Geophysical fingerprint of the 4–11 July 2024 eruptive activity at Stromboli volcano, Italy

Paroxysmal eruptions, characterized by sudden and vigorous explosive activity, are frequent at open-vent volcanoes. Stromboli volcano, Italy, is well known for its nearly continuous degassing activity and mild explosions from the summit craters, occasionally punctuated by short-lived paroxysms. Here, we analyse multiparameter geophysical data recorded at Stromboli in early July 2024 during a period of activity that led to a paroxysmal eruption on 11 July. We use seismic, infrasound and ground deformation data, complemented by visual and unoccupied aircraft system observations, to identify key geophysical precursors to the explosive activity and to reconstruct the sequence of events. Elevated levels of volcanic tremor and very long period seismicity accompanied moderate explosive activity, lava emission and small collapses from the north crater, leading to a major explosion on 4 July 2024, at 12:16 UTC. Collapse activity from the north crater area continued throughout 7 July, while effusive activity occurred from two closely spaced vents located within Sciara del Fuoco, on the northwest flank of the volcano. On 11 July, a rapid increase in ground deformation preceded, by approximately 10 min, a paroxysmal event at 12:08 UTC; the explosion produced a 5 km high eruptive column and pyroclastic density currents along Sciara del Fuoco. Our observations suggest that the early activity in July was linked to eruption of resident magma within the shallowest parts of the volcano plumbing. This was followed by lowering of the magma level within the conduit system as confirmed by the location of newly opened effusive vents. Rapid ground deformation before the paroxysmal explosion on 11 July is consistent with the expansion of a gas-rich magma rising from depth, similar to past energetic explosive events at Stromboli. Our findings offer valuable insights into Stromboli's eruptive dynamics and other open-conduit volcanoes, highlighting the importance of integrated geophysical observations for understanding eruption dynamics forecasting, and associated risk mitigation.

1 Introduction

Stromboli is an open-conduit stratovolcano located in the Tyrrhenian Sea, off the northern coast of Sicily; its activity is characterized by continuous degassing and frequent, small-to-moderate explosions occurring every few minutes from the summit craters, the well-known Strombolian activity. However, activity at Stromboli can rapidly escalate into more energetic events, referred to as major explosions, which eject centimetre-to-metre-sized ballistic projectiles; at times, sustained explosive activity is accompanied by partial collapses of the crater rim (Gurioli et al., 2013; Di Traglia et al., 2024). Since 2019, major explosions at Stromboli have occurred with a frequency of about 4–5 events per year, ejecting pyroclastic material to heights of over a hundred metres, which can travel beyond the summit crater area and potentially affect tourist paths (Rosi et al., 2013; Gurioli et al., 2013). During periods of heightened activity, Stromboli may also experience paroxysms, i.e. highly energetic eruptions that generate eruptive columns exceeding 4 km in height, ballistics of up to 2 m in diameter and significant collapse activity from the summit crater areas (Fig. 1). Paroxysms can be accompanied by the emplacement of pyroclastic density currents (PDCs) along the Sciara del Fuoco (SdF, Fig. 1a), which can enter the sea and travel up to 2 km from the shoreline with demonstrated potential to trigger tsunamis (Rosi et al., 2006; Calvari et al., 2006; D'Auria et al., 2006; Ripepe and Lacanna, 2024). Although paroxysms are less frequent than major explosions, with an average occurrence of just one every 4 years since 2003, they are the most impactful hazard for the island of Stromboli (Rosi et al., 2013). For instance, the recent paroxysm that occurred on 3 July 2019 resulted in a fatality (Giudicepietro et al., 2020; Giordano and De Astis, 2020; Andronico et al., 2021).

Figure 1 (a) Map of the monitoring network at Stromboli, showing the locations of seismo-acoustic, seismic and infrasound sensors. The inset shows the location of Stromboli volcano in Italy (MATLAB Mapping Toolbox). (b) Details of the summit area of Stromboli (Civico et al., 2024a), corresponding to the dashed white square in (a), showing the north (N) and centre–south (CS) summit crater areas.

