Introduction

Stromboli volcano is well known for its mild, persistent explosive activity at the summit crater characterised by discrete bursts every 5–20 min (e.g. Chouet et al. 1974; Patrick et al. 2007). With a frequency of about twice a year between 1999 and 2002, major explosions occur at the summit (Bertagnini et al. 1999; Landi et al. 2008; Andronico and Pistolesi 2010), causing ejection of bombs for several hundred metres around the crater area. Paroxysms are rarer and greater explosive events, throwing blocks up to 4 m wide to 2-km distance, causing destruction of buildings and severe damage to the inhabited areas, located less than 2 km away from the summit (e.g. Rittmann 1931; Calvari et al. 2006, 2011c). The last two paroxysmal events occurred on 5 April 2003 (Calvari et al. 2006; Harris et al. 2008) and 15 March 2007 (Calvari et al. 2010; Bonaccorso et al. 2012; Andronico et al. 2013). The persistent strombolian activity is fed by a degassed, high-porphyritic (HP) magma residing in the upper part of the volcanic conduit (Métrich et al. 2001; Lautze and Houghton 2005; Landi et al. 2008), whereas paroxysms usually involve the gas-rich and crystal-poor low-porphyritic (LP) magma of deeper source (Bertagnini et al. 1999; Métrich et al. 2005; Landi et al. 2008). Effusive activity occurs every 5–15 years (Barberi et al. 1993), mainly with lava flows from flank eruptive fissures spreading along the Sciara del Fuoco (SdF), a depression that cuts the NW flank of the island (Fig. 1). Overflows from the summit crater or intracrater lava flows (i.e. lava flows confined within the summit crater depression) are quite rare, normally small (less than 500 m long) and of short duration (hours), and their occurrence indicates a very high magma level within the feeder conduit that has often preceded major flank effusive events (e.g., Calvari et al. 2005a, b; Burton et al. 2008; Di Traglia et al. 2014). Thus, when overflows happen, there is an increasing concern for the possibility of an impending flank eruption (Bertolaso et al. 2008). Between the end of the 2007 effusive eruption, in April and December 2012, the output of lava flows has increased significantly, with 28 episodes (Table 1) bringing the average to 4.8 effusive events/year. In this paper, we describe this activity and relate it to the structural context of the volcano, deeply modified by the flank effusive eruptions and lateral failure that occurred along the SdF after 2002 (Bonaccorso et al. 2003; Tommasi et al. 2005; Acocella et al. 2006; Neri et al. 2008; Neri and Lanzafame 2009; Di Traglia et al. 2013, 2014). We describe the eruptive activity using the Istituto Nazionale di Geofisica e Vulcanologia (INGV) network of fixed cameras (Fig. 1) and field surveys detailing the morphological changes of the crater area over time. We use SO2 and CO2 flux measurements representative of the bulk plume (e.g. Aiuppa et al. 2009) and continuous temperature monitoring of low temperature (t < 100 °C) fumaroles (e.g. Madonia and Fiordilino 2013) to relate the crater and eruptive activity changes with the deep magma supply. We then present an interpretative model that allows some insight on the future behaviour and that might have important implications for the hazard evaluation at this and other open-conduit volcanoes.

Fig. 1
figure 1

a Location of the Aeolian Volcanic Arc (AVA) in the southern Tyrrhenian Sea, with the easternmost Stromboli island (arrow); (b) map of Stromboli showing the Sciara del Fuoco depression outline (SDF, blue line), with the square showing the area magnified in “c”, the Vallonazzo fault system (orange line) with location of the fumarole temperature station (VLZ, orange star), and the position of the stations for the measurements of SO2 gas flux (blue dots) and of the CO2/SO2 MultiGAS stations (yellow dot); c NW sector of Stromboli showing its summit crater and the three vent areas (NEC NE Crater area, CC Central Crater area, SWC SW Crater area), the location of the INGV monitoring cameras (green: SPI, SPT and SPV at Il Pizzo; SQT400 and SQV400 at 400-m elevation on the E flank of the Sciara del Fuoco); d view of the summit crater from Il Pizzo, and distinction of the crater areas, with the position of the main vents

Full size image
Table 1 List of the effusive events (LF) and major explosions (ME) that occurred at Stromboli between 1 April 2007 and 31 December 2012
Full size table

