Introduction

Pyroclastic density currents (PDCs) are multiphase flows made up of gas and fragmented solid particles and represent one of the deadliest volcanic manifestations1,2. Generally, PDCs are associated with explosive activity in volcanoes erupting from intermediate to felsic magma compositions, where the formation of huge pyroclastic flows is mainly caused by the collapse of the eruptive column during Plinian and sub-Plinian eruptions (e.g. Mt. Vesuvius, Italy3,4; Mt. Pelée, Lesser Antilles5), or from lateral blasts induced by the decompression of the volcanic edifice following a sector collapse (e.g. Mt. St. Helens, USA6). However, the most common type of PDC is produced during the gravitational collapse of unstable lava-domes and lava flows during effusive eruptions with andesitic to dacitic compositions7,8.

Additionally, volcanoes fed by basaltic to intermediate magma compositions have recently shown the ability to generate destructive PDCs and have garnered considerable interest within the scientific community9,10,11,12,13. Indeed, even these kinds of volcanoes have proved able to generate high-particle concentrated granular flows. These flows derive mainly from the gravity-induced collapse of proximal pyroclastic deposits accumulated on volcanic slopes close to the angle of repose of dry granular volcanic material (31–33°)14, though the agglutination of spatters would exhibit a higher angle of repose compared to dry granular deposits. In such cases, PDCs are also termed pyroclastic avalanches15,16. Such processes have been documented at various basaltic to andesitic volcanoes, including Stromboli (Italy)9, Arenal (Costa Rica)10 and Llaima (Chile)11. Some of these granular flows had great mobility, reaching unusually long runout distances despite their high static friction. This was clearly observed for the PDC formed during the 2018 eruption at Fuego Volcano, Guatemala, where 0.01-0.03 km3 of perched pyroclastic deposit on the upper portion of the volcano fed a granular flow which traveled for more than 12 km, causing fatalities and the devastation of the small villages located nearby12,13.

Many efforts have been made to reveal the mechanism of generation and propagation of block and ash flows deriving from lava-dome collapses. These studies have focused on the slope stability analysis of lava-domes17, the correlation between their extrusion rate and emplacement of PDCs18, and the role of fluidization processes in reducing the friction forces to explain the high mobility of some block and ash flows8,19,20,21,22,23. By contrast, few studies have focused on PDCs resulting from the collapse of pyroclastic deposits in basaltic volcanoes9,24,25, which are still not well constrained. Identifying the factors leading to the formation and propagation of PDCs is crucial to re-evaluate the hazard in those volcanoes which are generally perceived of as being quiet, since they are characterized by less energetic eruptive activity.

Mt. Etna (Italy), one of the most active basaltic volcanoes worldwide, whose monitoring is managed by the Istituto Nazionale di Geofisica e Vulcanologia - Osservatorio Etneo (INGV–OE), effectively proves an ideal volcano to unravel the predisposing conditions triggering these PDCs. Several PDCs have been observed during the last 20 years, for which different mechanisms have been invoked25. Most of them involved the South East Crater (SEC; Fig. 1), the youngest and most active among the summit craters of Mt. Etna (North East Crater: NEC; Voragine: VOR; Bocca Nuova: BN; Fig. 2). Formed in mid-May 1971, the SEC has since grown into a complex of overlapping composite cones25,26,27,28,29, due to frequent Strombolian activity and almost 270 paroxysmal episodes mostly clustered in sequences30,31. Paroxysms are typically dominated by lava fountains, as well as accompanied by lava flows, and form eruption plumes that have been previously observed to rise up to 14 km above sea level32 (a.s.l.). This intense activity contributed to the rapid growth of the SEC, with the opening of eruptive vents and fissures which repeatedly affect deep portions of the growing cone, consisting of a continuous accumulation and alternation of lava and tephra layers33. The largest PDCs were produced on 16 November 200634,35,36, and on 11 February 201437, but minor events also occurred during the 2011-2013 paroxysmal sequences38. Moreover, it is noteworthy that, at the beginning of 2020, a warning of potential slope instability was issued because the summit portion appeared perforated by several pits and showed crests and crags of strongly altered volcanic rock39. This intrinsic weakness of the cone was subjected to further stress after December 2020, when a high number of paroxysms, some of which considered large-scale lava fountains30, occurred, in particular over several months during the two main sequences in February–April 2021 and May–October 2021 (which reached more than 60 episodes40), plus 6 more episodes in 2022–2023. This frequent activity caused a notable increase in episodic collapses of portions of the upper slopes of the SEC and generated avalanches, making them one of the most hazardous volcanic phenomena in the summit area, given Mt. Etna is not only one of the most studied volcanoes by volcanologists, but also among the most frequented by tourists. The first episode, on 13–14 December 2020, led to the collapse of the unstable southern rim and generated three minor avalanches, the most energetic of which advanced nearly 2 km. Others occurred throughout the 2020–2022 activity period41, with the largest and most hazardous collapse event occurring during the 10 February 2022 paroxysmal eruption42.

