Geophysical Research Letters

Integrated petrochemical and geophysical data reveals thermal distribution of the feeding conduits at Stromboli volcano, Italy



[1] Samples of scoriae erupted at Stromboli volcano during its persistent strombolian activity were collected between 2005 and 2008. Chemical and mineralogical compositions were obtained on products erupted from the three main crater sectors (SW, Central and NE). Small chemical variations indicate a different degree of evolution coupled with small difference of magma temperature <10°C. Analysis of the acoustic data for the same time period as the scoria sampling, indicates that puffing (a persistent overpressurized bubble degassing) was, on average, mainly observed at the central craters and at times moved to the NE sector. The cross-check of the two independent data sets allowed us to assess correlation between composition of products and puffing activity at vents. The hotter products are always erupted from the vents where puffing occurs indicating that slightly higher temperature can be the expression of an enhanced two-phase bubble flow dynamics.

1. Introduction

[2] Stromboli, one of the iconic volcanoes on Earth, exhibits an extraordinary steady state eruptive behaviour dominated by persistent, moderate magmatic explosions and degassing. Persistent activity consists of two contrasting types: i) ordinary strombolian explosions, i.e., shortly sustained jets of gas and incandescent lava fragments that occur at intervals of tens of minutes from different vents and ii) continuous degassing activity which includes passive degassing from scattered fumarolic areas and open vents, and active degassing (“puffing” activity) from one or more vents. Puffing consists of discrete overpressurized puffs of gas/vapour that occur at a few seconds interval radiating excess pressure in the atmosphere and eventually ejecting small amount of incandescent lava fragments [Ripepe et al., 1996]. The amount of gas released to the atmosphere as continuous gas streaming and gas puffing accounts for about 2/3 of the total output [Harris and Ripepe, 2007]. Studies on explosive dynamics and composition of products of the present-day Stromboli activity have been numerous over the past two decades (see Bertagnini et al. [2008] and Ripepe et al. [2008] for a review). However, no attempts have ever been made to investigate possible compositional differences of products ejected at the same time from different vents and examine possible relationships between composition and eruptive dynamics.

[3] In order to assess possible relationships between degassing modality and magma chemistry, samples of scoriae ejected by different vents were collected between 2005 and 2008. Systematic analyses of the glassy matrices and minerals allowed estimation of the magmatic physical parameters, such as crystal content and equilibrium temperature of mineral phases. During the time of scoria sampling, we tracked from the back-azimuth of infrasonic waves the location of puffing across the craters area. This has allowed us to identify the sector of more intense active degassing activity. We finally discuss the data in terms of magma dynamics of the shallow conduits feeding the ordinary strombolian activity.

2. Sampling

[4] Activity of Stromboli occurs from vents situated within an elliptical area located at 750 m a.s.l., with the major axis of 300 m stretching in the NE–SW direction (Figure 1). Conventionally vents are spatially grouped into the so-called SW, Central and NE craters or crater sectors. Before March 2007 the southern part of the crater terrace was bounded by a nearly continuous crater rim, several meters higher than the active craters (Figure 1a). Depending on the energy of the explosion, scoriae were occasionally thrown out the crater terrace eventually falling on the crater rim. Only in that case the sample was collected immediately after its emission. An alternative sampling strategy was that of collecting the lapilli and scoriae that blanketed the southern rim of the crater terrace, since the products erupted from the SE, Central and NE craters fell on the corresponding sector of the crater terrace rim (Figure 1a). Scoriae and lapilli were sharp-edged and showed iridescent surfaces, suggesting that they were erupted within few days before the sampling. During the 2005 and 2006 we collected seven sets of samples representative of the products erupted from distinct craters in the same day or during the same interval of few days. In March 2007, during an effusive eruption, the crater terrace collapsed and the southern crater rim was largely destroyed (Figure 1b). After the 2007 eruption, only five scoriae, erupted from SW and Central craters were collected immediately after their emission.

Figure 1.

View of the Crater Terrace of Stromboli from Pizzo Sopra la Fossa (a) in June 2005 and (b) after the effusive eruption in March 2007.

3. Chemistry, Mineralogy and Glass Composition

[5] All the crystal-rich products erupted from Stromboli in the past decades have a virtually constant basaltic-shoshonitic composition associated with a poorly variable crystal content (45–55 wt%). Crystals consist of zoned plagioclase An60-90, zoned clinopyroxene Fs6-15 and olivine Fo70-75 set in a shoshonitic matrices (K2O 3.4–4.4, CaO 7.3–8 wt%). The rims of euhedral crystals in textural equilibrium with the groundmass, show a significant unvarying composition (plagioclase An65-70, clinopyroxene Fs12-14, olivine Fo71-74). The nearly constant characteristics of the products indicate that the physical and chemical conditions, which control the solid-liquid equilibrium of the magma remain virtually stable over time. The crystal-rich magma feeding the normal strombolian activity resides in a shallow, degassed reservoir, located within the volcano edifice [Métrich et al., 2001; Francalanci et al., 2004; Landi et al., 2004, 2009].

