Corresponding author: D. Patanè, Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo - Sezione di Catania, Catania 95125, Italy. (firstname.lastname@example.org)
 Since the second half of the 1990s, the eruptive activity of Mount Etna has provided evidence that both explosive and effusive eruptions display periodic variations in discharge and eruption style. In this work, a multiparametric approach, consisting of comparing volcanological, geophysical, and geochemical data, was applied to explore the volcano's dynamics during 2009–2011. In particular, temporal and/or spatial variations of seismicity (volcano-tectonic earthquakes, volcanic tremor, and long-period and very long period events), ground deformation (GPS and tiltmeter data), and geochemistry (SO2 flux, CO2 flux, CO2/SO2 ratio) were studied to understand the volcanic activity, as well as to investigate magma movement in both deep and shallow portions of the plumbing system, feeding the 2011 eruptive period. After the volcano deflation, accompanying the onset of the 2008–2009 eruption, a new recharging phase began in August 2008. This new volcanic cycle evolved from an initial recharge phase of the intermediate-shallower plumbing system and inflation, followed by (i) accelerated displacement in the volcano's eastern flank since April 2009 and (ii) renewal of summit volcanic activity during the second half of 2010, culminating in 2011 in a cyclic eruptive behavior with 18 lava fountains from New Southeast Crater (NSEC). Furthermore, supported by the geochemical data, the inversion of ground deformation GPS data and the locations of the tremor sources are used here to constrain both the area and the depth range of magma degassing, allowing reconstructing the intermediate and shallow storage zones feeding the 2011 cyclic fountaining NSEC activity.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The acquisition of multiparametric data is a necessary step in the attempt to better understand the behavior of a highly active volcano such as Mount Etna, in Sicily. Since the 1990s, more than 150 paroxysmal summit episodes (from lava fountaining to subplinian eruptions) and numerous effusive events have occurred. Some of these summit effusive eruptions, such as that of 1991–1993, produced large lava flows. In some cases, lava flows contemporaneously issued from both the northern and eastern flanks of the volcano, as during the 2001 and 2002–2003 multivent eruptions. The most common explosive activity of Etna ranges from mild Strombolian explosions to violent lava fountaining episodes, consisting of vigorous, continuously sustained jets of magma and gas, often accompanied by large ash emissions. This activity normally takes place at the summit of the volcano, where four craters are currently located (Figure 1b)—Northeast Crater (NEC), Voragine (VOR), Bocca Nuova (BN; BN1 and BN2), and Southeast Crater (SEC)—though lava fountaining was also recorded from the eruptive fissures near the summit during both the 2001 and 2002–2003 flank eruptions. The frequency of Etna's paroxysmal episodes has been very high during the last 16 years. Before the 2001 flank eruption [e.g., Patanè et al., 2004; Behncke et al., 2006; Alparone et al., 2007a], sequences of 22 and 64 lava fountain episodes occurred from the SEC, between September 1998 and February 1999 and from January to June 2000, respectively. More recently, between 2006 and 2008, an additional number of 25 eruptive events, among which three intense paroxysms of lava fountains, have taken place at SEC (18 from July to December 2006, six during 2007, and one on 10 May 2008, just before the onset of the 2008–2009 eruption) [Andronico et al., 2008; Patanè et al., 2008]. This explosive activity generated widely dispersed ash plumes and fallout deposits and also caused severe hazards for aviation with repeated temporary closures of the Catania International Airport [e.g., Alparone et al., 2007a; Andronico et al., 2008]. Given this high rate of eruptive events, Etna constitutes an important natural laboratory for the understanding of eruptive processes and magma ascent in basaltic volcanic environments.
 Flank eruptions are more hazardous than the more continuous summit activity, owing to potentially extensive lava flows capable of invading the densely populated lower slopes. Most of these eruptions are typically preceded by significant premonitory signals such as increases in volcano-tectonic (VT) earthquake activity, changes in volcanic tremor and long-period (LP) event features, and variations in ground deformation time series [e.g., Patanè et al., 2004, 2006; Andronico et al., 2005; Mattia et al., 2007; Di Grazia et al., 2009; Alparone et al., 2012]. Conversely, the smaller eruptions commonly occurring at the volcano's summit (Strombolian activity, lava fountains, lava flows, ash emissions, and explosive events) are preceded by less evident precursory phenomena. However, the dense multiparametric (seismic, GPS, geochemical) sensor network installed since 2007 around the summit area and the development of increasingly sophisticated systems for data analysis and modeling have now opened the way to understand the mechanisms of even these summit eruptions and eventually to predict their occurrence. Indeed, the improved multiparametric monitoring system has demonstrated its ability to capture small but significant variations in ground deformation, seismovolcanic activity, and geochemistry [e.g., Patanè et al., 2008; Di Grazia et al., 2009; Aiuppa et al., 2010a]. Such an approach was followed in some volcanoes, where the monitoring system is comparable to the one at Mount Etna, such as Piton de La Fournaise [e.g., Peltier et al., 2008], Merapi [e.g., Surono et al., 2012], Eyjafjallajokull [Sigmundsson et al., 2010], and Kilauea [e.g., Wright and Klein, 2008]. As for Mount Etna, Aiuppa et al. (2010a) have given first evidence on the effectiveness of an integrated approach combining continuous geochemical measurements (i.e., CO2/SO2 ratios of the volcanic gas plume) together with seismic and ground deformation data. They revealed a systematic trend in volcanic gas plume chemistry, seismicity, and deformation, accompanying the episodic activity of Etna during 2007–2008, when seven paroxysmal eruptions of the SEC were observed, suggesting that it is possible to quantitatively track these cycles of magma accumulation, degassing, and preeruptive to syneruptive ascent leading to SEC eruptions. However, the mechanisms leading to such recurrent paroxysmal activity, which also occurred more recently in 2011 and 2012, are still poorly understood [La Delfa et al., 2001; Allard et al., 2005; Behncke et al., 2006; Allard, 2010].
 Here, we report on the results of an integrated geophysical (seismicity, deformation), volcanological, and geochemical characterization of the recent (2009–2011) activity of Etna, a period during which 18 paroxysmal eruptive episodes took place at the SEC. Such an activity carried on also during 2012 with further seven lava fountain episodes that are not treated in this work. We show that this multidisciplinary approach allows quantitatively tracking the cycles of magma accumulation, degassing, and preeruptive to syneruptive ascent leading to SEC eruptions, and identifying the intermediate-shallow plumbing system feeding this activity in better detail than ever before.
2 The Mount Etna Volcano Multidisciplinary Monitoring Network
 Mount Etna monitoring system, consisting of geophysical, geochemical, and volcanological networks, is managed by the Istituto Nazionale di Geofisica e Vulcanologia, Osservatorio Etneo (INGV-OE) and by Sezione di Palermo (INGV-PA) for several activities regarding the geochemical monitoring. This permanent multidisciplinary network, with its 165 stations, aims at a rapid evaluation of volcanological, geophysical (i.e., seismic, ground deformation, gravimetric, magnetic), and geochemical parameters for surveillance purposes, as well as the acquisition of high-quality data for research. Particular emphasis has been given to real-time data transmission, network robustness and redundancy, and multisensor remote stations. Almost all data from the remote sites are transmitted via mixed radio-satellite or wireless links to the INGV monitoring center in Catania, where they are stored with a different sampling rate depending on data type. This permanent multidisciplinary network, with its 165 stations, is currently one of the largest worldwide.
2.1 The Ground Deformation GPS and Tilt Networks
 Since the 1980s, a variety of techniques such as leveling, clinometry, global positioning system (GPS), and, most recently, interferometric synthetic aperture radar (InSAR) have been used to measure the ground deformation of the edifice and to map out spatial-temporal deformation patterns at Etna. The continuous GPS (CGPS) monitoring of ground deformation on Etna started in November 2000. During the last 5 years, the spatial density of the CGPS network (hereafter, Etn@net) has been considerably enhanced—from 28 stations operating during 2007 (31 during 2008, 35 during 2009–2010) to the present (2011) configuration of 36 stations—allowing today a complete coverage of the volcanic edifice and in particular of the summit area, with an average spacing between the stations of about 5–6 km (Figure 1c). Continuous GPS (CGPS) data are collected with a sampling rate of 30 s (for the daily solution). The data were processed with the GAMIT/GLOBK software packages [Herring et al., 2006] with IGS (International GNNS Service; http://igscb.jpl.nasa.gov) precise ephemerides and Earth orientation parameters from the International Earth Rotation Service (http://www.iers.org) to produce loosely constrained daily solutions, according to the strategy described by Mattia et al. . Through the use of GLOBK software [Herring et al., 2006], the daily solutions were analyzed in order to estimate the average site velocity in the local “Etn@ref” reference frame [Palano et al., 2010]. In GPS processing with GLOBK, noise is commonly added as a time-dependent random walk error in the velocity estimation. In our case, we added a random walk error of 1 mm/yr1/2 which represents an appropriate error model for Etna CGPS data, as described in Palano et al. .