Fig. 1: Images of the 10 February 2022 paroxysmal eruptions and PDC.
figure 1

a Photo of the 10 February 2022 lava fountain with the 12 km a.s.l. eruption column (photo by B. Behncke). b Photo of a minor PDC flowing on the flank of SEC (photo by B. Behncke). c–f Frames taken from INGV-OE thermal cameras showing the main PDC emplaced between 21:26 and 21:27 UTC.

Full size image
Fig. 2: PDC deposit observed from satellite images and in the field.
figure 2

a Skysat scene acquired on 14 February 2022, where the area covered by the PDC deposit (brown layer) is visible. The positions of the 2002–2003 craters, the wooden shelter, and the summit craters (North East Crater: NEC; Voragine: VOR; Bocca Nuova: BN) are also reported. The inset (b) shows the RGB composite (Bands 12, 8, 4) of the Sentinel 2 image acquired on 11 February 2022. c, d Photos showing the PDC deposit seen from a proximal and distal point, respectively; the height of the vehicles in the (c) is about 3 meters (photos by D. Andronico). e Details of the burnt wooden shelter (photo by F. Ciancitto).

Full size image

Here, we first provide a multidisciplinary analysis of the pre-eruptive conditions of the source which generated the granular flow of tephra along the flanks of the SEC, rarely described on basaltic volcanoes. To do this, we combine field work, video-recordings from volcanic surveillance and private videos, remote sensing data, and petrological studies on deposits and products of the 10 February 2022 PDC. We then evaluate the propagation dynamics supported by a numerical modeling that has already been calibrated43, thus providing robust constraints for its interpretation. Our findings have relevant implications for the hazard assessment related to PDC inundation in the summit area of Mt. Etna, as well as for the volcanic surveillance of other volcanoes exhibiting comparable eruptive behavior which could potentially be destabilized by producing similar PDCs.

Results and discussion

The 10 February 2022 paroxysm, collapse, and PDC

During the early morning of 10 February, following several days of ash emissions and sporadic Strombolian explosions, explosive activity gradually increased at the SEC. On the same day, at about 14:20 UTC, a slow lava effusion from the summit area started towards the south. Strombolian activity evolved into powerful lava fountaining after 20:30 UTC (Fig. 1a), and after 20:35 UTC, the first PDC descended the south flank of the cone down to its base, followed by minor PDCs directed to the SE (Fig. 1b). An eruption column formed above the SEC reaching an estimated height of 12 km a.s.l.42, before being blown by the wind to the NW. Heavy fallout of large bombs often obscured the entire southern side of the SEC, so that the exact succession of events that followed is not entirely clear, but can be reconstructed based on the stratigraphic relationships between the products of different eruptive phases and processes. At about 21:20 UTC, an eruptive fissure opened on the S flank of the SEC cone, and a lava flow started directed towards the W side of the upper scoria cone of the 2002-03 craters. At 21:26 UTC, the southern flank of the SEC cone collapsed, triggering an avalanche of debris with meter-sized blocks in the south and southeast direction (Fig. 1c–g and Supplementary Movies 1-4). A PDC formed and enveloped the uppermost of the two 2002–2003 craters (Fig. 2a, b), descending SE to about 2730 m elevation and reaching up to 1.4 km distance in about 1 min. The debris resulting from the avalanche buried part of the snow cover and the lava flow emplaced before the collapse event (Fig. 2c, d). Afterward, lava continued to be emitted. Videos recorded by surveillance cameras show a wide glowing cloud above the PDC, which was being pushed towards the W-NW of the PDC by the intense winds blowing on the SEC, and that did not cover the deposit. When the cloud lifted, it left an incandescent deposit on the ground, which burned a small shelter of the Sicilian Civil Protection (Fig. 2e) and partially carbonized wooden poles along the tourist paths in the 2002–2003 craters area (Supplementary Fig. 1). The collapse formed a deep scar in the cone, and the PDC left a fan-shaped deposit down from the base of the SEC, filling low-lying areas and wrapping ridges of the pre-existing morphology. Part of its western portion was eventually covered by a lava flow during the later stages of the paroxysm. Explosions and pulsating lava fountains continued from the lower section of the collapse scar for about 15 min. After 21:50 UTC, lava fountaining diminished to Strombolian activity and, after 23:00 UTC, only some lava flows were still weakly fed.

