Geophysical (tilt, seismic tremor and gravity signals), geochemical (crater SO2flux) and infrared satellite measurements are presented and discussed to track the temporal evolution of the lava fountain episode occurring at Mt Etna volcano on 10 April 2011. The multi-disciplinary approach provides insight into a gas-rich magma source trapped in a shallow storage zone inside the volcano edifice. This generated the fast ascending gas-magma dispersed flow feeding the lava fountain and causing the depressurization of a deeper magma storage. Satellite thermal data allowed estimation of the amount of erupted lava, which, summed to the tephra volume, yielded a total volume of erupted products of about 1 × 106 m3. Thanks to the daylight occurrence of this eruptive episode, the SO2 emission rate was also estimated, showing a degassing cycle reaching a peak of 15,000 Mg d−1 with a mean daily value of ∼5,700 Mg d−1. The SO2 data from the previous fountain episode on 17–18 February to 10 April 2011, yielded a cumulative degassed magma volume of about ∼10.5 × 106 m3, indicating a ratio of roughly 10:1 between degassed and erupted volumes. This volumetric balance, differently from those previously estimated during different styles of volcanic activities with long-term (years) recharging periods and middle-term (weeks to months) effusive eruptions, points toward the predominant role played by the gas phase in generating and driving this lava fountain episode.
 Within the range of types of activity at a basaltic volcano, lava fountains, also called ‘fire fountains’, are an intriguing eruptive phenomenon. They are powerful, continuous but normally short-lived (from less than one to a few hours) gas jets that emit lava fragments to heights extending tens to hundreds of meters [e.g.,Wolff and Sumner, 2000]. These eruptive phenomena often occur at an advanced state of magma recharge and usually represent a prelude to bigger effusive eruptions [e.g., Parfitt et al., 1995]. To explain the mechanisms of explosive basaltic eruptions two different conceptual models have been proposed [Parfitt, 2004]. The first model explains this activity as generated by the disruption of fast-rising bubbly melt. The second model, deriving from a combination of theoretical approaches and laboratory experiments [Jaupart and Vergniolle, 1988, 1989], considers a separate ascent of a bubble foam layer previously accumulated at depth (known as ‘foam collapse’ model). During recent decades, spectacular lava fountains have often occurred also at Mt Etna during recharging phases preceding several flank eruptions [e.g., Alparone et al., 2003; Bonaccorso, 2006]. Such activity is accompanied by the formation of dispersal ash plumes and fall-out deposits, which often pose severe hazards to aviation and repeated temporary closures of the Catania international airport [e.g.,Scollo et al., 2009; Bonaccorso et al., 2011b]. Several studies have been conducted using different mono-disciplinary approaches, such as volcanic tremor analyses [Alparone et al., 2003], geochemical analyses of volcanic gases from remote sensing spectroscopy [Allard et al., 2005], tilt changes [Bonaccorso, 2006], acoustic records [Vergniolle and Ripepe, 2008], studies on fallout deposits, on compositional/textural features of the erupted product [Andronico and Corsaro, 2011] and on the rheological properties of magma [Giordano et al., 2010]. More recently, multi-disciplinary approaches have been undertaken at Mt Etna to better investigate lava fountaining phenomena.Aiuppa et al. used tremor analysis, crater gas composition and summit area deformation to infer a shallow magma storage volume at depth of 1–2.8 km a.s.l., i.e., in the last 2 km below the base of the South-East Crater (SEC), as the source where lava fountain events are generated.Bonaccorso et al. [2011b]studied in detail the 10 May 2008 lava fountain at Mt Etna, which preceded by only three days the onset of the long-standing 2008–2009 flank eruption starting on 13 May 2008. Petrological data from the erupted tephra together with a wide range of geophysical data recorded continuously during the eruption were taken into account. The lava fountain was modeled as a violent release of a bubble-rich magma layer previously trapped at the top of a shallow reservoir located between −0.5 and 1.5 km a.s.l. A new sequence of lava fountains started in 2011, with several episodes that occurred from SEC [internal reports atwww.ct.ingv.it]. From 12 January to 09 July, five lava fountains episodes occurred with a near month time interval (12 January, 18 February, 10 April, 12 May, and 09 July). Our investigation focused on the 10 April event, because it was one of the strongest episodes in the 2011 eruptive sequence and it also occurred during daylight, thus allowing the ultra-violet spectrometer network (named FLAME) to retrieve the SO2 flux from the eruptive crater. For the first time in the history of monitoring of Mt Etna, we considered the continuous geophysical signals (seismic tremor, gravity, tilt) in conjunction with the geochemical data on volcanic gases (SO2) measured by terrestrial remote sensing and the thermal activity data acquired by satellite (SEVIRI sensor).
