The 2007 eruption of Stromboli volcano: Insights from real-time measurement of the volcanic gas plume CO2/SO2 ratio

Journal of Volcanology and Geothermal Research 182 (2009) 221–230

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research

j ourna l homepage: www.e lsev ie r.com/ locate / jvo lgeores

The 2007 eruption of Stromboli volcano: Insights from real-time measurement of thevolcanic gas plume CO2/SO2 ratio

Alessandro Aiuppa a,b,⁎, Cinzia Federico b, Gaetano Giudice b, Giovanni Giuffrida b, Roberto Guida b,Sergio Gurrieri b, Marco Liuzzo b, Roberto Moretti c, Paolo Papale d

a CFTA, Università di Palermo, Palermo, Italyb Istituto Nazionale di Geofisica e Vulcanologia, sezione di Palermo, Italyc Istituto Nazionale di Geofisica e Vulcanologia, sezione Osservatorio Vesuviano, Napoli, Italyd Istituto Nazionale di Geofisica e Vulcanologia, sezione di Pisa, Italy

⁎ Corresponding author. CFTA, Università di Palermo,E-mail address: [email protected] (A. Aiuppa).

0377-0273/$ – see front matter © 2008 Elsevier B.V. Adoi:10.1016/j.jvolgeores.2008.09.013

a b s t r a c t

Article history:

Keywords:Stromboliplume chemistrymagma degassing

omboli in February–April 2007 offered a unique chance to test our currentunderstanding of processes driving the transition from ordinary (persistent Strombolian) to effusive activity,and the ability of instrumental geophysical and geochemical networks to interpret and predict these events.Here, we report on the results of two years of in-situ sensing of the CO2/SO2 ratio in Stromboli's volcanic gasplume, in the attempt to put constraints on the trigger mechanisms and dynamics of the eruption. We showthat large variations of the plume CO2/SO2 ratio (range, 0.9–26) preceded the onset of the eruption (sinceDecember 2007), interrupting a period of relatively-steady and low ratios (time-averaged ratio, 4.3) lastingfrom at least May to November 2006. By contrasting our observations with numerical simulations of volcanicdegassing at Stromboli, derived by use of an equilibrium saturation model, we suggest that the pre-eruptiveincrease of the ratio reflected an enhanced supply of deeply-derived CO2-rich gas bubbles to the shallow-plumbing system. This larger-than-normal ascent of gas bubbles was likely sourced by a 1–3 km deep gas–melt separation region (probably a magma storage zone), and caused faster convective overturning ofmagmas in the shallow conduit; an increase in the explosive rate and in seismic tremor, and finally thecollapse of the la Sciara del Fuoco sector triggering the effusive phase. The high CO2/SO2 ratios (up to 21)observed during the effusive phase, and particularly in the days and hours before a paroxysmal explosion onMarch 15, 2007, indicate the persistence of the same gas source; and suggest that de-pressurization of thesame 1–3 km deep magma storage zone could have been the trigger mechanism for the paroxysm itself.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

The persistent emission of a volcanic gas plume on Stromboli(Fig. 1), in the Aeolian Islands (Southern Italy), is the “continuous”counterpart of the “discrete and rhythmic” explosions characteristic ofthe volcano's world-well known mild Strombolian activity. Thisvolcanic gas plume consists of an atmospheric dispersion ofvolcanogenic gaseous volatiles (H2O, CO2, SO2 and HCl in roughly1:0.2:0.02:0.02 proportions; Allard et al., 1994, in press; Burton et al.,2007a) and metal-rich volcanic aerosols (Allard et al., 2000).

Volcanic gases represent a source of otherwise inaccessibleinformation on the volcano's dynamics, and it is no doubt they playa central role on the most various aspects of Stromboli's behaviour. Ithas recently been pointed out that Stromboli volcano emits through

Palermo, Italy.

ll rights reserved.

its summit plume more gas than potentially contributed by degassingof the erupted magma (Allard et al., 1994). This has been taken as anevidence of degassing-driven continuous magma convection into arelatively shallow (b1 km) magma reservoir, with degassed non-erupted magma sinking back into the conduit and being replaced bythe ascent of gas-rich less-dense magma (Harris and Stevenson, 1997;Stevenson and Blake, 1998). Volcanic gases have also been demon-strated to account for a large volumetric fraction of individualStrombolian explosions (Chouet et al., 1974; Ripepe et al., 1993); andthe ascent, accumulation and coalescence of gas bubbles in theconduit is thought to be the source mechanism for the formation ofthe gas slugs triggering the rhythmic Strombolian explosions and therelated very-long period seismicity (Ripepe et al., 2002; Chouet et al.,2003). The magmatic gas phase is also heavily implicated in thegeneration of Stromboli's paroxysms: these gas-drivenmore energeticevents (Barberi et al., 1993; Rosi et al., 2006) being likely triggered bythe fast ascent of highly-vesicular magma (Bertagnini et al., 2003;Métrich et al., 2005) or gas slugs (Allard, 2007) into the shallowvolcano reservoir. Notably, the post-hoc interpretation of volcanic gas

Fig. 1. The persistent emission of a volcanic gas plume on Stromboli volcano.

222 A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

composition data has offered the only precursory observation to aparoxysm, hitherto (Aiuppa and Federico, 2004).

Despite a better knowledge of volcanic gas plume compositions isthus essential to put constraints on degassing and eruptive phenom-

Fig. 2. Map of Stromboli volcano, showing the location of the automatic Multi-GAS on

ena, Stromboli's volcanic gas emissions have only occasionally beencharacterised in the past (Allard et al., 1994, 2000, in press; Aiuppaand Federico, 2004; Burton et al., 2007a). This paucity of volcanic gasmeasurements has been determined by many factors, including the

the summit crater's area at Pizzo Sopra la Fossa (triangle). SDF: Sciara del Fuoco.

223A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

inherent risks associated to accessing the open vents of the volcano,the absence of representative high-temperature fumaroles in an areaof safe access, and the current analytical limitations in remotely-sensing the two main volatiles, H2O and CO2, in a volcanic gas plume.

Very recently, we have demonstrated the application of a newly-developed device for the automatic continuous measurement of CO2

and SO2 concentrations in a volcanic gas plume (Aiuppa et al., 2007),based on the Multi-GAS technique (Aiuppa et al., 2005; Shinohara,2005; Aiuppa et al., 2006); and we have shown that increasing CO2/SO2 ratios in the plume allowed to forecast the onset of the recenteruptive events of Mount Etna volcano in 2006 (Aiuppa et al., 2007),highlighting the potential of this new technique in volcano monitor-ing. A similar Multi-GAS device has been recently installed on thesummit crater area of Stromboli (Fig. 2), and has been measuringvolcanic plume composition on regular basis since May 2006. Here,we report on the volcanic gas measurements carried out at Stromboliby the automatic Multi-GAS, during two years of observationsencompassing the recent eruptive phase of February to April 2007(this issue). We show that that volcanic gases emitted immediatelyprior to and during the eruption were remarkably different incomposition (richer in CO2) from those typical of ordinary Strombo-lian activity; and we interpret these anomalous high CO2/SO2 plumeratios as an evidence of enhanced supply of deeply-derived CO2-richgas bubbles to the shallow-plumbing system, triggering the transitionfrom ordinary (mild Strombolian) to anomalous (effusive to highly-explosive) volcanic activity. We also compare our experimental datawith results from an equilibrium saturation model (Moretti et al.,2003), which provides a quantitative interpretation of degassingprocesses on Stromboli, and contributes to extend further ourunderstanding on the volcano pluming system.

