Impacts of the 2014-2015 Holuhraun eruption on the UK atmosphere

Volcanic emissions, specifically from Iceland, pose a pan-European risk and are on the UK National Risk Register due to potential impacts on aviation, public health, agriculture, the environment and the economy, both from effusive and explosive activity. During the 2014-2015 fissure eruption of the Holuhraun in Iceland, the UK 25 atmosphere was significantly perturbed. This study focuses on the first four months of the eruption (September to December 2014). During this period there was one major incursion in September 2014, affecting the surface concentrations of both aerosols and gases across the UK, with sites in Scotland experiencing the highest sulfur dioxide (SO 2 ) concentrations. At the two UK EMEP supersite observatories (Auchencorth Moss, SE Scotland and Harwell, SE England) significant alterations in sulfate (SO 42- ) content of PM 10 and PM 2.5 during this event, 30 concurrently with evidence of an increase in ultrafine aerosol, most likely due to nucleation and growth of aerosol within the plume, were observed. At Auchencorth Moss, higher hydrochloric acid (HCl) concentrations during the September event (max = 1.21 µg m -3 , c.f annual average 0.12 µg m -3 in 2013), were assessed to be due to acid displacement of chloride (Cl - ) from sea salt (NaCl) to form HCl gas rather than due to primary emissions HCl from Holuhraun. The gas and aerosol partioning at Auchencorth moss of inorganic species by 35 thermodynamic modelling, confirmed the observed partioning of HCl. Volcano plume episodes were observed by the majority of the UK air quality monitoring networks during the first 4 months, at both hourly and monthly resolution. In the monthly networks, SO 2 concentrations were significantly elevated at remote “clean” sites in NE Scotland and SW England, with record high SO 2 concentrations for some sites. For sites which are regularly influenced by anthropogenic emissions, taking into account the underlying trends, the eruption led to statistically unremarkable SO 2 concentrations (return probabilities >0.1, ~10 months). However for a few sites, SO 2 concentrations were clearly were much higher than has been previously observed (return probability <0.005, >3000 months). The Icelandic eruption has resulted in a unique study providing direct evidence of atmospheric 5 chemistry perturbation of both gases and aerosols in the UK background atmosphere. The measurements can be used to both challenge and verify existing atmospheric chemistry of volcano plumes. If all European data sets were collated this would allow improved model verification and risk assessments for future volcanic eruptions.


Introduction
Volcanic emissions perturb atmospheric composition in the troposphere (Bobrowski et al., 2007;Horrocks et al., 10 2003;Martin et al., 2008;Oppenheimer et al., 2010;Oppenheimer et al., 2006;von Glasow, 2010) via emissions of ash and/or gases and aerosols to the atmosphere, particularly during active eruptions. These emissions can directly impact humans and ecosystems (Thordarson and Self, 2003) as well as have indirect effects on climate (Gettelman et al., 2015;Schmidt et al., 2012;Schmidt et al., 2014). Sulfur dioxide (SO2) and sulfate (SO4 2-) aerosol injection into the stratosphere is a well-documented form of atmospheric perturbation and climate 15 forcing; however, tropospheric atmospheric and surface effects, both local and regional, can only be studied serendipitously. In particular there are very limited detailed atmospheric observations available where both the physical characteristics and the chemical composition of volcanic plumes are probed in the distal plume, long distances away from the eruption source. In this case the distal plume was ~ 1000 km from its source in Iceland.
Volcanic plumes contain elevated quantities of reactive sulfur species, primarily in the form of SO2. Quantifying 20 the relative emission abundance of SO2 and SO4 2and the oxidative aging of the plume converting SO2 to SO4 2has been attempted previously, for example by Satsumabayashi et al. (2004) but there a very limited number of studies (Hunton et al., 2005;Rose et al., 2006;Mather et al., 2003;Kroll et al., 2015;Boulon et al., 2011;Satsumabayashi et al., 2004) which have quantified gas and aerosol composition beyond sulphur species and provided evidence of tropospheric chemistry of distal plumes including halogen chemistry and particle 25 growth (Boulon et al., 2011).
The recent eruption within the Holuhraun volcanic system in Iceland (August 2014 -February 2015) was the largest Icelandic eruption in terms of erupted magma and gas volume since the 1783-1784 CE Laki event, producing 1.6 km 3 of lava and total SO2 emission of 11±5 Mt during a period of 6 months (Gíslason et al., 2015).
It was almost purely effusive, hence producing negligible amounts of ash, but repeatedly causing severe air 30 pollution events in populated areas of Iceland due to high gas and aerosol concentrations. The ground level concentration of SO2 exceeded the hourly health limit (350 µg m -3 ) over much of the country for periods of up to several weeks (Gíslason et al., 2015). In Europe, anthropogenic emissions of sulfur have been declining over the past few decades and hence lower concentrations are observed widely (Fowler et al., 2007). EU-28 annual emissions of sulfur oxides in 2010 and 2011 were ~4.6 Mt (http://www.eea.europa.eu/data-and-35 maps/daviz/emission-trends-of-sulphur-oxides#tab-chart_1) and therefore the Holuhraun volcanic eruption added more than twice the EU-28 annual sulfur emissions to the atmosphere in just six months (Schmidt et al., 3 2015). This eruption provided a unique opportunity in Europe to study the impact of a large point source SO2 emission.
This paper studies the volcanic impact on the UK atmosphere in the first 4 months of the Holuhraun eruption (September to December 2014) and provides the first evidence of wide scale effects, based on the measurements from the UK air quality monitoring networks which deliver data at both high (hourly) and low (monthly) 5 temporal resolution. These observations provide information on the chemical composition of the distal plume, ~ 1000 km downwind of Iceland. Because Icelandic air arrives at the UK on northerly trajectories, the background air is clean and there is little interference from anthropogenic emissions when the air arrives at the Northern UK.