Satellite data mapping and morphological changes

Satellite imagery proved valuable in retrieving timely information on the extension of the deposit and the associated thickness distribution. We obtained the first cloud free Skysat acquisition on 14 February at 11:30 UTC (Fig. 2a). The reddish deposits in Fig. 2 include both the PDC, the cone collapse, and the lava flow. The detachment area measures about 50 ± 5 × 103 m2, the thickest deposits 270 ± 30 × 103 m2, while the thinnest deposits area is about 450 ± 50 × 103 m2. The hottest parts of the deposit covered by the lava flow are also evident from the Sentinel 2 image acquired on 11 February 2022 (Fig. 2b).

Pléiades and WorldView-1 images were acquired before (in July 2021) and after (in July 2022) the eruption and used to derive Digital Surface Models (DSMs), which were compared to obtain the thickness distribution of the volcanic deposits. The positive thicknesses cover an area of 3.78 km2, with a median thickness of 4.33 m, and reach a maximum value of 40 m in the SSE part downstream of the SEC. These deposits include seven eruptive events31,44 that emitted 12.7 ± 3.8 × 106 m3 of lava and increased the SEC cone volume by 7.8 ± 0.9 × 106 m3. The negative thicknesses are related to the mass failure associated with the collapse events and cover an area of 0.13 km2, with a median value of −17.75 m and a minimum value of −60 m. They are concentrated around the SEC, in the south-east and south-west directions, reaching to the south the altitude of 3075 m a.s.l. (Fig. 3a). The collapse occurring on 10 February 2022 is in correspondence with the area of greatest growth around the SEC, which had reached a maximum thickness of 110 meters, as measured by the DSM difference between August 2020 and July 2021 (Fig. 3b). The total volume of products between July 2021 and July 2022 was calculated by integrating the negative thickness distribution over the collapse area and the positive thickness distribution in the deposits accumulation area, while the uncertainty was quantified as the product of the area and the standard deviation of terrain residuals outside of the deposits. We found a volume of 1.6 ± 0.4 × 106 m3 and 3.2 ± 0.4 × 106 m3, respectively for the collapsed material from the cone and the thickest deposit, which includes the blocks from the cone and the lava flow. The thinnest deposit is not evident from the DSM difference, being less than 1 m thick. Assuming a thickness between 5 and 70 cm (i.e., the minimum and maximum thicknesses measured in the area affected by the thinnest deposit), we estimated a volume between 22.5 and 316.3 × 103 m3 for this thinnest deposit.

Fig. 3: Topographical variations of the Mt. Etna summit computed by DSM difference.
figure 3

a Thickness distribution (in meters) of the deposits obtained from DSM difference between June 2022 and July 202144. The legends show both negative values related to the collapse of different portions of the SEC, and positive values related to the growth of the cone, emplaced lava flows, and PDCs (green marks the change from negative to positive values). b Thickness distribution (in meters) of the deposits obtained from DSM difference between July 2021 and August 2020.

