1 Dynamics of mild strombolian activity on Mt. Etna 2 Pering T.D.a*, Tamburello G.b, McGonigle A.J.S.a,c, Aiuppa A.b,c, James M.R.d, Lane 3 S.J.d, Sciotto M.e, Cannata A.e, Patanè D.e 4 *Corresponding author: T. D. Pering, Department of Geography, University of Sheffield, 5 Sheffield, South Yorkshire, S10 2TN, UK. (ggp12tdp@sheffield.ac.uk) 6 a 7 b 8 c 9 90146, Palermo, Italy University of Sheffield, Dept. of Geography, Winter Street, S10 2TN, United Kingdom DiSTeM, Università di Palermo, via Archirafi, 22, 90123 Palermo, Italy Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Palermo, Via Ugo La Malfa, 153, 10 d 11 e 12 11 Catania, Italy 13 ABSTRACT 14 Here we report the first measurements of gas masses released during a rare period of 15 strombolian activity at the Bocca Nuova crater, Mt. Etna, Sicily. UV camera data acquired for 16 195 events over a ≈ 27 minute period (27th July 2012) indicate erupted SO2 masses ranging 17 from ≈ 0.1 to ≈ 14 kg per event, with corresponding total gas masses of ≈ 0.1 to 74 kg. Thus, 18 the activity was characterised by more frequent and smaller events than typically associated 19 with strombolian activity on volcanoes such as Stromboli. Events releasing larger measured 20 gas masses were followed by relatively long repose periods before the following burst, a 21 feature not previously reported on from gas measurement data. If we assume that gas 22 transport within the magma can be represented by a train of rising gas pockets or slugs, then Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, UK Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo, Piazza Roma, 2, 95125 1 23 the high frequency of events indicates that these slugs must have been in close proximity. In 24 this case the longer repose durations associated with the larger slugs would be consistent with 25 interactions between adjacent slugs leading to coalescence, a process expedited close to the 26 surface by rapid slug expansion. We apply basic modelling considerations to the measured 27 gas masses in order to investigate potential slug characteristics governing the observed 28 activity. We also cross correlated the acquired gas fluxes with contemporaneously obtained 29 seismic data but found no relationship between the series in line with the mild form of 30 manifest explosivity. 31 Mild Strombolian Activity, Ultra-Violet imaging, Volcanic Gas Measurements, Slug 32 Dynamics, Coalescence, Trailing Wake Interaction 33 34 1. Introduction 35 Strombolian eruptions are thought to arise from the rise, expansion and bursting of over- 36 pressured gas slugs, also termed Taylor bubbles (e.g., Chouet et al., 1974; Blackburn et al., 37 1976; Wilson, 1980; Vergniolle and Brandeis, 1994; 1996; Ripepe et al., 2008). The 38 behaviour of single slugs, where the rising bubbles are sufficiently separated from one 39 another to behave independently, has received considerable attention in the volcanological 40 and fluid dynamical literature (e.g. Davies and Taylor, 1950; Wallis, 1969; James et al., 2008, 41 2009; Llewellin et al., 2012). Indeed, theoretical frameworks have been developed to link 42 observed geophysical signals to the characteristics of single volcanic slugs (James et al., 43 2009; Llewellin et al., 2012; Lane et al., 2013). In contrast, only a few studies have addressed 44 the behaviour of multiple slugs in volcanic regimes (Seyfried and Freundt, 2000; James et al 45 2004; Pioli et al. 2012) given the additional complexities involved. 2 46 Recently developed UV camera technology (e.g., Mori and Burton, 2006; Bluth et al., 2007; 47 Tamburello et al., 2011a) has provided considerably enhanced spatial and temporal resolution 48 (≈1 Hz) in the acquisition of volcanic SO2 degassing time-series, relative to previously 49 applied spectroscopic approaches (Edmonds et al., 2003; Galle et al., 2003; Burton et al., 50 2009; Boichu et al., 2010). The acquired data have therefore led to increased understanding 51 of a number of explosive and passive degassing volcanic phenomena, for example, the 52 degassing mechanism in the Santiaguito lava dome, Guatemala (Holland et al., 2011), the 53 links between gas flux trends and seismicity during passive degassing (Tamburello et al., 54 2013; Pering et al., 2014), the relationship between gas emissions and very-long-period 55 seismicity at Mt. Asama, Japan (Kazahaya et al., 2011), and ties between gas emissions and 56 generated infrasonic energy (Dalton et al., 2010). 57 UV camera imagery, in addition to FTIR (Fourier Transform Infrared) spectroscopy have 58 also been used to investigate the dynamics of gas release from single slug driven strombolian 59 activity on targets such as Stromboli (Aeolian Islands, Italy) (e.g., Burton et al., 2007; Mori 60 and Burton, 2009; Tamburello et al., 2012; La Spina et al., 2013). This has led to constraints 61 on the gas mass released per event, and the slugs’ source depth. In contrast to Stromboli, 62 where this activity is quasi-continuous, such behaviour occurs only sporadically on Mt. Etna 63 (Sicily, Italy). 64 Here we report on the first application of UV camera imaging to measure gas masses from 65 strombolian activity on Mt. Etna, during a very rare period of this style of activity at the 66 Bocca Nuova (BN) crater. Indeed, prior to our observations, on the 27th of July 2012 there 67 had only been two previous episodes of strombolian activity from BN in the preceding 68 decade, in 2002 and 2011, respectively (GVP 2013). The acquired degassing data were 69 analysed within the physical framework developed by previous studies concerning slug flow, 70 in order to seek new insights into the conduit fluid dynamics. 3 71 72 2. Bocca Nuova activity, 27th July 2012 73 During the measurement period, activity on Etna was dominated by strombolian explosions 74 from a vent in the south-west corner of the BN crater (Fig. 1, ≈ N 37.7503°, E 14.9936° see 75 supplementary materials for a .kmz file containing all relevant measurement locations). Each 76 event lasted < 4s, was ash-free, involving a single audible bang, ballistic ejection of only a 77 small number of visible pyroclasts (e.g. see supplementary video), and the subsequent rapid 78 emission of gases. The largest clasts were observed to deform in a ductile fashion in flight. 79 Between explosions, the vent passively degassed (e.g., see Fig. 2a and video in 80 supplementary material). This vent generated explosions throughout the majority of July 81 2012, in addition to small lava flows (GVP, 2013). During the measurement period, 82 prevailing winds at the crater edge carried the gas emissions in an E-SE direction (see Fig. 1). 