 The permanent tilt network at Etna comprises 13 biaxial electronic instruments installed in shallow boreholes (Figure 1c). These instruments are installed at about 3 m (AGI 722 and 510 models with a resolution of 0.01–0.1 µrad) or at 10–30 m depth (AGI Lily model with a resolution of 0.005 µrad) [Bonaccorso and Gambino, 1997; Ferro et al., 2011]. However, tilt real precision, due to temperature and thermoelastic effects [Bonaccorso et al., 1999], is about 0.2–0.5 µrad for shallow and 0.05 µrad for depth devices. The instruments have a tilt component, named radial, directed toward the summit craters and a positive signal change indicating crater up, while a second component (tangential) is oriented orthogonally. The sampling rate is 1 or 10 min.
2.2 The Seismic Network
 The seismic permanent network configuration [e.g., Patanè et al., 1999; Patanè et al., 2004] has been considerably enhanced since 2005 (Figure 1d) by 24 bit digital stations equipped with broadband (Nanometrics Trillium 40s) sensors, and today, thanks to a configuration of 32 broadband and 12 short-period stations, it ensures a very good coverage of the volcanic area. The minimum and maximum distances station center of the summit area are ~0.9 and 30 km, respectively. Data latencies are mostly in the range 0–10 s, with the upper limit constrained by satellite channel sharing. At the INGV monitoring center in Catania, data are stored with a sampling interval of 0.01 s over consecutive, 2 min long digital archives. These new broadband observations have resulted in the detection of previously unobserved seismovolcanic signals.
 Finally, depending on the location of seismicity and in particular during eruptive crises, a mobile network of 10 stations, equipped with three-component broadband (20 s) sensors, is also available. Part of the stations can be telemetered and they are normally deployed to integrate the permanent network, in order to lower the threshold of detectable magnitudes.
2.3 The Volcanic Video Surveillance Network
 Eruptive activities occurring at the summit of the volcano (Strombolian activity, lava fountains, lava flows, ash emissions, and associated fallouts) are monitored by a network of video cameras integrating information derived from field survey. Images of Etna are transmitted in real time from the seven live cameras deployed around the volcano (four visible and three thermal; Figure 1e). The visible cameras are located in Catania at the “CUAD” (26 km S from the SEC and 35 m above sea level (asl)), Nicolosi (15 km S and 730 m asl), Schiena dell'Asino (5 km SSE and 2030 m asl), Milo (10 km ESE and 770 m asl), and Montagnola (3 km S and 2610 m asl), while the thermal cameras are located in Nicolosi, Montagnola, and (since the summer of 2011) Monte Cagliato (7.8 km ESE and 1115 m asl). Each camera acquires an image every 1–2 s and transfers the video signal in real time to the operations room, where all the images are stored and published (at a lower frame rate) on the website www.ct.ingv.it.
2.4 The Geochemical Networks/Monitoring
 The volcanic gas data used in this work were derived from the fully automated permanent MultiGAS (operated by INGV-PA) and Flux Automatic Measurements ((FLAME, operated by INGV-OE) gas networks (Figure 1e).
 The MultiGAS network consists of a series of (2–3 simultaneously operating) multicomponent gas analyzer instruments, which are permanently deployed at Etna's summit craters to measure (at 0.1 Hz) the in-plume concentrations of CO2 (by infrared spectroscopy) and SO2 (by specific electrochemical sensor) [see Aiuppa et al., 2007, 2010a for details]. The acquired data, temporarily stored in an internal memory board, are telemetered daily via radio-modem bridge back to Palermo, where they are postprocessed (using custom-made software) to calculate the in-plume CO2/SO2 ratios. The MultiGAS data used here are averages of plume compositions obtained from the two permanent Central Crater (CC) instruments, deployed at the currently active degassing vents of the VOR and BN, respectively (Figure 1b). We also obtained data at the NEC, though more randomly, and the compositions we observed were on the whole consistent with those obtained for the CCs.
 The FLAME (Flux Automatic Measurements) consists of nine autonomous ultraviolet scanning spectrometer stations spaced ~7 km apart and installed at an altitude of ~900 m asl on the flanks of Etna [Salerno et al., 2009a; Campion et al., 2010, Figure 1e]. Each scanner station consists of an Ocean Optics S2000 spectrometer connected via fiber optic cables to a telescope mounted in a rotating head. Each device operates independently from weather conditions scanning the sky for ~9 h (during daylight), so as to intersect the plume at a mean distance of ~14 km from the summit craters, and acquiring a complete scan in ~5 min. The light exposure of the detector is automatically adjusted throughout the day to optimize the signal. SO2 column amounts are retrieved in real time on site from the recorded ultraviolet open-path spectra applying the differential optical absorption spectroscopy (DOAS) technique but using a modeled clear-sky spectrum rather than a measured spectrum as in the classic DOAS technique [Salerno et al., 2009a; La Spina et al., 2010]. SO2-scanned volcanic-plume profiles are then transmitted in real time to the INGV-OE in Catania, where they undergo data-quality evaluation. Data that have passed the evaluation process are then automatically reduced in SO2 flux and available in real time for surveillance purposes [Salerno et al. 2009b]. By combining the MultiGAS-derived in-plume CO2/SO2 ratios with the FLAME-sensed SO2 fluxes, time series (daily averages) for Etna's CO2 flux were also obtained.
2.5 Other Geophysical Measurements
 In addition to the methods described above, geomagnetic and gravimetric monitoring is also systematically carried out at Etna. Notwithstanding the many (nine) geomagnetic stations installed near the summit area [Napoli et al., 2008], the temporal series of data indicate that magnetic signals are less reliable than the other geophysical data in terms of their relation with the physical and mechanical phenomena occurring underground ahead of most, generally less energetic, eruptions such as the paroxysms recorded during 2011, and only small coeruptive variations are usually measured. Conversely, gravimetric measurements could provide important constraints in understanding the dynamics of the shallow plumbing system. However, at Etna, because of the logistic difficulties involved in conducting gravity surveys, these measurements are usually conducted fairly infrequently, typically at 1 year intervals between surveys, by using relative spring gravimeters, which highlight spatiotemporal gravity variations with respect to a fixed reference site [Carbone and Greco, 2007]. The lengthy periods between absolute gravity surveys mean a low resolution in the ability to monitor subsurface events over time, even though the gravity measurements are very accurate when made. Since in our analysis we will consider continuous data, we prefer not to use discrete measurements, such as those mentioned above or also GPS surveys. Moreover, here we do not consider data coming from three continuous gravity stations, which are also available to investigate high-frequency gravity changes (in the order of a few hours), for their irregular working.
3.1 Volcanic Activity
3.1.1 Volcanic Activity During 2009–2010
 During the first half of 2009, slow, low-rate lava effusion from the eruptive fissure on the upper eastern flank of Etna, which had opened on 13 May 2008, continued to feed short lava lobes and tongues onto a large apron of lavas formed during this eruption. The activity was entirely effusive until 8 March, when ash emissions occurred from a vent located on the lower part of the eruptive fissure, heralding a short phase (11–14 March) of Strombolian activity and increased lava output. New lava flows were emplaced along the northern margin of the lava apron, reaching up to 1 km in length. After the Strombolian activity ceased, lava effusion decreased to very low rates, continuing to drop slowly until the night of 5–6 July, when the eruption ended after 417 days.
 Four months later, on the late afternoon of 6 November 2009, a bright glow marked the opening of a new pit on the eastern flank of the Southeast Crater (SEC) cone. This roughly 10 m wide pit formed on the eastern rim of SEC, which had been the source of a long series of eruptive events between summer 2006 and spring 2008. Intermittent glowing from this pit was observed through mid-March 2010, but no magmatic products were ejected from the vent. A strong, phreatomagmatic explosion took place from the pit crater on the afternoon of 8 April 2010, causing the rims to collapse and enlarging the pit by a few tens of meters. Further collapses occurred intermittently over the following months, most notably on the morning of 19 June, again enlarging the pit. After two minor ash emissions, on 5 July and 12 August 2010, a powerful phreatomagmatic explosion took place from the BN on the afternoon of 25 August, producing a dark ash plume that rose 1–2 km above the volcano summit. Further, less energetic explosions occurred frequently over the following week, before the interval between such events increased (Table 1 ash emissions from BN, August–December 2010). The latest ash emission from the BN occurred on 22 December, one day before the focus of Etna's activity shifted to the abovementioned pit crater.
Table 1. Significant Eruptive and Collapse Events at Etna's Summit Craters During 2009–2010a
aThe very short lived events are characterized by missing end times. For some events, the exact start and end times are not available.