Structural framework

The volcanism of the Aeolian Volcanic Arc (AVA, Fig. 1a) is the result of the subduction of the Ionian crust beneath the Calabrian Arc (e.g. Barberi et al. 1974). Stromboli is the easternmost island of the arc, and recent eruptive activity has occurred along a prevailing NE-SW direction (Hornig-Kjarsgaard et al. 1993; Keller et al. 1993). This is also the orientation of the Vallonazzo tectonic structure, an eruptive fissure ∼8.7 ka old and 700 m long, striking N40° E (Fig. 1b) and located between 270 and 100 m a.s.l. on the NE flank of Stromboli (Hornig-Kjarsgaard et al. 1993; Keller et al. 1993; Calvari et al. 2011a; Wijbrans et al. 2011; Francalanci et al. 2013) and which has weak fumarolic activity at its upper end. The present persistent explosive activity occurs from an open conduit elongated NE-SW (Hornig-Kjarsgaard et al. 1993; Keller et al. 1993; Acocella et al. 2006; Neri and Lanzafame 2009) and located in the uppermost portion of the SdF depression, which formed by several sector collapses during the last 13 ka (e.g. Tibaldi 2001). The conduit feeds three vent zones aligned NE-SW and located at ∼750 m a.s.l. within the summit crater (Fig. 1c): the NE Crater zone (NEC), Central Crater zone (CC) and SW Crater zone (SWC) (Fig. 1d). Several vents are within each of the crater areas. The vents have changed in number and size as a function of the magma level within the conduit (e.g. Spampinato et al. 2008), but their position (Fig. 2a) has been remarkably constant over more than a century (Anderson 1905; Washington 1917; photos Fondo Ponte, www.ct.ingv.it). The source of explosions and tremor is located at depths shallower than 200 m beneath the summit crater (Chouet et al. 1997) and shifted NW towards the SdF (Martini et al. 2007), with most of the seismic energy radiated by a small volume at ∼100-m depth beneath the SdF (Chouet et al. 2003). Thus, the uppermost part of the volcanic conduit is close to the SdF surface and is prone to be intersected and/or structurally conditioned by shallow landslides, as happened during the 2002–2003 and 2007 flank eruptions (Bonaccorso et al. 2003; Calvari et al. 2005a, b; Martini et al. 2007; Neri and Lanzafame 2009; Zanon et al. 2009; Casagli et al. 2010; Di Traglia et al. 2014).

Fig. 2
figure 2

a Helicopter photo of Stromboli summit taken from SW on 20 January 2006 and showing the summit crater area before the start of the 2007 flank effusive eruption. The floor of the summit crater area is close to the crater rim and displays several degassing vents and cinder cones. Photo INGV. b Helicopter photo of Stromboli taken from SW on 7 March 2007 showing the summit collapses and the opening of large arcuate fractures surrounding the crater depression. Photo by A. Bonaccorso

Full size image

At the onset of the 2002–2003 eruption, several fractures opened around the summit crater (Calvari et al. 2005a; Finizola et al. 2009). The spreading of an eruptive fissure triggered a hot avalanche that expanded to the sea (Calvari et al. 2005a, b; Pioli et al. 2008), and a failure along the SdF caused two large landslides and tsunami (Bonaccorso et al. 2003; Tommasi et al. 2005; Chiocci et al. 2008). Tommasi et al. (2005) considered the shallow intrusion of a magma body as the trigger of the two landslides. The eruptive fissure drained completely the lava from the summit crater, which collapsed into a NE-SW elliptical depression. The NW displacement of the SdF slope due to the landslides formed on the upper SdF a number of step-like blocks (Bonaccorso et al. 2003). A few NW-SE aligned vents opened during this eruption, which Acocella et al. (2006) attributed to dike with this orientation. The unstable eastern portion of the SdF then fragmented into large blocks and caused several small collapses (Tommasi et al. 2005; Falsaperla et al. 2008), but most of the movements ceased after the eruption, and the lava flow field had instead the effect of stabilising the SdF slope from further erosion (Chiocci et al. 2008; Falsaperla et al. 2008). No major deformation was observed until 2007 (Bonaccorso et al. 2008), when during the initial phases of this next eruption the summit crater collapsed (Calvari et al. 2010; Fig. 2b), displacing 1–3 × 106 m3 rocks (Neri and Lanzafame 2009; Casagli et al. 2010). This was a crucial turning point for the volcano, because for the first time ever, the ground deformation monitoring system, emplaced in 1992 (Bonaccorso 1998) and implemented in 2003 (Mattia et al. 2004) and comprising permanent GPS and tilt networks, detected a clear deflation of the entire volcanic edifice (Bonaccorso et al. 2008). The overall volcano deflation was strongly amplified in the summit crater area, where a massive collapse dramatically changed the morphology of the summit crater (Fig. 2a, b). Such a fast and impressive structural change had not been observed at this volcano in over a century (e.g. Anderson 1905; Washington 1917; photos Fondo Ponte, www.ct.ingv.it).