[6] The products analyzed in this work have features identical to those of the crystal-rich products erupted in the past decades. Despite being erupted during a 3-year-long time period and from the three different sectors of the crater terrace they have constant bulk chemistry within the analytical error. However, the glassy matrices highlight compositional differences that, even if small, appear to be systematically associated with the different crater sectors (Figure 2a): while the SW glassy matrices (K2O = 4.1–4.3, CaO 7.3–7.6 wt%) are systematically more evolved than those of the Central crater (K2O = 4.0–4.1, CaO 7.6–7.8 wt%), the NE glassy matrices show a somehow larger scattering (K2O = 4.0–4.2, CaO 7.4–7.6 wt%). Because of the constant bulk rock and mineral composition, the variations observed in the glassy matrices must be related to different crystal contents. Using least-square mass balance calculations the most evolved glass can be derived from the less evolved one by crystallization of 6–8 wt% crystals of plagioclase (70–77 wt%), clinopyroxene (17–21 wt%) and olivine (6–10 wt%), in proportion close to that of the modal values commonly observed in the crystal-rich products of Stromboli. It implies a maximum variation of crystal content of about 6–8 wt%, eventually linked to different volatile content dissolved in the liquid and/or different temperature of the erupted magma.

Figure 2.

(a) Average composition of the glassy matrices of the products erupted from SW, Central and NE craters between June 2005 and December 2008 plotted in K2O versus CaO diagram (20 quoted analyses on each sample. Analyses performed with JEOL-JXA-8200 electron microprobe -WD/ED combined microanalyzer- at INGV, Rome). (b) Relationship between glass composition of the ejecta and density of puffing activity from the NE crater sector at the time of the sampling.

4. Thermal Effect on Crystal Content

[7] In order to establish what is the dominant phenomena inducing the different crystal contents (volatile content or temperature), the rims of plagioclase and olivine in textural equilibrium with the groundmass were analyzed. It is well known that plagioclase composition strongly depends on the water content dissolved in the liquid phase. Thus, a systematic variation of water content in the groundmass would imply a systematic variation of the plagioclase rims in equilibrium with the groundmass. Indeed, plagioclase in samples erupted from SW, Central and NE craters does not show variations related to the emission crater. Plagioglase rims, several tens to less then 10 μm thick, have always a nearly constant composition (An65-70) associated with degassed conditions [Métrich et al., 2001; Landi et al., 2004].

[8] Conversely, olivine-liquid can be used as magma thermometer [Putirka, 2008]. While rims <20 μm of olivine have similar composition (Fo71-73) in the SE and NE products, they are slightly richer in magnesium (up to Fo75) in the Central crater products. Olivine-liquid thermometers, calculated in the different samples, yield temperatures from 1100°C to 1180°C, depending on the applied model [Putirka, 2008]. However, whichever is the employed thermodynamics model, Central products with less evolved glassy matrix and olivine Fo75 result hotter than SW products of ∼10°C.

[9] We conclude that the slight compositional variations observed between the products erupted during the normal strombolian activity are mainly controlled by small changes in the magma temperature, which is related to the different crater sectors. Hotter magmas are always emitted from Central craters, products from the SW craters show colder compositions, while NE products are more dispersal (Figure 2a).

5. Mapping the Gas Outflow Distribution

[10] The precise position of the degassing area within a magma feeding system is still difficult to locate. However, when the gas reaches the shallow portion of the magma column with a pressure (Pgas) larger than the external one (Patm + Pmagma) it tends to explode. The breaking of the shallow magma film and/or the mass outflow of gas/fragments mixture spray generate pressure perturbations which can be detected as acoustic waves and located by infrasonic antennas (array) [Ripepe et al., 2007]. Infrasound has allowed us to identify at Stromboli a persistent degassing activity driven by small (<0.5 m in diameter) gas bubbles bursting every 1–2 sec (puffing) [Ripepe et al., 1996]. This degassing activity related to a persistent overpressurized bubbly flow regime can be more easily mapped by using infrasonic array techniques, and it is thought to be responsible for ∼45% of the total gas budget of Stromboli volcano [Harris and Ripepe, 2007].

[11] At Stromboli, puffing activity is generally stable at a single crater for time periods ranging from hours to months, but can shift from one crater to another (Figure 3). The position of puffing activity has been suggested to reflect the position of the main gas bubbly flow channel within the shallow feeding system, which should coincide with the section of the feeding conduit more fluidized by the gas [Ripepe et al., 2009].

Figure 3.

