Magma convection, where low-viscosity, gas-rich magma ascends, degasses, and crystallizes before sinking down the same conduit in either annular or side-by-side flows, has been proposed for active basaltic volcanoes, where excess gas fluxes relative to erupted lava volume can be observed. Experimental studies show that convection is produced by buoyant ascending gas-rich magma and descending degassed magmas following density difference contrast, while geophysical studies point to the endogenous growth of active volcanoes through magma accumulation in plutons. However, many aspects of the convection process remain unclear, in particular, the depth to which magma ascends before overturning. Models have been proposed where overturn occurs near the surface and also at depths greater than 2 km from the top of the magma-filled conduit. The long-term monitoring of volcanic gas compositions may reveal new insights into the convection process, as each gas has a unique solubility-pressure profile. We report measurements of SO2 and HCl gas fluxes from Etna between October 2007 and May 2011, in which an ~90% collapse in halogen flux was observed together with an effusive eruption. This observation indicates that the halogen fluxes, during quiescent periods on Etna, require both magma supply to the shallowest levels and a period of residence. The lava effusion has the effect of reducing the shallow residence time, drastically reducing the halogen flux. These results provide a new interpretative framework for the degassing process and gas composition monitoring to explain subtle variations in magma supply and residence times in basaltic volcanism.

Density-driven magma convection was proposed by Francis et al. (1993) to explain observed excessive degassing relative to lava eruption rates at several persistently active volcanoes (e.g., Shinohara, 2008; Edmonds et al., 2022, and references therein). The fluid mechanics controlling magma convection were investigated by Kazahaya et al. (1994). They found that convection consists of the concentric flow of upwelling bubble-rich, low-density magma in the core of the cylinder and downwelling degassed magma within the conduit. They showed that the gas flux is a function of the density difference between gas-rich and degassed magmas, conduit diameter, and magma viscosity. The magma convection model was invoked by Allard (1997) to explain excessive degassing at Etna (Sicily, Italy) in the absence of shallow intrusions. Experimental data on convective circulation (Beckett et al., 2011; Cardoso and Woods, 1999; Molina et al., 2012) supported the degassing model of Kazahaya et al. (1994). Ferlito et al. (2014) proposed a process of fluid flushing in a basaltic open-conduit system, invoking the permeable flow of an exsolved volatile phase to explain the persistent gas plume emission and excess volatile budget at Etna.

Several questions regarding how magma convection produces persistent volcanic degassing remain open: What is the mechanism of gas loss from ascending magma? Is there any chemical interaction between the down- and upwelling magmas? Does the magma convection reach the top of the magma column or stop at greater depths? The gas loss mechanism was investigated by Burton et al. (2007) and Becket et al. (2014) at Stromboli, who found that the relatively low vesicularity of eruption products from Strombolian activity could be explained by the decoupling of gas from magma and flow through the permeable magma filling the conduit from pressures of ~45 MPa, ~1.6 km below the top of the magma column. Witham (2011) suggested a potential volatile recharge of the downwelling degassed magma from the volatile-rich ascending magmas to explain broad H2O-CO2 concentration variations observed in melt inclusions (Spilliaert et al., 2006). This hypothesis was revised by Métrich et al. (2010), who proposed a CO2 flushing-enrichment process to explain melt inclusion variations. Following on that work, Stevenson and Blake (1998), Métrich et al. (2010), and Aiuppa et al. (2010) proposed that magma convection at Stromboli takes place up to 2 km below the top of the magma column and then ceases, with the shallow conduit filled with stagnant magmas.

The question of the extent to which convection may take place, and how the mechanism works, at active basaltic volcanoes is the subject of this article, which we addressed by exploring the SO2 and HCl emission data from Etna before, during, and after the 2008–2009 effusive eruption.