Fig. 3. TheMulti-GAS installed on the Pizzo Sopra la Fossa area at Stromboli. The configuration1: IR spectrometer; 2: electrochemical sensors; 3: datalogger; 4: Radio-modem; 5: pneuma

2. Study area, materials and methods

2.1. Stromboli and its 2007 eruption

Stromboli, the northernmost island of the Aeolian archipelagos, risesfrom a depth of 1500–2000 m from the Tyrrhenian Sea floor, withits emerged top, less than 100 ky old, rising up to 900 masl. The volcaniccomplex lies on a 15–20 km thick metamorphic continentalcrust (Morelli et al., 1975), and it is composed of volcanic productswhich exhibit a variety of petrologic features, from calc-alkaline toshoshonitic (Francalanci et al., 1989; Peccerillo, 2001; Corsaro et al.,2005; Landi et al., 2006).

Stromboli historical activity is characterised bya persistent degassingand mild Strombolian activity (Rosi et al., 2000), consisting of rhythmicand short-lived (seconds) explosions driven by the fast ascent of gasslugs within the magma filled conduits. These “normal” Strombolianexplosions (occurring at an average frequency of 1 event every fewminutes) consist of gas jets transporting scoriae, molten lava fragmentsandbombs, deriving from the partly-degassed crystal-richmagmafillingthe upper part of the conduit (Bertagnini et al., 1999; Métrich et al.,2001). Usually, the emitted products reach between tens to a fewhundred meters of height and are deposited in proximity of the craterarea. The chemical composition of gases emitted during Strombolianexplosions differ from those released during quiescent (passive) degas-sing, suggesting a deep (between 2.7 and 0.8 km below the vents) sourceregion for the gas slugs producing the explosions (Burton et al., 2007a).

Rarer more powerful explosions (major explosions or paroxysm,following Barberi et al., 1993) produce high eruptive columns andemit huge amount of blocks, lapilli and ash, covering the summit areawith a fallout deposit. These more energetic explosions occur with a

of theMultiGAS, and its main internal components, are detailed on the upper-left panel:tic group (pump, filter, electro-valve); 6; electrical-connectors.

Fig. 4. (a) An example of 30 minute cycle acquired from the automatic MultiGAS.Parallel variations of the concentrations of CO2 (right scale, black line) and SO2 (leftscale, grey line) are observed; (b) Time-averaged ratio for the 30 minute measurementcycle, obtained from the gradient of the best-fit regression line in a CO2 vs. SO2 scatterdiagram.

224 A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

frequency lower than 1 event/year, and erupt nearly-aphyric volatile-rich pumice fragments simultaneously with the crystal-rich scoriae(Bertagnini et al., 2003). Both magmas have the same potassium-rich(HK) basaltic composition.

Effusive eruptions occur even less frequently (3 eruptions from 1985to 2007), and are systematically characterised by failure of thegravitationally-unstable NE volcano's flank (the Sciara del Fuoco), givingrise to the draining of lava (typically a few tens of Mm3 are emittedduring each effusive eruption; e.g., 11∙106 m3 during the 2002–2003eruption, Landi et al., 2006) from the crystal-rich upper feeding system.

Themost-recent effusive eruption at Stromboli started on the 27th ofFebruary 2007with the opening of a fracture field on the eastern edge ofthe Sciara del Fuoco (at the base of the NE crater) at the altitude of about550masl (see Ripepe et al. and Barberi et al., this issue, for a descriptionof the eruption). The initial stages of the eruptionwere characterised byexceptionally-higheffusion rates (up to 20m3/s; Calvari et al., 2007) andby fast mass movements within the Sciara del Fuoco. In the followingdays, lava effusion rates progressively declined, and the eruptive ventmoved to around 400 masl, with the formation of a lava field along theSciara del Fuoco, quickly reaching the coast. After a virtual exhaustion oflava effusion, a new effusive vent opened at an altitude of about 500 mon March, 9; although the lava flow was well sustained, it lasted only24 h. In the following days, the eruption continued from the 400mvent,with a decreasing trend of the effusive rate. This trend of apparentexhaustion of volcanic activity was interrupted on March, 15 at 20:37GMT, when a paroxysmal explosion occurred at the summit craters,giving rise to a 3000–3500masl high ash column (NOAA-AVHRR data—INGV report) and the deposition of bombs, blocks, and coarse ash on theupper flanks of the volcano. Preliminary analysis of glass fragmentsrevealed a basaltic composition for the pyroclasts emitted during theparoxysmal event of March 15 (http://www.ct.ingv.it/Report/RPTVGPTR20070317.pdf), in analogy with the products emitted duringthe paroxysmal event of April, 5, 2003 (Métrich et al., 2005; Rosi et al.,2006). Lava effusionwasmore vigorous in the hours after the paroxysm,and persisted, though at a declining rate, until April, 2. During the entireeffusive phase, the summit craters showed avirtually-total exhaustionofthe explosive activity but the persistence of VLP seismicity, consistentlywith what observed during the 2002–2003 effusive eruption (Ripepe etal., 2005). The entire eruptive event was also accompanied by aprogressive contraction of the volcanic edifice, as recorded by the GPSand clinometric network (Bonaccorso et al., 2007 and this issue).

2.2. Measurements: the Multi-GAS

This paper reports on a set of systematic measurements of the CO2/SO2 ratio in Stromboli's plume, acquired by a fully-automated Multi-GAS device installed on May 2006 on the Pizzo Sopra la Fossa area(Figs. 2 and 3). A few measurements were occasionally taken duringJanuary–May 2006 in the same site, using a similar portableinstrumentation. The Pizzo Sopra la Fossa area site is located ~300 mSE of Stromboli's summit open vents, and is often fumigated by thevolcanic gas plume (Aiuppa and Federico, 2004) because of thedominant SE-trending winds (Brusca et al., 2004) and the peculiarmorphology of the volcano's summit. The Multi-GAS (Multi-sensorGas Analyzer System; Aiuppa et al., 2007), developed at the IstitutoNazionale di Geofisica e Vulcanologia, Sezione di Palermo, measuresreal-time the concentrations of CO2 and SO2 in a volcanic gas plume byintegrating an infrared spectrometer (for CO2 determination; modelGascard II, calibration range, 0–4000 ppmv; accuracy, ±2%; resolu-tion, 0.8 ppmv) and an electrochemical sensor specific to SO2 (SO2-S-100 model, Membrapor®, calibration range, 0–200 ppmv; accuracy,2%; resolution, 0.5 ppmv). In order to chemically-sense the two abovespecies, the volcanic gas plume is actively pumped during themeasurements (at an average flow rate of 0.6 l/m) into infrared andelectrochemical cells (working in series). Both sensors are connectedto a data-logger board enabling data capture and logging. Because of

power consumption requirements (~870 mA are required to operatethe device, which is fed by 2 wet 100 A/h batteries connected to aphotovoltaic module), the Multi-GAS is daily operated at Stromboliduring 4 cycles per day, each lasting 30min. During operation, data arecaptured from the sensors every 9 s (in fact each measurementcaptured and logged represents the median of 9 samples collectedwith 1 s frequency), for a total of 200 determinations of CO2 and SO2

concentrations during each 30-minute cycle (Fig. 4a). At the end ofeach cycle, a radio link operates automatic data transfer from theremote Multi-GAS to the base station in Palermo, where data areelaborated. The data are also stored in a non volatile memory insidethe Multi-GAS data logger, and the memory is erased normally once aweek. The post-acquisition elaboration involves the calculation of theCO2/SO2 ratio from the acquired data (note that the actualconcentration of each volatile is more dependent on atmosphericprocesses controlling plume transport and dispersion rather than onvolcanic processes). At this aim, the procedure adopted as defaultinvolves calculating a time-averaged ratio for each 30 minutemeasurement cycle (overall uncertainty in the derived ratios,≤20%). This is obtained by taking the gradient of the best-fitregression line in a CO2 vs. SO2 scatter diagram (Fig. 4b; see alsoAiuppa et al., 2006). A measurement cycle is considered null (e.g., noratio is calculated) when CO2 and SO2 concentrations are below fixedthreshold values (e.g., SO2b6 ppm): this occurs any time the volcanicgas plume is not fumigating the Pizzo Sopra La Fossa area because ofunfavourable wind conditions, and gives rise to many gaps in theacquired dataset.