2. The 10 April 2011 Eruptive Episode
 In 2011, eruptive activity resumed at Mt Etna after the end of the long-lasting 2008–2009 flank eruption. Several paroxysmal events occurred from the SEC, one of the four summit craters of the volcano and the most active one in the last years. The 10 April event was well observed and monitored through the network of sensors set up by the Istituto Nazionale di Geofisica and Vulcanologia (INGV). Two days before an increase in strombolian activity started inside the SEC, and in the early afternoon of 9 April the intra-crater explosive activity was accompanied by the effusion of a significant lava flow that rapidly filled the inner part of the crater. In the late afternoon of the same day, lava started pouring from the crater to form a 1.5 km-long flow that advanced towards the SE. In the morning of 10 April, the intensity of explosions increased markedly and between 08:00 and 09:00 (all times are UT) it developed into a lava fountain. Between 09:15 and 09:30 the height of the lava fountain steadily reached over 200 m above the crater rim. It began producing a significant emission of ash and lapilli that rapidly formed an eruptive column that rose convectively about 2 km above the volcano summit. From 11:30 to 13:30 lava flows reached their maximum expansion and maximum length (about 2.5 km). Lava jets sometimes reached up to 300 m in height until decreasing visibly to a maximum height of 100 m after about 13:10, when the fountaining activity decreased rapidly. After 13:30 only mild discontinuous magma jets were observed, reaching only some tens of meters in height. The eruption ceased at 14:00, but marked ash fall (made up of both coarse and fine lava fragments) was reported from 14:00 until 17:00 over the SE flank of the volcano down to the Ionian Sea coast.
3. Geophysical Data
 The Mt Etna permanent tilt network consists of 13 bi-axial electronic instruments, which are installed in shallow boreholes at about 3 m depth (AGI 722 model) or at 10–30 m depth (AGI Lily model). They all have a resolution on the order of 0.1 microradians or less, which is mainly detectable during rapid tilt changes [Bonaccorso and Gambino, 1997]. The instruments have a radial component directed towards the central crater VOR and a tangential component oriented orthogonally. The sampling rate is usually once every 10 minutes. During the 10 April paroxysm (from 09:00 to 13:30), several tilt stations recorded significant variations. In particular, stations MSC, MDZ, MSP, CDV and CBD (Figure 1) showed changes of about 0.2–0.5 μrad (Figure 2); the other stations measured less visible changes, in some cases masked by daily thermoelastic effects [Bonaccorso et al., 1999]. The most evident variations are visible on the radial components and indicate a general deflation of the edifice during the lava fountain. In Figure 2 radial tilt components recorded at CDV, MDZ, MSP and CBD signals are raw data. The MSC signal has been filtered from daily thermal noise using a linear correlation filter. The CBD signal is N40E oriented and shows a different pattern, which may be caused by a sliding effect of the eastern flank [Bonaccorso et al., 2011b]. This behavior is very similar to that observed during the 10 May 2008 paroxysm [Bonaccorso et al., 2011b], but in the 10 April case changes are smaller and detected clearly in fewer stations. These tilt changes have been recorded over a wide area (the detecting stations are at a distance of 6–8 km from the summit craters), implying that the depth of the deflation source is not very shallow. It is reasonable to suppose that a depressurizing source located at depth of 3–4 km b.s.l., as inferred by Bonaccorso et al. [2011b] for the 10 May 2008 episode, may also have acted for the 10 April episode.