Unrest and eruption at Stromboli generate a broad range of geophysical signals. Nucleation and coalescence of gas bubbles into gas slugs (Sparks, 2003; Burton et al., 2007; Caricchi et al., 2024) and their ascent within the conduit generate characteristic seismic and deformation signals (Marchetti et al., 2009); gas slug bursting at the top of the magma column produces infrasound waves (Colò et al., 2010). Real-time detection and monitoring of these signals are crucial for risk mitigation at Stromboli; in the recent past, major explosions and paroxysms have been anticipated by detectable changes in geophysical signals from tens of seconds to tens of minutes before their occurrence (Giudicepietro et al., 2020; Ripepe et al., 2021a; Longo et al., 2024). Except for the 2019 eruptive activity (the most intense in recent years), Stromboli's paroxysms are typically preceded by periods of lava effusion or a general increase in surface activity that lasts for several days (Ripepe et al., 2009; Valade et al., 2016). Several studies have suggested that effusive eruptions may act as a trigger for paroxysmal explosions through a mechanism of decompression of the volcano plumbing system, evidenced by a drop in magma levels within the conduit (Aiuppa et al., 2010; Calvari et al., 2011; Ripepe et al., 2017). The most significant effusive event in terms of its volume occurred between December 2002 and July 2003 (Ripepe et al., 2017), which caused landslides, triggered a partial collapse of the SdF and culminated in a paroxysm on 5 April 2003; this was the first large-scale paroxysmal event recorded since 1985 (Calvari and Nunnari, 2023). However, it should also be noted that effusive eruptions are not necessarily followed by paroxysms. An example is the November 2014 effusive eruption, which did not lead to paroxysmal activity (Rizzo et al., 2015). At the other end of the spectrum lies the paroxysm of July 2019, for which no clear increase in activity prior to the main event was recorded. As highlighted by Laiolo et al. (2022), thermal and gas flow levels had slightly increased but remained below the alert thresholds.

Multiparameter data are crucial to understand unrest at Stromboli and to detect transitions between low-to-moderate activity and more explosive phases (Pistolesi et al., 2011; Andronico et al., 2021). A variety of models account for the occurrence and characteristics of seismic signals recorded at Stromboli and similar volcanoes (e.g. Chouet et al., 2008; Suckale et al., 2016; Ripepe et al., 2021b). Petrological analyses suggest Stromboli's conduit is stratified, with two types of magma: highly porphyritic (HP) and low-porphyritic (LP) magma (Bertagnini et al., 2003; Francalanci et al., 2004, 2005). Eruptions are believed to result from gas slugs rising through the HP magma, which acts as a viscous plug controlling their ascent and explosion (Sparks, 2003; Burton et al., 2007; Aiuppa et al., 2010; Caricchi et al., 2024). A recent model by Caricchi et al. (2024) suggests that instability of gas-rich foam layers at the base of the magma column could also trigger paroxysmal explosions.

In this study, we report on the most recent paroxysm at Stromboli, which occurred on 11 July 2024, following a month of unrest at the summit craters, as reported by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) (INGV, 2024d). We analyse the precursory geophysical activity leading up to the paroxysm based on seismic, infrasound and ground deformation data gathered by the INGV monitoring network, complemented by observations conducted with unoccupied aircraft systems (UASs) during the study period. The UAS imagery provides a valuable tool to interpret geophysical data and understand the conditions leading up to the paroxysm on 11 July, offering a high-resolution reconstruction of the eruptive events and associated morphological changes at the volcano. Unless otherwise stated, all descriptions of surface activity in this paper are from direct field observations by the authors during the study period.

2 Chronology of eruptive activity

The activity bulletins issued by INGV (see the “Data availability” section), from 24 May until the early days of July, reported an increase in surface activity at Stromboli, particularly from the north (N) crater area (Fig. 1b), characterized by continuous and intense spattering, i.e. quasi-continuous emission of pyroclastic material through sequential, small-to-moderate explosions ejecting ballistics at heights of ∼ 10–20 m above the vent (Harris and Ripepe, 2007; Giudicedipietro et al., 2020) (Fig. 2a). The average frequency of explosions fluctuated between 13 (medium) and 16 (high) events per hour with spattering occasionally leading to lava flows along the SdF. On 23 and 28 June, lava flows began, following intense spattering from the N crater, converging into a canyon-like structure created by previous PDC activity in October 2022 (Di Traglia et al., 2024). Sulfur dioxide (SO₂) and carbon dioxide (CO₂) emissions remained at average levels, as did the carbon-to-sulfur (C $/$ S) ratio (INGV, 2024a, b).

Figure 2 Timeline of the observed surface activity and key visual observations at Stromboli between late May and mid-July 2024. (a) Timeline showing the chronology of activity, which marks periods of activity characterized by lava flows (green), collapses (blue) and spattering (red). Significant events are labelled, such as intense spattering, a major explosion on 4 July, opening of new vents and the paroxysm on 11 July. (b–e) Sequence of images gathered at the times indicated by the dashed yellow lines in panel (a). From left to right: spattering activity on 3 July, a PDC event reaching into the sea on 4 July, continued lava flow on 8 July and the paroxysmal explosion on 11 July (photo in e is courtesy of Giuseppe​​​​​​​ De Rosa – OGS).