Methods

Morphology changes of Stromboli’s summit are herein investigated using images recorded by fixed cameras and sporadic photos taken during field surveys at Il Pizzo (Fig. 1c). Photos have been taken using one of the two Canon PowerShot advanced compact/bridge cameras (S3 and SX1IS), which obtains panoramic shots from multiple images (Fig. 3). The quantification of explosive activity was carried out using the INGV visual monitoring system. This comprises thermal infrared and visual cameras located at Il Pizzo (SPI, SPT, SPV, Fig. 1c), at ∼250 m from the craters, plus one visual and one thermal infrared camera (SQV400 and SQT400) at 400-m elevation on the E flank of the SdF (Fig. 1c). The latter two cameras are located at ∼800 m from the crater terrace and allow a view from NE of the NEC and of the upper eastern sector of the SdF. All images also record date in the format dd/mm/yy and time in UT format hh:mm:ss. To obtain a semi-quantitative description of the eruptive activity, we used the images recorded by the SPI, SPT and/or SPV cameras, viewing the entire crater area. We manually counted the total number of events that occurred during each day of cloud-free observation and divided by 24 h, plotting the results as an integer. On average, less than 5 % of the days were affected by clouds. When the Il Pizzo cameras were not available, the counting was carried out using the SQT400 and SQV400 cameras.

Fig. 3
figure 3

Photos collected from Il Pizzo showing a view from SE of the crater area, and the morphology changes occurred between 13 May 2007 and 29 September 2012. View is like in Fig. 1d, with orientation of the crater terrace SW-NE (left-right) and field of view ∼300 m. The numbers on the top right of each frame represent date in the format yyyymmdd. Photos by T. Ricci

Full size image

SO2 was measured by means of the FLAME network (Burton et al. 2009; Salerno et al. 2009a, b), which comprises four scanners installed in late 2004 at a distance of ∼2 km from the summit of the volcano (Fig. 1b). Each scanner comprises a S2000 spectrometer from Ocean Optics Inc. with optical resolution of 1.1 nm FWHM in the 290–380 nm UV wavelength region. An anemometer measures wind speed and direction. SO2 was measured daily and observations spanned from 6:00 to 14:30 and from 5:30 to 17:30 UT in winter and summer time, respectively. Error in SO2 flux ranged between −22 % and +36 % (Salerno et al. 2009b). CO2 flux is calculated by combining the measurements of the CO2/SO2 plume ratio from the MultiGAS network and the SO2 flux obtained from the FLAME network using a procedure described by Aiuppa et al. (2010). The MultiGAS network consists of two fully automated instruments for the determination of CO2 and SO2 concentrations in the volcanic plume, located at Il Pizzo (Fig. 1b), that are equipped with a Gascard NG spectrometer for CO2 (measurement range 0–3,000 ppm; accuracy ±2 %; resolution 0.1 ppm) and an electrochemical sensor (CiTicel, City Technology Ltd., calibration range 0–30 ppmv; repeatability 1 %; resolution 0.1 ppm) for SO2. Fumarole temperature, representative of water vapour flux variations (Madonia and Fiordilino 2013), was monitored at Vallonazzo (Fig. 1b) using an Onset Micro-Station logger equipped with a 12-bit temperature smart sensor operating in the range −40–85 °C with a resolution of 0.03 °C, installed at 2-m depth inside the fracture. Sampling period was 4 min, and an hourly value records the average of 15 measurements.