1 January to 6 July 2009
Upper E flank
[13 May 2008]
Before 04:00 on 6 July 2009
Continued, slow, and diminishing lava outflow from 13 May 2008 fissure
6 November 2009
Opening of a collapse pit 10 m wide and showing bright glow at night
8 April 2010
A series of explosive ash emissions, small pyroclastic flow about 100 m long to NE, widening of the 6 November 2009 pit
19 June 2010
A series of ash emissions caused by collapse of a large section of the western rim of the 6 November 2009 pit
5 July 2010
A single small ash emission
8 August 2010
A single small ash emission
25 August 2010
Strong explosion generating ash column 1 km high, followed by a series of smaller ash emissions; light ash falls to SE as far as Catania; collapse of the W rim of the crater; further small ash emissions at 17:29 and 23:25
27 August 2010
Small ash emission
28 August 2010
Small ash emissions
29 August 2010
Small ash emissions
30 August 2010
Small ash emission
Sporadic small ash emissions, often only seismically detected (due to poor weather)
7 October 2010
Moderate ash emission, light ash falls to south (to Rifugio Sapienza and little beyond)
22 October 2010
Moderate ash emission; three explosion signals (seismicity)
1 November 2010
Ash emission, several explosion signals
12 November 2010
Dull, roaring sounds from 6 November 2009 pit; light emissions of rather dilute ash
14-15 November 2010
Near-continuous ash emission accompanied by loud roaring sounds (audible in summit area); the ash contains a significant proportion of juvenile material
22 December 2010
Moderate-strong ash emission generating a plume a few hundred meter high; ash to NE to as far as Linguaglossa
23 December 2010
Strong incandescence from 6 November 2009 pit, possibly Strombolian activity; spike in volcanic tremor amplitude
29 December 2010
Sporadic, very weak emissions of hot gas (and possibly minor amounts of ash) from 6 November 2009 pit
 An increase in the volcanic tremor amplitude and intense glow at the SEC pit crater on the late afternoon of 23 December 2010 marked the first significant change in the activity leading to the 2011 eruptive phase.
3.1.2 Volcanic Activity During 2011
 The latest, and ongoing, phase of eruptive activity started in earnest during the first few days of January 2011 with a brief episode of mild Strombolian activity from the pit crater of the SEC cone. Ten days later, a much more powerful, paroxysmal eruptive episode occurred from the same pit crater, the first in a series of similar events whose count, at the end of 2011, had reached 18 (Table 2). For the first half of 2011, these events were separated by quiescent intervals ranging from 4 to 8 weeks, but in July, the paroxysmal episodes became more frequent with the quiescent interval dropping to as little as 5.5 days (between the episodes of 25 and 30 July). From late July, the quiescent intervals became progressively longer again (Table 2). The last episode of 2011, on 15 November, was followed by 50 days of quiescence; seven further paroxysmal episodes occurred again between 5 January and 24 April 2012. The most striking result of this series of paroxysmal eruptive episodes was the growth of a large new pyroclastic cone around the former pit crater formed on 6 November 2009, now named “New Southeast Crater” (NSEC). By mid-December 2011, the tallest portions of this cone stood approximately 3200 m above sea level, about 200 m over the pre-January 2011 surface (Figure 2), and 90 m lower than the top of old SEC cone. The southeastern rim of the crater was cut by a deep breach coinciding with the NW-SE-trending eruptive fissure that had erupted during all episodes since episode #12 and which channeled all lava flows issuing from the crater. That fissure extended from the southeastern flank of the cone across the crater and for a short distance beyond its northern rim.
Table 2. Eruptive Episodes During 2011 at the New Southeast Crater
Lava Flow Length
Tephra Fall Direction
Small Strombolian explosions started early on 11 January
Poor visibility due to inclement weather
Violent explosive interaction between lava flows and snow/ice
Two days after this episode, magmatic activity starts at the Bocca Nuova and lasts 7 days
Long series of powerful explosions with loud detonations at end of paroxysm; small rheomorphic lava flow on SE flank of new cone
Long series of powerful explosions with loud detonations at end of paroxysm
Eruption column 3–5 km tall
A portion of the SE flank of the new cone collapses at the height of the paroxysm and a small lava flow extrudes from the collapse scar
A new eruptive fissure about 200 m long opens on the SE flank of the new cone at the height of the paroxysm (04:20 GMT)
A fissure opens on the north flank of the new cone, generating small lava flows and phreatomagmatic explosions from various vents; eruption column possibly 6–8 km tall; a portion of the cone is rafted by flowing lava and then rotated into a vertical position forming a conspicuous rock pinnacle
Lava fountaining from a vent within the crater and one on the southeast flank of the cone
Lava fountains from the main vent within the crater rise up to 800 m high; minor fountains from fissure on SE flank of cone and from a vent on its N flank
Lava fountains from main vent within crater and minor fountains from fissure on SE flank of cone
Lava fountains from main vent within crater and from a vent on SE flank of cone
Lava fountains from main vent within crater up to 800 m high, and minor fountains from fissure on SE flank of cone
 In Figure 3, the maps of the lava flows emitted during the 18 eruptive episodes of 2011 at NSEC are shown. Preliminary estimates of the volume of the volcanic products erupted in 2011 give about 25–30 × 106 m3 of lava, and about 20 × 106 m3 of pyroclastic material, most of which constitutes the newly formed cone, whose volume by the end of the year exceeded 10 × 106 m3. The bulk volume of all products erupted during 2011 is at least 30 × 106 m3. The durations of the most intense phases of lava fountaining, lava effusion, and intense tephra emission varied from 20 min to a few hours; in general, the episodes were shorter during the latter part of the year. Likewise, the buildup periods (from the first signs of renewed activity to sustained lava fountaining) varied from a few days to less than 2 h, shorter buildup periods becoming more common with time. Most paroxysms ended rather abruptly, though two (#6 and #7) ceased with long-drawn-out sequence of Strombolian and vulcanian explosions lasting for several hours after the end of lava fountaining. Episodes 1 through 11 were characterized by eruptive activity from two or three vents, all located within the NSEC crater rim; one of these vents was located near the east rim of the crater, where a deep notch remained open throughout all paroxysmal episodes due to the vigorous outflow of lava. Minor deviations from this pattern were observed during episode #3 (10 April), when a series of explosions occurred from a vent at the SE base of the newly forming cone, followed by the emission of a small lava flow. Likewise during episode #6 (19 July), when heavy fallout of fluid spatter onto the south flank of the cone led to the formation of a thick, rheomorphic lava flow, though it cannot be excluded that an eruptive vent opened at the same time on the SE slope of the cone. From episode #10 (12 August) on, the SE flank of the cone showed increasing signs of structural instability, starting with a minor collapse after the end of episode #10, apparently caused by flowage of still-hot material on the steep flank of the cone. The same area underwent more significant collapse during episode #11 (20 August), caused by the intrusion of lava through the cone's flank; a more voluminous lava flow was emitted from the collapse area in this event. Finally, during episode #12 (29 August), the SE slope of the cone was cut by a short eruptive fissure, along which several vents became active, and the focus of effusive activity shifted to this location. Subsequently, lava flows took a more southerly course than those emitted during the previous 10 episodes (Figure 3). The NSEC was not the only summit vent of Etna to show eruptive activity in 2011. Ash emissions from the BN started on 14 June, heralding a short but vigorous phase of Strombolian and effusive activity from a single vent on the eastern crater floor, which lasted from 11 to 17 July. This was the first significant magmatic activity in that crater since the spring of 2001, coinciding timely with the acceleration in the frequency of paroxysmal episodes from the NSEC mentioned above.
3.2 Geochemical Data
3.2.1 Plume CO2/SO2 Ratios
 After nearly 3 years (2006 to 2008) of substantial changes in its chemical nature (observed in coincidence with—prior and during—the main eruptive events; see Aiuppa et al. [2007, 2010a] for detailed reports), the volcanic gas plume issuing from Etna's central summit craters (CCs; BN and VOR) showed the first protracted (~12 months) period of compositional constancy since 2005 (when the MultiGAS network became operational). Figure 4d shows that the CCs volcanic gas plume underwent a period of relatively stable CO2/SO2 ratios (range, 0.4–7; all ratios quoted here and below are molar) from late May 2009 (thus anticipating by > 1 month the termination of the 2008–2009 effusive period) to early June 2010. For this ~12 month period, we calculated an average CO2/SO2 ratio of ~2.3, well below the 2006–2008 (~8.5) and 2005–2011 (~6.6) time averages, respectively. This unusual (for Etna) steady degassing behavior was interrupted, in the second half of 2010, by two large cycles of CO2/SO2 ratio increase-decrease (Figure 4d), similar in magnitude and duration (~60 and ~90 days, respectively) to those observed prior to the March 2007 to May 2008 paroxysmal episodes. Both cycles consisted of an initial phase of CO2 increase (relative to SO2), culminating with very high (>25) CO2/SO2 ratios in the median portion of the cycle, followed by a phase of CO2/SO2 ratio decrease, which, as observed previously (in 2007–2008) [Aiuppa et al., 2010a], anticipated the resumption of eruptive activity at the summit craters. More specifically, the former of these cycles anticipated the reawakening of the BN in late August–September 2010 (see section 3.1.1), while the latter was a prelude to the resumption of paroxysmal activity at the Southeast Crater in January 2011 (NSEC episode #1). Etna's repeatedly explosive activity in 2011 resulted in a highly dynamic pattern of variation of the CC plume CO2/SO2 ratio (Figure 4d). The 2011 CO2/SO2 ratio average (5.9) remained well above the 2009 average (~2.3) (a statistically significant difference), indicating that fresh CO2-rich gas/magmas persisted in replenishing the shallow plumbing system. While extremely cold and snowy conditions on Etna's summit during the winter of 2010–2011 sometimes caused the MultiGAS malfunctioning, leading to numerous data gaps, we were still able to detect noticeable CO2/SO2 ratio increases at the CCs days to weeks prior to episodes #2–#4 of the NSEC (Figure 4d). After a period of relative CO2/SO2 ratio steadiness in May–June 2011 (associated with a pause in eruptive activity at the NSEC), a new change (CO2/SO2 ratio >15) in early July announced the renewal of paroxysmal activity (NSEC episode #5) (Figure 4d). From July onward, when paroxysmal episodes became more frequent (see section 3.1.2), changes in plume compositions were consistently more rapid (large day-to-day variations in the CCs were recurrent), and clear preevent CO2/SO2 ratio increases at the CCs were captured by the MultiGAS network prior to some NSEC episodes (e.g., episodes #9–#12), whereas they were less visible on other occasions (e.g., episodes #6 and #13).