3. Results

Fig. 5 is a plot of CO2/SO2 ratios in Stromboli's volcanic gas plumein the 2006–2007 period. In the following sections, the time evolutionof the ratio is outlined distinguishing between four distinctive stagesin the most-recent eruptive activity of the volcano.

Fig. 5. Time evolution of the volcanic plume CO2/SO2 ratio during 2006–2007. The timing of the main volcanic events is also shown.

225A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

3.1. Ordinary Strombolian activity (January–November 2006)

The year 2006 was a period of relatively “normal” activity atStromboli, with eruptive activity being confined to the summit areaand consisting of mild Strombolian activity from the open vents on thecrater's terrace. In this time interval, measurements of the CO2/SO2

ratio were sporadic until May 2006 and more systematic from thenonward, after the permanent Multi-GAS was installed. Both discrete(pre-May) and permanent (May to November) Multi-GAS measure-ments point to relatively-steady and low CO2/SO2 ratios (range, 1.6 to10.3; mean, 4.3) during this period (Fig. 5). In particular, CO2/SO2

ratios of ~5 were observed in May and August 2006, when about 15–20 Strombolian events per hour were observed on average onStromboli (INGV-CT report, http://www.ct.ingv.it/Report). CO2/SO2

ratios then declined to about 3 in late August to September 2006,when a slight decrease in the frequency of Strombolian events(averaging at ~10 events/h; INGV-CT report, http://www.ct.ingv.it/Report) were consistently observed. A slight increasing trend of theCO2/SO2 ratios was observed in early November 2006 (up to 5).

The time-averaged CO2/SO2 ratio of 4.3 for January–November2006 is lower than the ratio of 8 (Fig. 5), proposed by Burton et al.(2007a,b) as distinctive of Stromboli's time-averaged emissions, basedon repetitive FTIR measurements between mid-2000 and September2002 (the CO2-rich [CO2/SO2 ratiosN20] gas phase emitted duringindividual explosions representing a negligible contributions tovolatile budget, dominated by quiescent degassing between explo-sions; Allard et al., 1994, in press). The low CO2/SO2 ratio of 4.3 for theJanuary–November 2006 period supports that plume emissions weresustained by degassing of a relatively-degassed (e.g., CO2-depleted)magmatic system at that time.

3.2. The pre-eruptive phase (December 2006–February 2007)

A volcanic gas plume with previously-unrecorded high CO2/SO2

ratios was emitted by Stromboli's vents during early December 2006(Fig. 5). This cycle of increase and than decrease of the ratio lasted~20 days and reached a maximum on December, 5 (when a ratio of 26was measured; Fig. 5), preceding a major explosion occurring on thesummit craters on December, 15 2006 at 12.29 GMT by ~10 days(http://www.ct.ingv.it/Stromboli2002). Mild Strombolian activity

continued in late December 2006 to late January 2007, ourinterpretations of this phase of activity being however hampered bya gap of data lasting almost one month (Fig. 5), due to malfunctioningof the MutiGAS. Since late-January early-February 2007, volcanicactivity gradually intensified, however, with a 3-fold increase in theexplosive rate at the craters (Ripepe et al., 2007 and this issue), a 4-fold increase in the amplitude of seismic tremor (Ripepe et al., 2007and this issue; Patanè et al., 2007), and a significant acceleration inrate of displacement of the Sciara del Fuoco sector (Casagli et al., thisissue). MultiGAS measurements taken since early February 2007showed large fluctuations of the CO2/SO2 ratios (range, 2–24), and thepersistence of periods of anomalously high ratios; the most striking ofwhich occurred on February 13 (CO2/SO2 ~24; Fig. 5), just one daybefore a sharp acceleration in the seismic tremor (N8 μm/s; Ripepeet al., 2007 and this issue) and in mass displacements on the volcanoon February 14, as a prelude to the imminent failure of the Sciara delFuoco and onset of the effusive phase on February, 27.

3.3. The effusive eruption (February 27–April 2, 2007)

The onset of the effusive phase on February 27, with a lava flowoutpouring from a fracture at the base of theNE crater (Ripepe et al. andBarberi et al., this issue), was associated with the lowest CO2/SO2 ratiosin our dataset (from 1.3 at 17 GMT to 0.9 at 23 GMT; Fig. 5). In thefollowing days, however, as lava effusion was still persisting inside laSciara del Fuoco,with significantpulsations in theeffusion rates, CO2/SO2

started to increase again (note the large increase in the CO2/SO2 ratiobaseline, being on average at ~11 during the effusive phase). CO2/SO2

ratios were particularly high in two main phases (Fig. 5): (i) between 7and 9 March (CO2/SO2 in the 15.6–18.3 range), marking a phase ofrenewal of eruption intensity with opening of a new eruptive vent at500 masl in the la Sciara del Fuoco (Barberi et al., this issue); and (ii) inthe days (March 13, CO2/SO2 ~21) and hours (March 15 at 17.30 GMT,CO2/SO2 ~19) before the paroxysmal explosion of March 15 at 20:37GMT (cfr. 2.1). The latter observations, together with a precursor 10-foldincrease in the CO2 flux from the summit craters before the paroxysm,will be detailed elsewhere (Aiuppa et al., in preparation). Immediatelyafter the paroxysm, measured CO2/SO2 ratios declined to ~8, and thenoscillated between minima (4) and maxima (25) during the conclusivephase of the effusive eruption (lava effusion ended on April, 2).

226 A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

3.4. The post-eruptive phase (April–December 2007)

As already observed after the 2002–2003 effusive event (Ripepeet al., 2005), the transition from effusive activity (exhausted by April2) to Strombolian activity was not abrupt but gradual, and “ordinary”Strombolian explosions completely resumed only in mid-June 2007.This transition stage of activity of the volcano in April toMay 2007wasassociated with the persistence of rather-unusual pulsating ashemissions from the volcano's summit craters (which mostly emittedaltered lithic fragments); and by an exceptional seismic crisis in thefirst half of May 2007, when N70 very long period (VLP) events perhour were observed (Martini et al., this issue). CO2/SO2 ratiosaveraged at ~9 during April to May 2007, but high ratios (15 to 25)occasionally occurred throughout the whole period (Fig. 5). Only frommid May a new steady state condition seemed to be attained; withrather-constant CO2/SO2 ratios at ~9, but still a factor ~2 larger than inthe January–November 2006 period. As “ordinary” Strombolianactivity resumed at summit craters on June 2007, the frequency ofexplosions maintained rather low (5–10 explosions/h) until Septem-ber 2007, when it increased to 10–15 explosions/h (http://www.ct.ingv.it/Report). This more intense Strombolian activity was antici-pated, since August 2007, by CO2/SO2 ratios in the plume slightlyhigher than in the months before; with peaks of the ratio up to 17,occurring shortly-before more powerful explosions (as revealed bythe detection by the INGV-OV seismic network of medium to highamplitude explosion quakes; http://www.ct.ingv.it/Report).