3.2. Seismic Tremor
 At Mt Etna seismic tremor is continuous in time and its close relationship with eruptive activity allows for tracking of the changes in the eruptive state of the volcano [e.g., Falsaperla et al., 2005; Alparone et al., 2003]. This can be particularly useful when visual observation or field survey are hindered by poor weather conditions. A permanent network of 32 digital broadband seismic stations, equipped with Nanometrics Trillium seismometers with a corner period of 40 s, is operating. We investigated the volcanic tremor amplitude (RMS) associated with the 10 April lava fountain episode by analyzing the seismic signal recorded at the EBEL station (Figure 1), the closest one to the SEC. Figure 3shows the temporal variation of the RMS tremor amplitude, calculated on 5-minute-long sliding windows in the frequency band 0.5–5 Hz. After the end of the lava fountain of 18 February, volcanic tremor amplitude stayed, on average, at a low level (less than 0.7 × 103 nm/s) with moderate fluctuations. Starting from 07:00 of 9 April, the amplitude progressively increased. In the late afternoon of the same day, the RMS values were about 7 × 103 nm/s, and in the early morning of 10 April the tremor amplitude reached values of about 1.4 × 104 nm/s (Figure 3). The similarity between the RMS trend with that detected for paroxysmal episodes occurring in the past [Alparone et al., 2003; Bonaccorso et al., 2011b] clearly suggested a new impending and strong phase of volcanic activity. Indeed, between 07:30 and 08:30 a further sharp increase occurred, marking the beginning of the paroxysmal phase of the tremor, which at 10:30 culminated in a relative maximum of about 105 nm/s . Such a maximum was shortly followed by a small decrease and then, between 11:25 and 12:25, by a phase in which the tremor amplitude reached the highest values (about 1.2 × 105 nm/s) during this lava fountaining. Afterwards, as already observed in the other lava fountaining episodes of 2011, an abrupt drop of the RMS amplitude at 13:30 marked the end of the paroxysm. A few hours later, RMS was at low steady values, comparable with those recorded before the eruptive episode. Even the tremor source locations, calculated according to the method described by Di Grazia et al.  showed a time pattern which resembled that observed for previous paroxysms [Bonaccorso et al., 2011b]. The tremor sources progressively shifted from a position located north-northeast of the summit area at depths from 0.5 to 1.5 km (a.s.l.), on the days immediately before the lava fountain, to new positions toward the southeast, and moved to shallower depths towards the SEC. On 9 April, the tremor source was located at shallower depth (above 1.5 km a.s.l. and below SEC). After the end of the 10 April eruption, tremor source location shifted back northward and deepened to reach its pre-eruptive depth (0.5 - 1.5 km a.s.l.).
 Two continuously recording gravity stations (SLN and BVD) on the southern flank of the volcano (at 1740 and 2920 m a.s.l., respectively) were operative during the 10 April fire fountain (Figure 1). These stations are equipped with LaCoste & Romberg spring gravimeters and acquire at one data per minute sampling rate. Figure 2shows the raw gravimetric sequences, after removing the effects of Earth tide and instrumental drift. Both reduced temporal gravity series, from about 09:22 displayed rapid and marked changes that are time-correlated with the most intense phases of lava fountaining. Gravity variations reached an amplitude of about −200μGal at BVD and about 35 μGal at SLN. Subsequently, at the end of the paroxysmal event, both signals returned to their original levels. As already observed and inferred for the May 2008 episode [Bonaccorso et al., 2011b], these gravity variations can be mostly ascribed to the fast ascent of a very low-density mixture of gas and dispersed magma clots (dispersed flow) within the conduit generated by the collapse of a foam layer at the top of the magmatic source. As the gas-magma mixture ascends, it replaces a more degassed and denser magma that was already residing inside the conduit, thus causing a localized mass decrease that induced a negative gravity variation at BVD and a positive change at SLN. The signs and the amplitude ratio between the gravity changes at the two stations allows for constraining of the position and the size of the gravity source. To model the gravity changes we have to consider both the expansion of the bubble foam layer previously accumulated and the passage of the dispersed flow through the conduit. FollowingBonaccorso et al. [2011b]we modelled the foam effect using a spherical source located at a height of about 1.7 km a.s.l., with a radius of 170 m, and the SEC conduit as a cylindrical-shaped body, assuming a radius of about 10 m, as inferred byBonaccorso , set at a height ranging between 1.9 and 2.9 km a.s.l. (i.e., from the top of the foam source). A density variation of 2.2 g/cm3 from the resident magma (2.7 g/cm3) to the gas-magma dispersed flow (0.5 g/cm3) results in a magma with a 75% vesicularity [Bonaccorso et al., 2011b]. The model matches the pattern of observations quite well and shows that at BVD the main contribution is given by the foam source rather than by the conduit, whereas at SLN the contribution of the foam is negligible, since it is at the same altitude of the station, and the small positive change arises from the density variation within the conduit.