3.2.2 SO2 Fluxes
 Figure 4b shows the daily SO2 flux measurements from the bulk volcanic plume of Mount Etna between January 2009 and December 2011. During this 3 year period, the emission rate was characterized by a mean value of 2500 Mg d−1 (standard deviation δ 755 Mg d−1), ranging from a minimum value of 300 Mg d−1 (19 October 2010) to a maximum of 11,000 Mg d−1 (22 May 2011). The diagram shows that SO2 emission rates rose significantly in two periods, from October 2009 to February 2010, and from late September 2010 until the end of 2011, with mean emission rates approximately twice the mean long-term ones. In the first period of increased emission rates, the trend in the SO2 flux fluctuated around a mean value of ~3600 Mg d−1, punctuated by peaks, the highest of which occurred on 8 November 2009 and on 2 and 11 February 2010 (mean value of ~8500 Mg d−1). Throughout this 5 month period, the SO2 flux increase went through three main wide fluctuations of almost 3 month periodicity, and then the signal leveled off sharply to mean values of ~2000 Mg d−1 by the end of February 2010. In the second period of sustained SO2 emission, the mean flux was just below that measured in the first period (mean value of ~2700 Mg d−1), and the emission rate peaked several times with values between 8000 and 11,000 Mg d−1.
3.2.3 CO2 Fluxes
 The dramatic variations of both SO2 fluxes (Figure 4b) and CO2/SO2 ratios (Figure 4d), recorded at Etna during 2009–2011 and described above, are ultimately reflected in the highly dynamic nature of the CO2 flux record (Figure 4c), which is obtained by combining the two above datasets. We show (Figure 4c) that the CO2 flux remained stable and low (<13,000 Mg d−1) for nearly 1 year, from the summer of 2009 (when the 2008–2009 eruption ended) to the spring of 2010. From the summer of 2010, however, the rate of CO2 flux release increased dramatically, with two main pulses of unusually large CO2 degassing in June–July 2010 and, particularly, in September–December 2010 (these mirroring the two cycles of CO2/SO2 ratio increase is described above). The CO2 flux persisted at remarkably high (>20,000 Mg d−1) levels until January 2011, accompanying the onset of episodic eruptive activity at the New SEC: the >100,000 Mg d−1 CO2 fluxes measured in January–March 2011, when put into the context of previous works [Allard et al., 1991; Aiuppa et al. 2006, 2008], emerge as by far the highest ever reported for Etna volcano to date.
3.3 Seismic Data
 Notwithstanding the great variety of seismic signals recorded at Mount Etna, we can divide seismovolcanic signals into two large groups. The first group includes volcano-tectonic (VT) earthquakes [e.g., Chouet, 1996; Patanè et al., 1997, 2004], generated by tectonic stress and/or by stress arising from magma upwelling in the Earth's crust. The second group is related to the seismic manifestations of fluid dynamics, which includes hybrid, long-period (LP), and very long period (VLP) events, volcanic tremor, and explosion quakes, which clearly differ from VT earthquakes in their seismic and spectral signatures [e.g., Del Pezzo et al., 1993; Patanè et al., 2004, 2008].
3.3.1 VT Seismicity
 At Mount Etna, about 50% of VT earthquakes are very shallow (focal depth <5–10 km) and affect almost the whole area, though prevalently the eastern flank. Similarly to what has been observed in many volcanic areas [McNutt, 2005], these earthquakes show high-frequency content (mostly above 4–5 Hz) and sharp arrivals, comparable to those of typical earthquakes of tectonic areas, and magnitudes generally lower than 4 [e.g., Ferrucci and Patanè, 1993]. They mostly occur in the form of swarms, whereas foreshock-mainshock-aftershock sequences are rarely recorded [Patanè et al., 2004]. During recharging phases, VT seismicity mostly affects the western and southern sectors of the volcano, with foci generally in the medium-lower crust (10–30 km) [Patanè et al., 2004].
 VT earthquakes at Etna during 2009–2011 were studied obtaining the following information: daily number of VT earthquakes and a corresponding cumulative seismic strain release curve (Figure 4e); the time variation of VT earthquake depths (Figure 4f); seismic strain release maps for different subsets of VT earthquakes, chosen according to their times and source depths (Figure 5a); and seismic strain release sections highlighting the seismicity distribution in depth during three different time periods (Figure 5b).
 Most of the VT earthquakes were shallow (< 10 km depth), and their epicenters were mainly located below the eastern flank (Figures 5a and 5b). The deep VT earthquakes (> 10 km depth) mostly took place below the northwestern (~20–30 km below sea level (bsl)) and southern (~10–15 km bsl) flanks (Figures 5a and 5b). Considering the time variations of seismicity, the main deep swarms involving the northwestern flank took place in three different seismic sequences in December 2009, June 2010, and May 2011 (Maletto area; Ma in Figures 1a, 4e, and 4f). The deeper activity affecting the southern flank occurred during December 2008 (one month before the studied period), January and September 2009, July 2010, and March 2011 (Parmentelli-Ragalna area; PR in Figures 1a, 4e, and 4f). For the strongest seismic swarm, a significant variation in the seismic strain release can be recognized (Figure 4e). On the other hand, during the studied period, also important shallow VT swarms took place in the eastern flank, the most significant on May 2009 (along the S. Leonardello fault; SLf in Figure 4e) and at the beginning of April 2010 in the Pernicana fault area (PF in Figure 4e). It is noteworthy that during the 2011 eruptive period, the seismic activity returned to a low level, as during the past eruptive periods, such as the months following the onset of the 2008 eruption (May–December 2008).
3.3.2 Volcanic Tremor and LP and VLP Events
 As for the volcanic tremor at Etna, a peculiar aspect is its continuity in time, as also observed at other basaltic volcanoes with persistent activity such as Stromboli (Italy) [Langer and Falsaperla, 1996]. Another interesting feature of the volcanic tremor is its close relationship to eruptive activity, highlighted by variations in amplitude, spectral content, wavefield features, and source location at the same time as changes in volcanic activity [e.g., Alparone et al., 2007b; Cannata et al., 2008; Patanè et al., 2008; Cannata et al., 2009a]. In order to obtain information about the time changes of tremor energy, the RMS of the seismic signals recorded at the vertical component of EBEL station (Figure 1d) was calculated in the band 0.5–5.5 Hz within 10 min long sliding windows (Figure 4g).
 The tremor source locations were retrieved by following the approach described by Patanè et al.  and Cannata et al. , inverting the spatial distribution of the tremor amplitude at 8–19 broadband stations located at altitudes higher than 1000 m asl (black triangles in Figure 1d) using a grid-search approach (Figures 4h and 6). On the basis of seismic volcanic tremor source locations, we also evaluated the reduced displacement (RD) from body waves by using the signal recorded by station EBEL during 2011 (Figure 7). RD is a measure of the tremor amplitude corrected for geometrical spreading and instrument magnification, given by Aki and Koyanagi , useful to make comparisons among different activities at the same volcano as well as at distinct volcanoes.