4. Discussion

The large variations of CO2/SO2 ratios (range, 0.9–26) during twoyears of MultiGAS observations are evidence of the dynamic nature ofvolcanic degassing at Stromboli (Allard et al., 1994; Harris andStevenson, 1997; Stevenson and Blake, 1998; Allard et al., in press).Our large range of measurements support the idea that CO2 and SO2

undertake differential degassing from basaltic magmas upon theirascent and accumulation within the crust (Bertagnini et al., 2003;Métrich et al., 2004, 2005; Spilliaert et al., 2006a,b; Aiuppa et al., 2006,2007; Burton et al., 2007a,b), because of their contrasting solubilitiesand pre-eruptive contents in magmas (e.g., Carroll and Halloway,1994). They also demonstrate that the emission of a high CO2/SO2

ratio gas-phase from basaltic volcanoes is not an exclusive feature ofsyn-eruptive explosive degassing (Allard et al., 2005; Burton et al.,2007a,b), but can also take place during quiescent (passive) degassing(e.g., Aiuppa et al., 2006, 2007).

The changing nature of volcanic activity at Stromboli during ourobservation periods is probably the cause for the large variations in

Fig. 6. (a, left) Model dissolved CO2 (thin curves) and SO2 (thick curves) contents in Strombeither 0.2 (grey curves) and 2.1 (black curves) wt.% are shown; (b, right) Model CO2/SO2 rarange.

volcanic gas compositions; whilst a much narrower compositionalrange seems to characterise the “ordinary” activity of the volcano (e.g.January–November 2006; Fig. 5). In order to quantitatively interpretour observations, and in the attempt to provide evidence for thiscause-effect dependence between the volcano's activity state and thechemistry of the volcanic gas plume, we first use the equilibriumsaturation model of Moretti et al. (2003) to numerically calculate thepressure-dependent evolution of the CO2/SO2 ratio in the gas phase atequilibriumwith Stromboli's magmas (in a range of pressure relevantto Stromboli's magmatic plumbing system; Fig. 6). We then combinethe results of these numerical simulations with our present under-standing of Stromboli's shallow plumbing system (Fig. 7) to constrainthe sort of key informations on volcanic processes that are potentiallyaccessible from the observation of the CO2/SO2 ratios in the volcanicgas plume. We finally apply our dataset and the results of numericalmodelling to derive inferences on the key source mechanismscontrolling the triggering and the evolution of the 2007 unrest ofStromboli.

4.1. Modelling the CO2/SO2 ratio in Stromboli's gas emissions

The Moretti et al. (2003) saturation model is used here to calculatethe equilibriumcomposition of themagmatic gasphase (CO2–H2O–SO2–

H2S system) coexistingwith Stromboli's shoshoniticmelt at a given setofP–T–X conditions. The model, which has been updated at various steps(Moretti and Papale, 2004; Moretti, 2005; Moretti and Ottonello, 2005),has recently beenusedbyAiuppa et al. (2007) tomodel volcanic gasdatafrom Mount Etna. In brief (the reader can refer to the above papers formore detailed descriptions), theMoretti et al. (2003)model is based on:i) regular mixture approach (H2O–CO2–melt saturation surface; Papale,1999; Papale et al., 2006); ii) polymeric treatmentof silicatemelts (sulfursolubility and speciation in melts, oxidation state of both melt and gasphase); iii)multiple gas-phase species equationof state (BelonoshkoandSaxena, 1992); iv) mass balances for H2OTOT, CO2

TOT, STOT, where thesuperscript refer to the total (exsolved+dissolved) content of the givenvolatile. All the information are exchanged among calculating modules,modifying composition and redox quantities up to getting modelfunctions zeroed by means of globally convergent non-linear method(Press et al., 1991). This procedure ensures internal consistency anddisplays all the non-linearities implicit into a complex modellingaccounting for component interactions.

The model has been initialized with a set of values distinctive ofStromboli's volcanism. Melt volatile-free composition was taken byaveraging data in Bertagnini et al. (2003). Degassing was modelled asan isothermal process at T=1150 °C, as from average melt inclusionhomogenization temperatures (Métrich et al., 2001; Bertagnini et al.,

oli's melt in the 300 to 0.1 MPa range. Model results for model runs with CO2TOT

fixed attios in the equilibrium gas phase coexisting with Stromboli's melt in the 300 to 0.1 MPa

Fig. 7. Schematic illustrative section of Stromboli's shallow plumbing system in the 0–4 km depth range. The evolution of the modelled CO2/SO2 ratios over the same depth-pressurerange (calculated by the equilibrium saturation model as described in text) are also shown by the vertical bars on right part of the plot (for model runs with CO2

TOT of 0.2 wt.% and2.1 wt.%, respectively). The dyke-conduit system imaged bymodelling of seismic (Chouet et al., 2003; La Rocca et al., 2004) and GPS (Mattia et al., 2004) data is thought to occupy theshallowest part of the feeding system, with a transition from dyke to conduit at ~300–400 m below summit vents (as derived from the source depth of VLP seismicity; Chouet et al.,2003; Ripepe et al., 2005). The system is filled with a high-porphyric (HP) magma emitted as scoriae during persistent Strombolian activity, and is in a state of persistent convection(Harris and Stevenson, 1997; Stevenson and Blake, 1998). Assuming that gas–melt separation occurs at the top of this convecting system (at atmospheric pressure) fixes an upperlimit for the CO2/SO2 ratio in the degassed magmatic gas phase at 0.8–8 (as supported by the pressure-depth dependent evolutions of modelled CO2/SO2 ratios). On the other hand,CO2/SO2 ratios as high as 26, recurrently measured in the period December 2006–April 2007 cannot be produced by degassing at atmospheric pressure, the depth range (1.5–3.5 kmbelow summit vents) of their derivation (actually the depth of gas separation from the melt) being shown in the diagram. This coincides with the volcano–crust interface, which hasbeen interpreted (Allard et al., 1994; Burton et al., 2007a; Allard et al., in press) as a key structural and geological transition in Stromboli's feeding system, also favouring the formationof gas slugs sustaining Strombolian explosions during “normal” activity at Stromboli (Burton et al., 2007a). The location of hypocenters of deep volcano–tectonic earthquakes (Patanèet al., 2007) supports the presence of a magma storage zone in the same depth range, as early invoked by many other authors based on geochemical and petrological considerations(Allard et al., 1994; Francalanci et al., 1999; Vagelli et al., 2003; Allard et al., in press). This sill-like magma storage zone is also thought to favour extensive gas–melt separation(Menand and Phillips, 2007), supplying the conduit with CO2-rich gas bubbles. We therefore suggest that increasing gas contributions from the low-porphyric (LP) magma, possiblyoccupying this deep magma storage zone, was the cause of the anomalous composition of the volcanic gas plume emitted at Stromboli in the pre- and sin-eruptive period.