Basics of MARGA operation
The Measurement of Aerosols and Reactive Gases Analyser (MARGA, Metrohm Applikon B.V, NL) provides 20 hourly resolution measurements of water soluble inogranic aerosol speciation ( SO4 2-, Cl -, NO3 -, NH4 + , Na + , K + , Ca 2+ and Mg 2+ ) and gases (SO2, HCl, HNO3, HONO and NH3). At the two field sites Harwell and Auchencorth Moss (Figure 1), the instruments are configured to have two sample boxes, one for PM10 and on for PM2.5. The instruments use wet rotating denuders (WRD) (Wyers et al., 1993) and steam jet aerosol collectors (SJAC) (Khlystov et al., 1995) for sampling of gases and aerosols respectively. Analysis is carried out online by ion 25 chromatography (both anion and cation) at an hourly resolution. A detailed description for the instrument and QA/QC procedures used by both instruments are given in Twigg et al. (2015). There is one operational difference between Auchencorth and Harwell instruments, where Auchencorth Moss uses preconcentration columns (Metrosep A PCC 1 HC IC preconcentration column (2.29 mL) for anions and a Metrosep C PCC1 HC IC pre-concentration column (3.21 mL) for cations) on the IC to achieve lower detection limit (DL) compared to 30 the Harwell instrument which uses fixed loops (250µL for anions and 480µL for cations) and therefore has a magnitude higher DL as described by Makkonen et al. (2012). Data from both MARGA instruments are available in the UK-Air (http://uk-air.defra.gov.uk/data/) and EBAS (http://ebas.nilu.no/default.aspx) databases.

SMPS
At Harwell aerosol number size distributions were measured using a scanning mobility particle sizer (SMPS)

35
(Electrostatic classifier 3080, differential mobility analyser 3081, and condensation particle counter 3775, all TSI Inc.). Air was sampled at 4 m above ground level, through a PM1 cyclone before entering the analyser via a drier which ensured the relative humidity of the sample air was kept below 45%. The aerosol sample flow rate was set to 0.3 L min -1 and the Classifier sheath flow was maintained at 3 L min -1 ; a detailed description of the method and set-up employed at Harwell can be found in Beccaceci et al. (2013) and data is freely available through the UK-Air website.