On 3 July, at 16:35 UTC, intense spattering was observed from a vent located within the N crater sector, leading to a sequence of partial collapses of the N crater rim, which also remobilized material that had been erupted in the preceding days. These collapses mostly consisted of cold material with a minor contribution of hot deposits. At 17:02 UTC, a lava flow began from the same vent, accompanied by spattering and moderate explosions (Fig. 2b). The activity continued throughout the night, with lava fronts moving down to an elevation of 550–600 m a.s.l.

On 4 July, at 12:11 UTC, a major explosion occurred from the N crater and, at 14:10 UTC, a new lava flow emerged at the base of the N crater area at ∼ 700 m a.s.l., advancing towards Bastimento and Filo di Fuoco, located along the northeast boundary of SdF. After about 1 h, a second lava flow started at an elevation of ∼ 580 m a.s.l., which reached the sea. At 16:15 UTC, another vent opened at ∼ 510 m a.s.l., producing a third lava flow accompanied by PDCs that rapidly descended the SdF into the sea (Fig. 2c). During the evening of 4 July and throughout the following night, lava flow activity continued, accompanied by occasional collapses of pyroclastic material.

Between 5–6 July, 83 landslide events were observed, effusive activity fluctuated and lava emission moved further downslope – originating from two new eruptive vents at ∼ 485 m a.s.l. (Fig. 2d). The flow formed a delta at the shoreline, and steam plumes were observed, caused by magma–seawater interaction. Explosive activity from the summit craters halted at the beginning of the effusive phase.

On 11 July, at 12:08 UTC, a paroxysmal eruption occurred from the N crater area, producing an ash plume ∼ 5 km high, which dispersed towards the southwest (Fig. 2e). Shortly after, a pyroclastic flow rapidly advanced along the SdF, which triggered a small-scale tsunami wave. The paroxysmal phase ended with a series of secondary and less-intense PDCs.

In the following hours, effusive activity ceased, and no further explosions were observed, except for a minor event on 12 July, at 08:28 UTC (Fig. 2a), which was followed by a small collapse event in the N crater area.

3 Geophysical observations

In this study, we use data recorded by the geophysical monitoring network deployed and maintained on Stromboli by INGV (Fig. 1a). The network includes two seismic broadband stations equipped with Nanometrics Trillium (0.02–40 s) three-component seismometers and Trident digital acquisition systems (IST3 and ISTR stations), as well as another four broadband stations employing two GURALP CMG-3ESPC 120 s and two GURALP CMG-40T-60S seismometers (STR6-STRE and STRC-STRG, respectively). All the data recorded are digitized at 100 Hz.

The infrasound network includes five Chaparral microphones at stations STRA, STRC, STRG, STRE and STR6 and a Geco srl sensor at STRV. Infrasound data are digitized at 100 Hz (only STRA at 50 Hz) and recorded with 24-bit resolution using Guralp Affinity and Gaia2 digitizers (https://eida.ingv.it/, last access: 25 November 2024; https://www.ov.ingv.it/index.php/ricercanew/stromboli, last access: 25 November 2024). An additional infrasound station, called PISA (Fig. 1a), was deployed on 4 July at 13:35 UTC, 35 min before the onset of the effusive activity. PISA was equipped with an IST-2018 broadband microphone, and the data were sampled at 100 Hz using a DIGOS DATA­CUBE³ 24-bit digitizer (e.g. Gheri et al., 2024).

3.1 Seismic characterization of eruptive events

Volcanic tremor is traditionally thought to reflect magma movement within the conduit (McNutt and Nishimura, 2008; Chouet et al., 1997; Ripepe and Gordeev, 1999); at Stromboli, volcanic tremor is routinely monitored by means of the root mean square (rms) of the continuous seismic signal (5 min moving window) in the 1–3 Hz frequency band (Giudicepietro et al., 2023). Figure 3a shows rms tremor amplitude values of the order of 10⁻⁶ m s⁻¹ (recorded at the IST3 site), which correspond to a tremor classified by INGV as high. A marked and short-lived increase in seismic rms tremor amplitude was observed after the major explosion at 12:11 UTC on 4 July (Fig. 3a). During this period, the signal reached unprecedented levels, peaking at 10⁻⁴ m s⁻¹ at 17:00 UTC. Short-lived increases in rms tremor amplitude values were still noted throughout 5 July, although the rms amplitude exhibited an overall decline to values of the order of 10⁻⁷ m s⁻¹, lower than those recorded at the beginning of July. In the following days (6–11 July), the rms tremor amplitude was marked by a series of short-duration peaks during lava flow activity. This behaviour changed again on 11 July, when the onset of paroxysmal activity coincided with a new increase in rms tremor amplitude (Fig. 3a). After the paroxysm, the rms tremor amplitude decreased again with only sparse and brief intervals of increased amplitudes between 12–13 July (Fig. 3a). From late 13 July onwards, the amplitude stabilized around 10⁻⁷ m s⁻¹, indicating that volcanic activity had reduced and returned to background levels. Additional details of the signals recorded during 4–11 July are shown in the Supplement (Fig. S1a).