Intracrater lava flows and overflows before 2007

Between the 1975 and 1985–1986 flank eruptions (Capaldi et al. 1978; De Fino et al. 1988) and the 1990s, documentation of eruptive activity at Stromboli is rather incomplete, and therefore, only the period since 1990 is considered here. Intracrater lava flows and minor overflows between 1990 and 2002 are documented only in five cases: (1) April–May 1993, (2) August 1994, (3) August 1996, (4) sometime between May 2000 and May 2001, and (5) November 2002. The April–May 1993 lava flow followed a phase of vigorous spattering from the NEC and culminated with the emission of two lava flows ∼100 m long exiting the NEC area (Smithsonian Institution 1993). During the second half of August 1994, a small lava flow extruded from the NEC and descended a few tens of metres down the NE portion of the SdF. The August 1996 lava flow was emitted from SWC and advanced only a few metres within the crater depression (Smithsonian Institution 1997). A small lava flow was emplaced sometime between May 2000 and May 2001 within the crater (Smithsonian Institution 2001), and another small lava flow on the N outer flank of the crater depression was observed on November 2002 (Burton et al. 2008). No further effusive episodes were detected between the end of the 2002–2003 eruption and the start of the 2007 flank eruptive event (Calvari et al. 2010).

The 2007–2012 eruptive activity

Morphological changes and eruptive activity

A list of the major eruptive events characterising the 2007–2012 period is given in Table 1, and they are also represented in Fig. 4a. Figure 3 shows the morphology changes of the summit, and Fig. 4b displays the trend of explosive activity at the summit crater obtained from the INGV camera network. The graph shows four eruptive phases (E1 to E4, Fig. 4b). The first (E1) heralded the 2007 flank effusive eruption. After its end on 2 April, Stromboli’s crater remained obstructed (Fig. 3a) until 1 July 2007, when strombolian activity resumed, producing small cinder cones on the crater floor (Fig. 3a, b). From July to the end of 2007, explosive activity was mild, without major explosions or lava flows (Fig. 4a, b). It gradually increased in 2008, including three major explosive events (Table 1, Fig. 4a), and contributing significantly to the growth of the cinder cones, especially on the NEC (Fig. 3c, d). On 13 September 2008, the first effusive event within the crater depression occurred, ∼1.5 years after the end of the 2007 flank eruption (Table 1, Fig. 4a). The explosive activity decreased again between September and November, and by December 2008, ejecta and rubble filled most of the summit depression (Fig. 3d, e). The eruptive activity during early 2009 increased significantly (E2 in Fig. 4b, where uncertainty in time is due to a data gap), with several effusive events confined within the crater area and lasting a few days rather than just a few hours (Table 1, Fig. 4a). The growth of the cinder cone on the NEC right above the crater rim (Fig. 3e, f) caused frequent landslides of ejecta falling on the steep SdF slope. A major explosive event occurred on 3 May 2009 (Table 1), during which LP magma was erupted (La Felice and Landi 2011) for the first time since the 15 March 2007 paroxysm (e.g. Pistolesi et al. 2011; Bonaccorso et al. 2012).

Fig. 4
figure 4

a Time distribution of major explosions (grey vertical bars) and cumulative lava flows (red dots) occurred at Stromboli between January 2007 and December 2012 and listed in Table 1. The 2007 eruption and the 15 March 2007 paroxysm are comprised within the first yellow vertical band. b Average daily number of explosions/hour at Stromboli summit crater against time. c 95th percentile of the SO2 flux from the summit crater in Mg/day with polynomial best-fit line. d 95th percentile of the CO2 flux from the summit crater in Mg/day with polinomial best-fit line. e 95th percentile of the temperature (°C) recorded at the Vallonazzo fumarole. See text for further explanation