227A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

2003). Pressure was decreased stepwise (from 280 to 0.1 MPa) tosimulate volatile partitioning to the vapor phase upon magma ascentand decompression from the ~10 km deep magma storage zone ofBertagnini et al. (2003). Oxidation state was fixed at NNO+0.3 inagreement with Burton et al. (2007a). Total water and sulfur contents(H2OTOT: 3.4 wt.%; STOT: 0.2 wt.%, respectively) were derived fromvolatile contents in primitive glass inclusions from Stromboli (Métrichet al., 2001; Bertagnini et al., 2003; Métrich et al., 2005; Burton et al.,2007b). For CO2, direct use of dissolved contents in primitive meltinclusion is questionable: due to low solubility of carbon dioxide insilicate melts and its deep exsolution, even the highest measured CO2

contents of glass inclusions might underestimate CO2TOT (see Papale,

2005 for a discussion). We thus performed two set of distinctcalculations, in which CO2

TOT was evaluated at either: (a) 0.2 wt.%, thehighest measured (dissolved) CO2 content in the parental meltinclusions, entrapped in primitive olivine crystals at ~280 MPa(Métrich et al., 2001; Bertagnini et al., 2003); or at 2.1 wt.%, asevaluated by Burton et al. (2007a) accounting for the abundance(~2.5 wt.%) of the CO2-rich gas phase coexisting with the samemelt at280 MPa (the authors used the Newman and Lowenstern, 2002 codefor their calculations). The two above conditions represent lower andupper limits for CO2

TOT, respectively, and give rise to two paralleldegassing model trends (Fig. 6) encompassing the “real” naturalconditions. Both model runs have been calculated in “closed system”

conditions, implying that total (melt+gas) volatile contents areconserved on decreasing pressures.

A more-in-depth discussion of model results, comprehensive of adescription of water degassing trends, will be given elsewhere. Herewe focus on the results of model calculations pertinent to theinterpretation of the CO2/SO2 ratio, which are given in Fig. 6. TheFigure shows equilibrium CO2 and S dissolved contents in the silicatemelt (Fig. 6a) and CO2/SO2 ratios in the coexisting gas phase (Fig. 6b),respectively, both as a function of pressure (in MPa). The decliningdissolved melt contents upon pressure decrease (Fig. 6a) testify forthe increasing extents of volatile degassing upon migration ofStromboli's magmas from their source region to the surface (seeAllard et al., in press for a review). However, a contrasting behaviourbetween the two volatiles is evident in Fig. 6a: according to modelpredictions, CO2 is extensively degassed at depth (~98% of the originalCO2 content in the parental melt is already lost from the melt andseparated into the gas phase at 100MPa), and is released at a virtually-constant rate throughout the whole degassing path (Fig. 6a); whilstSO2 degassing is minor at depth (note that modelled dissolved Scontents are relatively-constant for PN200 MPa, particularly in theCO2

TOT=0.2 wt.% case), and becomes significant only at Pb100 MPa.This much deeper exsolution of CO2, compared to more soluble SO2,implies that the former will by far dominate the volcanic gas phase inequilibrium with the melt at high pressure (e.g., the modelled CO2/SO2 ratio is ~150 at 240 MPa; note that the melt is not volatilesaturated at PN240 MPa for CO2

TOT=0.2 wt.%; Fig. 6b). The modelledCO2/SO2 ratio will then decrease upon pressure decrease (Fig. 6b):this demonstrates that, in spite of the modelled values of the ratio at

Fig. 8. Changes over time in the feeding rate of CO2-rich bubbles to shallow plumbing asa potential source mechanism for Stromboli's mutable activity state in 2006–2007. (a)In the period January–November 2006, the rate of supply of CO2-rich gas bubbles to theshallow plumbing system is low, as supported by CO2/SO2 ratios in the volcanic gasplume b10. Close-to-surface gas–melt separation from the shallow-convecting magmalikely sustains the surface gas emissions; (b) An increase in the supply of deep-rising(separation pressureN40 MPa; depthN1.5 km below summit vents) gases is suggestedto occur since December 2006, as reflected by the overall increase of CO2/SO2 ratios inthe bulk volcanic gas plume. The increasing gas content of the magma induces a fasterrate of magma convection, which in turn reflects into increasing explosive rate, RMStremor amplitude and rate of deformations in the Sciara del Fuoco, finally triggering theonset of the effusive phase in late February; (c) The relatively-fast empting of theconduit-dyke system, triggered by the high effusive rates during the eruption, causes adepressurization of the deep-feeding system, which effect is a slow (increasing CO2/SO2

ratios from 7 to 15March, 2007) and than sudden ascent of CO2-rich gas pockets and/orgas-rich magmas, triggering the paroxysm on March 15 at 20:37 GMT; (d) after April,the rate of CO2-rich bubble supply to the system progressively declines. Steady-stateconditions and persistent Strombolian are re-established by June 2007.

228 A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

each pressure being substantially dependent on the assumed CO2TOT,

the CO2/SO2 ratio is a sensible indicator of the source region ofvolcanic gases (Aiuppa et al., 2007; Burton et al., 2007a), and most-specifically of the depth (pressure) of gas–melt separation (with theshallower that depth, the lower the ratio).

4.2. CO2/SO2 ratios and Stromboli's plumbing system

The long-lived degassing and eruptive activity (Allard et al., 1994;Rosi et al., 2000), the relatively-uniform composition of the eruptedproducts (Francalanci, 1993; Francalanci et al., 2004; Landi et al.,2006), the large “volatile excess” (i.e., that Stromboli emits more gasthat can be produced by the erupted magma; Allard et al., 1994), andheat-flow computations (Harris and Stevenson, 1997; Stevenson andBlake, 1998), collectively suggest that convective magma overturndrives the steady-state activity of Stromboli volcano (Fig. 7). Thecontinuous convective supply of less-dense volatile-rich magma hasbeen evaluated to take place at an average rate of 0.3 m3·s−1 (Allardet al., 1994; Ripepe et al., 2005), and is driven by the sinking of more-dense degassedmagmas at the top of themagmatic column,which arerecycled back into the conduits and within the volcanic pile (Allardet al., 1994; Kazahaya et al., 1994; Stevenson and Blake, 1998;Kazahaya et al., 2004; Allard et al., in press). Albeit the existence ofa shallow convective magmatic system is suggested by heat-flowcomputations (Harris and Stevenson, 1997), no seismic evidencesupports the existence of a large shallow magma accumulation zone;and the depth, geometry and volume of the shallow convectingplumbing system are still poorly constrained. There is howeverconvincing evidence for the shallowest part of the system consistingof a NE–SW-trending dyke-like structure (dipping ~60° to NW;Chouet et al., 2003; La Rocca et al., 2004; Mattia et al., 2004; Fig. 7),with a dyke-to-conduit transition at ~500–600 masl, being the most-likely source for persistent VLP seismicity at Stromboli (Chouet et al.,2003; Ripepe et al., 2005). The magma storage capacity of the dyke-conduit conduit system, sustaining Stromboli's persistent activity, hasbeen recently evaluated at ~6×106 m3 (Gauthier et al., 2000), of thesame order of magnitude of the total magma volumes emitted duringrecent flank eruptions (De Fino et al., 1988; Landi et al., 2006). Highly-porphyric vesicular magmas, erupted as scoriae during ordinaryactivity, occupy the shallowest (b1 km; Landi et al., 2004) portions ofthe dyke-conduit system (Fig. 7).

Accepting that continuous magma overturning within the dyke-conduit system takes place up to the surface in a perfect closed-systemcondition, with magma-gas separation at 0.1 MPa, our calculations(Figs. 6b and7) put the upper limit for theCO2/SO2 ratio in the releasedvolcanic gas phase at ~0.8 (CO2

TOT=0.2 wt.%) to ~8 (CO2TOT=2.1 wt.%).

This 0.8–8 range fits the composition of the quiescent volcanic gasplume emitted during period of “ordinary” activity of the volcano (e.g.,January–November 2006; Fig. 5), providing further support for theconvective overturn model and demonstrating that degassing frommagmas filling the upper conduits dominates persistent gas emissions(at least during repose periods between individual explosions; Burtonet al., 2007a). Our computations also suggest that time-changes in theCO2/SO2 ratio during “ordinary” activity may likely reflect changes intotal CO2 content (CO2

TOT) in the feeding magmas; with higher CO2TOT

reflecting into larger supply rate of CO2-rich gas bubbles to theshallow-convectivemagma plumbing system, and thus to higher CO2/SO2 ratios in surface emissions.