The 2008–2009 eruption lasted 421 days, starting on the morning of 13 May 2008, when an eruptive fissure opened east of Etna’s summit craters. The fissure opening was accompanied by an intense swarm of volcano-tectonic events, together with an increase in seismic tremors and a rapid deformation of the summit area (Aloisi et al., 2009; Cannata et al., 2009; Di Grazia et al., 2009). Effusive activity from the eruptive fissure continued without interruption until 6 July 2009 and consisted of ephemeral vents and the overlapping of secondary flows forming a compound lava field (Bonaccorso et al., 2011; James et al., 2012).

In this study, we examined the SO2 and HCl fluxes during the period 2007–2011 to gain insights into magma dynamics in Etna and the roles of magma convection and shallow magma storage in halogen gas emissions. The results, which highlight the high solubility of HCl in melts, are indications of the dynamics of shallow magma storage, ascent, and eruption.

We performed measurements of SO2 and HCl column amounts using a Bruker OPAG-22 open-path Fourier transform infrared (OP-FTIR) spectrometer, equipped with a ZnSe beam splitter and LN2-cooled mercury-cadmium-telluride (MCT) detector. Absorption spectra were collected between 1000 and 6000 cm–1 with a resolution of 0.5 cm–1, using the Sun as a source of infrared radiation (e.g., Francis et al., 1998; see the Supplemental Material1). We collected approximately three measurements per week from sites located 16 km from the volcano summit at an altitude of ~1000 m. Column amounts of SO2 and HCl were retrieved from each spectrum using the forward model-fitting techniques described in Francis et al. (1998). The composition of the bulk plume was measured over ~20–30 min by collecting 100–200 spectra. This allowed accurate measurements of the SO2/HCl molar ratio by plotting SO2 amounts against HCl amounts for each spectrum and performing linear regressions. Typical errors on the gradients (ratios) of the linear regressions were 5%. Retrieved molar ratios were converted to mass ratios and multiplied by SO2 flux to determine the flux of HCl.

Bulk SO2 flux from the volcanic plume of Etna was measured by the permanent network of scanning ultraviolet (UV) spectrometers called FLAME (Fig. 1; e.g., Salerno et al., 2009b, 2018). Each station measures UV spectra in scans across the sky during daylight hours, intersecting the plume at a mean distance of ~14 km from the summit craters and acquiring a complete scan in ~5 min. Raw spectra are reduced on site using the DOAS (differential optical absorption spectroscopy) technique and a modeled clear-sky spectrum (e.g., Burton et al., 2009; Salerno et al., 2009a). Uncertainty in SO2 fluxes ranges between –22% and +36% (Salerno et al., 2009b).

From October 2007 to May 2008, the SO2/HCl ratio trend by weight (Fig. 2A) decreased from 5.0 to 3.2, followed by a relatively stable variation until the eruption onset. During the first days of eruption, the summit SO2/HCl trend increased to a value of ~4.9; in contrast, data collected while Strombolian activity was taking place at the eruptive fissure coupled with lava effusion displayed a value of ~1.5. As explosive activity declined, the SO2/HCl value started a rapid increasing trend, reaching its maximum value of ~10 on 26 March 2009. Figure 2B shows the daily SO2 flux measurements from the bulk volcanic plume of Etna between October 2007 and May 2011; the mean flux was 2400 Mg/d (1δ = 1500 Mg/d), encompassed between a minimum value and maximum value of 200 and 20,000 t/d, respectively. The emission rates displayed a steady and moderate trend marked by three waxing-waning stages recorded during phases of eruptive unrest. The first stage occurred between October 2007 and December 2008, and the second occurred from January 2009 to February 2010. From March 2010, a downward phase started, which leveled off the SO2 emission rate at mean values of ~2000 Mg/d.

The calculated HCl flux determined by combining the ratio in weight with bulk SO2 flux measurements is shown Figure 2B. The HCl flux displayed a moderate increase (with peak of ~1300 Mg/d) connected to the main explosive event, which occurred during the opening of the effusive fracture in the initial explosive phase of the 2008–2009 eruption. After that, the HCl flux decreased from ~600 to ~300 Mg/d.