Figure 3 (a) Seismic tremor or rms tremor amplitude (denoted RMSA) calculated every minute using a moving time window of 5 min, within the volcanic tremor frequency band of Stromboli (1–3 Hz), from 2 to 18 July. (b) Spectrogram of the E component from the IST3 seismic station for the same period.

The spectrogram in Fig. 3b shows nearly continuous energy in the 1–3 Hz range, typically associated with tremor signals at Stromboli (Ripepe et al., 1996). Energy levels in this band change throughout the pre-, syn- and post-explosive activity periods, peaking on 4 July (dark red in Fig. 3b) at 17:15 UTC, following the major explosion, which coincides with the rms peak (see also Fig. S1c, e). A pulsating phase was observed from 6–11 July, with another peak during the paroxysm. Explosive activity between 4–11 July exhibited a broader frequency range in the 0.5–15 Hz band. It is worth noting that the eruptive event on 4 July was preceded by a high-energy signal in the narrow frequency band of 0.2–0.3 Hz (Fig. 3b). We also observe that this very low frequency signal was not recorded before the paroxysm on 11 July. Finally, on 10 July at 05:09 UTC and on 11 July at 02:26 and at 15:21 UTC, high-energy signals were observed around 0.05–0.08 Hz, exhibiting a dispersive spectrum typical of teleseismic events as reported by USGS (for further information, see: https://earthquake.usgs.gov/earthquakes/search/, last access: 25 November 2024).

Figure 4 (a) Hourly rates of very long period (VLP) events from the INGV catalogue. Vertical dashed red lines indicate the major explosion and paroxysm that occurred on 4 and 11 July, respectively. (b) VLP waveform events (> 5 × 10⁻⁶ m s⁻¹) recorded on 3 July, at station STRE, normalized with respect to maximum amplitude (light grey). The red waveform represents the average of all high-amplitude waveforms. (c) Continuous waveform recorded at station STRE (east–west (EW) component) on 3 July 2024, filtered between 0.03–0.3 Hz. (d) Extract from (c) showing a sequence of VLP events recorded on 3 July over a 7 min period by the STRE station on the same horizontal component.

We have also analysed the occurrence of very long period (VLP) earthquakes that have traditionally been associated with pressure disturbances and the dynamics of gas-rich magma within fluid-filled structures (Chouet et al., 1997, 1999; Marchetti and Ripepe, 2005; Legrand and Perton, 2022), and this analysis is one of the main tools used to monitor unrest at Stromboli. VLP events at Stromboli are thought to be generated by a pre-eruptive expansion due to rising pressure in the magma column, followed by a post-eruptive contraction as pressure decreases. Final oscillations in the VLP signal may be caused by fluctuations in the conduit or edifice. (Legrand and Perton, 2022). An increase in the frequency of occurrence of these signals is typically a precursor to periods of elevated eruptive activity (Ripepe et al., 2009; Delle Donne et al., 2017). Figure 4a derived from information sourced from the INGV bulletins (INGV, 2024d) provides an overview of the rates of VLP seismicity at Stromboli between the end of May and mid-July 2024, after the 11 July paroxysm. From May until mid-June, VLP event rates remained stable, fluctuating around high values between 12 and 15 events per hour. A mean rate of ∼ 13 events per hour is defined, at Stromboli, as “normal activity” (Ripepe et al., 2009), and this suggests that an efficient degassing mechanism of the magma column is established (Ripepe et al., 2021b). A significant peak is observed around mid-June, with the number of VLP events reaching 19 events per hour on 16 June. This peak is followed by a slight decrease in event rates, although the number of events remained elevated compared to previous days. Figure 4b shows the characteristic compression–decompression cycle of VLP events at Stromboli; this waveform represents the normalized stack of all VLP events with maximum amplitude greater than 5 × 10⁻⁶ m s⁻¹ at station STRE. Figure 4c, and more specifically Fig. 4d, shows a 1 d filtered (0.03–0.3 Hz) seismic record illustrating the occurrence of VLP events as recorded at station STRE, the closest seismo-acoustic station to the eruptive area, located on the east flank of SdF at 495 m of elevation (see Fig. 1).