Our model calculations suggest instead that the high CO2/SO2

ratios measured at Stromboli since late November–early December2006, which persisted throughout the majority of the February 27–April 2 effusive event, cannot be produced by close-to-surface gas–melt separation, but require a deeper source region for the emittedvolcanic gas phase. We propose that gas–melt separation at ~40–100 MPa (Fig. 7), followed by fast bubble ascent without further re-equilibration with the shallower magma, is required in order to

produce the volcanic gas CO2/SO2 ratio of ~25, which we measuredseveral times since December 2006. The estimated gas–melt separa-tion depth (1 to 3 km below the crater vents, assuming an averagerock density of 2700 kg·m−3; Fig. 7) is in agreement with the sourcedepth of gas slugs sustaining “ordinary” Strombolian activity (CO2/SO2 ratio of ~20), which composition is determined by FTIR-sensing ofthe syn-eruptive gas phase emitted upon Strombolian explosions(Burton et al., 2007a). It is also in agreement with the hypocenters ofdeep earthquakes detected at Stromboli during the 2007 effusiveevent (Patanè et al., 2007), and interpreted as the potential evidenceof a 3 km b.s.l. deep magma storage zone. This deep source region forvolcanic gases emitted at Stromboli during December 2006–April2007 has two obvious implications: (i) first, it requires an efficientmechanism for gas–melt separation, and thus reinforces earlierconclusions (Allard et al., 1994; Burton et al., 2007a; Allard et al., inpress) of the presence of a significant physical or geologicaldiscontinuity in the 1 to 3 km depth range; (ii) secondly, it supportsthe idea that an increase in the feeding rate of deep CO2 bubble-richmagmas to the Stromboli's shallow plumbing system was the triggerfor the 2007 eruption, as detailed below.

4.3. The trigger and evolution of the 2007 eruption of Stromboli from avolcanic gas plume perspective

A period of substantial “ordinary” activity in early 2006 reflected aphase of relativelyminormagma feeding rate to the shallow plumbingsystem (Fig. 8a). Degassing was likely dominated by close-to-surfacegas separation from the convective magmas (CO2/SO2 in the volcanicgas plume was b8); whilst the relatively-minor ascent of gas bubblesfrom a deeper source (N1 km deep, but probably at 3 km b.s.l)

229A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

contributed to the generation of gas slugs and to the individualStrombolian explosions (Burton et al., 2007a,b). As the rate of supplyof CO2-rich gas bubbles from depth increased since late November–early December (Fig. 8b), as supported by our increasing CO2/SO2

ratios in the volcanic gas plume to values of 20–25, the overalldecrease of magma density in the conduits promoted a fasterconvection (as suggested by an increase in the SO2 flux; Aiuppaet al., 2007; Burton et al., 2008, this issue), an increase in the explosiverate (Ripepe et al., 2007, and this issue), and in the rate of seismicenergy release (e.g., mostly in the form of seismic tremor in LP band;Patanè et al., 2007). The causes for an increasing supply of CO2-richgas bubbles are unknown, but the composition of the emitted gassuggests it could be related to magma replenishment of the deepmagma storage zone of Patanè et al. (2007) with more primitive (e.g.,CO2-rich) magmas. Whatever the cause, however, and in analogy withwhat already observed in 2002 (Ripepe et al., 2005), the inevitableconsequence of this sustained convection of bubble-rich magmas inthe conduits was the collapse of the gravitationally-unstable Sciara delFuoco, and the onset of the effusive phase.

After the eruption started, the release of a CO2-rich volcanic gasplume persisted, however; and CO2/SO2 ratios increased up to thehigh values recorded in the days and hours before the March 15paroxysms (Fig. 5). In particular, the rate of CO2 emission from thesummit craters was in the period March 8–15 about 10 times higherthan ever measured before (Aiuppa et al., in prep.), suggesting aneven-larger supply of CO2-rich bubbles before the paroxysm. Wepropose that the rapid emptying of the dyke-conduit system, as due tothe opening of the fracture systemwithin the Sciara del Fuoco and thesustained lava effusion, provoked a general depressurization of thedeep magmatic system. This could ultimately facilitate the suddenrelease and ascent of CO2-rich gas bubbles (Allard, 2007) and aphyricprimitive magmas (erupted as light-coloured pumices; Bertagnini etal., 2003; Métrich et al., 2005) from depth, triggering the paroxisticevent of March 15 (Fig. 8c; Aiuppa et al., in prep.). The recurrentassociation between effusive events and paroxysms at Stromboli, as2002–2003 testifies for feedback mechanisms between effusion anddepressurization of a bubble-rich deep magma.

The relatively-high CO2/SO2 ratios measured during April to Mayand July to August 2007 (Fig. 5) suggest that the supply of CO2-richdeep gases persisted after the eruption termination, but at a decliningrate (e.g., the CO2 flux from the summit craters progressively declinedafter April 2, Aiuppa et al., in prep.). However, the steady-state time-averaged CO2/SO2 ratios for the post-eruptive phase were ~9 (Fig. 5),still a factor ~2 larger than in the January–November 2006 period,suggesting a potential temporary modification in the rates/mechan-isms of gas ascent within the plumbing system (Fig. 8d).

5. Conclusions

Contrasting compositions of the emitted volcanic gas plume havebeen observed during two years of Multi-GAS observations atStromboli encompassing the recent effusive event of February–April2007. Between January and November 2006, a phase of relatively-ordinary activity at Stromboli, CO2/SO2 ratios in the volcanic gasplume were relatively-constant and systematically below ~10; whilsthigh CO2/SO2 ratios (as high as 26) persisted since December 2006and throughout most part of the effusive and post eruptive (April–May 2007) phase. Numerical calculations, performed by use of athermodynamic saturation model, support the idea that theseincreasing CO2/SO2 ratios in the pre-eruptive and eruptive phasewere due to the supply of CO2-rich gas bubbles to the shallow dyke-conduit Stromboli's feeding system. The ascent of this CO2-rich gasphase, which source degassing pressure is estimated at 40–100 MPaby contrasting field-measured and modelled CO2/SO2 ratios, probablyreflected replenishment of a ~1–3 km deep magma storage zone witha more primitive (CO2-rich) magma, which eventually erupted during

the March 15 paroxysmal event as light-coloured aphyric pumices.This supply of CO2-rich gases, which we suggest started in December2007, triggered an increase in the rate of magma convection, and hadthe effect of increasing the explosive rate and volcanic tremoramplitude in January–February 2007, up to the final failure of theSciara del Fuoco on February, 27.

Our measurements emphasize the role of volatiles in drivingvolcanic systems toward critical volcanic activity states; and reinforcethe importance of CO2/SO2 ratio observations in volcanic plume forpredicting volcanic eruptions (Aiuppa et al., 2007).

References

Aiuppa, A., Federico, C., 2004. Anomalous magmatic degassing prior to the 5th April2003 paroxysm on Stromboli. Geophys. Res. Lett. 31. doi:10.1029/2004GL020458(L14607).

Aiuppa, A., Federico, C., Giudice, G., Guerrieri, S., 2005. Chemical mapping of a fumarolicfield: La Fossa Crater, Vulcano Island (Aeolian Islands, Italy). Geophys. Res. Lett. 32.doi:10.1029/2005GL023207 (L13309).