83 84 3. Methodology 85 SO2 fluxes from the BN vent were measured between 09:32:58 and 09:59:58 GMT on July 86 27th, 2012, with two PC-synchronised Apogee-Alta U260 UV cameras, each fitted with a 16 87 bit 512 × 512 pixel Kodak KAF-0261E thermo-electrically cooled CCD array detector. Each 88 camera had a Pentax B2528-UV lens with a focal length of 25 mm, providing a ≈ 24° field of 89 view. A filter was placed in front of each lens, one centred on 310 nm and the other on 330 90 nm, and each of 10 nm full width at half maximum transmission bandwidth. As SO2 absorbs 91 in the 310 nm wavelength region, but not at 330 nm, a pair of simultaneously acquired 92 images from the cameras can be processed to yield absorbance values. The data capture and 93 analysis were achieved using the Vulcamera code (Tamburello et al., 2011b) and full details 94 on the methodology are covered in Kantzas et al. (2010). 4 95 The UV camera apparatus was located as denoted in Fig. 1, ≈ 250 m from the vent (N 96 37.7525°, E 14.9950°), providing the view of the BN crater shown in Fig. 2a and care was 97 exercised to position the cameras away from potential contamination by gases from other 98 sources; the acquisition frequency was ≈ 1 Hz. Given this close proximity to the source we 99 anticipate that error arising from light dilution was small; e.g., from scattering of radiation 100 from outside of the instrumental field of view to within it, i.e., between the camera and the 101 measured vent area, an error source which could potentially lead to an underestimation in 102 measured column amount values. This being said, it is not possible at this stage to assign a 103 definitive characterisation of measurement error from this effect, as radiative transfer has yet 104 to become a routine component of UV camera retrievals (e.g., Kern et al., 2009, 2010). The 105 same is true of light scattering within the plume, which could potentially act to cause 106 overestimation in concentration values. 107 3.1 Camera calibration 108 To calibrate the system, cells of known concentrations (100, 200, 400, 1600 ppm m with 109 manufacturer stated error budgets of ± 50 ppm m, and ± 100 ppm m for the 400 ppm m and 110 1600 ppm m cells, respectively) were placed in front of the cameras in sequence, and the 111 absorbances determined. In our measurements, the image background was the basaltic rock 112 face of the BN crater wall, as opposed to the sky, which is more conventionally used for such 113 observations. Hence, the calibration, vignetting correction (an essential step in removing the 114 inhomogeneous illumination of the detector across the field-of-view) and reference image 115 acquisition steps (see Kantzas et al., 2010 for full details) of the measurement were 116 performed by viewing the crater wall though air with minimal SO2 concentration, adjacent to 117 the rising gas plume. A rock-reflectance light source approach is also commonly used in the 118 study of planetary surfaces bodies (e.g. Hendrix et al., 2003) and in our case, this provided 119 around 40% of the UV light intensity of the background sky immediately above the crater, 5 120 e.g., a sufficiently strong source for our observations. The measurement location was also 121 free from fumarolic contamination and unaffected by gases sourced from other craters. 122 Given the variation in light scattering orientation from the background basaltic rock across 123 the camera field of view, we also investigated whether any angular dependency in cell 124 calibration across the image might be introduced due to this effect. This was achieved by 125 imaging an SO2 free region with a basaltic rock background in the Etnean summit area with 126 illumination conditions as similar as possible to those during the measurements (e.g., there 127 was a thin strip of sky in the uppermost region of the images). In particular we tested whether 128 calibration could be skewed over the angular difference between the plume gases and the 129 adjacent background rock viewing orientations in our measurements (≈ 12) by determining 130 calibration lines for a number of data points in the SO2 free image within this diameter of the 131 image centre (Fig. 2b). Plotted together (Fig. 2b) the calibration data points reveal very 132 similar calibration gradients in all cases, with an overall R2 = 0.99, leading us to exclude the 133 possibility of this effect introducing significant error. 134 135 3.2 Data Processing 136 The data analysis firstly involved detecting strombolian explosion events in the UV camera 137 records by identifying when the gas emission speed markedly increased and solid ejecta were 138 identifiable. For each such event SO2 gas masses were derived from the processed UV 139 camera SO2 concentration images using the integrated volume amount (IVA) technique 140 (Tamburello et al., 2012). With this approach, gas concentrations were integrated within an 141 appropriately chosen 2D subsection of the image immediately above the vent, of sufficient 142 size to encompass the explosive clouds to generate the IVA (Fig. 2c). Fig. 3 shows the gas 143 cloud propagation over five consecutive images following one such explosion, showing 6 144 wireframe sketches (Fig. 3a-e) of the advancing cloud, the cloud vector of motion and the 145 IVA integration area. 146 These IVAs require correction for background SO2 levels associated with the collection of 147 gases within BN following emission, as the spatial location of these varied temporally 148 throughout the acquisition in response to changing atmospheric conditions. Background 149 correction was achieved by determining integrated SO2 concentrations for two subsections of 150 the image, adjacent to the explosion, and of identical dimensions to the area used in the 151 explosion cloud 2D integration (Fig. 2c). The explosion IVA was then corrected by 152 subtracting the average of the masses within these two background areas which typically 153 agreed with one another to within ≈ 6%. For each event, the temporal peak in the corrected 154 IVA record was identified, then integration was performed between the event onset and event 155 termination to yield the explosive gas mass. For reference, video material is provided in the 156 auxiliary materials showing two acquired UV camera image time series. 157 These data were then applied to investigate total slug masses, using contemporaneously 158 acquired Multi-GAS (Aiuppa et al., 2007) gas ratio data from a unit deployed by INGV 159 (Istituto Nazionale di Geofisica e Vulcanologia) sezione di Palermo. The Multi-GAS unit was 160 located on the crater’s edge at the site shown in Fig. 1. (N 37.7409°, E 14.9953°) at a 161 distance of ≈ 200 m from the active vent and away from possible contamination sources; the 162 wind direction and speed were E-SE and 10-14 m s-1, respectively. Averaged over the 163 acquisition period, the measured Multi-GAS molar ratios were: CO2/SO2 ≈ 2.8; H2O/SO2 ≈ 164 8.5; and H2O/CO2 ≈ 3. Temporal averaging was applied due to the difficulty of isolating 165 individual explosive events in the Multi-GAS record resulting from the spatial separation of 166 the vent and the Multi-GAS unit and the time resolution of the Multi-GAS data (0.5 Hz). 