Full size image

The explosive activity was rather mild in 2010, but punctuated by six effusive events and five major explosions (Table 1, Fig. 4a), with the largest flow on 12 December spreading on the SW portion of the SdF (Fig. 5) and reaching an elevation as low as ∼400 m a.s.l. During 2010, a hornito started to grow on the SW rim of the crater terrace (Fig. 5). Explosive activity increased early on November (E3, Fig. 4b), followed by a pulsing decline. The January–March 2011 period was characterised by an overall declining trend that increased again in the second half of the year. Nine effusive episodes and eight major explosions characterised the year (Table 1). From April, an increasing explosive trend (E4, Fig. 4b) was followed by a large lava flow erupted between 1 and 2 August from NEC (Fig. 5) which spread along the NE part of the SdF and had the most advanced flow front at ∼500-m elevation. The summit crater depression was filled, and a large hornito (Figs. 3i, j and 5) formed on the SWC rim. During 2012, explosive activity was rather stable. Six major explosions and five effusive events occurred in this time, with four flows along the SdF (Table 1, Fig. 4a). Summarising, between January 2007 and December 2012, we observed 28 major explosions and 29 lava flows, with an average of 4.7 and 4.8 episodes per year, respectively. This displays a substantial increase when compared to the frequency of ∼2 major explosions per year recorded before 2002 (Bertagnini et al. 1999) and to 1 effusive event every 5–15 years on average over the last three centuries (Barberi et al. 1993).

Fig. 5
figure 5

Helicopter photo taken from NNW on 19 August 2011, showing the hornito growing on the crater rim, the 12 December 2010 (bounded by a yellow dotted line) and 1–2 August 2011 (bounded by a red dotted line) lava flows. Photo courtesy of Italian national Dipartimento di Protezione Civile

Full size image

Degassing regime

The different pressure-related solubility of CO2, H2O and SO2 (Lesne et al. 2011) and the consequent fluxes released by the summit crater and fumarolic field give important signals when gas-rich magma enters the feeder system from the source region (e.g. Allard et al. 1994; Aiuppa and Federico 2004; Aiuppa et al. 2009; Salerno et al. 2009b; Calvari et al. 2011b; Madonia and Fiordilino 2013). Daily fumarole temperature and CO2 and SO2 fluxes for the period January 2007-December 2012 (temperature data available since April 2007) are represented in Fig. 4c–e. We plotted the 95th percentile of the measurements, calculated in backward moving weekly windows (1-day step), in order to filter the signals for high-frequency outliers; long-term trends of CO2 and SO2 signals are evidenced by polynomial best-fit lines. The thermal signal is characterised by a very stable background value, oscillating 1–2 °C around a mean value of 27 °C, interrupted by three anomalies (T1 to T3 in Fig. 4e). The first (T1) occurred between August and October 2007 and coincided with the renewal of explosive activity from the summit vents. The second (T2) occurred from September 2009 to July 2010. Despite its larger amplitude (maximum value 42 °C), T2 occurred in a period of mild explosive activity. The last (T3, October 2010–July 2011) was the most intense (maximum temperature 65 °C) and accompanied the most turbulent period of volcanic activity (explosion peaks E3 and E4, major explosion and effusive event clusters M2 and F2–F3, Fig. 4b, a, respectively). The CO2 (Fig. 4d) and SO2 (Fig. 4c) fluxes show a generally similar pattern. Their main peaks (C1, over 10,000 Mg day−1, and S1, over 1,000 Mg day−1, respectively) were recorded at the beginning of the 2007 eruption and followed by a rapid decay, modulated by minor oscillations. The second SO2 peak (S2, over 700 Mg day−1) occurred on January 2009, at the same time as the E2 explosion peak (Fig. 4b) and major explosion cluster M1 (Fig. 4a). The apparent lack of correspondence with a preceding CO2 anomaly, as expected due to its lower solubility (Lesne et al. 2011), could be ascribed to the paucity of CO2 measurements during the entire year 2008 (Fig. 4d). A dissimilarity between the two fluxes is evidenced at the end of 2009, when a CO2 spike (C2, over 6,000 Mg day−1) was observed on November 2009, in the same period as the onset of thermal anomaly T2 (Fig. 4e), and not followed by a corresponding SO2 anomaly. The two signals coupled again few months later and showed a long-lasting (14–16 months) oscillation. This was more evident in the SO2 flux (S3, Fig. 4c) with a culmination on March 2011, which was preceded by a weak relative maximum in the CO2 flux about 2 months earlier. It is noteworthy that the most intense thermal anomaly T3 (Fig. 4e) occurred at the same time of these oscillations.