 A number of papers deal with the relation between eruptive activity and LP events at Mount Etna [Patanè et al., 2008; Di Grazia et al., 2009; Cannata et al., 2009b]: it was shown how occurrence rate, energy, spectral content, and/or source location of LP events often change before, during, and after eruptive activities. LP events recorded during the period from January 2009 to December 2011 were investigated through several parameters: (i) occurrence rate, (ii) peak-to-peak amplitude, and (iii) source location. The ECPN station, located just southwest of the summit craters, was used as the reference station because of its longest record (Figure 1d). About 1,060,000 LP events, whose daily number and peak-to-peak amplitudes are displayed in Figures 8a and 8c, were detected during the analyzed period. A subset of 230,000 LP events with high peak-to-peak amplitudes was selected to perform location analysis. LP events were located by following a new grid-search method adopted here for the first time, based on the joint computation of two different functions: (i) semblance, used to measure the similarity among signals recorded by two or more stations [e.g., Neidell and Taner, 1971] and (ii) R2 values, calculated on the basis of the spatial distribution of seismic amplitude [Patanè et al., 2008; Cannata et al., 2010]. The nine stations nearest to the summit area were used (the summit station ring, made up of seven stations, and two stations belonging to the middle ring), and the 3-D grid of possible locations was 6 × 6 × 4.25 km3, centered within the volcanic edifice and with a vertical extent from −1 km to 3.25 km asl (~ the top of the volcano). The horizontal and vertical grid spacing was 250 m. The space distributions of both semblance and R2 values were determined. The two grids of values were normalized by subtracting the minimum value and dividing by the maximum one. Thus, the values belonging to two grids ranged from 0 to 1, and the same weights were assigned to semblance and R2. Then the two normalized grids were summed node by node. The source was finally located in the node with the largest composite semblance-R2 value. The surfaces enclosing all the grid nodes with more than 800 LP events during the whole period (Figure 9a), as well as during three distinct time intervals, were obtained and drawn in Figures 9b–9d. Furthermore, VLP events, whose waveforms at Mount Etna are generally made up of a couple of cycles lasting ~20 s [e.g., Cannata et al., 2009b], were also studied. About 1,330,000 events were detected, whose daily number is given in Figure 8b. Also in this case, a subset of about 180,000 events was extracted and located. The mostly radial wavefield of these events makes the radial semblance method [Kawakatsu et al., 2000] suited to locate them. Also in the case of VLP events, the surface enclosing all the grid nodes with more than 800 VLP events during the whole period was obtained and drawn in Figure 9e.
 The volcanic tremor sources were located at shallow depth below the summit area (from 0 km asl to the top of the volcano; Figure 6). LP and VLP events, whose daily number ranged from a few to more than 3000, were also located at very shallow depths, generally above 2 km asl (Figures 9a and 9e).
 Focusing on the time variations of the seismic signals, from January 2009 to November 2010, volcanic tremor sources were located east of the summit area at altitude 1.5–3.0 km asl, roughly below the fissure of the 2008–2009 eruption (Figures 4h and 6). At the same time, LP sources were located below the BN at 2–3 km asl (Figure 9b). A first significant change took place from May to November 2010 when a gradual increase in the LP peak-to-peak amplitude occurred (Figure 8c). In November 2010, several changes in the seismic signals were observed: (i) volcanic tremor sources shifted toward the NEC at altitudes ranging from 1 to 2 km asl (Figures 4h and 6), (ii) LP peak-to-peak amplitudes sharply decreased (Figure 8c), (iii) LP sources moved toward the NEC and to higher altitude (>2.5 km asl) (Figure 9c), and (iv) seismic strain release, associated with shallow VT earthquakes, sharply decreased. Subsequently, another period characterized by many changes in seismic signals was June–July 2011 when (i) a slight increase in seismic release, mainly due to shallow VT earthquakes located below the summit area, took place (Figures 4e); (ii) seismic RMS increased (Figure 4g); (iii) the volcanic tremor source became shallower (up to 3 km asl; Figure 4h); (iv) LP peak-to-peak amplitudes increased (Figure 8c); and (v) LP sources moved again toward the BN at altitudes comprised between 2–3 km asl (Figure 9d).
 It is noteworthy that other important and short-lived changes, time related to the lava fountaining episodes, affected the volcanic tremor. Indeed, as already observed by other authors analyzing lava fountains at Etna [e.g., Patanè et al., 2008], the energy of the volcanic tremor increased and its source moved toward the NSEC to very shallow depth (roughly 3 km asl), in coincidence with the lava fountains. Further, RD reached maximum values of almost 150 cm2 and for all the paroxysmal episodes overcame 30 cm2, which at Mount Etna, in literature, is associated to intense Strombolian activity and lava fountain [e.g., Alparone et al., 2003; Sciotto et al., 2011].
3.4 Ground Deformation Data and Modeling
 Since 2008, the increased spatial density of the CGPS network and the better coverage in the summit area has enabled a detailed analysis of the geophysical processes acting at different spatial and temporal scales on this volcano. Therefore, it is now possible to consider for surveillance purposes, beyond the lengthening-shortening of GPS baselines to track the evolution of inflation-deflation cyclic phases [e.g., Patanè et al., 2006, 2008; Di Grazia et al., 2009, Aiuppa et al., 2010a], also the analysis of areal deformation, in order to monitor deformation patterns and investigate their spatial and temporal variations in different sectors of the volcanic edifice.
 Between 2008 and 2011, the areal variations calculated for two triangles (Figure 4i), one related to three summit GPS stations (EPDN-EPLU-ECPN) and the other related to three intermediate altitude GPS stations (EMCN-ESLN-EMEG), showed different trends and revealed well-defined patterns of ground deformation. On the basis of these trends, Aloisi et al. [2011a] modeled ground deformation sources, choosing the following periods (see Figure 4i; see also Aloisi et al., 2011a): (1) 14 May to 02 August 2008 (deflation of the entire GPS network), (2) 03 August 2008 to 14 June 2009 (deflation of the summit area and inflation at lower altitudes), (3) 15 June 2009 to 21 May 2010 (inflation of the entire GPS network), and (4) 22 May to 31 December 2010 (inflation at medium and low altitudes and end of the inflation in the summit area). In this study, we further consider the following three periods for 2011: (5) 01 January to 20 May 2011 (minor deflation in the summit area and end of the inflation at lower altitudes), (6) 21 May to 16 July 2011 (inflation of the entire GPS network), and (7) 17 July to 17 October 2011 (deflation of the entire GPS network followed by a new inflation observed at the whole GPS network scale, just as recorded during the sixth period). Each period is characterized by an almost uniform pattern of ground deformation, as determined by an in-depth analysis of time series and velocities maps.
 From January 2011, when the lava fountaining activity started, until 20 May 2011 (period 5, Figure 10), the entire GPS network recorded displacements smaller than about 0.005 [m]. Therefore, it is not possible to propose a well-constrained source model for this period. Subsequently, during the sixth period (21 May to 16 July 2011), the entire GPS network showed a rather fast inflation. At the end of this period, the fountaining episodes became more frequent and the GPS network began recording clear deflation (17 July 2011–17 October 2011). Thereafter, the volcano edifice was affected by a renewed rapid inflation, which continued during 2012.
 Differently from Aloisi et al. [2011a] that included the effects of the topography using the varying-depth model of Williams and Wadge , here we modeled all the volcanic sources related to the seven periods described above using the topographically corrected method of Williams and Wadge . This approach is a more precise method to model the combination of horizontal and vertical deformations compared to the varying-depth model of Williams and Wadge , which assumes a different elevation for each recording station corresponding to the relative elevation. The Williams and Wadge  approach provides deeper pressure sources than the ones calculated using the approach proposed in Williams and Wadge  correction. The reason of this result is due to the different approach to the mitigation of the effects related to the topography: indeed, in Williams and Wadge , the correction is applied considering the digital elevation model (DEM) of the volcano, while in Williams and Wadge , the correction is applied only to the heights of the stations, assuming a different reference elevation for each point at which the solution is desired. In this way, this last approach is a so-told “varying depth model,” while the one proposed in Williams and Wadge  is a “topographically corrected model.”
 Regarding the first four periods, starting from the results of Aloisi et al., 2011a, we performed an analytical inversion of CGPS data, describing the pressure source with the arbitrarily oriented, finite, prolate, and spheroidal cavity embedded in an elastic half-space of Yang et al.  and, as said before, applying the topographically corrected method of Williams and Wadge . The results are reported in Table 3. The pressure source is determined by eight parameters: the coordinates xc and yc of the spheroid center, the coordinate zc of the spheroid center depth, the azimuth θ measured counterclockwise from the positive y direction around the z axis, the dipping angle ϕ measured clockwise from the positive y direction around the x axis, the major semiaxis a, the ratio b/a between the minor and major axes, and the intensity of the pressure P on the surface of the spheroid.