(a) Azimuthal coverage of summit craters from the 5-elements small aperture infrasound array (PSF: Pizzo Sopra la Fossa). (b) Map distribution of 8.8 millions of infrasonic location during the 2005–2009 time period reveals that puffing is mainly located at the Central crater. (c) Multi-channel semblance of infrasonic signals across the array as a function of back-azimuth and time. Vertical white line indicate the sampling episodes, black lines indicate the backazimuth relative to the NE and Central craters, respectively.

[12] Since 2003 puffing has mostly occurred at the Central crater and seldom moved to the NE crater [Ripepe et al., 2007], while was rare at the SW crater (Figure 3). The puffing location seems then to reflect composition of scoria samples, with hotter products emitted from the Central crater, and colder products from the SW craters.

[13] This relation between puffing location and chemical composition is more complex at the NE crater. During the week before the scoria sampling, the density of puffing location at the NE crater sector was calculated as the ratio between the number of infrasonic puffing locations at this crater and the total puffing activity at the whole crater terrace. A large variation of puffing density (from <4% to ∼70%) is observed. Colder samples (higher K2O and lower CaO in the glassy matrix) are associated with low values of location density (<5%), indicating that during the week preceding the sampling, puffing was located in a different sector of the crater terrace. Hotter samples (lower K2O and higher CaO) are associated with high values of location density (>60%), indicating that scoria were collected when puffing was actually located at the NE crater (Figure 2b). Chemical composition of scoria from the NE crater is then changing according to the location density of puffing activity indicating a strong link with the active degassing.

6. Insights Into the Shallow Feeding System

[14] Small chemical variations of the products erupted from the different crater sectors point to a different degree of evolution coupled with small difference of magma temperature <10°C. Despite being close to the analytical errors, these differences are systematically coupled with the location of active degassing in the different vents (Figures 2 and 3). In particular the hotter products are always erupted from the vents where puffing occurs indicating that slightly higher temperature can be the expression of an enhanced two-phase liquid/gas flow dynamics. A slightly higher content of water dissolved in the melt, undetectable by petrological data, might be induced by overpressure along the degassing path and could participate to hinder the crystallization of the magma.

[15] These results contribute to improve the conceptual model of the shallow magmatic system of Stromboli volcano. The NE-SW elongated feeding conduit is sealed at the top by tens of meters thick rock slab, virtually representing the so-called crater terrace. Explosive vents located in the crater area are connected to the main conduit by small magmatic and gas pathways whose number and position can vary through time. Vents located in the central part of the crater terrace tend to host puffing, whereas lateral vents preferentially host regular strombolian explosions. Whenever puffing is observed to move to lateral vents we do also notice an increase of the magmatic temperature (Figure 4).

Figure 4.

Cartoon of the shallow feeding system of Stromboli. The unsteady and vortical bubble flux drives efficient convection, responsible for the general compositional homogeneity within the magma column. The dynamics of the bubble column remains virtually stable over time, leading to stable degassing location associated with slightly hotter magmas. The upper part of the conduit possibly works as a buried lava lake. Vertical/horizontal dimension ratio not in scale.

[16] The structure and degassing dynamics of Stromboli's shallow plumbing system shows interesting similarities with experimental results from bubble columns, where gas rises through a conduit filled with liquid. Here, for low gas regimes and elongated column geometries (height ≫ width), the flow is unsteady and vortical, leading to convective cells and to an oscillatory bubble plume efficiently mixing the liquid column [Diaz et al., 2006]. The efficient convection within the feeding system is responsible for a general homogeneity of the magma conduit in agreement with the homogeneous chemistry of magma erupted at Stromboli. Conduit convection at Stromboli has also been invoked to explain both the mass unbalance between gas (SO2) emitted and magma erupted [Allard et al., 1994] and changes in magma vesicularity following the explosive intensity [Colò et al., 2010]. Moreover, convection might explain the different temperature of products erupted from the different sectors of the crater terrace. In agreement with thermal investigations of lava lake dynamics, where the hotter sectors corresponds to spots of magma up-rise and crust disruption [Calkins et al., 2008], larger temperature is expected where gas is preferentially rising through the liquid. Finally, convection might also explain why the location of puffing activity is more often stable in the Central crater [Ripepe et al., 2007] whereas the migration to the NE crater sector could be consistent with a drift of the convection cell centroide towards NE [Ripepe et al., 2009].

[17] We believe that this scenario closely agrees with the high stability of the explosive centres of Stromboli volcano and with the observed geochemical variation of erupted material linked to the migration of the degassing centres. In fact, location and major explosive features of the three active sectors (NE, Central, SW) remains basically stable through time within the crater terrace [e.g., Marchetti and Ripepe, 2005] also after the collapse of the crater terrace as occurred during the 2002/2003 and the 2007 lava eruptions.


[18] This work has been supported by INGV-DPC programs and DevNET grant of Department of Civil Protection. The paper has been improved by the critical reading of Alessandro Aiuppa.