167 During the entire acquisition, the ratios were relatively stable (with errors on gas ratios of ≈ 4 168 – 15% e.g. Pering et al., [2014]), and total gas masses were calculated based on the 7 169 assumption that H2O, CO2, and SO2 dominated the plume composition (e.g., Aiuppa et al. 170 2007). The molar plume composition was therefore taken to be 8% SO2, 22% CO2 and 70% 171 H2O from the Multi-GAS measurements, on which basis the explosive SO2 gas masses were 172 converted, via multiplication, using the respective mass ratios, to total gas release per event. 173 However, it is likely, as per previous studies at similar targets (e.g. Burton et al. 2007; 174 Tamburello et al. 2012), that the gas compositions from the passive and explosive 175 contributions were non-identical. Our determined total gas masses are therefore best- 176 estimates given the data available. 177 A gas flux time series was also constrained by summing the image concentrations over a 178 cross section above the vent (Fig. 2c) to generate an integrated column amount (ICA) data- 179 stream, then multiplying this by the plume speed, projected onto a vector perpendicular to 180 this cross section. The inter-event plume rise speed was determined using a cross correlation 181 technique on ICA data derived from two parallel sections of the rising plume, in periods after 182 the increase in emission speed associated with gas explosions had subsided (e.g., McGonigle 183 et al., 2005; Williams-Jones et al., 2006), with results of ≈ 5 m s-1. During the explosions 184 themselves the plume speed was constrained by frame by frame tracking of the cloud front 185 across the camera field of view. 186 187 3.3 Seismicity 188 The potential relationship between gas flux and seismic RMS (root-mean-square) was 189 investigated using signals recorded by three seismic stations (EBCN N 37.752365° E 190 14.986281°; ETFI N 37.738195°, E 15.000649°; and EBEL N 37.740238° E 15.008239°; see 191 Fig. 1 for EBCN location) belonging to the permanent network, run by INGV, Osservatorio 192 Etneo – sezione di Catania. Since these stations are located close to the summit craters (≈ 1 8 193 km away from the centre of the summit area), the seismic RMS patterns were mostly affected 194 by the temporal variations of volcanic tremor, long period (LP) and very long period (VLP) 195 events. The seismic RMS was calculated over windows of 2, 5, 10 and 30 s in two distinct 196 frequency bands: 0.05-0.5 Hz and 0.5-5.0 Hz. These bands were chosen because they contain 197 most of the energy of the seismo-volcanic signals (volcanic tremor, LP and VLP events) at 198 Mt. Etna (e.g., Cannata et al., 2013). Fig. 2d shows the seismic RMS time series preceding, 199 accompanying and following the UV camera acquisition period. The comparison between 200 seismic RMS and the gas flux data was performed using the method of Martini et al. (2009) 201 and Zuccarello et al. (2013), based on “randomised correlations”. In particular, this involved 202 considering both a zero time difference between the seismic and emission rate time series, 203 and testing different possible time lags (ranging from -10 to 10 minutes). Infrasonic signals, 204 recorded by the permanent infrasonic network, run by INGV, Osservatorio Etneo, were also 205 analysed. However, wind noise at the sensors, obscured the volcano-acoustic signals to such 206 an extent than no meaningful use of these data could be made. 207 4. Results 208 We measured 195 events over the acquisition period, which ranged ≈ 0.1 – 14 kg in SO2 mass 209 corresponding to ≈ 0.1 – 74 kg in total gas mass per event, such that we estimate that ≈ 183 210 kg of SO2 and ≈ 9.7 x 102 kg in gas overall were released explosively in this time window. In 211 contrast, the total passive SO2 release was ≈ 360 kg in this interval, calculated by integrating 212 the gas flux record over the time period, then subtracting the total explosive SO2 release. The 213 ratio of passive to active degassing was therefore ≈ 67% passive: 33% active. 214 A histogram of total gas masses for the explosions is shown in Fig. 4a, revealing a strong bias 215 towards smaller masses, with a population of > 150 in the ≈ 0.2 – 20 kg range. The interval 216 between event onsets ranged ≈ 1 – 46 s, with a modal value of ≈ 4 s and median of ≈ 5 s (Fig. 9 217 4b) and the duration of each event was <4 s, Fig. 4c shows a plot of time from burst onset to 218 that of the following slug, vs. total gas mass for each of the explosive events, revealing that 219 for a given gas mass, there is a fixed time below which no subsequent gas burst was observed 220 to occur (e.g., the shaded area in Fig.4c), In contrast, Fig 4d, a plot of time between burst 221 onset and that of the preceding slug vs. total slug mass, reveals no such feature (Fig. 4d). 222 Furthermore, no significant link was found in the between the seismicity and gas flux time 223 series data, suggesting that pressure and force change of the magma/gas mixture, within the 224 conduit, were not strongly coupled to the edifice. 225 5. Modelling 226 The first step in exploring the sub-surface processes driving the observed surficial activity is 227 to consider which conduit flow regime might be operating in this case. By combining our 228 estimated total gas masses with the ideal gas law (�� = ���, where P is gas pressure, V is 229 230 volume, � the number of moles, � the universal gas constant [≈ 8.314 J K-1 mol-1] and � temperature, respectively) at an atmospheric pressure of ≈ 69 kPa and temperature of 231 1273.15 K (e.g., an appropriate value for just above the magma surface), bubble volumes 232 ranging ≈ 0.4 – 411 m3 are derived. Assuming a conduit radius of ≈ 1 m, that the bubbles are 233 approximately as wide as the conduit, and that burst overpressure is of order one atmosphere, 234 bubble lengths of ≈ 0.1 – 53 m are generated. Given that a bubble becomes a gas slug when 235 bubble lengths exceed the conduit diameter (Davies and Taylor, 1950; Wallis, 1969), and a 236 maximum film thickness is reached (e.g. Llewellin et al., 2012), criteria which the observed 237 activity meet, we can potentially model the observed activity as being driven by bursting gas 238 slugs. 