Aiuppa, A., Federico, C., Giudice, G., Gurrieri, S., Liuzzo, M., Shinohara, H., Favara, R.,Valenza, M., 2006. Rates of carbon dioxide plume degassing from Mount Etnavolcano. J. Geophys. Res. 111. doi:10.1029/2006JB004307 (B09207).

Aiuppa, A., Moretti, R., Federico, C., Giudice, G., Gurrieri, S., Liuzzo, M., Papale, P.,Shinohara, H., Valenza, M., 2007. Forecasting Etna eruption by real time evaluationof volcanic gas composition. Geology 35 (12), 1115–1118. doi:10.1130/G24149A.

Aiuppa, A., Burton,M., Caltabiano, T., Giammanco, S., Giudice, G., Guerrieri, S., Liuzzo, M.,Murè, F., Randazzo, D., Salerno, G., in preparation. Magmatic CO2 flux monitoringreveals precursors to explosive volcanic activity.

Allard, P., 2007. A CO2-rich gas trigger of explosive paroxysms at Stromboli volcano, Italy.Geophys. Res. Abstracts, 9, 08044, Proc. EGU 2007 Assembly, Vienna April 2007.

Allard, P., Carbonelle, J., Métrich, N., Loyer, H., Zettwoog, P., 1994. Sulphur output andmagma degassing budget of Stromboli volcano. Nature 368, 326–330.

Allard, P., Aiuppa, A., Loyer, H., Carrot, F., Gaudry, A., Pinte, G., Michel, A., Dongarrà, G.,2000. Acid gas and metal emission rates during long-lived basalt degassing atStromboli volcano. Geophys. Res. Lett. 27, 1207–1210.

Allard, P., Burton, M., Murè, F., 2005. Spectroscopic evidence for a lava fountain drivenby previously accumulated magmatic gas. Nature 433, 407–410.

Allard, P., Aiuppa, A., Burton, M., Caltabiano, T., Federico, C., Salerno, G., La Spina, A., inpress. Crater Gas Emissions and the Magma Feeding System of Stromboli Volcano.In: Stromboli Volcano, AGU Geophysical Monograph.

Barberi, F., Rosi, M., Sodi, A., 1993. Volcanic hazard assessment at Stromboli based onreview of historical data. Acta Vulcanol. 3, 173–187.

Belonoshko, A.B., Saxena, S., 1992. A unified equation of state for fluids of C–H–O–N–S–Ar composition and their mixtures up to very high temperatures and pressures.Geochim. Cosmochim. Acta 56, 3611–3626.

Bertagnini, A., Coltelli, M., Landi, P., Pompilio, M., Rosi, M., 1999. Violent explosions yieldnew insights into dynamic of Stromboli volcano. EOS Trans. AGU 80 (52), 633–636.

Bertagnini, A., Métrich, N., Landi, P., Rosi, M., 2003. Stromboli volcano (AeolianArchipelago, Italy): an open window on the deep feeding system of steady statebasaltic volcano. J. Geophys. Res. 108, 1–15.

Bonaccorso, A., Aloisi, M., Bonforte, A., Gambino, S., Guglielmino, F., Mattia, M., Palano,M., Pugliesi, G., 2007. Ground deformation monitoring at Stromboli: insights onrecent eruptions. Proceedings of the XXIV IUGG Conference, 2–13 July 2007.

Brusca, L., Inguaggiato, S., Longo, M., Madonia, P., Maugeri, R., 2004. The 2002–2003eruption of Stromboli (Italy): Evaluation of the volcanic activity by means ofcontinuous monitoring of soil temperature, CO2 flux, and meteorological parameters.Geochem. Geophys. Geosys. 5 (12). doi:10.1029/2004GC000732 1525–2027.

Burton,M., Allard, P., Muré, F., La Spina, A., 2007a.Magmatic gas composition reveals thesource depth of slug-driven Strombolian explosive activity. Science 317, 227–230.

Burton, M.R., Mader, H.M., Polacci, M., 2007b. The role of gas percolation in quiescentdegassing of persistently active basaltic volcanoes. Earth Planet. Sci. Lett. 264, 46–60.

Calvari, S., Bertagnini, A., Cristalli, A., Landi, P., Lodato, L., 2007. Volcanological andthermal monitoring of the 2007 Stromboli eruption. Proceedings of the XXIV IUGGConference, 2–13 July 2007.

Carroll, J.R., Halloway (Eds.), 1994. Volatiles in Magmas. Rev. in Mineralogy, vol. 30.Chouet, B., Hamisevicz, N., Mcgetchin, T.R., 1974. Photoballistics of volcanic jet activity at

Stromboli, Italy. J. Geophys. Res. 79 (32), 4961–4976.Chouet, B., Dawson, P., Ohminato, T., Martini, M., Saccorotti, G., Giudicepietro, F., De

Luca, G., Milana, G., Scarpa, R., 2003. Source mechanisms of explosions at Strombolivolcano, Italy, determined from moment-tensor inversions of very-long-perioddata. J. Geophys. Res. 108 (B1), 2019.

Corsaro, R.A., Miraglia, L., Zanon, V., 2005. Petrologic monitoring of glasses in thepyroclastites erupted in February 2004 by the Stromboli volcano, Aeolian Islands,Southern Italy. J. Volcanol. Geotherm. Res. 139, 339–343.

De Fino, M., La Volpe, L., Falsaperla, S., Frazzetta, G., Neri, G., Francalanci, L., Rosi, M.,Sbrana, A., 1988. The Stromboli eruption of December 6, 1985–April 25, 1986:volcanological, petrological and seismological data. Rend. Soc. Ital. Mineral. Petrol.43, 1021–1038.

Francalanci, L., 1993. Mineral chemistry of Stromboli volcanics: some evidence onprocesses of magma evolution. Acta Vulcanol. 3, 99–113.

Francalanci, L., Manetti, P., Peccerillo, A., 1989. Volcanological and magmatologicalevolution of Stromboli volcano (aeolian Islands): the role of fractional crystallization,

230 A. Aiuppa et al. / Journal of Volcanology and Geothermal Research 182 (2009) 221–230

magma mixing, crustal contamination and source heterogeneity. Bull. Volcanol. 51,355–378.

Francalanci, L., Tommasini, S., Conticelli, S., Davies, G.R., 1999. Sr isotopic evidence forshort magma residence time for the 20th century activity at Stromboli volcano,Italy. Earth Planet. Sci. Lett. 167, 61–69.

Francalanci, L., Tommasini, S., Conticelli, S., 2004. The volcanic activity of Stromboli inthe 1906–1998 period: mineralogical, geochemical and isotope data relevant to theunderstanding of Strombolian activity. J. Volcanol. Geotherm. Res. 131, 179–211.

Gauthier, P.J., Le Cloarec, M.F., Condomines, M., 2000. Degassing processes at Strombolivolcano inferred from short-lived disequilibria (210Pb–210Bi–210Po) in volcanicgases. J. Volcanol. Geotherm. Res. 102, 1–19.

Harris, A.J.L., Stevenson, D.S., 1997. Magma budgets and steady-state activity of Vulcanoand Stromboli volcanoes. Geophys. Res. Lett. 24, 1043–1046.

Kazahaya, K., Shinohara, H., Saito, G., 1994. Excessive degassing of Izu-Oshima volcano:magma convection in a conduit. Bull. Volcanol. 56, 207–216.

Kazahaya, K., Shinohara, H., Uto, K., Odai, M., Nakahori, Y., Mori, H., Iino, H., Miyashita,M., Hirabayashi, J., 2004. Gigantic SO2 emission from Miyakejima volcano, Japan,caused by caldera collapse. Geology 32, 425–428.