4. Gas Geochemistry: SO2 Flux Remote Sensing
 At Mt Etna, SO2 flux is measured during the daylight hours by the FLAME (FLux Automatic MEasurement) scanning ultraviolet spectrometer network [Salerno et al., 2009a, 2009b]. Three out of nine stations of FLAME were used in this study (Figure 1). During daylight, each device scans the sky intersecting the plume at a distance of ∼14 km from the summit craters [Salerno et al., 2009a]. FLAME automatically compute in real-time the SO2 flux, whose uncertainty ranges between −22 and +36% [Salerno et al., 2009b]. Between 8 and 9 April 2011 and from 07:20 to 09:00 of 10 April, the SO2 flux was characterized by low steady values around the mean flux of ∼1500 Mega gram per day (Mg d−1, mean standard deviation, 1σ = 500). This behavior abruptly changed after 09:00, showing an increase in the SO2 emission rates that peaked at 10:54 reaching the maximum value of 15,500 Mg d-1 (Figure 2). This was followed by a general decreasing trend punctuated by two sharp drops in the emission rates at 11:23 and 11:47 (10300 and 5500 Mg d−1, respectively). Overall, during the eight hours of observation, the SO2 flux showed a degassing cycle encompassed between minima of 1340 (09:05) and 2200 Mg d−1 (15:35), with the mean daily emission rate of ∼5700 Mg d−1 (1σ = 3600). Using the SO2 measurements, we calculated the volume of cumulative degassed magma (Dm) in the period between the two lava fountaining episodes of 17–18 February and 8–10 April. We chose this temporal window since we assumed that the April episode was supplied by magma that had accumulated in the volcano shallow feeder system since the end of the previous February lava fountaining event. Following Allard , the total volume of Dm was calculated using mean crystal fraction of 30% and mean original sulphur content for Etnean magma of 0.3 wt% [e.g., Spilliaert et al., 2006]. This yielded a cumulative magma volume of ∼10.5 × 106 m3, of which 4% had degassed over the three days of the eruptive activity. The error on Dm is from 22% to 36% since it derives from the SO2 flux measurement. Errors associated with the parameters used for estimating Dm have such a small error compared to the uncertainty in SO2 flux, having no significant influence on the error of Dm.