239 Slugs consist of a quasi-hemispherical nose and a base of morphology (e.g. Fig. 5) dependent 240 on the fluid dynamical regime (e.g., Davies and Taylor, 1950; Bendiksen, 1985; Campos and 10 241 Guedes de Carvalho, 1988; Nogueira et al., 2006; Araújo et al., 2012). During the ascent 242 process, the slug base has a relatively constant velocity, in contrast to the nose, which 243 accelerates due to depressurisation induced volumetric expansion (James et al., 2006, 2008, 244 2009). An annular film of falling fluid surrounds the slug body, and is important in forming 245 the trailing wake behind the slug, a feature that influences the coalescence of neighbouring 246 slugs (Pinto et al., 1996) and contributes to the generation of turbulence (Krishna et al., 247 1999). Slug characteristics are controlled by conduit and magmatic parameters, which also 248 determine the likelihood of bubble stability. The dimensionless inverse viscosity, Nf , can be 249 used to investigate the properties of slugs as follows: 250 �� = 251 where � is magma density, � � √� �� (1) magma dynamic viscosity, g the acceleration due to gravity 253 and �� the conduit radius. We assign a magmatic density of 2600 kg m-3 in line with the 254 column (without slugs). Whilst we measured the vesicularity of a single ejectile clast (34%; 255 collected during similar activity from the same vent on the 25th of July) we abstained from 256 using this single datum to modify the above density estimate, given that this provided no 257 constraint on vesicularity at depth. Furthermore, we found that our model runs were rather 258 insensitive to uncertainty in density. For the remaining parameters we apply 259 Pa s, g = 9.81 m s-2 and �� = 0.5 - 1.5 m, in keeping with existing literature estimates for 252 literature estimate of James et al. (2008) as being broadly representative of the bulk magma = 100 - 1000 260 similar activity (e.g. Seyfried and Freundt, 2000), resulting in an Nf range of 8 – 423. 261 According to Campos and Guedes de Carvalho (1988), for Nf values <500 wakes will be 262 closed and axi-symmetric such that turbulence is limited. 263 Another aspect to consider is the net magma motion and hence the validity of assuming a 264 stagnant magma column as has been the case in previous volcanic slug flow models (e.g. 11 265 James et al., 2008; 2009; Del Bello et al. 2012, in both cases concerning Stromboli). Based 266 on visible observations of the activity (see visible imagery in supplementary material), the 267 magmatic flux from the vent was negligible, hence, in common with the prior models, we 268 also assume there was no net vertical magmatic flux in this case. 269 In the absence of a previously developed model to characterise near-surface multi-slug flow, 270 we resort to the single slug model of James et al. (2008), to probe first order estimates of the 271 slug parameters. Following James et al. (2008) the position and length of an ascending slug 272 as a function of time can be derived by numerically solving: 273 � � + �′ � = � � �−� ℎ− − �� − �ℎ− − ���− (2) 275 where h is the height of magma overlying the slug nose, � is the ratio of specific heats of the 276 conditions and dots representing time derivatives. The initial gas pressure, P0, is set to 277 � �ℎ + � where ℎ is the initial liquid height above the slug and P is atmospheric pressure 274 278 279 280 281 282 gas (here we use a value of 1.4) and L is slug length, with zero subscripts indicating initial at the vent exit. �′ is the squared ratio of the conduit and slug (� ) radii: �′ = � � . (3) where � is calculated by determining the thickness of the falling film al., (2012) and subtracting this from �� ; ′ = . + . tanh . − . ′ ′ from Llewellin et is found from: log �� . 283 h, within equation 2, is a function of the constant rise velocity usl of the slug base: 284 � = ��√ ��� , (4) (5) 12 285 where the Froude number, Fr, appropriate for the given inertial-viscous regime is determined 286 using the simplification of Llewellin et al., (2012): [ +( . − . 287 �� = . 288 The range of determined Nf values, 8 – 423, therefore gives estimates of film thickness of ≈ 289 0.13 to 0.43 m, and slug base velocities of ≈ 0.24 – 1.82 m s-1. 290 We calculate the depth at which the ascending bubbles are sufficiently long to be considered 291 as slugs by initialising the model at depths greater than this point (e.g., where bubble length is 292 twice the conduit radius). Using mid-point values of 1 m for conduit radius and 500 Pa s-1 for 293 viscosity (e.g. Nf = 46, 294 170 m for the largest slugs, and only ≈ 5 m for the vast majority of bursts within the median 295 mass range (e.g. Fig 5a). Following this, we generate estimates of slug lengths at burst, using 296 equation 2, of ≈ 3 – 27 m. By combining these constraints with estimates for slug rise speeds, 297 we infer minimum rise times of ≈ 93 – 708 s from the slug transition depths to the surface for 298 the largest slugs. 299 In a multi slug regime, the dynamics will clearly be rather more complex than for single slugs 300 (e.g. Krishna et al., 1999; Pinto et al., 1998, 2001). As such, there are a number of limits to 301 using single slug models in our case, including the possibility that the rising slugs might not 302 become conduit filling until closer to the surface than predicted by these models. 303 Furthermore, slugs will be affected by pressure variations and magma motions induced by 304 other slugs, and may coalesce with their neighbours. In a multi-slug system, slug base 305 velocities can also exceed those predicted for single-slug systems (Krishna et al., 1999), with 306 velocity fluctuations between individual slugs likely, which will further enhance slug 307 interaction and the possibility of coalescence. Furthermore, whether the slug wakes are open �� ) . ′ ] . (6) = 0.28 m, and � = 1.1 m s-1) this gives slug transition depths of ≈ 13 308 or closed will play a significant role in determining whether turbulence occurs and whether 309 rising slugs interact with their neighbours. Pinto and Campos (1996) provide the following 310 relation (appropriate to the above Nf values) to characterise the distance beyond which no 311 interaction occurs between rising slugs, termed the wake interaction length (e.g. see Fig. 5), 312 and hence within which, inter-slug coalescence becomes likely: 313 � 314 � = �� . + . × − �� . (7) This gives estimates of wake interaction lengths of ≈ 1.5 to 10.4 m, over the Nf range 8 - 423. 