5
At Auchencorth Moss aerosol size distributions in the range of 14-673 nm were set to be measured using a scanning mobility particle sizer (SMPS) (Electrostatic classifier 3081, differential mobility analyser 3080 and condensation particle counter 3775, all TSI, Inc.). Air was sampled at 2 m above ground level through a PM10 head and PM2.5 cyclone before entering the analyser via a drier which ensured the relative humidity of the sample air was kept below 45%. The aerosol sample flow rate was set to 0.3 L min -1 and the classifier sheath 10 flow was maintained at 3 L min -1 as set out in Wiedensohler et al. (2012). In October 2015, the Auchencorth Moss SMPS took part in an intercomparison organised by the EU Horizon 2020 ACTRIS 2 (aerosol, clouds and trace gases research infrastructure), held at the world aerosol calibration centre (TROPOS, Leipzig, Germany).
During this exercise the classifier used at Auchencorth was found to have an offset and was starting a scan at 35 nm instead of 14 nm, though it is unclear if this may have slowly drifted over the 18 months since installation at 15 the site. Therefore data presented from Auchencorth Moss is a qualitative indicator of an increase in ultrafine particles as the size distribution could not be verified.

AGANet DELTA and Precip-Net
The DEnuder for Long-Term Atmospheric sampling (DELTA), used in AGANet across the UK, is described by Sutton et al. (2001). The sampling system consists of a series of coated denuders (to capture gases) and filters (to 20 capture the aerosol). Air is sampled at a flowrate of 0.2 -0.4 L min -1 , with the sampling inlet at a height of 1.5 m.
The first pair of denuders (15 cm) after the inlet are coated with K2CO3/glycerol to capture acidic gases (HNO3, SO2 and HCl). The next pair of denuders are coated with citric acid to capture gaseous NH3. A filter pack is situated at the end of the sampling train, containing two cellulose coated filters: the first is impregnated with K2CO3 to capture and retain NO3 -, SO4 2-, Cland Na + , Ca 2+ and Mg 2+ aerosol. The second filter is impregnated 25 with citric acid to capture NH4 + . Downstream of is a gas meter, to record the volume of air sampled and an air pump. A DELTA sampling train is exposed for 1 month and samplers are extracted with deionised water.  Precip-Net uses bulk precipitation samplers at 39 non-urban sites with fortnightly sample collection. Samples are analysed for cations (Na + , Ca 2+ , Mg 2+ , K + , NH4 + ) and anions (PO4 3-, NO3 -, SO4 2-, Cl -) using ion chromatography (further details of both the sample method and analysis can be found in Irwin et al. (2002)). Data from both AGANet and Percip-Net are freely available from UK-Air.

35
The GOME2 instrument on MetOp-B is a nadir viewing UV/visible spectrometer with a spatial resolution of 40 x 80 km 2 . SO2 column densities are retrieved using a Differential Optical Absorption Spectroscopy approach

5
including a non-linearity correction for SO2 saturation effects (Richter, 2009). As no corrections are made for the effects of deviations from the assumed plume height of 10 km, the data shown should be used as qualitative indicator only.

EMEP4UK chemical transport model
The EMEP4UK model rv4.3 (Vieno et al., 2016), is a chemical transport model which is the regional application 5 of the EMEP MSC-W model (Simpson et al., 2012), which is used in this study to identify and investigate the

Statistical analysis of AGA-Net data
As well as high resolution analysis of the volcanic plume, the trends in the SO2 from AGANet measurements were analysed to assess the impact of the fissure eruption on the background atmosphere in the UK. The 25 likelihood of a reoccurrence of the observed concentrations in the UK background was calculated. For most sulphur compounds the AGAnet observations at many sites show decreasing trends over time, both for annual mean concentration and the annual maximum concentration. A high concentration superimposed on a downward trend would appear to be a less unlikely observation at the end of the time series than at the beginning, so the data were adjusted to remove any underlying trend before further analysis. Exceedances over a threshold follow 30 a Pareto distribution. The threshold was chosen by fitting an 85% quartile regression using a smoothing spline for each site individually. The fitted Pareto distribution was used to assess the probabilities of the concentrations associated with the volcanic eruption occurring, expressed as a return probability and return time, which is the statistical likelihood of a similar concentration to be observed again based on the long term trend of SO2 at each site expressed in the resolution of the measurements (Table 1)

