SO 2 and ozone levels in the plume are found to be strongly anticorrelated. Ozone losses reach of up to about 10 ppbv. Accounting for the distance from the source at which these measurements were taken (7-21 km downwind from 645 the summit), and using modelled wind speeds, the ozone destruction rate is estimated at approximately 1.3 × 10 −5 mol O 3 per mol SO 2 per second. This value is similar to observation-derived estimates reported very close to the Mt Etna vents (<500 m downwind) (Surl et al., 2015), indicating continual ozone loss in the plume up to 10's km downwind.
The aircraft observations are analysed with the WCV model forced by emission fluxes of volcanic gases, including SO 2 , mercury, and halogens (HBr, HCl), each set to values within typical ranges observed for Etna in a passive degassing state. The 650 model initialization also includes a representation of high-temperature radicals (Br, OH, NO) and a volcanic sulfate particle emission. The WCV mechanism includes photolytic, gas-phase and multi-phase reactions of bromine and chlorine, as well as gas-phase oxidation of SO 2 . WCV was run using two-way nested grids, enabling 1 km resolution for the simulation of plume processes close to the volcano -the focus of this study. with a very similar value from an analysis of the instantaneous rates of reactions. In summary, the WCV model shows apparent skill in reproducing plume halogen chemistry and impacts on tropospheric ozone.
Inspecting the bromine chemical system, we found that HBr, the dominant form of bromine at emission, is converted to other forms within the first few minutes of plume evolution. These forms undergo a continuous cycling in the plume, with with BrO 665 and HOBr being the dominant daytime forms. We found that a lower plume density, caused by a greater wind speed, slowed the evolution of the bromine chemical system with more bromine residing in HBr, HOBr, and BrONO 2 . The balance between BrO and HOBr varies moderately with time of day due to diurnal variations in HO 2 , yielding slightly lower BrO/SO 2 around solar noon. Inspection of the rates of reaction find that although the overall proportions of bromine in different forms stabilises after a few 10s of minutes of plume evolution, bromine is constantly cycling between forms, and this process includes the ozone 670 destructive Br + O 3 BrO reaction. Overall the ozone loss depends on several different reactions that reduce oxidised bromine without recreating ozone. We find that, for young plumes (<1 hour old), the most important reactions are those of BrO with halogen monoxides (BrO or ClO).
The conversion of volcanic HBr into reactive bromine forms occurs by the heterogeneous reaction of HOBr. Once plume HBr is depleted, this reaction acts to convert HCl into reactive chlorine, leading to the release of methane-oxidising chlorine radicals.

675
As a result the lifetime of methane is reduced in the plume core, however at the plume edge methane lifetime increases due to lower OH. The CH 4 oxidation initates organic chemistry processing that results in formaldehyde being elevated in the plume compared to the background. Cl radicals also generate ClO and OClO -species that have also been detected in the plumes of some volcanoes including Mount Etna. Modelled near-source OClO/SO 2 ratios are of approximately similar magnitude (10 −5 mol/mol) to those measured by Gliß et al. (2015).

680
The model plume chemistry is investigated over a range of emission scenarios. If the radical species expected to be produced in the high temperature volatile-air mix are excluded, the evolution of the halogen chemistry is greatly slowed and delayed.
This result highlights the importance of understanding these very early processes in order to have an accurate picture of the overall chemistry. A sensitivity study of the model response to variations in the emissions of both halogens and sulfur finds that ozone loss depends on the bromine emission flux with both linear and quadratic components, reflecting the complexities 685 of the plume chemistry.
Finally, the model outputs are inspected to identify halogen impacts on HO x , sulfur, NO x and mercury chemistry.
Despite the volcano being an initial source of high-temperature OH radicals, for the early (<1 hr age) plume considered in this study the in-plume instantaneous lifetime of SO 2 in the model is substantially increased (from about 2 days to about 2 weeks) due to depletion of OH. These modelling results therefore strengthen the case for using SO 2 as a plume tracer on these scales.

690
The substantial depletion of OH is attributed to both SO 2 and halogen chemistry which further reduces OH concentrations within the plume. Halogen chemistry also causes depletion of HO 2 .
The SO 2 oxidation that does occur nevertheless produces sulfate aerosol mass and surface area within the plume. Secondary aerosol is formed more quickly in a simulation that excludes volcanic halogens. This result demonstrates that volcanic halogen chemistry can critically influence sulfur oxidation processes, and emphasizes the need to include halogens in studies of volcanic 695 sulfate aerosol impacts. Despite the volcano being modelled as an initial source of high-T NO, modelled in-plume NO x levels are lower than the surrounding air due to plume chemistry destroying these species. In-plume HNO 3 was found to be elevated for two reasons: bromine chemistry converts NO x to HNO 3 , and background aerosol-phase nitrate is displaced into the gas phase by the acidic plume.

700
The model includes a very simple mercury scheme which includes photolysis reduction of mercury halides. In this passively degassing case, WCV predicts some early-stage oxidation of mercury by the initial high-temperature region Br radicals, butin contrast to previous model studies for Mount Etna passive degassing -predicts a very slow net oxidation by halogens in the downwind plume. Further model-observation studies of volcanic mercury are warranted.
Overall, the WCV model appears to show reasonable skill in replicating observed in-plume phenomena of ozone loss specific 705 to this case study and established downwind trends in BrO/SO 2 for minutes-old passive degassing plumes more generally. This skill gives credence to the assessments of the chemical processes occurring within the plume. WCV operated using nestedgrids enables to reach 1 km resolution, however, we suggest caution in using results from the model for processes occurring at sub-km scale within the very early plume. This study investigated the chemical processes occurring in the passively degassing plume of Mount Etna. We caution against extrapolating these results to stronger (more dense) eruption plumes, as such plumes 710 may experience phenomena out of scope of this study, such as near-total ozone depletion that perturbs the halogen chemistry.
WCV is being applied in a follow-up study to investigate the plume chemical processes in such dense plumes from volcanic eruptions to the troposphere, as a contrast to this passive degassing case. In future, WCV can also be applied to assess the tropospheric impacts of volcanic halogen chemistry in plumes as they disperse and may remain chemically active for up to regional scales.

715
Code availability. Availability of model code The code of WCV is available on GitHub (Surl, 2020). This repository is being actively maintained. The version of the code used to generate the results of this study are included in this repository as a static branch (etna2012).
The modifications to the PrepChem utility have been submitted to the maintainers of this software for consideration, and are available from the authors on request.

720
Data availability. Data availability WRF-Chem generates NetCDF files as output. The output relating to the innermost domain (d04 on Figure 3) are uploaded to a Zenodo online repository (Surl, 2021). This repository also contains the input settings files used for the runs.
Author contributions.