Table 3. Modeled Volcanic Sources by Using the Topographically Corrected Method of Williams and Wadge 
500485 ± 230
500524 ± 157
498432 ± 160
497104 ± 323
498917 ± 122
497085 ± 600
499130 ± 155
499651 ± 82
4178453 ± 291
4178625 ± 158
4178070 ± 240
4179778 ± 343
4178522 ± 511
4179777 ± 1164
4179099 ± 82
4177185 ± 95
−1981 ± 365
−1483 ± 239
−4105 ± 176
−8573 ± 70
−4579 ± 104
−8081 ± 974
−5532 ± 356
−4706 ± 229
62 ± 37
0 ± 66
133 ± 35
0 ± 74
62 ± 34
0 ± 56
122 ± 55
43 ± 60
105 ± 6
34 ± 64
97 ± 3
93 ± 34
181 ± 73
348 ± 109
763 ± 214
451 ± 208
1114 ± 251
411 ± 422
223 ± 19
200 ± 14
0.818 ± 0.167
0.143 ± 0.168
0.111 ± 0.220
0.267 ± 0.278
0.073 ± 0.016
0.489 ± 0.178
−4.0E + 09 ± 1.2E + 09
−4.5E + 09 ± 5.8E + 08
7.5E + 09 ± 6.3E + 08
−9.0E + 09 ± 4.0E + 08
7.9E + 09 ± 3.3E + 09
5.3E + 09 ± 6.7E + 08
2.4E + 09 ± 4.9E + 08
−2.8E + 09 ± 5.7E + 08
−1.0E + 07
−2.5E + 06
2.6E + 07
−3.7E + 07
3.7E + 07
5.5E + 07
1.7E + 07
−1.4E + 07
 Regarding the following two periods, we image the deformation pattern recorded by the permanent GPS network and the tilt network during the sixth and seventh periods (Figure 10—the fifth period has not been modelized for the small variations observed, as already stated before), as induced by spherical pressure sources acting inside the volcano. The time series of the GPS permanent stations, due to an uncertainty in kinematic positioning of about ±0.020 (m) for the horizontal components and of about ±0.050 (m) for the vertical component, showed no meaningful displacements related to single lava fountain episodes, except for a small variation observed at the summit station ETFI during the first episode on 12 January 2011. Conversely, tilt stations recorded systematic variations for each single paroxysmal episode, due to their higher level of sensitivity. The stations generally recorded changes of fractions of microradians more clearly on the radial components, indicating a general deflation of the edifice during the fountains. It is, however, interesting to observe that the fountaining activity, as a whole, induced the deflation recorded at the GPS network during the seventh period. To model the two last periods, we used the entire GPS network, except for the stations located away from the volcano edifice and the stations located on the eastern flank of the volcano, which is affected by an almost constant ESE-ward motion whose origin is still debated and controversial (for a review, see, e.g., Aloisi et al. [2011b]). Moreover, we used the variation recorded at the CDV tilt station, which was the only benchmark that, corrected by seasonal drifts, clearly showed the overall inflation and deflation periods (sixth and seventh periods). The pressure source was modeled using the finite spherical magma body solution of Mctigue , obtaining a plausible data fit (Figure 11). In particular, we chose this source model among the other models published in literature since it shows a good trade-off in this case between the number of degrees of freedom and the obtained data fit. The Mctigue  pressure source is determined by five parameters: the coordinates “xc,” “yc,” and “zc” of the sphere center, the radius “a,” and the pressure “P” on the surface of the model.
 To estimate the model parameters for all the periods, we performed an analytical inversion using the Genetic Algorithms [Goldberg, 1989] and, subsequently, applying the Pattern Search technique [Lewis and Torczon, 1999]. Finally, to estimate the uncertainty of each optimized model parameter, a Jackknife resampling method [Efron, 1982] was adopted. As mentioned before, the effects of the topography were included using the varying-depth model of Williams and Wadge . The medium was supposed homogeneous and isotropic with a Young modulus of 75 GPa and a Poisson ratio of 0.25 [Aloisi et al., 2011b].
 Final optimal solutions for each eruptive episode and their spatial distribution are shown in Figures 10 and 12 and Table 3. In Table 3, we also reported the misfit between the observed and modeled data as the value of the WRMSE (Weighted Root Mean Square Error):
where Di is the measured variation with an error ϵi at the i-th benchmark (0.005 m for horizontal displacements, 0.02 m for vertical displacements, 0.5 µrad for tilt observations) and DCi is the respective modeled value.
 It is noteworthy that the pressure sources explaining the inflation period (sixth period from 20 May to 16 July 2011) and the deflation period (seventh period from 16 July to 17 October 2011) are characterized by a similar location below the summit area, at a depth of about 5.0 [km] (bsl). Regarding the modeled variation in magma volume (Table 3), we estimated about +17 x 106 m3 for the inflation period (sixth period) and about −14 × 106 m3 for the deflation period (seventh period) with an uncertainty of about ± 7 × 106 m3.
4 Combined Interpretation and Discussion
 Magma intrusion into the shallow crust (or into the volcanic pile) at Etna may be followed by summit and/or lateral eruptions, after several months of internal replenishment. Intrusive episodes at shallow depth (~3–5 km bsl) affecting the central feeding system are usually followed by a reactivation of the summit craters [i.e., Bonaccorso and Patanè, 2001; Patanè et al., 2004; Di Grazia et al., 2009; Aiuppa et al., 2010a].
 The long-lasting 2008–2009 flank eruption was followed by a quiescent period and, in 2010, by the reactivation of the BN summit crater. These events were precursory to the onset of a new eruptive phase characterizing Mount Etna's summit craters in 2011, with a series of 18 powerful lava fountaining episodes of short-lived and high intensity. In the following sections, we will discuss the results achieved by the combined analysis of the data presented in the previous sections, with the attempt to test the ability of a multidisciplinary approach to quantitatively track the cycles of magma accumulation, degassing and preeruptive to syneruptive ascent, which led to the 2011 summit eruptions at the NSEC.
4.1 Evidence of Magma Ascent During 2009–2010 (Phases I to III)
 As already observed at Mount Etna during previous eruptions [e.g., Patane et al., 2005], the outpouring of magma after the onset of the 13 May 2008 flank eruption [e.g., Aloisi et al., 2009; Di Grazia et al., 2009] was accompanied by intense deflation of the volcanic edifice [see Aloisi et al., 2011a, Figure 2] and very scarce seismicity (VT earthquakes). Deflation phenomena, due to the magma flowing out and consequently to the pressure decrease inside the plumbing system, have often been observed in many volcanoes around the world (e.g., Piton de La Fournaise) [Peltier et al., 2008; Kilauea, Poland et al., 2009]. However, in the case of the 2008–09 eruption, the deflation did not continue throughout the entire eruptive event, and in its course, some significant changes in the ground deformation were recognized, starting only a few months after its onset. In particular, deflation phenomena regarded the entire GPS network during period 1 (see section 3.4), and successively during period 2 only the summit area (see Figure 10 and Aloisi et al. [2011a]). Seismic activity remained at a very low level until 16 December 2008, when a new strong seismic swarm (116 earthquakes; Mmax = 3.7) affected the central-southern part of the volcanic edifice between 9 and 14 km of depth (Parmentelli-Ragalna area; PR in Figures 1a, 4e, and 4f). After this swarm, a significant increase in the number of VT earthquakes and a modification of the associated seismic strain were observed during the first months of 2009. During 2009–2010, we can distinguish three main phases (Phase I to III), thanks to the comparison of the several monitored signals.
4.1.1 Phase I (January–September 2009)
 Seismicity during the first months of the year was at very low levels. It was only from March 2009 that a significant increase in the cumulative energy release was observed which, as seen before at Mount Etna, evolved during the following months with more or less intense periods of seismic activity (Figures 4e and 5). The new inflation phase, which had started at lower altitudes during August–September 2008 (I1 in Figure 4i), accelerated from the first few days of April 2009 (A1 in Figure 4i) and continued at nearly the same rate until November 2010 (Figure 4i). Between late April and the first days of May 2009, GPS measurements at the stations located above 2500 m of altitude (Figure 1c) recorded an inversion of the deflation trend, and the start of a new inflation phase (I2 in Figure 4i). In May 2009, the very shallow seismicity at the S. Leonardello fault (SLf in Figures 1a, 4e, and 4f), combined with increased aseismic creep phenomena along the Trecastagni-Acitrezza fault (TAf in Figure 1a), represent the response to the sudden acceleration of the eastern flank (as evidenced by ground deformation since April at lower elevations on the volcanic edifice). Indeed, horizontal GPS measurements showed a brief period (approximately 2 months) of intense acceleration (~15 cm/yr) in the movement of the southeastern flank and in particular of the easternmost stations. During the first months of 2009, the degassing activity from the summit craters remained regular and constant. The CCs' volcanic gas plume underwent a period of relatively stable and low CO2/SO2 ratios, with values of the CO2 flux oscillating around 10,000 Mg d−1 and average values of the SO2 flux of 2000 Mg d−1. In the light of previous research [e.g., Aiuppa et al., 2010a, and references therein], these time constant and surprisingly low (also compared with results dating back to the 1980s) [Allard et al., 1991] CO2/SO2 ratios support the hypothesis that while the deep magmatic system was undergoing inflation only a very minor (if any) magmatic recharge of the volcano's shallow (0–3 km asl) plumbing system was occurring. Thus, the volatile-poor [Collins et al., 2009] and chemically evolved magma, resident in Etna's upper conduit system, was the main (exclusive) source of gas venting at the summit craters from May 2009 to June 2010.