We employed UV spectroscopy to determine the SO2 fluxes and OP-FTIR spectroscopy measurements to determine the gas composition in terms of SO2/HCl ratio. These data can be compared with volatile concentrations measured in Etna melt inclusions to provide insights into degassing processes. During the growth of olivine crystals, small amounts of melt can be trapped, recording the melt composition at the pressure of crystal formation (e.g., Métrich and Mandeville, 2010; Edmonds and Wallace, 2017; Moretti et al., 2018). Spilliaert et al. (2006) argued that little or no sulfur (S) is exsolved at pressures greater than ~140 MPa at Mt. Etna. On the contrary, Cl concentrations increase in residual melts as crystallization occurs (e.g., 20 MPa; Moretti et al., 2018), showing a limited degassing during ascent (Spilliaert et al., 2006). This is supported by decompression experiments on water-rich rhyolitic melts (Gardner et al., 2006) and by gas phases driving the Etna 2001 lava fountain sequence (La Spina et al., 2015). Moreover, Cl does not reach a saturated concentration in the melt (Carroll and Holloway, 1994) but is, instead, partitioned between fluid and melt (Shinohara, 2009), favoring shallow degassing, where the exsolved gas mass is greatest. Petrologic studies on melt inclusions and whole-rock samples at Etna indicate that ³90% of S and up to 55% of Cl are lost during the degassing process during rapid magma ascent (Spillieart et al., 2006).

The dissolved concentrations of S and Cl in melt inclusions allowed us to calculate the bulk degassing composition produced by magma during decompression and differentiation. Bulk degassing entails that the ratio of volatile components is identical to that originally in the melt phase corrected for the relative quenched amount in crystallized bulk rocks. Combining the original content of S = ~0.35 wt% and Cl = 0.23 wt% in recent products (Moretti et al., 2018) with the residual volatile concentrations (S ~0.0145 wt% and Cl ~0.07 wt%), the bulk degassing process will produce an integrated gas sample with SO2/HCl weight ratio of ~4.2. Therefore, the observation of a lower or higher weight ratio implies a non-bulk-degassing process. In Figure 2, we show the daily SO2 flux and HCl flux from October 2007 to May 2011. There is a coupling between the two records before the onset of effusive activity. Then, gradually, the two geochemical signals decouple, with HCl gradually decreasing in terms of SO2 from November 2008. In order to acquire insights into the decoupling and associated degassing process, we considered the cumulative amounts of SO2 and HCl, scaling the upper limits of the y-HCl axes to reproduce the bulk degassing weight ratio (~4.2; Fig. 3). We note that, between October 2007 and May 2011, Etna emitted 3.4 × 106 Mg of SO2 and 0.65 × 106 Mg of HCl. If a closed-degassing system is assumed, in which all volatile components escape from the melt with continuous re-equilibration, then the weight ratio in the gas phase might be identical to that present in the melt phase. Therefore, the cumulative amount of SO2 and HCl records should match. Instead, as shown in Figure 3, after following a coincident temporal pattern, the two series decoupled from November 2008. This evidence suggests that in the quiescent, noneruptive stage, when SO2-HCl series are coupled, magma at shallow depth may promote HCl outgassing. Nevertheless, in the eruptive stage featured by lava outflow, this degassing process might be potentially perturbed due to a reduction in the residence time of magma in the shallow plumbing system.