5. Satellite Thermal Data
 Data acquired by SEVIRI sensor (spatial resolution of 3 km2 at Nadir, temporal resolution of 15 minutes, downloaded from EUMETSAT, www.eumetsat.int), onboard the meteorological satellite MSG-2, were processed by the HOTSAT system to monitor high-temperature surface anomalies over Mt Etna [Ganci et al., 2011]. During the night between 9 and 10 April, we measured an almost continuous thermal activity, with the first hotspot detected on 9 April at 19:27 (see Figure S1 in the auxiliary material). After this, we registered an increase in the heat flux, with a peak of about 6.5 GW on 10 April at 10:12. During 10 April, from 10:27 until 16:42, discontinuous ash emissions occurred, as shown in Figure S1. The presence of ash could have covered the emitted lava and, hence, hidden the thermal anomalies, thus leading to an underestimation of the radiative power. At 17:00 a new peak of 6.2 GW occurred after the intensive ash emission stopped. From 17:12 we observed a gradual decrease of the thermal activity and the last hotspot was detected on 11 April at 5:12. We converted the total thermal flux estimated from SEVIRI thermal infrared images to time average discharge rate (TADR), following Harris et al. . Since the conversion from heat flux to volume flux depends on many lava parameters (such as density, specific heat capacity, eruption temperature, etc.) and it has to be determined as a function of flow conditions [Harris et al., 2010], we defined a range of variability for each parameter adopting the extreme values found by Harris et al. , which proved reasonable in calibrating satellite thermal data for Etna lavas [e.g., Vicari et al., 2011]. The peak values of TADR were estimated on 10 April, before and after the ash plume emission, at 10:12 and at 17:00, and ranged between 7 and 20 m3 s−1. By integrating separately minimum and maximum estimates of TADR, we computed two cumulative curves of the erupted lava volume. Over the entire period of thermal emission, we estimated erupted lava volumes in the range between 0.25 and 0.75 × 106 m3. It is worth noting that TADR does not reflect the instantaneous at-vent effusion rate, but it depends on the previous volume flux, integrated over some period prior to satellite overpass and that this computation is affected by uncertainties and assumptions, such as ash cloud attenuation and/or inability to distinguish between lava draining channels and cooling phenomena [Vicari et al., 2011]. To reduce the ash cloud attenuation effect, we discarded ash covered images and linearly interpolated the peak values between 10:12 and at 17:00. The new lava volumes thus estimated range between 0.35 and 0.97 × 106 m3. Assuming that lava stopped flowing around 18:30 on 10 April (at the end of the period of more intense thermal activity), the final satellite-derived volume of lava is estimated in the range 0.3–0.9 × 106 m3 (i.e., a mean value of 0.6 ± 0.3 × 106 m3). The uncertainty in satellite-derived effusion rate estimates is quite large, up to about 50%, and mainly arises from the lack of constraints on the lava parameters used to convert thermal flux into lava effusion rate. However, the uncertainty is comparable to the error in field-based effusion rate measurements [e.g.,Harris et al., 2007].
6. Discussion and Conclusions
 In agreement with previous recent observations [Bonaccorso et al., 2011b], continuous geophysical signals (tilt, seismic tremor, gravity) recorded during this investigation confirm interesting aspects of lava fountain mechanisms. In particular, the ground tilt detected on the flanks of the volcano largely suggests deflation of a source located at 3–4 km b.s.l. (i.e., deeper than the shallower one related to seismic tremor, which is located at depth between 0.5 and 1.5 km a.s.l.). This finding again confirms the existence of an intermediate magma storage volume, consistent with the pressuring/depressuring storage volumes modeled since 1990's [e.g., Bonaccorso et al., 2006; Bonforte et al., 2008]. During the recharge phase of 2010, the volcanic tremor was located stably within a crustal volume contained between −0.5 and 1.5 km a.s.l., almost in the same position as that of the recharging phase preceding the 2008–2009 eruption [Bonaccorso et al., 2011a]. The top portion of this volume coincides with the location of the gas bubble layer inferred by Allard et al.  from volcanic gas analysis carried out during a lava fountain episode in June 2000. During 9–10 April, the increasing volcanic tremor signals marked very well the preparatory phase of the lava fountain (Figure 3). The gravity changes fit with the effects produced by a volume-source where a low density mass moved from the same top