Discussion

The structural changes of Stromboli summit crater that occurred between 2002 and 2007 are unprecedented, having not been observed over the past century (e.g. Anderson 1905; Washington 1917; photos Fondo Ponte, www.ct.ingv.it). In addition, 2007 was the first time ever that volcano-wide ground deformation was measured since the installation of our monitoring network in 1992 (Bonaccorso 1998). A clear overall deflation of the entire volcanic edifice in 2007 (Bonaccorso et al. 2008) was strongly amplified in the summit crater area, where a massive collapse dramatically changed the morphology of the summit crater (Fig. 2a, b; Calvari et al. 2010). We believe that these structural changes have modified the shallow feeder conduit from a narrow, straight and efficient pipe, to a wider zone full of hot debris and scoria, partially melted (as represented by Del Moro et al. 2013 in their figure 12), with lava fingering in between. These changes would affect the eruptive style of the volcano by increasing the number of major explosive events and lava overflows from the crater rim and decreasing the possibility of flank effusive eruptions and paroxysms. To support this hypothesis, we use the trends of SO2 and CO2 from the plume, and temperature recorded from the Vallonazzo fumarole, which demonstrate that the several-fold increase in the number of lava flows and major explosions is not related to anomalously high supplies of gas-rich magma rising along the conduit. Instead, it relies on shallower processes related to the increased capacity of the uppermost conduit to accommodate a larger amount of magma not in a single, wider volume, but in an articulated fingering system developed beneath the craters.

After the end of the 2007 flank eruption in April, the drainage of magma from the uppermost conduit produced a graben-like summit collapse (Fig. 7a) and a very low magma level. The volcano entered a state of mild activity, with no lava flows or major explosions and a significant decrease in the number of explosions. This was matched by a decreased output of CO2 (Fig. 4d) and SO2 (Fig. 4c) from the summit crater. The almost stable temperature measured at the Vallonazzo fumarole (Fig. 4e), interrupted by only a small anomaly (T1), seems related to changes in the degassing partition ratio between the conduit and the fractures feeding the fumarole (De Gregorio et al. 2007; Madonia and Fiordilino 2013). It took almost 1 year to observe an increase of eruptive activity. The first major explosion occurred on 29 February 2008, followed by the first effusive event inside the crater on 12 September 2008 (Table 1, Fig. 4a). The major explosive events of end 2008–early 2009 (M1 in Fig. 4a) are associated with a degassing anomaly mainly evidenced by the SO2 peak S2 (Fig. 4c), indicating a normal, periodic recharging phase of the shallow magmatic system. The M1 events did not erupt LP magma that was instead detected during a major explosion on 3 May 2009 for the first time since 2007 (Table 1). The involvement of gas-rich LP magma and the increasing tendency of effusive episodes to spread along the outer slope of the crater in 2009 is indicative of a new phase of deep magmatic recharge, clearly remarked by the CO2 spike C2 (Fig. 4d) and by the multi-step increase in fumarole temperature (T2 and T3, Fig. 4e). This is accompanied by the filling of the crater depression and by the growth of cinder cones and hornitos on the NE and SW edges of the crater terrace (Fig. 3). This increment was mirrored by a significant increase of radiative power detected by satellite (Coppola et al. 2012). In addition, inflation of the summit during 2009 became detectable by ground-based interferometric synthetic aperture radar (GBInSAR; Di Traglia et al. 2013), and significantly increased before major explosions or effusive events, suggesting the involvement of greater magma volumes in the uppermost conduit when compared to the previous year. With effusive events spreading increasingly outside the crater terrace, also the erupted lava flow volume grew, passing from hundreds to thousands of cubic metres, and with the episodes of 12 December 2010 and 1–2 August 2011 being the largest (Fig. 5).

The intense volcanic activity here described displays some anomalous features when compared to the period preceding the last two flank eruptions: (i) no paroxysmal events accompanied the observed supply of gas-rich LP magma from the source, given that the last paroxysm occurred during the 2007 eruption (e.g. Calvari et al. 2010; Bonaccorso et al. 2012; Andronico et al. 2013); (ii) during its final degassing stage at shallow depth, an upwelled magma batch massively releases SO2 (see peaks S1–2 in Fig. 4c), but this did not happen during the 2010–2011 activity, characterised instead by a relatively quiet, long-lasting degassing; (iii) the ash and bombs erupted during the 2008 persistent explosive activity displayed mingling between LP and HP magma (D'Oriano et al. 2011; Gurioli et al. 2014), whereas before 2002, LP magma was erupted only during paroxysmal or major explosions (Bertagnini et al. 1999; Francalanci et al. 2004; Lautze and Houghton 2005). Following the above considerations, another process should be considered to explain the recent several-fold increase in eruptive activity.