315 316 6. Discussion 317 6.1 Modelling and Activity Dynamics 318 The modelled slug wake interaction lengths (lmin) of ≈ 1.5 to 10.4 m are suggestive that 319 individual slugs could rise in the conduit separated by relatively little melt without 320 interacting, so long as the slugs and their wakes retain stability. As a mass of gas rises 321 through a conduit it will undergo decompressional expansion due to the reduction of 322 overhead magma. When the gas mass transitions to become a slug, at a point when the slug 323 length approaches the conduit diameter (Davies and Taylor, 1950; Wallis, 1969) and the 324 maximum film thickness has been reached (e.g. Llewellin et al., 2012), decompressional 325 expansion of the slug length continues. The slug base rises at a constant velocity (Viana et al., 326 2003) while the nose accelerates towards the magma surface. Acceleration of the slug nose 327 increases on approaching the magma surface. This process therefore enhances the chance of 328 coalescence between slugs, with slug interaction initiating around the interaction length, 329 within which the whole of a trailing slug will accelerate into the base of a leading slug, 330 whereby the slug base velocity, in tandem with the slug nose, will increase (e.g. Pinto et al., 14 331 1996) before complete capture at the point of coalescence. By combining our modelled slug 332 interaction lengths of ≈ 1.5 to 10.4 m with estimates for slug base rise velocity of ≈ 0.24 – 333 1.82 m s-1, and the mean delay between events of ≈ 4 s, we can estimate a separation distance 334 between rising slugs of ≈ 0.96 – 2.2 m, clearly within the modelled slug interaction lengths. 335 It is therefore feasible that the observed rapid activity could occur with potential inter-slug 336 interactions leading to slug coalescence events. 337 With a closed and axisymmetric wake, there will be little disruption of fluid following the 338 passage of a rising slug. This could therefore allow the occurrence of the observed high 339 frequency explosive activity via the bursting of individual gas slugs. However, it is possible 340 that in a multi-slug environment, instability could still be generated by the extension of fluid 341 disturbance beyond the estimated wake interaction length (e.g. Krishna et al., 1999). Given 342 the inherently necessary estimates and assumptions for a number of parameters in our 343 analysis, it is possible that the degree of turbulence has been under-represented, and that 344 turbulent interaction of the magma-gas mixture with rising gas masses could lead to 345 instability in rising masses causing homogenous bubble morphology alterations. Despite this, 346 the majority of bubbles, in the observed activity, are estimated to transition into slugs at 347 relatively shallow depths in the conduit and �� numbers of ≈ 423 suggest limited turbulence 348 and hence relatively stable bubble morphology. Furthermore, our estimated final slug lengths 349 of ≈ 3 – 27 m for the majority of bursts are acquired through volumetric expansion, such that 350 the largest masses, which have the greatest expansion, will be most prone to coalescence 351 events. 352 In the supplementary video data and Fig. 3 there is clear evidence of events occurring in very 353 rapid succession, e.g., every few seconds around 09:55:33 GMT. In such cases, the gases 354 from adjacent bubbles are propelled from the vent in markedly different directions. Whilst we 15 355 cannot rule out the influence of factors such as vent geometry, atmospheric transport (e.g., 356 eddy generation) and the magma surface itself (e.g., topographic alterations due to vent 357 collapse or pyroclast deposit) in driving the explosive direction, we suggest that this 358 observation could hypothetically be evidence of interaction of the trailing slug with the wake 359 of the leading slug. This process can cause asymmetric deformation of the trailing slug’s nose 360 (e.g., Nogueira et al., 2006; Figueroa-Espinoza and Fabre, 2011), leading to a displacement in 361 the explosive gas release vector. 362 Fig. 4c portrays a repose gap, such that the largest slugs are characterised by relatively long 363 delays before the onset of the following event; no such feature is observed in terms of time 364 before the bursts (Fig. 4d). We also suggest here that the most likely causative mechanism is 365 slug coalescence, such that when a slug enters into the wake of the preceding Taylor bubble, 366 it is accelerated towards the bubble base (Pinto et al., 1998, 2001). Therefore, during the high 367 frequency strombolian activity reported on here, larger coalescence generated slugs could 368 form from closely spaced rising Taylor bubbles. This would then leave a longer delay before 369 the onset of the following event, e.g., explaining the repose gap. The absence of this feature 370 prior to such bursts also supports this, in the sense that a slug has no influence on those 371 preceding it. 372 We also considered whether other processes associated with strombolian volcanic dynamics 373 might provide alternate explanations for this repose gap. In particular, the rise velocities of 374 the base of slugs in a stagnant fluid are independent of mass (Viana et al., 2003), and are 375 rather defined by conduit width (notwithstanding the effects of complex geometries and 376 rheology). It is therefore unlikely that the rise speed-dependent model (Wilson, 1980; Parfitt 377 and Wilson, 1995) could account for this phenomenon. This is of course unless the slug 378 arrival times could be effectively pre-determined by the volume-related behaviour of bubbles 379 in the melt before the transition to slugs, given the estimated shallow transition depths. 16 380 The collapsing foam model (Jaupart and Vergniolle, 1988; Vergniolle and Brandeis, 1994), 381 where bubbles in traps, or accumulated as a foam, collapse to generate slugs at variable 382 temporal intervals was also considered, e.g., release of a large slug from a foam could lead to 383 a longer period of stability before the next foam collapse event. However, as the foam 384 collapse model is strongly related to storage, it could be more logical to expect this to cause 385 longer inter-event durations before the largest eruptions, to allow sufficient gas accumulation 386 in the foam/trap to take place, and as shown in Fig. 4d no such behaviour is evident. In view 387 of all of the above we cautiously suggest that the repose gap is related to the coalescence of 388 gas slugs, although, regardless of the precise driving mechanism, this observation does stand 389 as both novel and intriguing. 390 391 6.2 Mass Considerations and Comparisons 392 Whilst 195 events were measured, we can of course only discuss the implications of our work 393 with respect to the observation period, given the relatively limited acquisition duration. 