Formation of sulfate aerosols
Current understanding of volcanic emissions is that the major fraction of observed SO4 2is not directly emitted from the magma but is formed as secondary aerosol through oxidation of SO2 in the atmosphere (Mather et al., 2013), though there are some reports suggesting primary emissions are possible (Allen et al., 2002;Zelenski et al., 2015). As shown in Figure  aerosol was confirmed by the presence of fine aerosols measured by the SMPS instruments at both Auchencorth and Harwell. The volcano plume event was characterised by the high particle number density at low diameters, increasing in diameter with time (initiating at ~1200 hours GMT on the 21st, Figure 5). The feature of increasing particle numbers, or "banana", starting with high particle numbers at the detection limit of the SMPS is characteristic of particle nucleation and growth; however, as it is not a Lagrangian measurement, and because the 5 nucleation does not represent a wide-spread regional phenomenon (as it probably does, e.g., in the nucleation studies conducted in the Boreal environment; (Kulmala et al., 1998)), the evolution of the size distribution with time needs to be interpreted with caution: only if trajectories and wind speed do not change with time can the temporal change at the fixed site be translated into the temporal change within the plume. It is possible that a population of ultrafine H2SO4 particles were emitted or formed at source, however, it is highly unlikely due to 10 the transport time that aerosol would have remained in the ultrafine fraction observed as they would have undergone further growth by coagulation and further condensation of condensable vapours. It is much more likely that, sulfuric acid was formed during transport through oxidation of the high concentrations of SO2 by the OH radical the production of which is linked to solar radiation. With increasing time of sunrise, the measurements at Auchencorth reflect particles whose nucleation was initiated further and further away from the 15 site and had increasingly time to grow during transport. The SMPS at Harwell also recorded similar events as the plume passed over. This is the first evidence of boundary layer surface-level particle growth observations in a distal volcanic plume for the UK and complements observations from the 2010, which is the only previous report of nucleation and secondary aerosol formation event reported for a distal plume during the explosive, ash-rich plume (Eyjafjallajökull in 2010) at an elevated free tropospheric atmospheric station in Europe (Puy de Dôme 20 observatory, France) (Boulon et al., 2011). At that station, the free tropospheric conditions and size range of measurements allowed the clear interpretation of particle nucleation. In addition to the particle population changes observed, the measurement indicates that there was an unquantified air quality impact during the 2014 eruption in addition to the SO2 air quality impacts discussed in the recent study of  due to particles. 25

Modification of the chemical composition within a volcanic plume
The chemical composition of PM2.5 and the gas concentrations observed during the event at Auchencorth are summarised in Figure 6. It is clear that the aerosol was dominated by SO4 2-. Whilst the aerosol at this site is normally neutralised, with free ammonia (NH3) available (Twigg et al., 2015), during the plume event the aerosol turned acidic. During the event the measurements at the background site clearly showed that there was an 30 increase not only in the sulfur species but also in hydrochloric acid gas (HCl) and a variety of other chemical species in both gas and aerosol phase ( Figure 6). HCl peaked at 1.21 μg m -3 during the event compared with an annual average of 0.12 μg m -3 in 2013. As discussed in Aiuppa (2009), Pyle and Mather (2009) and summarised in Witham et al. (2015) and the literature cited therein, primary emissions of HCl from volcanoes can vary enormously depending on the magma type and the particular eruption characteristics (Aiuppa, 2009;Aiuppa et 35 al., 2009;Pyle and Mather, 2009). The near-source measurements of the gas composition from the Holuhraun eruption indicated that the gas phase in the plume was proportionally very low in halogen content, with a molar HCl/SO2 ratio of <1%. It is unlikely that HCl would persist longer in a plume than SO2 given the high solubility of HCl and comparably low reactivity of SO2. However, given that the SO4 2aerosol is highly acidic, the HCl 8 would need to be scavenged onto other non-sulfate aerosol or into cloud droplets. Hence the elevated HCl observed in the plume event is either due to transport of primary HCl or displacement of HCl from background sea salt aerosol or a combination of the two. It is hypothesised that the most likely explanation for the observation of HCl coinciding with the plume is the oxidation of SO2 to sulfuric acid which then displaced HCl in pre-existing sea salt aerosol (NaCl) in the air mass. The thermodynamic model ISORROPIA-II (Fountoukis 5 and Nenes, 2007) was used to calculate the theoretical partitioning between the gas and aerosol phase. The model clearly reproduces the HCl peak which is attributed to the displacement of Clfrom sea salt ( Figure 6).
Further evidence was found when the ratio of Na + and Clwas compared to the known ratio of sea water, where a large relative depletion of aerosol Clwas found during elevated SO4 2- (Figure 7) at Auchencorth Moss. It is noted, that between 09:00 (GMT) on 21/09 and 03:00 (GMT) on 22/09, the Na + was known to be 10 underestimated, attributed to acidic composition of the aerosol resulting in a reduction in the performance of the cation column (concentration of the Li + internal standard decreased). Whilst correction based on the Li + standard is possible, this assumes that the retention was similarly depressed for all cations. The data therefore have been flagged as invalid during the QA/QC procedures of data submission to UK-Air and EMEP but have been presented here as it is thought to be useful data for research purposes. As such the depletion of Clis thought to 15 be even greater than that demonstrated in Figure 7.