portion (about 1.7 km a.s.l.) through a conduit towards the surface at about 2.9 km a.s.l., where the SEC is located. Therefore, as a first conclusion, the different geophysical signals support the idea that a gas–rich magma could be trapped in a shallow storage zone (indicated by the top portion of the tremor volume). Such magma would generate the fast ascending gas-magma dispersed flow that fed the lava fountain when an overpressure threshold was exceeded and caused a depressurization in the deeper magma storage volume at depth of 3–4 km b.s.l. SEVIRI data help build a detailed chronology of the thermal activity during the paroxysmal event. HOTSAT calculations detected the first thermal anomalies about twelve hours before the beginning of the lava fountain (9 April, 19:30), when lava outpouring was enough to be detectable from the satellite sensor. Considering that the sluggish flow continued through the evening of 10 April, we estimated the satellite-derived dense rock equivalent (DRE) volume ranging from 0.3 to 0.9 × 106 m3. This value refers to the lava emission and it does not include the volume of tephra ejected in the same period. We estimated the balance between degassed and erupted magma (Dm and Em, respectively) based both on the SO2 flux measurements and on the thermal activity recorded by satellite. The calculated Dm value during the time between the two lava fountaining episodes of February and April 2011 was ∼10.5 × 106 m3, while the 10 April lava fountaining Em, derived from satellite thermal data, yielded a mean volume of ∼0.6 × 106 m3. Based on the average value from literature, we assumed that the estimated tephra volume is about 50% of the effusive volume [Behncke et al., 2006; Allard et al., 2006; Calvari et al., 2011]. Therefore, the total erupted volume is ∼1 × 106 m3, which yields a volumetric ratio Dm / Em of about 10:1. This ratio is significantly greater than previous estimates [Allard, 1997; Allard et al., 2006]. Allard  and Allard et al. argued that over a broad time-scale, i.e., between 1975 and 1995 and between 2001 and 2006, only about one fourth of the total degassed magma volume was eventually erupted, with the majority of magma remaining confined at depth within the crust [e.g.,Francis et al., 1993; Caracausi et al., 2003]. A similar 4:1 ratio between stored and erupted magma was also observed from gravimetric data [Bonaccorso et al., 2011c]. Indeed, during the period 1994–2000, the observed positive gravity changes were associated with a magma recharge through intrusion of an estimated magma volume of 800 × 106 m3, of which only 200 × 106 m3 was erupted during the following 2001 and 2002–03 eruptions.
 The general mechanism to explain the long-term volumetric imbalance between Dm and Em and, hence, the excess degassing, at active volcanoes tends to envision magma recycling within the volcano's shallow feeder system. It suggests that degassed magma would sink down in the conduit, being replaced by fresh undegassed and ascending magma [e.g., Oppenheimer and Francis, 1997; Stevenson and Blake, 1998; Harris et al., 1999]. This model has also been proposed at Mt Etna by Allard . However, such volumetric balances have normally been estimated for long-term (several years) volcanic degassing. Recently,Steffke et al. investigated effusive flank eruptions occurring at Mt Etna between 2002 and 2006. They found that over short-time scales (weeks to months), the volumetric balance between Dm and Em has a ratio close to 1 during the effusive phase of the 2002–03 eruption, and during the 2004–05 and the 2006 eruptions. The significant imbalance between influx of degassing magma and efflux of degassed lava (Dm / Em ∼ 10) observed in the case of the 10 April eruption, therefore, highlights that this lava fountaining episode was driven by a predominant gas phase (at least 70% [Cashman et al., 2000]) previously decoupled from melt and accumulated at the top of a shallow magma storage at about 2 km a.s.l. [Vergniolle and Ripepe, 2008; Allard et al., 2005; Bonaccorso et al., 2011b], hence supporting the mechanism of “foam collapse” proposed by Jaupart and Vergniolle .
 Thanks are due to the personnel of INGV-Catania employed in the volcano monitoring of the Mt Etna for making the multi-disciplinary data available. In particular, GS, TC and SG acknowledge F. Mure' and D. Randazzo for their technical assistance in the FLAME network. We are grateful to EUMETSAT for SEVIRI data. We thank S. Conway for revising the English language of this paper. We thank an anonymous referee for the constructive comments and the AE R. Harris for the helpful assistance.
 The Editor wishes to thank an anonymous reviewer for their assistance evaluating this paper.