394 During the measurements, the captured SO2 masses for individual bursts ranged ≈ 0.1 – 14 395 kg, somewhat lower than those reported for strombolian explosions at other targets e.g., 396 Stromboli ≈ 15 – 40 kg (Mori and Burton, 2009) and ≈ 2 – 55 kg (Tamburello et al., 2012); 397 and Pacaya (3 – 29 kg) (Dalton et al., 2010). Our Etnean measurements demonstrate ratios of 398 passive to active degassing of 67%: 33%) rather lower than those reported for Stromboli 399 (77%: 23%; by Tamburello et al., 2012; 97-92%:3-8% by Mori and Burton, 2009), in line 400 with the rather higher strombolian eruptive frequency in the former case e.g., on timescales of 401 seconds vs. minutes. Indeed, strombolian activity on Mt. Etna, whilst relatively rare in 402 comparison to the quasi-constant activity on Stromboli, does often manifest these rather 17 403 shorter inter-eruptive periods (GVP, 2013), perhaps hinting at distinct mechanisms driving 404 the eruptions in the two cases. 405 The relatively low gas masses released per event are also likely related to the weak seismic 406 strength manifested at the time of observations (Fig. 2d), consistent with a mild form of 407 strombolian activity and reduced gas supply from depth, in contrast to the stronger seismic 408 events registered in the preceding hours (see Fig 2d). Moreover, at the time of measurement 409 the volcanic tremor source centroid was roughly located beneath Etna’s North East crater at ≈ 410 2 km a.s.l. which likely masked any signal from the waning BN activity. Hence, whilst clear 411 relationships between explosive gas masses and seismic signals have been reported 412 previously at Mt. Etna (e.g. Zuccarello et al., 2013) and elsewhere e.g., on Stromboli and 413 Asama volcanoes (McGonigle et al., 2009; Kazahaya et al., 2011) no correlation is evident 414 here where the gas slugs are smaller. This is of course consistent with the model that seismo- 415 volcanic signals (such as volcanic tremor, LP and VLP events) are generated by the slug 416 and/or displaced magma moving within the conduit to generate a gas volume related seismic 417 signal, possibly in a resonant manner (O’Brien and Bean, 2008), and adds credence to the 418 near surface development of the observed activity. 419 7. Summary and Conclusions 420 Here we report the use of UV cameras to constrain erupted gas masses during strombolian 421 activity on Mt. Etna for the first time. Total gas masses per event of ≈ 0.2 – 74 kg were 422 captured, rather less than those found for this explosive style on other volcanoes, due to the 423 mild, yet very frequent (i.e. every ≈ 4 s), form of activity. This is corroborated by the 424 generally poor correlation with seismic signals, in contrast to the robust connections, evident 425 elsewhere, for instance at Stromboli (Ripepe et al., 2005; McGonigle et al., 2009). 18 426 A broad consideration into the fluid dynamical regime intimates the potential for wake 427 interaction between adjacent rising slugs, given their relatively modest separation in the 428 conduit. We also report on an observed repose gap, in which the larger slugs have longer 429 repose intervals than the smaller ones, before the following explosion. This could be 430 indicative of slug coalescence, with the larger slugs being formed by the interaction between 431 two or more slugs, leaving a relatively long delay before the arrival at the surface of the next 432 distinct slug. We estimate that these bubbles transition to full slug flow at shallow depths of < 433 170 m and that wake interaction becomes important in the upper portion of the conduit in the 434 region of greatest vertical slug expansion, hence promoting coalescence. 435 Acknowledgements 436 T. D. Pering and A. J. S. McGonigle acknowledge the support of a NERC studentship, the 437 University of Sheffield and a Google Faculty Research award. A. Aiuppa acknowledges 438 support from the European Research Council Starting Independent Research Grant 439 (agreement number 1305377). We are finally grateful to Ed Llewellin and two anonymous 440 reviewers for their reviews which have greatly improved the quality of this paper. 441 442 References 443 Aiuppa, A., Moretti R., Federico C., Giudice G., Gurrieri S., Liuzzo M., Papale P., Shinohara 444 H., Valenza M., 2007. Forecasting Etna eruptions by real-time observation of volcanic gas 445 composition. Geology 35 (12), 1115–1118, DOI: 10.1130/G24149 446 Araújo, J. D. P., Miranda, J. M., Pinto, A. M. F. R., Campos, J. B. L. M., 2012. Wide-ranging 447 survey on the laminar flow of individual Taylor bubbles rising through stagnant Newtonian 448 liquids, International Journal of Multiphase Flow 43, 131-148 19 449 Bendiksen, K. H., 1985. On the motion of long bubbles in vertical tubes, International Journal 450 of Multiphase Flow 11 (6), 797-812 451 Blackburn, E. A., Wilson, L., Sparks, R. S. J., 1976. Mechanisms and dynamics of 452 strombolian activity, Journal of the Geological Society of London 132, 429-440 453 Bluth, G. J. S., Shannon, J. M., Watson, I. M., Prata, A. J., Realmuto, V. J., 2007. 454 Development of an ultra-violet digital camera for volcanic SO2 imaging, Journal of 455 Volcanology and Geothermal Research 161, 47-56 456 Boichu, M., Oppenheimer, C., Tsanev, V., Kyle, P. R., 2010. High temporal resolution SO2 457 flux measurements at Erebus volcano, Antarctica, Journal of Volcanology and Geothermal 458 Research, 190, 325-336, DOI:10.1016/j.jvolgeores.2009.11.020 459 Burton, M., Allard, P., Muré, F., La Spina, A., 2007. Magmatic Gas Composition Reveals the 460 Source Depth of Slug-Driven Strombolian Explosive Activity, Science 317, 227-230 461 Burton, M. R., Caltabiano, T., Mure, F., Salerno, G., Randazzo, D., 2009. SO2 flux from 462 Stromboli during the 2007 eruptions: Results from the FLAME network and traverse 463 measurements, Journal of Volcanology and Geothermal Research 183 (3-4), 214-220, 464 DOI:10.1016/j.volgeores.2008.11.025 465 Campos, J. B. L. M., Guedes de Carvalho, J. R. F., 1988. An experimental study of the wake 466 of gas slugs rising in liquids, Journal of Fluid Mechanics 196, 27-37 467 Cannata, A., Di Grazia, G., Aliotta, M., Cassisi, C., Montalto, P., Patanè, D., 2013. 