Before the major explosion on 4 July, we observed a clear drop in the occurrence of VLP events (Fig. 4a) from 10–15 to 7–10 events per hour. The rates of VLP events remained stable until the 11 July paroxysm, peaking again at 12 events per hour on that day. After the paroxysm, a further decrease in VLP rates was observed with hourly counts ranging from 6 to 10 events.

3.2 Infrasound location of the 11 July 2024 paroxysm

We have also analysed infrasound data recorded by the INGV acoustic monitoring network and an additional microphone installed during the period of activity (Fig. 1). The infrasonic record before 4 July shows a typical background of moderate Strombolian activity occasionally interspersed with larger explosions (see Fig. S2a). The major explosion on 4 July generated an infrasonic transient with a pressure of 5 Pa (Fig. S2b) at station STR6, ∼ 750 m from the CS crater area. Following this event, a marked decrease in acoustic energy was observed until the 11 July paroxysmal event, which produced infrasonic waves with a peak amplitude of 115 Pa at the STR6 site (at ∼ 750 m from the source; see Figs. 1a and S2c).

By analysing infrasound data, we located the source of the paroxysmal eruption on 11 July 2024. We employed the RTM-FDTD (reverse time migration finite difference time domain) method of Fee et al. (2021), which implements waveform back-projection over a grid of candidate source locations. Travel times between potential source locations and all stations in the network are calculated via FDTD modelling (Kim and Lees, 2014; Fee et al., 2017; Diaz-Moreno et al., 2019) to account for the effect of topography on the propagation of the acoustic wave field. In the RTM-FDTD method, waveforms are back-projected, and a detector function (e.g. network stack, network semblance) is evaluated for each candidate source, with the detector maximum corresponding to the most likely location. For FDTD calculations of travel times, we employed a UAS-derived digital elevation model (DEM) of the SdF and the summit crater areas (Civico et al., 2024b, c), conducted on the morning of 4 July with initial individual resolutions ranging between 20 and 50 cm per pixel. This DEM was merged with a reference elevation model (Civico et al., 2021) of the rest of the island, resampled and parsed into a 5 × 5 m grid for the purpose of FDTD modelling. For FDTD modelling, the source time function was approximated by a Blackman–Harris function with a cutoff frequency of 5 Hz (high enough to include the dominant frequency of the explosion signals (between 0.2 and 2 Hz) while still allowing for time-efficient computing), and the acoustic wave field was propagated along the discretized topography using 15 grid points per wavelength (Wang, 1996). We used a constant sound velocity of 330 m s⁻¹ (estimated from the signal propagation across the network) and a stratified atmosphere model based on density and temperature data obtained from the Reanalysis v5 (ERA5) dataset (see the “Data availability” section), produced by the European Centre for Medium-Range Weather Forecasts of the Copernicus Climate Change Service. We used data corresponding to the ERA5 grid node closest to Stromboli, at 12:00 UTC on 11 July 2024 (coordinated universal time, UTC). The inferred source location for the paroxysmal explosion on 11 July 2024, along with a record section of the infrasound waveforms used and the detector function, is shown in Fig. 5. The location identifies a source located approximately 50 m below the rim of the N crater (Fig. 5a) at an elevation of ∼ 685 m. The estimated origin time for the event is 12:08:52 UTC.

Figure 5 Infrasound location of the 11 July 2024 paroxysmal event using the RTM-FDTD method (see text for details; DEM of 14 July 2024 from Civico et al., 2024b, c). (a) Map view of the network semblance maximum around the Stromboli crater region. RTM-FDTD semblance location is indicated by a white star; (b) record section of the filtered infrasound waveforms (bandpass filter 0.01–15 Hz) used for locating the event. The offset corresponds to source–station distance; (c) normalized network detector function (i.e. maximum network semblance amplitude over time).

3.3 Tilt and eruptive events

Ground tilt at Stromboli has frequently been inferred to reflect processes like slug coalescence, slug ascent and conduit emptying (Marchetti et al., 2009; Genco and Ripepe, 2010; Bonaccorso, 1998). Over the last decade, tilt has become central to real-time monitoring and eruption early warning at Stromboli. Ripepe et al. (2021a), for example, demonstrated the scale invariance of tilt at Stromboli; that is, all explosions, regardless of their intensity, follow the same ground inflation–deflation pattern. A significant tilt was reported on 4 July (INGV, 2024c). The major explosion at 12:00 UTC was accompanied by a characteristic inflation–deflation pattern (Longo et al., 2024), followed by a pronounced deflation trend that began at 16:20 UTC and continued until 19:50 UTC (INGV, 2024c).