We have shown that the 2002–2003 and 2007 flank eruptions have significantly modified the uppermost conduit of the volcano (e.g. Acocella et al. 2006; Calvari et al. 2005a, b, 2010; Neri and Lanzafame 2009). In fact, the eruptive vents directly connected to the feeder conduit that formed between the crater and the lowest elevation of 400 m a.s.l. along the SdF drained the magma from the upper conduit and caused several collapses of the summit area (Fig. 6). The fractures developed during the initial phases of the 2002–2003 eruption and the landslide scars occurred along the SdF, although shallow, have intersected the outer part of the central conduit (Bonaccorso et al. 2003; Calvari et al. 2005a, b; Tommasi et al. 2005; Acocella et al. 2006) that is located close to the SdF surface (Chouet et al. 2003, 2008; Martini et al. 2007). The further gradual NW displacement of the SdF triggered the opening of two parallel eruptive fissures during the initial phases of the 2007 eruption (Calvari et al. 2010) that evolved into a graben-like collapse structure (Figs. 2b and 6; Neri and Lanzafame 2009), widened by the drainage of ∼7–10 × 106 m3 of magma (Calvari et al. 2010; Marsella et al. 2012). The plug of solid rock that likely formed within the uppermost conduit produced the downward shift (∼100 m deeper from the previous ∼500 m a.s.l.) of the seismicity associated to the persistent explosive activity (Martini et al. 2007). Also, the final digestion of this plug by the rising magma was detected by seismicity that moved upwards returning to its previous depth after the end of the 2007 eruption (Martini et al. 2007). The summit collapses, comprising the nested ring fractures (Fig. 2b) and the funnel-like vertical failure of the summit zone, have modified the uppermost conduit from a narrow, straight and efficient pipe to a wider zone full of hot debris and scoria with lava fingering in between, increasing its volume and enhancing its capability to store and keep hot a greater amount of magma (Fig. 7). An upward migration of the isotherm has been recently postulated on the basis of the composition of thermally altered ejecta erupted during the 15 March 2007 paroxysm (Del Moro et al. 2013), displaying the partial melting of the tephra accumulated within the crater during the persistent strombolian activity. Petrologic studies on ash and bombs erupted in 2008 revealed for the first time mingling between degassed (HP) and gas-rich (LP) magma occurred in the uppermost 250 m of the conduit even during persistent, mild strombolian activity (D'Oriano et al. 2011; Gurioli et al. 2014). It is here worth noting that LP magma is normally erupted only during paroxysms or major explosions, not during the mild, persistent strombolian activity (e.g. Bertagnini et al. 1999; Métrich et al. 2005). Thus, a new process must have caused the mingling of LP and HP magma in the uppermost conduit during the persistent strombolian activity. We propose a model based on the observations that suggest the absence of a unique shallow magma storage where critical overpressures can generate paroxysms. Conversely, the accumulation of magma in a finger-like system composed of a network of smaller volumes might have allowed the observed mingling between HP magma resident in the shallow system and the LP magma rising from depth (D'Oriano et al. 2011; Gurioli et al. 2014), thus improving its ability to produce several major explosions rather than few paroxysms. This wider and hotter upper conduit (Fig. 7b) might have favoured lava flow inside the crater depression and along the SdF slope, whereas the presence of hot debris within the conduit would have allowed fingering of the rising lava and the growth of cinder cones and hornitos on the edges of the crater. It is worth noting that normally fingering of magma into the country rock produces thin sills that quickly cool down. The process described here instead, involving the uppermost feeder conduit, promotes heat accumulation and promote re-melting of the ejecta previously accumulated within the crater depression during the persistent strombolian activity. This is made possible by the fact that fingering of lava occurs within a volume of hot scoria and ejecta that comprise the uppermost conduit and that is bounded by country rock (Fig. 7b).