The potential emplacement of a magma batch at shallow level was corroborated by ground deformation data. Aloisi et al. (2009) argued that the magma intrusion started from the central conduit system at a depth ~2000 m below the summit area. Moreover, further evidence of a shallow reservoir is provided by the chemistry of the lava erupted in the initial stage of the event, which displays features similar to those erupted during summit crater activity (principally plagioclase-bearing trachybasalts; Corsaro and Miraglia, 2009). Nevertheless, Corsaro and Miraglia (2014) argued that a change in magma composition occurred over the eruption because of the mixing between evolved and primitive magma. Moreover, La Spina et al. (2010), in their summit survey during (July 2008) and after the end of effusive activity (August 2009), reported an increase in the SO2/HCl ratio for the central crater, highlighting a clear depletion of HCl emission. Considering an amount of 90% of S lost during the degassing magma process (and original amount of S = 0.35 wt%; e.g., Moretti et al., 2018) and assuming a density of 2800 kg m–3 for erupted lavas and a crystal content of 30% (typical values for basalt), consequently, each 1 Mg of measured SO2 would require the degassing of ~81 m3 of vesiculated lavas. Therefore, during the 421 d of effusive eruption, the total SO2 released was ~0.84 × 106 Mg, which implies the degassing of ~68 × 106 m3 of lavas. This volume agrees with the estimated lava emission volume of 70 ± 20 × 106 m3 reported by Corsaro and Miraglia (2014). Following the volume of magma degassed, we obtained an averaged magma degassing rate of ~2 m3 s–1, which matches the effusion rate of 2.2 m3 s–1 reported by Bonaccorso et al. (2011). Evidence of a shallow magma batch, as the reservoir for the 2008–2009 eruptive activity, also arises from the balance between degassed and erupted lava calculated on the basis of the measured SO2 flux (Allard, 1997). Eventually, the lava flow drainage of magma gradually led to a break in the magma convective overturn within the shallow reservoir, as suggested by the decoupling of S-Cl emissions from November 2008.

Finally, considering the period of ~6 months between the onset of the effusive eruption (13 May 2008) and the onset of decoupling in gas fluxes (November 2008), together with the averaged lava effusion rate of 2.2 m3 s–1, we estimated a magma volume of ~34 × 106 m3. This volume represents the amount of magma involved in the convection process before being perturbed by the effusive activity.

Another consideration emerges when observing the coupled gas trend in relation to the volcanic activity. The coupled periods might suggest a more efficient extraction of the gas phase from magma, as a consequence of increased permeability due to circulation in interconnected networks of bubbles. This gas phase–melt partition probably promoted an increase of melt viscosity, in turn increasing pressure in the plumbing system with the progressive growth and coalescence of gas bubbles in the shallow magma, which eventually promoted the explosive activity. Nevertheless, this observation needs to be further explored by integrating gas geochemical and petrological data in a numerical model.

In this work, we presented bulk SO2 and HCl fluxes recorded between October 2007 and May 2011, and during the long-stage effusive 2008–2009 eruption at Mt. Etna, the mechanism of which was closely connected with the central conduit of Etna. By integrating remotely sensed bulk fluxes, fundamental insights into the shallow degassing system at Etna have been achieved. These S and Cl data allow us to hypothesize that significant magma residence time is required to sustain an efficient halogen degassing process in the quiescent stage. Synchronous changes, both in terms of time and magnitude, in SO2 and HCl fluxes were observed throughout the study period, with the exception of the period after November 2008. At that point, halogen degassing dropped by 90% and gradually regained its efficiency beginning in December 2010. Our results provide a novel interpretative framework for the degassing process, in which subtle variations in HCl with respect to SO2 might reflect fluctuations in the magma supply and residence times in the shallow plumbing system of Etna. Remote bulk SO2 and HCl observations by UV and FTIR spectroscopy highlight the essential role of monitoring at active volcanoes and for the decryption of eruptive mechanisms.

1Supplemental Material. Description of volcanic gas plume measurements by the OP-FTIR solar occultation method. Please visit to access the supplemental material, and contact with any questions.

This study benefited from fruitful discussions with Istituto Nazionale di Geofisica e Vulcanologia (INGV) colleagues. We acknowledge F. Mure for the technical management of the FLAME network. The research leading to these results received funding from the INGV Departmental Strategic Project IMPACT. The research was funded by the Dipartimento of Protezione Civile Italiana. Thorough reviews by Ryunosuke Kazahaya and Robin Campion are gratefully acknowledged.

Gold Open Access: This paper is published under the terms of the CC-BY license.