For the paroxysm on 11 July 2024, Fig. 6 shows the seismic-derived tilt, reconstructed from the EW horizontal component record at station STRE. The relationship between displacement and tilt sensitivity is a function of the long-period corner frequency of the seismometer used. By applying a magnification factor (e.g. Aoyama and Oshima, 2008; Genco and Ripepe, 2010; De Angelis and Bodin, 2012), which is constant around the natural period of the seismometer, we were able to convert the seismometer's output from displacement to ground tilt. Slow inflation is observed, starting ∼ 600 s before the explosion (Fig. 6b); the seismic-derived tilt sharply accelerates approximately 1 min before reaching its peak of 1.5 µrad at the onset of the explosion, followed by rapid deflation. This pattern is consistent with previous observations of tilt at Stromboli before paroxysms and major explosions (e.g. Genco and Ripepe, 2010; Ripepe et al., 2021a). We note that this tilt signal is derived from an individual seismic record of an instrument that is not likely oriented in the direction radial to the source; for this reason, we will focus on the interpretation of the deformation trend and will not use the measured tilt amplitude for modelling purposes.

Figure 6 (a) Radial tilt recorded by STRE broadband seismic station on 11 July 2024; (b) detail of tilt recorded before the 11 July paroxysm: the signal shows a marked amplitude increase starting ∼ 10 min before the onset of the explosion.

4 Discussion

In this paper, we have presented geophysical data recorded between early and mid-July 2024 at Stromboli; the period of unrest included a major explosion on 4 July, significant collapse activity in the N summit crater area, emplacement of lava flows and a paroxysmal event on 11 July. Surface activity at Stromboli intensified late in May with a marked increase in the occurrence of Strombolian explosions, the onset of effusive activity from SdF and increasing volcanic tremor. Using multiparameter observations, we reconstructed the chronology of the eruptive activity, which culminated in the paroxysmal explosion on 11 July 2024.

4.1 Eruptive activity during the first week of July 2024

In the first week of July, we observed a steady increase in volcanic tremor reaching unprecedented amplitudes on 4 July (see Figs. 3a and S1). Volcanic tremor at Stromboli has typically been linked to the coalescence of gas bubbles from layers of smaller bubbles and their ascent along the shallower conduit (McNutt and Nishimura, 2008; Chouet et al., 1997; Ripepe and Gordeev, 1999), suggesting that variations in tremor intensity are controlled by changes in gas flow within the conduit.

It has been frequently speculated that an increase in volcanic tremor reflects an increase in the volume of gas within the magma (Ripepe et al., 1996), which in turn is linked to a higher occurrence of explosions at the top of the magma column. Field observations of increasing spattering in early July (Fig. 1) support a model of increased surface activity linked to the ascent of gas-rich magma within the shallow conduit. The spattering activity observed at the start of our study period represents an intensified form of puffing. Spattering results from the quasi-continuous bursting of small gas pockets within a bubbly flow regime, which generates pyroclasts (Rosi et al., 2013). This activity typically marks the initial stages of unrest and eruption at Stromboli, where gas-rich magma is actively degassed through continuous explosive bursts (Del Bello et al., 2012). The high rates of VLP events observed during the same period further support the hypothesis of gas-rich magma migration within the shallow plumbing system. These events are traditionally linked to the rapid expansion of gas bubbles rising through the liquid melt in the shallow conduit (Chouet et al., 2003; James et al., 2006); more recently Ripepe et al. (2021b) suggested that VLP waveforms at Stromboli are generated at the top of the magma column, mainly after the onset of Strombolian explosions; they showed that the occurrence of VLP event can be linked to explosive magma decompression in the uppermost ∼ 250 m of the conduit. The recorded VLP events showed similar waveforms (Fig. 4b), suggesting a stable source mechanism and location; locations in the shallow parts of the conduit can be linked to magma accumulation at a shallow depth, close to the surface. While the number of VLP events did not show any significant variation before the major explosion on 4 July, volcanic tremor increased slowly but steadily (Fig. 3a). Coinciding with strong ground deflation after the major explosion (INGV, 2024c), volcanic tremor reached an unprecedented peak amplitude of ∼ 8 × 10⁻⁵ m s⁻¹ at ∼ 17:00 UTC, associated with the opening of a new effusive vent at ∼ 510 m elevation within SdF (Fig. 2a) and the occurrence of numerous mass-wasting events linked to collapse activity within the lower N crater area and upper section of SdF. We suggest that these signals reflect the emptying of the shallowest parts of the conduit system and the overall lowering of the magma level within the shallow volcano plumbing, reflected in the opening of new effusive vents at progressively lower elevations. The transition between explosive and effusive regimes was also marked by a clear decrease in the occurrence of VLP events (Fig. 4) and a migration of their source deeper within the conduit (as reported by the automatic seismic monitoring of INGV Osservatorio Vesuviano: http://eolo.ov.ingv.it/eolo/, last access: 25 November 2024) and as already observed during past unrest by Ripepe et al. (2015). This contrasts with the flank eruptions of 2007 and 2014 (Ripepe et al., 2009, 2015) when VLP rates remained high during effusion; in July 2024, it appears that effusion reduced the overall explosivity through progressive degassing of the shallow magma rather than recalling fresh, gas-rich magma from depth. The new effusive regime, indeed, was characterized by a substantial lack of Strombolian explosive activity at the surface between 4–11 July, as observed in the field by our research team. The quasi-continuous collapse activity, observed from 13:00 UTC on 4 July, appeared to be linked to instabilities in the crater area around newly created vents; this instability persisted in the following days, with the number of events peaking on 5 July (83 recorded occurrences recorded in a single day; INGV, 2024c). The collapse activity recorded along the N crater rim, adjacent to the SdF, resulted in significant changes to the morphology of this sector of the volcanic edifice (Fig. 7).