Fig. 6
figure 6

Plan view (a, b, c) and associated frontal sections (a', b', c') of the shallow feeder system of Stromboli volcano between 2002 and 2007, displaying changes following dike intrusions, landslides and graben-like collapse formation

Full size image
Fig. 7
figure 7

Aerial photos of Stromboli taken from N, displaying a the graben-like collapse formed during the 2007 eruption, and in pink the corresponding narrow, pipe-shaped shallow feeder conduit, and b the same situation as on June 2013, with the much wider shallow feeder conduit where lava (in pink) is fingering in between the debris and scoria. Eruptive vents are in red. Grey lines in (b) indicate the boundaries of the December 2010 and August 2011 lava flows. SdF Sciara del Fuoco

Full size image

The existence of a shallow magma storage articulated in a network of small volumes is well supported by degassing data. Temperature data show large anomalies (T2 and T3, Fig. 4e), indicating a shallow source for advective heat, very efficient in transferring water vapour from the magma to the surrounding country rock. This is compatible with the higher surface/volume ratio of a finger-like degassing system rather than of a simple cylindrical conduit. The large, impulsive and short-lasting C2 anomaly (Fig. 4d), associated to thermal anomaly T2 (Fig. 4e), can be interpreted as the early clue of the upwelling of a gas-rich LP magma batch. This is later accommodated in the shallower, finger-like storage, where it mingles with an older, degassed magma. During this process, it “quietly” releases its volatile content. The signature of this degassing style is given by the behaviour of CO2 and SO2 flux curves (S3 in Fig. 4c and associated CO2 signal in Fig. 4d), characterised by mild, long-lasting oscillations rather than by short-lasting, high amplitude spikes.

Concluding remarks

The data presented in this paper show a remarkable increase in effusive activity by overflows from the summit at Stromboli volcano during the more than 5 years spanning the end of the last flank eruption (April 2007) and December 2012. This frequency has increased more than 25 times, passing from one episode every 5–15 years (Barberi et al. 1993) to an average of 4.8 episodes per year between January 2007 and December 2012 (Table 1). Also, during this time period, the number of major explosions increased, passing from the 2 events per year before 2002, to an average of 4.7 episodes per year between January 2007 and December 2012. A supply of gas-rich magma from the source region was detected at the end of 2009 and perhaps also at the end of 2008, by peaks of fumarole temperature, and CO2 and SO2 signals (Fig. 4c, d). While this has typically been linked to triggering of paroxysms, no paroxysms occurred in the period of time here considered. This volcano is reckoned for being characterised by a remarkable steady supply (Calvari et al. 2011c; Francalanci et al. 2012), and we lack signs of a greater input from the source region, given that the supply detected between 2008 and 2009 is much less than that causing the 2007 eruption (Fig. 4c, d). Thus, we suggest that the increase of eruptive activity observed at Stromboli from April 2007 to December 2012 was not caused by a greater supply of gas-rich magma from the source, but instead resulted from a wider and hotter uppermost conduit, initiated by movements that occurred in the SdF after the 2002 landslide events (Acocella and Neri 2009; Falsaperla et al. 2008) and that followed the graben-like collapses that occurred during the 2007 eruption, which involved the entire summit crater zone (Neri et al. 2008; Neri and Lanzafame 2009; Zanon et al. 2009; Di Traglia et al. 2013). This is also confirmed by more recent GBInSAR results that indicate an increased magmastatic pressure within the shallow plumbing system causing its lateral expansion (Di Traglia et al. 2014). The summit of the volcano has evolved through the years to form a wider collapse depression in the uppermost conduit, elongated NE-SW. This allows a greater volume of magma to be stored (Fig. 7), which degas at quasi-equilibrium conditions (Fig. 4 and related discussion). This would keep magma hot for longer, thus enhancing the possibility for the HP magma here contained to mingle with the LP magma fed from depth even during the persistent strombolian activity (D'Oriano et al. 2011; Gurioli et al. 2014), rather than forming obstructions below which the LP magma can accumulate and increase its pressure up to the Vulcanian style disruption, typically observed during the paroxysmal events (e.g. Calvari et al. 2006, 2010). This process will probably eventually end with restoration of the previous central conduit path. The present geometric configuration of the shallow portion of the feeder conduit could also favour the release of the magma pressure along conduit walls, decreasing the probability of dike intrusions across the upper conduit walls and favouring an increase of activity at craters. Undetected summit morphology changes might occur at other open-conduit volcanoes; thus, it is extremely important to continuously monitor the summit activity in order to interpret their eruptive behaviour.