468 Monitoring seismo-volcanic and infrasonic signals at volcanoes: Mt. Etna case study. Pure 469 and Applied Geophysics, doi:10.1007/s00024-012-0634-x. 20 470 Chouet, B., Hamisevi, N., McGetchi, T. R., 1974. Photoballistics of volcanic jet activity at 471 Stromboli, Italy, Journal of Geophysical Research 79 (32), 4961-4976 472 Dalton, M. P., Waite, G. P., Watson, I. M., Nadeau, P. A., 2010. Multiparameter 473 quantification of gas release during weak Strombolian eruptions at Pacaya volcano, 474 Guatemala, Geophys. Res. Lett. 37 (L09303), DOI:10.1029/2010GL042617 475 Davies, R. M., Taylor, G. I., 1950. The mechanics of large bubbles rising through extended 476 liquids and through liquids in tubes, Proceedings of the Royal Society, London A200, 375- 477 390 478 Del Bello, E., Llewellin, E. W., Taddeuicci, J., Scarlato, P., Lane, S. J., 2012. An analytical 479 model for gas overpressure in slug-drive explosions: Insights into Strombolian volcanic 480 eruptions, Journal of Geophysical Research: Solid Earth 117 (B2), 481 DOI:10.1029/2011JB008747 482 Edmonds, M., Herd, R. A., Galle, B., Oppenheimer, C. M., 2003. Automated, high time- 483 resolution measurements of SO2 flux at Soufrière Hills Volcano, Montserrat, Bulletin of 484 Volcanology 65 (8), 578-586 485 Figueroa-Espinoza, B., Fabre, J., 2011. Taylor bubble moving in a flowing liquid in vertical 486 channel: transition from symmetric to asymmetric shape, Journal of Fluid Mechanics 679, 487 432-454 488 Galle, B., Oppenheimer, C., Geyer, A., McGonigle, A. J. S. Edmonds, M., Horrocks, L., 489 2003. A miniaturised ultraviolet spectrometer for remote sensing of SO2 fluxes: a new tool 490 for volcano surveillance. Journal of Volcanology and Geothermal Research 119 (1-4), 241- 491 254, DOI:10.1016/S0377-0273(02)00356-6 492 GVP, 2013. Etna. (Online) Available at: http://www.volcano.si.edu/volcano.cfm?vn=211060 21 493 Hendrix, A. R., Vilas, F., Festou, M. C., 2003. Vesta’s UV lightcurve: hemispheric variation 494 in brightness and spectral reversal. Icarus 162 (1), 1-9 495 Holland, A. S. P., Watson, M, I,, Phillips, J. C., Caricchi, L., Dalton, M. P., 2011. Degassing 496 processes during lava dome growth: Insights from Santiaguito lava dome, Guatemala. Journal 497 of Volcanology and Geothermal Research 202 (1-2), 153-166, 498 DOI:10.1016/j.volgeores.2011.02.004 499 James, M. R., Lane, S. J., Chouet, B., Gilbert, J. S., 2004. Pressure changes associated with 500 the ascent and bursting of gas slugs in liquid=filled vertical and inclined conduits, Journal of 501 Volcanology and Geothermal Research 129, 61-82, DOI:10.1016/S0377-0273(03)00232-4 502 James, M. R., Lane, S. J., Chouet, B. A., 2006. Gas slug ascent through changes in conduit 503 diameter: Laboratory insights into a volcano-seismic source process in low-viscosity 504 magmas, Journal of Geophysical Research 111 (B05201), DOI:10.1029/2005JB003718 505 James, M.R., Lane, S. J., Corder, S. B., 2008. Modelling the rapid near-surface expansion of 506 gas slugs in low-viscosity magmas, Geological Society, London, Special Publications 307, 507 147-167 508 James, M. R., Lane, S. J., Wilson, L., Corder, S. B., 2009. Degassing at low magma-viscosity 509 volcanoes: Quantifying the transition between passive bubble-burst and Strombolian 510 eruption, Journal of Volcanology and Geothermal Research 180, 81-88 511 Jaupart, C., Vergniolle, S., 1988. Laboratory models of Hawaiian and Strombolian eruptions, 512 Nature 331, 58-60 513 Kantzas, E. P., McGonigle, A. J. S., Tamburello, G., Aiuppa, A., Bryant, G., 2010. Protocols 514 for UV camera volcanic SO2 measurements, Journal of Volcanology and Geothermal 515 Research 194, 55-60 22 516 Kazahaya, R., Mori, T., Takeo. M., Ohminato, T., Urabe, T., Maeda, Y., 2011. Relation 517 between single very-long-period pulses and volcanic gas emissions at Mt. Asama, Japan, 518 Geophysical Research Letters 38 (L11307), DOI:10.1029/2011GL047555 519 Kern, C., Deutschmann, T., Vogel, L., Wöhrbach, M., Wagner, T., Platt, U. 2009. Radiative 520 transfer corrections for accurate spectroscopic measurements of volcanic gas emissions. B. 521 Volcanol. 72, 233-247, DOI: 10.1007/s00445-009-0313-7 522 Kern, C., Kick, F., Lübcke, P., Vogel, L., Wöhrbach, M., Platt, U., 2010. Theoretical 523 description of functinoality, applications, and limitations of SO2 cameras for the remote 524 sensing of volcanic plumes. Atmos. Meas. Tech. 3, 733-749, DOI:10.5194/amt-3-733-2010 525 Krishna, R., Urseanu, M. I., van Baten, J. M., Ellenberger, J., 1999. Rise velocity of a swarm 526 of large gas bubbles in liquids, Chemical Engineering Science 54, 171-183 527 Lane, S. J., James, M. R., Corder, S. B., 2013. Volcanic infrasonic signals and magma 528 degassing: First-order experimental insights and application to Stromboli, Earth and 529 Planetary Science Letters 377-378, 169-179, DOI:10.1016/j.epsl.2013.06.048 530 La Spina, A., Burton, M. R., Harig, R., Mure, F., Rusch, P., Jordan, M., Caltabiano, T., 2013. 531 New insights into volcanic processes at Stromboli from Cerberus, a remote-controlled open- 532 path FTIR scanner system, Journal of Volcanology and Geothermal Research 249, 66-76 533 Llewellin, E. W., Del Bello, E., Taddeucci, J., Scarlato, P., Lane, S. J., 2012. The thickness of 534 the falling film of liquid around a Taylor bubble, Proceeding of the Royal Society A 468, 535 DOI:10.1098/rspa.2011.0476 23 536 Martini, F., Bean, C. J., Saccorotti, G., Viveiros, F., Wallenstein, N., 2009. Seasonal cycles of 537 seismic velocity variations detected using coda wave interferometry at Fogo volcano, São 538 Miguel, Azores, during 2003-2004, J. Volcanol. Geotherm. Res. 181, 231-246 539 McGonigle, A. J. S., Hilton, D. R., Fischer, T. P., Oppenheimer, C., 2005. Plume velocity 540 determination for volcanic SO2 flux measurements, Geophys. Res. Lett. 32 (L11302), 541 DOI:10.1029/2005GL022470 542 McGonigle, A. J. S., Aiuppa, A., Ripepe, M., Kantzas, E. P., Tamburello, G., 2009. 543 Spectroscopic capture of 1 Hz volcanic SO2 fluxes and integration with volcano geophysical 544 data, Geophys. Res. Lett. 36 (L21309), DOI:10.1029/2009GL040494 545 Mori, T., Burton, M., 2006. The SO2 camera: a simple, fast and cheap method for ground- 546 based imaging of SO2 in volcanic plumes, Geophys. Res. Lett. 33 (L24804), 547 DOI:10.1209/2006GL027916 548 Mori, T., Burton, M., 2009. Quantification of the gas mass emitted during single explosions 549 on Stromboli with the SO2 imaging camera, Journal of Volcanology and Geothermal 550 Research 188, 395-400 551 Nogueira, S., Riethmuller, M. L., Campos, J. B. L. M., Pinto, A. M. F. R., 2006. Flow 552 patterns in the wake of a Taylor bubble rising through vertical columns of stagnant and 553 flowing Newtonian liquids: An experimental study, Chemical Engineering Science 61, 7199- 554 7212 555 O’Brien, G. S., Bean, C. J., 2008. Seismicity on volcanoes generated by gas slug ascent, 556 Geophys. Res. Lett. 25 (L16308), DOI:10.1029/2008GL035001 24 557 Parfitt, E. A., Wilson, L., 1995. Explosive volcanic eruptions – IX. The transition between 558 Hawaiian-style lava fountaining and Strombolian explosive activity, Geophysical Journal 559 International 121 (1), 226-232 560 Pering, T. D., Tamburello, G., McGonigle, A. J. S., Aiuppa, A., Cannata, A., Giudice, G., 561 Patanè, D., 2014. High time resolution fluctuations in volcanic carbon dioxide degassing 562 from Mount Etna. Journal of Volcanology and