4.1.2 Phase II (October 2009 to March 2010)
 The increase in seismic release rate recorded during phase I persisted until early September. After a brief period of very low seismicity, an intense deep earthquake swarm affected the northwestern sector of the volcano (Ma in Figures 1a, 4e, and 4f), on 19 December 2009, producing a further increase in the cumulative energy release. This swarm, with more than 400 recorded earthquakes (1.0 ≤ Ml ≤ 4.8) localized between 20 and 30 km of depth, represents one of the most important energy release events in the last two decades in this sector of the volcano. Deep earthquakes in Etna's western sector are thought [Gresta et al., 1990; Bonaccorso et al., 1996; Vinciguerra et al., 2001; Patanè et al., 2004] to be early markers of future volcanic activity, and this process, which acts with surprising regularity at Mount Etna, has already been observed before several eruptions. During the same period, inflation of the edifice continued almost at the same rate, and the CCs' volcanic gas plume registered a slight decrease of the CO2/SO2 ratios, reflecting a factor ~2 SO2 flux increase (~4000 Mg d−1) relative to phase I, at nearly constant CO2 flux (~10,000 Mg d−1). This increased degassing of SO2 flux from the shallow magmatic system (suggested by the fact that SO2 degassing was prevailing over CO2 degassing) resulted in the first significant event in terms of volcanic activity after the end of the eruption in July 2009. Notable change in SO2 emission rates has been observed elsewhere prior to the onset of eruptive activity, e.g., Mount St. Helens [Casadevall et al., 1983], Pinatubo [Daag et al., 1996], Kilauea [Sutton et al., 2001], and during the lava fountaining sequence at Pu'u ‘O'o vent -Kilauea [Chartier et al., 1988]. Indeed, on 6 November 2009, a small pit crater on the lower eastern slope of the SEC formed (Figure 4a), showing vivid incandescence for several months without producing any explosive activity. During the first months of 2010, the degassing activity from the summit craters showed alternating phases of higher or lower intensity without, however, culminating in any emission of solid material up to 8 April 2010.
4.1.3 Phase III (April–December 2010)
 A strong shallow seismic swarm on 2–3 April along the Pernicana fault hit the northeast of the volcano (Pf in Figures 1a, 4e, and 4f), marking the start of phase III. This swarm, comprising about 240 earthquakes with a main shock of Ml = 4.3, caused the largest energy release along this fault since the one accompanying the beginning of the 2002–2003 eruption.
 In the following months, the seismic activity remained low; a slight increase in its occurrence rate was recorded only from late October to early November, with the activation of several structures on the lower southeastern flank of the volcano. It is noteworthy that between March and October 2010, significant variations in the seismovolcanic activity occurred, with an increasing trend of several parameters such as the RMS of the volcanic tremor and peak-to-peak amplitudes of LP events (Figures 4g and 8c). More sustained seismovolcanic activity has commonly been interpreted as due to pressurization of volcanic plumbing systems, at Etna [Patanè et al., 2008; Di Grazia et al., 2009] and at other volcanoes as well such as Redoubt [Chouet et al., 1994], St. Helens [Moran et al., 2008], and Colima [Varley et al., 2010]. In this light, both the modest explosive event of 5 July 2010 and the much stronger event of 25 August 2010 at the BN can be considered as the response to slow but continuous pressurization of the shallower portion of the plumbing system. This hypothesis is particularly corroborated by geochemical data. Indeed, in comparison to the period of relatively stable (and low) plume CO2/SO2 ratios which had been observed from May 2009 to early June 2010 (all indicative of a very slight magmatic recharge to the volcano's shallow plumbing system, as well as the presence of volatile-poor magma, resident in Etna's upper conduit system), the volcanic gas plume issuing from the Etna's CCs showed two large cycles of CO2/SO2 ratio increase-decrease (the first one from the end of May to the first days of July and the second during September–December 2010; Figure 4d). These cycles are similar in magnitude and duration (~60 and ~90 days, respectively) to those observed prior to the March 2007 to May 2008 paroxysmal episodes at the Southeast Crater [Aiuppa et al., 2010a]. Overall, these large CO2/SO2 ratio variations are supportive of two pulses of more deeply rising CO2-rich gases and/or gas-rich magmas [Aiuppa et al., 2007; Shinohara et al., 2008], recharging the shallower portions of the plumbing system starting in May 2010, as also underlined by increase in SO2 flux since September (Figure 4b). Similar pulses of CO2 gas release have been observed at other volcanoes in different context, including for instance Stromboli [Aiuppa et al., 2010b], Kilauea [Poland et al., 2012], and Redoubt [Werner et al., 2012]. The former of these two cycles, occurring between the end of May and the first days of July, anticipated, together with the renewal of very shallow seismicity taking place at the beginning of April along the Pernicana fault (Figures 4e and 4f), the resumption of sporadic eruptive activity at the summit craters since the strong explosion taking place from the pit crater on 8 April 2010 (Figure 4a). The reactivation of the CCs occurred with a powerful phreatomagmatic explosion that took place from the western vent of the BN on the afternoon of 25 August, producing a dark ash plume that rose 1–2 km above the summit of the volcano. Further, less energetic explosions occurred frequently over the following weeks. The latest ash emission occurred on 22 December, a day before the focus of Etna's activity shifted to a new location, namely, the pit crater formed on 6 November 2009 on the lower east flank of the SEC cone. This marked the first sign of a significant change in the activity leading to the ensuing 2011 eruptive phase.
 Overall, if we consider the sources of ground deformation active during 2009–2010, after the deflating phase accompanying the 2008–2009 eruptive period (period 1 and part of period 2), the inflating source has deepened with time, migrating from about −4.0 km bsl to about −8.0 km bsl (periods 2, 3, and 4 in Table 3). As suggested by Aloisi et al. [2011a], this behavior may be closely related to gas boiling. In agreement with these authors, magma located in the high portion of the plumbing system, subjected to low confining pressure, reaches saturation first, whereas the deep residing magma, subjected to higher confining pressure, needs more time to reach saturation generating the overpressure necessary to trigger inflation. Therefore, the evidence of a progressive deepening of the inflation sources could indicate that magma residing in Etnean plumbing system, from the volcanic edifice down to about −8 km bsl, was progressively recharged in gas content indicating a greater likelihood of an imminent explosive eruptive cycle.
4.2 The 2011 Cyclic Eruptive Activity of Lava Fountains (Phases IV and V)
 On 12 January 2011, a powerful paroxysmal eruptive episode from the pit crater on the eastern flank of the old SEC cone marked the beginning of a series of 18 similar episodes that occurred during 2011 (Table 2). The intervals between these episodes varied greatly, from several weeks in the first half of 2011 to little more than 5 days at the end of July; after mid-August, the intervals became progressively longer (Table 2). The cyclic eruptive activity in 2011 reflected the highly dynamic pattern of variations in the CCs' plume CO2/SO2 ratio and in the bulk SO2 flux (Figures 4b and 4d).
 Seismic data and, in particular, volcanic tremor and LP and VLP events, were particularly useful to locate (and image) such shallower magmatic plumbing system, feeding both the persistent degassing at the CCs and, cyclically, the paroxysmal activity at the NSEC. By using volcanic tremor source locations, the shallower magma storage zone, feeding the 2011 fountaining activity, was located below the summit area at 1–2 km asl. As already observed by Aiuppa et al. [2010a], LP and VLP sources are located above this shallow magma storage zone. These events could be triggered by gas bubbles, released from the aforementioned magma batch and feeding surface gas emission. The 2011 Etna eruptive activity can be divided schematically into two main parts, as described below.
4.2.1 Phase IV (January–May 2011)
 The CO2 fluxes persisted at remarkably high values (>20,000 Mg d−1; Figure 4c) during the recurrent eruptive activity at NSEC in phase IV (four episodes recorded; Figure 4a), with the >100,000 Mg d−1 fluxes measured in January–March 2011 representing the highest ever reported for Etna, to date [Allard et al., 1991; Aiuppa et al. 2006, 2008]. Similarly, SO2 flux over this 5 month period indicated sustained emission rates, with a maximum value of the whole time series measured on 22 May (11,000 Mg d−1). The observations suggest that fresh (CO2-rich) magmas, having been accumulating below the volcanic edifice in the previous months and/or being eventually resupplied in between the NSEC eruptive episodes, sustained the cyclic paroxysmal activity. Notably, it was only when both CO2 and SO2 emissions slowed down that the NSEC activity finally paused (in May–June).
 During phase IV, the summit area (small triangle in Figure 10) recorded a continuous deflation, likely related to magma outpouring during the fountaining episodes (Figure 4i), and VT seismicity was low. This trend was reversed from mid-May, however, when a rapid inflation phase was recorded (Figures 4i, 10, and 11a). The inflation phase followed by only a few weeks a new intense, deep earthquake swarm affecting the northwestern side of the volcano area (Ma in Figures 1a, 4e, and 4f), producing a further modification of the cumulative energy release (Figures 4e). The Maletto swarm, with more than 170 recorded earthquakes (1.0 ≤ Ml ≤ 3.6), localized between 20 and 30 km depth bsl, represented the most important release of seismic energy during 2011. The simultaneous occurrence of deep seismicity and ground deformation confirms that earthquakes taking place in the western sector can be closely related to magma replenishment of the deep plumbing system, which may be rapidly followed by a renewal of volcanic activity, as occurred in the following phase V.