Long term perturbation of the UK atmosphere
The relative importance of the volcanic plume over the four months on the UK surface composition and the wider region with respect to air quality and acid deposition can only be assessed with measurements over a wider geographic region. The low-temporal resolution (monthly) measurements of gas and aerosol composition with  (Parry et al., 2014). The majority of the western UK received 35 less than 20% of the long-term average rainfall, hence the amount of sulfur deposited by wet deposition during this period was not important to the UK (Figure 9). It therefore has to be noted that the reported high SO4 2-Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-177, 2016 Manuscript under review for journal Atmos. could be the result of lack of dilution due to low precipitation and cannot be directly attributed to the volcanic plume.

Conclusions
The Holuhraun eruption perturbed all aspects of the UK atmosphere periodically during the latter part of 2014.
Elevated SO2 were observed by the networks at both high and low resolution. This complemented the study by 5  who reported similar observations for SO2 across Europe for the same period. This study, however, provides further details of the chemistry within the volcanic plume which are not addressed by . In this study high SO2 concentrations, were demonstrated to have resulted to an increase in tropospheric HCl due to the acid displacement of Clfrom sea salt at the EMEP supersite Auchencorth Moss.
Elevated particulate SO4 -2 and particle size distributions from the two EMEP supersites suggested that new 10 particle formation and growth were occurring as the plume passed over the UK. Future work now needs to be done investigate the direct and indirect effects of the perturbation of chemistry, specifically with regards to human health and crop yields.
The analysis also provides evidence to support the recent modelling undertaken which concluded that volcano eruptions in Iceland will intermittently affect the UK (Witham et al., 2015) with the effects varying both spatially 15 and temporally during an eruption, primarily driven by meteorology. There is a significant difference in effects on both human health and ecosystem effects between acidic-non-acid aerosol and this study presents the first evidence that volcanic aerosol reaching the UK can be acidic, however this will be highly dependent on the mixing of the plume with the background atmosphere. There are also further impacts which have not yet been fully assessed, for example the net effect on climate (Gauci et al., 2008;Gettelman et al., 2015) and ecosystem 20 function.
The study has highlighted even though anthropogenic SO2 concentrations have dramatically decreased in the UK over the last 30 years, there is still a need to maintain the network of analysers as it is not just needed to confirm recovery, but also provides a useful tool to track the progression and impact of volcanic plumes and other pollution events. High resolution chemical composition of aerosol are essential for the identification of the origin 25 of aerosol events observed concurrently with the SO2 plumes and to understand the atmospheric chemistry. This paper presents the first detailed observations of chemistry within a distal volcano plume at the surface in the UK.
This dataset is unique and can be used by modellers to test long term impacts of volcanic eruptions and the evolution of the plume chemistry.
While the 2014-2015 eruption in Holuhraun system was the largest eruption in Europe in over 200 years, there is 30 a potential for even larger events. For example, the 1783-84 Laki eruption was over 10 times larger in terms of erupted magma and gas volume. An event of this magnitude would cause significant and wide-spread pollution over Europe and even cause excess mortality (Schmidt et al., 2011). Though some work has been done on a limited set of the European air quality networks by  and Gíslason et al. (2015), a further study is required of the data from across the European compliance networks, as well as the EMEP and ACTRIS 35 networks to integrate both particle characterisation and gas chemical composition. This would allow the Holuhraun event to be fully characterised and quantified.