Landi, P., Métrich, N., Bertagnini, A., Rosi, M., 2004. Dynamics of magma mixing anddegassing recorded in plagioclase at Stromboli (Aeolian Archipelago, Italy). Contrib.Mineral. Petrol. 147, 213–227. doi:10.1007/s00410-004-0555-5.

La Rocca, M., Saccorotti, G., Del Pezzo, E., Ibanez, J., 2004. Probabilistic source location ofexplosion quakes at Stromboli volcano estimated with double array data.J. Volcanol. Geotherm. Res. 131, 123–142.

Landi, P., Francalanci, L., Pompilio, M., Rosi, M., Corsaro, R.A., Petrone, C.M., Nardini, I.,Miraglia, L., 2006. The December 2002–July 2003 effusive event at Strombolivolcano, Italy: an insight into the shallow plumbing system by petrochemicalstudies. J. Volcanol. Geotherm. Res. 155, 263–284.

Mattia, M., Rossi, M., Guglielmino, F., Aloisi, M., Bock, Y., 2004. The shallow plumbingsystem of Stromboli Island as imaged from 1 Hz instantaneous GPS positions.Geophys. Res. Lett. 31. doi:10.1029/2004GL021281 (L24610).

Menand, T., Phillips, J.C., 2007. Gas segregation in dykes and sills. J. Volcanol. Geotherm.Res. 159, 393–408.

Métrich, N., Bertagnini, A., Landi, P., Rosi, M., 2001. Crystallization driven bydecompression and water loss at Stromboli volcano (Aeolian Islands, Italy).J. Petrol. 42, 1471–1490.

Métrich, N., Allard, P., Spilliaert, N., Andronico, D., Burton, M., 2004. 2001 flank eruptionof the alkali- and volatile-rich primitive basalt responsible for Mount Etna'sevolution in the last three decades. Earth Planet. Sci. Lett. 228, 1–17.

Métrich, N., Bertagnini, A., Landi, P., Rosi, M., Belhadj, O., 2005. Triggering mechanism atthe origin of paroxysms at Stromboli (Aeolian Archipelago, Italy): the 5 April 2003eruption. Geophys. Res. Lett. 32. doi:10.1029/2004GL022257 (L10305).

Morelli, C., Giese, P., Cassinis, R., Colombi, B., Guerra, I., Luongo, G., Scarascia, S., Schutte,K.G., 1975. Crustal structure of Southern Italy. A seismic refraction profile betweenPuglia, Calabria and Sicily. Boll. Geofis. Teor. Appl. 17, 183–210.

Moretti, R., 2005. Polymerisation, basicity, oxidation state and their role in ionicmodelling of silicate melts. Ann. Geophys. 48 (4/5), 583–608.

Moretti, R., Ottonello, G., 2005. Solubility and speciation of sulfur in silicate melts: theConjugated Toop-Samis-Flood-Grjotheim (CTSFG) model. Geochim. Cosmochim.Acta 69 (4), 801–823. doi:10.1016/j.gca.2004.09.006.

Moretti, R., Papale, P., 2004. On the oxidation state and volatile behaviour inmulticomponent gas–melt equilibria. Chem. Geol. 213, 265–280.

Moretti, R., Papale, P., Ottonello, G., 2003. A model for the saturation of C–O–H–S fluidsin silicate melts. In: Oppenheimer, C., Pyle, D.M., Barclay, J. (Eds.), VolcanicDegassing. Geological Society, London, Special Publications, vol. 213, pp. 81–102.

Newman, S., Lowenstern, J.B., 2002. VolatileCalc: a silicate melt–H2O–CO2 solutionmodel written in Visual Basic for Excel. Comput. Geosci. 28, 597–604.

Papale, P., 1999. Strain-induced magma fragmentation in explosive eruptions. Nature397, 425–428.

Papale, P., 2005. Determination of total H2O and CO2 budgets in evolving magmas frommelt inclusion data. J. Geophys. Res. 110 (B3) (B03208).

Papale, P., Moretti, R., Barbato, D., 2006. The compositional dependence of themulticomponent volatile saturation surface in silicate melts. Chem. Geol. 229,78–95.

Patanè, D., Matti, M., Di Grazia, G., Cannavò, F., Gianpiccolo, E., Musumeci, C., Montalto,P., Boschi, E., 2007. Insights into the dynamic processes of the 2007 Strombolieruption and possible meteorological influences on the magmatic system. Geophys.Res. Lett. 34. doi:10.1029/2007GL031730.

Peccerillo, A., 2001. Geochemical similarities between the Vesuvius, Phlegraean Fieldsand Stromboli volcanoes: petrogenetic, geodynamic and volcanological implica-tions. Mineral. Petrol. 73 (1–3), 93–105.

Press, W.H., Teukolsky, S.A., Vetterling, W.T., Flannery, B.P., 1992. Numerical Recipes inFortran. Cambridge University Press (963 pp).

Ripepe, M., Rossi, M., Saccorotti, G., 1993. Image processing of explosive activity atStromboli. J. Volcanol. Geotherm. Res. 54, 335–351.

Ripepe, M., Harris, A.J.L., Carniel, R., 2002. Thermal, seismic and infrasonic evidences ofvariable degassing rates at Stromboli volcano. J. Volcanol. Geotherm. Res. 118,285–297.

Ripepe, M., Marchetti, E., Ulivieri, G., Harris, A., Dehn, J., Burton, M., Caltabiano, T.,Salerno, G., 2005. Effusive to explosive transition during the 2003 eruption ofStromboli volcano. Geology 33, 341–344.

Ripepe, M., Ulivieri, G., Genco, R., Lacanna, G., 2007. Monitoring the eruption by anintegrated Geophysical network. Proceedings of the XXIV IUGG Conference, 2–13July 2007.

Rosi, M., Bertagnini, A., Landi, P., 2000. Onset of the persistent activity at Strombolivolcano (Italy). Bull. Volcanol. 62, 294–300.

Rosi, M., Bertagnini, A., Harris, A.J.L., Pioli, L., Pistolesi, M., Ripepe, M., 2006. A casehistory of paroxysmal explosion at Stromboli: timing and dynamics of the April 5,2003 event. Earth Planet. Sci. Lett. 243, 594–606.

Shinohara, H., 2005. A new technique to estimate volcanic gas composition: plumemeasurements with a portable multi-sensor system. J. Volcanol. Geotherm. Res.143, 319–333.

Spilliaert, N., Allard, P., Métrich, N., Sobolev, A.V., 2006a. Melt inclusion record of theconditions of ascent, degassing, and extrusion of volatile-rich alkali basalt duringthe powerful 2002 flank eruption of Mount Etna (Italy). J. Geophys. Res. 111 (B4).doi:10.1029/2005JB003934 (B04203).

Spilliaert, N., Métrich, N., Allard, P., 2006b. S–Cl–F degassing pattern of water-rich alkalibasalt: Modelling and relationship with eruption styles on Mount Etna volcano.Earth Planet. Sci. Lett. 248, 772–786.

Stevenson, D.S., Blake, S., 1998. Modelling the dynamics and thermodynamics ofvolcanic degassing. Bull. Volcanol. 60, 307–317.

Vagelli, G., Francalanci, L., Ruggieri, G., Testi, S., 2003. Persistent polybaric rests of calc-alkaline magmas at Stromboli volcano, Italy: pressure data from fluid inclusions inrestitic quarzite nodules. Bull. Volcanol. 65, 385–404.