Figure 7 Multidirectional hillshade plots of Stromboli's crater area: (a) 4 July 2024 (Civico et al., 2025), (b) 14 July 2024 (Civico et al., 2024b, c), and (c) map of elevation difference (DEM of differences), highlighting morphological changes between 4 and 14 July 2024. Purple areas indicate material loss, whereas orange areas indicate material gain.

4.2 Eruptive activity during the second week of July 2024

The effusive regime, that began on 4 July, ended with the occurrence of the paroxysmal explosion on 11 July. The explosion generated an infrasonic pressure of 115 Pa at station STR6 with an associated VLP peak amplitude of 5.8 × 10⁻⁵ m s⁻¹ (see Fig. S3). The associated ash plume reached a height of 5 km above the vent, and pyroclastic flows moved down the SdF. After that, volcanic activity reduced its intensity, accompanied by low levels of tremor and VLP events; tremor increased again on 12 July, associated with emplacement of a small lava flow.

The eruptive crisis of July 2024, culminating into the 11 July paroxysm, is consistent with previous eruptions at Stromboli, such as those in April 2003, March 2007 and July-August 2019. The observations that we have presented in this paper can be used to inform a conceptual model of the entire sequence of processes responsible for the observed surface and eruptive activity, within the framework of previous studies (e.g. James et al., 2006; Chouet et al., 2008; Del Bello et al., 2012; Suckale et al., 2016; McKee et al., 2022).

At the more explosive end of the spectrum of Strombolian activity, major explosions and paroxysms are often explained invoking the “slug model” (James et al., 2006; Chouet et al., 2008; Del Bello et al., 2012). In this model, gas bubbles (slugs) form deeper in the magma column and gradually coalesce as they rise through the conduit due to an increase in the magma viscosity. As gas slugs ascend, they expand due to the decreasing confining pressure and eventually reach the surface. When they burst at the top of the magma column, they release gas explosively, fragmenting the magma and producing pyroclasts and feeding ash plumes of varying sizes. After the major explosion on 4 July, an effusive regime was established, characterized by lava flows, during which more degassed magma was erupted. Following the initial explosive activity driven by gas slugs, we infer that the transition to the effusive regime was controlled by depressurization of the shallow plumbing system similar to the model of Ripepe et al. (2017). The depressurization of the system caused by the initial explosive activity allowed magma to flow and reach the surface, forming lava flows, without further explosive activity. As the shallow volcanic conduit progressively emptied, it led to structural instability, causing collapses and landslides along the SdF.