Geothermal Research 270, 115-121, 563 DOI:10.1016/j.volgeores.2013.11.014 564 Pinto, A. M. F. R., Campos, J. B. L. M., 1996. Coalescence of two gas slugs rising in a 565 vertical column of liquid. Chemical Engineering Science 51 (1), 45-54 566 Pinto, A. M. F. R., Coelho Pinheiro, M. N., Campos, J. B. L. M., 1998. Coalescence of two 567 gas slugs rising in a co-current flowing liquid in vertical tubes, Chemical Engineering 568 Science 53 (16), 2973-2983 569 Pinto, A. M. F. R., Coelho Pinheiro, M. N., Campos, J. B. L., 2001. On the interaction of 570 Taylor bubbles rise in two-phase co-current slug flow in vertical columns: turbulent wakes, 571 Experiments in Fluids 31, 643-652 572 Pioli, L., Bonadonna, C., Azzopardi, B. J., Phillips, J. C., Ripepe, M., 2012. Experimental 573 constraints on the outgassing dynamics of basaltic magmas, Journal of Geophysical Research 574 117 (B03204), DOI:10.1029/2011JB008392 575 Ripepe, M., Harris, A. J. L., Carniel, R., 2002. Thermal, seismic and infrasonic evidences of 576 variable degassing rates at Stromboli volcano. Journal of Volcanology and Geothermal 577 Research 118 (3-4), 285-297 25 578 Ripepe, M., Marchetti, E., Ulvieri, G., Harris, A., Dehn, J., Burton, M., Caltabiano, T., 579 Salerno, G., 2005. Effusive to explosive transition during the 2003 eruption of Stromboli 580 volcano, Geology 33 (5), 341-344 581 Ripepe, M., Delle Donne, D., Harris, A., Marchetti, E., Ulivieri, G., 2008. Stromboli 582 Volcano: An Integrated Study of the 2002-2003 Eruption. In: Dynamics of Strombolian 583 Acitivity, Calvari, S., Inguaggiato, S., Puglisi, G., Ripepe, M., Rosi, M., (Eds). Geophysical 584 Monograph Series 182, 39-48, DOI:10.1029/182GM05 585 Seyfried, R., Freundt, A., 2000, Experiments on conduit flow and eruption behaviour of 586 basaltic volcanic eruptions, Journal of Geophysical Research 105 (B10), 23,727-23,740 587 Tamburello, G., Kantzas, E. P., McGonigle, A. J. S., Aiuppa, A., 2011a. Recent advances in 588 ground-based ultraviolet remote sensing of volcanic SO2 fluxes, Ann. Geophys. 54 (2), 199- 589 208 590 Tamburello, G., Kantzas, E. P., McGonigle, A. J. S., Aiuppa, A., 2011b. Vulcamera: a 591 program for measuring volcanic SO2 using UV cameras, Ann. Geophys. 54 (2), 219-221 592 Tamburello, G., Aiuppa, A., Kantzas, E. P., McGonigle, A. J. S., Ripepe, M., 2012. Passive 593 vs. active degassing modes at an open-vent volcano (Stromboli, Italy), Earth and Planetary 594 Science Letters 359-360, 106-116 595 Tamburello, G, Aiuppa, A., McGonigle, A. J. S., Allard, P., Cannata, A, Giudice, G., 596 Kantzas, E. P,, Pering, T. D., 2013. Periodic volcanic degassing behaviour: The Mount Etna 597 example. Geophysical Research Letters 40 (18), 4818-4822, DOI:10.1002/grl.50924 598 Vergniolle, S., Brandeis, G., 1994. Origin of the sound generated by Strombolian explosions, 599 Geophys. Res. Lett. 21 (18), 1959-1962 26 600 Vergniolle, S., Brandeis, G., 1996. Strombolian explosions: a large bubble breaking at the 601 surface of a lava column as a source of sound, Journal of Geophysical Research 101 , 20,433- 602 20,448 603 Viana, F., Pardo, R., Yánez, R., Trallero, J. L., Joseph, D. D., 2003. Universal correlation for 604 the rise velocity of long gas bubbles in round pipes, Journal of Fluid Mechanics 494, 379-398 605 Wallis, G. B., 1969. One-dimensional two-phase flow. New York, NY: McGraw-Hill 606 Williams-Jones, G., Horton, K. A., Elias, T., Garbeil, H., Mouginis-Mark, P. J., Sutton, A. J., 607 Harris, A. J. L., 2006. Accurately measuring volcanic plume velocity with multiple UV 608 spectrometers, B. Volcanol. 68, 328-332 609 Wilson, L., 1980. Relationships between pressure, volatile content, and ejecta velocity in 610 three types of volcanic explosions, Journal of Volcanology and Geothermal Research 8, 297- 611 313 612 Zuccarello, L., Burton, M. R., Saccorotti, G., Bean, C. J., Patanè, D., 2013. The coupling 613 between very long period seismic events, volcanic tremor, and degassing rates at Mount Etna 614 volcano, J. Geophys. Res. Solid Earth 118, 4910-4921, DOI:10.1002/jgrb.50363 615 Figure Captions (colour on web only) 616 Figure 1: Map of the Mt. Etna’s summit showing the BN vent (red circle), the UV camera 617 location (end of red arrow), the Multi-GAS location, the wind direction (grey arrow) and the 618 seismic station EBCN (black circle). 619 Figure 2: a) Strombolian activity from the vent at the south-west corner of Bocca Nuova; 620 image taken at the time and location of our acquisitions; b) gas free 310 nm camera image 621 showing four pixel regions used to investigate the angular variation in cell calibrations using 27 622 rock as the measurement background; the resulting plotted calibration data (cell concentration 623 vs. measured absorbance with points colour matched to the corresponding pixel region) show 624 good agreement between the four regions and a collective R2 >0.99; c) UV camera gas 625 concentration image of BN showing IVA1, the area used to determine erupted gas masses 626 with reference to two background areas: IVA2 and IVA3, and ICA1 and ICA2, which were 627 used to calculate gas emission rates as detailed in the main text; and d) Seismic RMS from 628 stations EBCN and EBEL throughout July 27th 2012 (ETFI omitted to provide greater figure 629 clarity), showing the period of intense strombolian activity. 630 Figure 3: A sequence of cropped UV camera gas concentration images to illustrate a single 631 strombolian event and determination of SO2 concentration (images 1-5); alongside are 632 wireframe representations of the burst front for each image (a-e); the red box indicates the 633 area used to produce the integrated volume amount (IVA) from Fig 2c; and red arrows 634 indicate two distinct burst vectors for the main burst in images 1-5 and a subsequent burst in 635 image 5, respectively which with points x1 and x2 denoting two burst origins. 636 Figure 4: histograms showing a) the mass distribution of the erupted slugs; b) the inter-slug 637 duration timing distribution (modal value of ≈ 4 s); log-log plots showing c) the inter-slug 638 duration after each burst vs. that burst’s gass mass, with a blank area indicated, termed the 639 repose gap (discussed more fully in the text), and d) the inter-slug duration before each burst 640 vs. that burst’s gas mass. 641 Figure 5: Morphology of a gas slug, including the most important features. In addition, two 642 possible slug formation theories are illustrated: 1) via coalescence of bubbles; and 2) via the 643 collapsing foam model. 644 28 645 Figure 1 646 647 648 649 650 651 652 653 29 654 Figure 2 655 656 657 658 659 660 661 662 663 664 30 665 Figure 3 666 667 668 669 670 671 672 673 31 674 Figure 4 675 32 676 Figure 5 677 678 33