4.2.2 Phase V (June–December 2011)
 The early May Maletto seismic swarm was followed by about 2 months of sudden increase in microseismicity, localized in the central-eastern part of the volcano at a depth of less than 10 km (Figures 4e and 4f). It is noteworthy that in June, most of the VT earthquakes were localized close to the summit crater at very shallow depth, the majority above the sea level. Between 20 May and 16 July, the entire GPS network showed a very fast inflation (Figures 4i and 10). In June, LP event and volcanic tremor activities intensified suggesting pressurization phenomena of shallow volcanic plumbing system, as observed during phase III. At the end of this phase, and after a hiatus of 2 months, the fountaining episodes restarted (on 9 July) and became more frequent; at the same time, the BN showed a brief period of Strombolian activity (the first over the last 10 years), and the GPS network started to record again a clear deflation of the summit area (16 July 2011 to 17 October 2011; Figures 4i, 10, and 11b). It is worth noting that both the inflation and deflation sources, modeled by ground deformation data, have a similar location below the summit area at a depth of about −5.0 km. This level is compatible with the magma storage volume revealed by our modeling in the plumbing system, at a depth of about 4 km (bsl). The renewed paroxysmal activity in phase V (Figure 4a) was also announced by a new change in plume composition (CO2/SO2 ratio >15) and in SO2 flux (~6000 Mg d−1) in early July. Therefore, geophysical and geochemical variations in concert heralded the renewal and intensification of the paroxysmal activity. Subsequently, from November 2011 onwards, the volcano edifice was affected by a new fast inflation and increase in SO2 flux release, persisting beyond the end of 2011. These processes, very similar to those recorded during May–June 2011, heralded the resumption of eruptive activity at the NSEC in early January 2012.
 The identification and interpretation of eruption precursors is a crucial aspect of modern volcanology. As the range of monitoring tools increases, the models required to interpret the acquired datasets, and identify precursory signals to eruption, need to become more and more sophisticated. Precise eruption forecasting requires clear understanding of the conduit and/or deeper processes, and the interaction between ascending fluids and the surrounding environment. It is now well accepted that the most promising approach to detect and investigate volcanic unrests is the joint monitoring and analysis of geophysical (first and foremost seismic and ground deformation) and geochemical signals [Scarpa and Gasparini, 1996; Tilling, 2008].
 With this premise, we have jointly analyzed volcanological, geochemical, and geophysical data, with the objective of identifying the most important variations in the monitored parameters, the time-space relationships among them, and to investigate the volcano dynamics and magma movements in both the deep and shallow portions of the plumbing system. The availability of high-quality data, derived from the modern, dense multiparametric sensor network, has allowed us to accurately characterize the intermediate-shallow plumbing system geometry and Etna's behavior in 2009–2011. This improved knowledge has enabled making forecasts and alerting authorities and Civil Defense days/hours before the onset of each paroxysm [e.g., Alparone et al., 2007a; Patanè et al., 2008; Di Grazia et al., 2009; Aiuppa et al., 2010a].
 The main results of this work can be summarized as follows:
 Concerning identification of deep magma recharge events, about three decades of detailed seismological studies have shown that the distribution of VT seismicity in the intermediate and lower crusts (10–30 km) beneath Etna can be key to infer the seismic response of local structures to the stress field induced from the episodic ascent of magmatic fluid. During the investigated period (2009–2011), several intense seismic swarms in the intermediate-lower crust have occurred before and during the eruptive cyclic phase of 2011, suggesting the periodic ascent of magma batches from depth. The occurrence of these intense seismic swarms, in the western and central-southern volcano's sectors, has repeatedly been observed in the past, and usually anticipated, during the intereruptive periods, the onset of volcanic activity (i.e., a few months or years before) [Bonaccorso et al., 1996; Vinciguerra et al., 2001; Patanè et al. 2004; Alparone et al., 2012]. It is noteworthy that we have here clearly observed, for the first time, that during an eruptive phase, this deep seismicity, closely related to magma replenishment of the deep-intermediate plumbing system, may also be rapidly followed by variation/rejuvenation of volcanic activity (after few weeks or months), such as in the case of the series of paroxysmal episodes occurred since July 2011 that followed the deep seismic swarm of early May.
 The deformation field, which we characterized thanks to GPS measurements, was mostly driven by pressurization of the intermediate-shallow magmatic plumbing system (between 8–9 km and 2 km depth). The continuous GPS network indicated, during 2009–2010, a nonuniform deformation style, with multiple sources acting at different times on different segments of a multilevel magma reservoir (sources of deformation 1–7 in Figure 12). Areal dilatation was also accompanied, from 2009 to the first half of 2010, by other geophysical and geochemical changes, including VT seismicity, SO2, and CO2 flux. Such variations were indicative for that (new and/or residing) magma in the intermediate plumbing system was progressively enriching its gas content. Such pressurization of the intermediate-shallow plumbing system had, as its final result, an enhanced degassing activity at the summit vents;
 The lava fountaining activity in 2011 was preceded by further significant geophysical and geochemical short-term variations during the second half of 2010, with a steady increase of SO2 flux since September and two main pulses of more deeply rising CO2-rich gases and/or gas-rich magmas [Aiuppa et al., 2007; Shinohara et al., 2008] refilling the upper volcano's plumbing system and ultimately leading to the reactivation of the Etna's summit craters. During the 2011 fountaining period, other significant variations took place in May–August 2011 supporting that there was a further ascent of gas-rich magma, which led to recharging of the surficial portions of the plumbing system.
 Seismic data, and in particular volcanic tremor and LP and VLP events, were used to image the shallower portions of the plumbing system at Mount Etna. Using volcanic tremor source locations, the shallower magma storage zone, feeding the 2011 episodic fountaining activity, was located below the summit area at 1–2 km asl (SMS in Figure 12). As already observed in recent studies [e.g., Patanè et al., 2008; Cannata et al., 2009b; Aiuppa et al. 2010a], LP and VLP sources are placed above this shallow magma storage zone. Gas bubbles, released from the underlying magma batch, feeding an intense gas emission from summit craters, could have triggered the 2011 eruptive events. In particular, the intermediate plumbing system imaged in this paper by the inversion of GPS data is depicted as an elongated magma reservoirs system extending from ~2.0 km bsl, beneath the volcanic edifice, down to about 8–9 km bsl, sloping slightly toward the NW, with the centers of the storage volumes located at about 8–9, 4–5, and 2.0 km bsl (Figure 12). It is noteworthy that such sources are located between the eastern border of the low-velocity zone and the western border of the high-velocity body (Figures 10 and 12) that many authors suppose to represent a path for rising and accumulating magma [e.g., Aloisi et al., 2002; Patanè et al., 2003, 2006].
 In conclusion, by making use of the multidisciplinary observations described in this work, we recognized that during 2009–2011, significant changes in volcanic activity state, mainly distinguished into five phases, were paralleled (and often anticipated) by systematic (somewhat cyclic) trends in the monitored geophysical (seismicity, deformation) and geochemical (plume composition) parameters. It is remarkable that for a medium-size eruptive cycle, we have succeeded in carefully defining the geometry and the evolution of the different portions of the plumbing system, both deep and shallow, by using ground deformation and seismic data, also constrained by plume composition and eruptive activity trends. An ongoing study, dedicated to the modeling of the magma dynamics during the lava fountain episodes, and the triggering mechanisms that regulate this peculiar cyclic activity, will the subject of a future publication.
 Seismic Data Analysis Group (S. Alparone, G. Barberi, G. Di Grazia, E. Giampiccolo, V. Maiolino, A. Mostaccio, C. Musumeci, A. Scaltrito, L. Scarfì, A. Ursino), GPS Group (M. Aloisi, V. Bruno, M. Palano, M. Rossi, D. Scandura), Cartographic Laboratory Group (S. Branca, E. De Beni, L. Lodato, C. Proietti, L. Spampinato), Video Suveillance Group (E. Biale, F. Ciancitto, E. Pecora, M. Prestifilippo), and Geochemical Group (T. Caltabiano and F. Murè) of INGV-OE are kindly acknowledged for providing VT earthquake and GPS, Volcanological, and Geochemical data, and for their technical assistance. F. Cannavò is thanked to implement the Williams and Wadge  algorithm. G. Giudice, M. Liuzzo, and G. Giuffrida of INGV-PA are gratefully acknowledged for their technical and scientific contribution to the MultigAS network. The research leading to these results has received funding from the European Research Council under the European Union's Seventh Framework Programme (FP7/2007/2013)/ERC grant agreement 1305377 (PI, A.Aiuppa). We are grateful to the Associate Editor Jean-François Lénat and the two anonymous reviewers for their useful suggestions that greatly improved the paper.