According to Ripepe et al. (2017), the emptying of the conduit creates a “vacuum” effect that draws more gas-rich magma from deeper within the system. As volatile-rich magma rises and experiences lower pressures, activity can be triggered, sometimes resulting in a paroxysmal event. The dynamics of the 11 July paroxysmal explosion shared similarities across seismic, acoustic and deformation parameters with past events (Genco and Ripepe, 2010; Ripepe et al., 2021a). This consistency further validates the established models of activity at Stromboli, where the largest explosions and energetic events, such as paroxysms, are driven by the same source mechanism. The scale-invariant conduit dynamics of ground deformation demonstrate that inflation amplitude and duration scale directly with the magnitude of the explosion (Ripepe et al., 2021a). Ground deformation observed on 11 July (Fig. 6) follows the same exponential inflation pattern as seen in previous paroxysms (Ripepe et al., 2021a). This behaviour is typically explained by bubble dynamics, where the pressure on the conduit walls increases due to the rapid volumetric expansion of gas in highly vesiculated magma. As gas rises and expands, it pushes the magma column toward the surface, often leading to precursory lava emissions from the vent. Ground deformation is likely caused by a combination of increasing magma static pressure and the pressurization of degassed magma at the top of the column, driven by the exponential expansion of the gas phase. When the pressure applied by the gas-rich magma exceeds the tensile strength of the viscous magma plug, fragmentation occurs, resulting in the explosive release of gas and pyroclastic material (e.g. paroxysm). Another possible mechanism, proposed by Suckale et al. (2016) and McKee et al. (2022), suggests that the explosion is triggered by the rapid expansion and release of gas when a partial rupture occurs in the plug at the top of the magma column.

4.3 Morphological changes of the crater area caused by the explosive activity

During the study period, we also collected UAS data and compiled very high resolution repeat DEMs (0.2–0.5 m per pixel), which allowed for quantifying topographical changes via DEM differencing. The difference between DEMs on 4 July (morning; Civico et al., 2025) and 14 July (Civico et al., 2024b, c) is shown in Fig. 7c. The data processing methodology follows procedures described in Civico et al. (2021, 2024b). The most notable morphological variations were observed in the afternoon of 4 July, while the paroxysm on 11 July did not lead to significant changes.

The summit craters experienced loss of material due to the opening of two eruptive vents at approximately 700 and 500 m a.s.l. While the CS crater sector showed a roughly circular crater floor deepening of about 84 m, the N sector was affected by the complete dismantling of its northern rim and external slope, marking the deepest morphological change observed at the summit craters in the last decade, with a maximum difference in altitude of 109 m. The total volume loss recorded in the summit craters sector was estimated at 3.3 Mm³ (Civico et al., 2024b).

Unlike the summit craters, the subaerial portion of the SdF slope was affected by both accumulation and erosion processes. Here, the main loss of material (2.74 Mm³; Civico et al., 2024b) was localized along the canyon formed in October 2022 (Di Traglia et al., 2024), which widened and deepened during the July 2024 eruption. On the other hand, accumulation processes instead were mainly due to PDC and lava flow deposits within the northeastern sector of the edifice. The maximum accumulation of lava occurred at the new lava delta (maximum difference in altitude of 45 m), located in the centre of the SdF shoreline (Civico et al., 2024b).

5 Conclusion

The eruptive activity at Stromboli, starting from 4 July and culminating with a paroxysm on 11 July 2024, provides a comprehensive case study of explosive volcanism at open-conduit volcanoes and offers valuable insights into its causative processes and mechanism.

The July 2024 paroxysm was preceded by a prolonged phase of heightened activity, characterized by increased volcanic tremor and VLP events. The elevated levels of seismicity, combined with observed crater rim collapses and lava flows, suggest a progressive destabilization of the volcanic edifice. In particular, the major explosion on 4 July and the subsequent paroxysm on 11 July highlight the role of magma gas dynamics, where increased gas volumes and pressure led to significant eruptive events.

Analysis of the seismic records reveals that the volcanic tremor intensity is linked to gas-rich magma movement, reaching in this eruptive sequence unprecedented values at Stromboli. However, the variability in VLP events indicates that, while useful for monitoring overall volcanic unrest, these signals alone may not serve as reliable precursors for major explosive events. Instead, the combined analysis of different geophysical parameters, including ground deformation, proved crucial for early warning and forecasting as previously suggested by Ripepe et al. (2021a).

Ground deformation patterns, specifically the inflation–deflation cycle observed before explosions, align with previous studies, confirming that such patterns reflect the occurrence of imminent explosions regardless of their magnitude. The exponential inflation observed before the paroxysm, caused by gas expansion and the rise of slugs within the magma column, is the same as in other paroxysmal events at Stromboli.

Through UAS data, Civico et al. (2024b) were able to estimate a total volume loss of about 6.0 Mm³ involved after the gravitational mass collapses occurred on 4 and 11 July. The partial collapses generated a reshaping of the summit crater area as well as a deepening of the 2022 canyon along SdF, thus increasing flank instability.

In conclusion, our results demonstrate how geophysical, visual observation and UAS-derived topographic data offer new and valuable insights for tracking and characterizing the processes that control the onset of volcanic explosive activity at Stromboli and other similar volcanoes. We suggest that multiparameter volcano monitoring will lead to further significant advances in volcanic hazard mitigation.