Full size image

Stratigraphy and grain-size distribution of the PDC units

The deposit of the 10 February 2022 PDC consists of a reddish-brown, fan-shaped body with an elongated lobe on its eastern side, composed of fine ash to lapilli in the lower part and fine lapilli to blocks up to tens of meters across in the upper part. The deposit extends from the base of the scar on the southern flank of the SEC up to a maximum distance of 1.4 km. It rests on a rather homogenous topography with an average inclination of ∼12–14° and is formed both by lava flows and by fallout beds of scoriaceous lapilli emplaced during the previous paroxysmal episodes (Fig. 4). At the margins of the flow deposit, few lobes were present, while levees were not so clearly observed.

Fig. 4: Locations of the stratigraphic sections of the PDC deposit.
figure 4

a A total of 16 stratigraphic sections were analyzed over the three recognized sectors (red dots: FS1; yellow dots: FS2; blue dots: FS3). The stars are the location of chosen sections associated with (b) FS1, (c) FS2, (d) FS3. The location of section 17, which has been qualitatively analyzed, is also reported (gray dot). Photos by B. Behncke.

Full size image

Sixteen stratigraphic sections were analyzed along the eastern lobe (Fig. 5 and Supplementary Table 1), where the total thickness ranges from 0.05 m to values greater than 1 m from distal to more proximal portions, respectively. Thicknesses of less than 0.5 m have been measured within the upper scoria cone of the 2002–2003 craters. No section was analyzed on very proximal areas close to the SEC, since the deposit was buried by lava flows after the PDC emplacement. Based on field observations of the PDC outcrops, we distinguished three flow sectors (FS1 to FS3) along the eastern lobe of the deposit (Fig. 4). FS1 is the most proximal sector to the source among those investigated and includes two sections at the top and base of the upper scoria cone of the 2002–2003 craters, respectively, and three sections northward at ∼50–100 m distance, where thicknesses from 0.7 m to greater than 1.1 m were measured. FS2 is the intermediate sector between FS1 and FS3, located eastward of the 2002–2003 craters, where four sections are characterized by a thickness ranging from 0.25 m to 0.65 m. FS3 extends from ∼100 m below FS2 until the more distal outcrops of the deposit; in seven analyzed sections, it varies from a few centimeters to a maximum thickness of 0.40 m.

Fig. 5: Stratigraphic log of the chosen sections and GSD diagrams.
figure 5

a Image showing representative stratigraphic log of the analyzed sections. b GSD diagrams for selected sections in each FS. The Unit A for FS1 has been selected from another section because it is not fully exposed. c The diagram reports the Md (φ) vs σ (φ) variations for all analyzed samples. d Frequency of reddish and gray clasts vs scoriaceous and juvenile particles.

Full size image

Integrating the observations of the stratigraphic sections with the analyses of the grain-size distributions (GSD; Supplementary Table 2) enabled us to recognize and define at least 4 main units (Fig. 5; Supplementary Fig. 2). The basal layer, Unit A, was not fully visible in some sections of FS1 owing to the greater thickness of the deposit in the more proximal areas which prevented us from observing the bottom. In FS2 and FS3, it is made up of very poorly-sorted massive reddish ash and fine lapilli, with thicknesses ranging from 5 cm to 30 cm. Unit A generally has a GSD peaked at 2φ (0.250 mm), but occasionally shows a bimodal distribution, with a fine mode at 2φ and a coarse mode at −2φ (4 mm). Unit B reaches a thickness of 20–40 cm and is a poorly-sorted layer of reddish coarse ash mixed with lapilli. In FS1 and FS2, Unit B is formed of coarser lapilli, with a peak of the coarse-grained population between −3φ (8 mm) and −5φ (32 mm). In the most distal sections of FS3 Unit B is usually characterized by 2 modes, with the finer peaked at 2φ and the other in the range from −2φ to −3φ. Unit C is 40–60 cm thick and has been observed in FS1 and to a limited extent in FS2. It has a coarser-grained matrix than the previous Units A and B, with the largest clasts concentrated in the −6φ class (64 mm); isolated blocks up to ~50 cm in size are also scattered within the layer. The uppermost Unit D is visible on the surface of the PDC deposit. It typically consists of a clast-supported, ∼10–20 cm-thick layer, containing mostly lithic fragments ranging from coarse lapilli to meter-sized blocks. No GSD analysis has been performed on this unit since the field survey was carried out a few months after the PDC emplacement, and the layer may have undergone potential remobilization.

A further section has been qualitatively analyzed on a thick portion of the deposit (>140 cm) on the west part (Fig. 5). It was not possible to fully expose Unit A during the excavation, while Unit B reaches more than 40 cm. In this outcrop, the stratigraphy also included Unit C (∼60 cm), and the thickest Unit D (40 cm).

The grain-size analysis reveals that most of the samples (∼54%) are coarse-grained, with more than 50% of the clasts > −1φ (2 mm), mostly associated with Units B and C in FS1 and FS2. Fine-grained populations with clasts < −1φ are mainly observed in Unit A and in few samples from Unit B in FS3, while none are observed in Unit C. In particular, the fraction <2φ constitutes 20-45% of the total in FS3, decreasing in FS1 and FS2 up to 2% in Unit C. These features are consistent with the calculation of the median grain-size Md (φ), which shows some differences depending on the unit and location of the analyzed sample (Fig. 5). The largest estimated Md (φ), representative of the finer clast population, is observed for Unit A in FS3, ranging from 0.82 to −1.16 with a minimum of -1.44 in FS1 where the bottom was visible. Unit B displays Md (φ) from −0.02 to −5.02, with notable variability in FS1 and FS2. Unit C, observed only in FS1 and FS2, shows the lowest Md (φ) values from −6.00 to −6.07. In general, almost all samples are very poorly sorted, with values of sorting σ (φ) = 2.04 -3.45, except for two samples in FS1 with σ (φ) = 1.88–1.89.

Textural analysis of the components

The components forming the coarse part > −1φ (2 mm) of the 10 February 2022 PDC deposits (A, B and C units) comprise: i) dark gray, very irregular fragments, ranging in size from centimeters to decimeters, rarely with smoothed and variably abraded surfaces; ii) coarse scoriaceous lapilli and bombs; iii) gray, dense lava blocks and chips with blocky and angular external morphologies; iv) reddish–orange, variably altered and abraded scoriaceous clasts.

Components of the ash matrix in the range from 1 to 0 φ (0.5–1 mm) comprise two main types of fragments distinguishable by their external morphologies and colors (Fig. 6a, b), and internal texture. The first clasts population represents 10–30 vol. % of the ash fraction and includes shiny black, mostly fluidal glassy particles. These are variably vesicular (ranging from 40 to 70 % of vesicularity) and show a variable content of microphenocrysts and microlites of plagioclase, clinopyroxene, olivine, and Fe-Ti oxides (Fig. 6c). We interpret these particles as representative of the juvenile fraction of the deposit, related to the lava fountain activity that occurred before the PDC. On the whole, we observe no significant variation in the content of this clast population from the base to the top of the deposit and with the distance from the vent (20–30 vol.%). The black, glassy particles are less abundant (10 vol.%) only in the case of the proximal sample at the base of the deposit. The second particle population constitutes 70–90 vol. % of the ash fraction; this includes gray to reddish/bright-orange particles with blocky to spongy shapes. The groundmass of these particles ranges from microlite-rich (with very minor glass) to holocrystalline (microcrystalline). Microlites in the groundmass comprise plagioclase, clinopyroxene, olivine, and Fe-Ti oxides (Fig. 6c). Particles are often altered on the external rims and inside vesicles, with devitrified glass and the formation of secondary minerals in cryptocrystalline masses. Microphenocrysts of the same mineral species occur. Olivine exhibits well-developed sub-solidus transformation into magnetite and Fo-rich portions (symplectite exsolution; Fig. 6d, e), and Fe-Ti oxides show trellis-type lamellae textures (Fig. 6f). Trellis-type lamellae in oxides have been experimentally reproduced at temperatures >700 °C and oxidizing conditions45, conversely, olivine sub-solidus modification occurs at T >1000 °C (see D’Oriano et al.46 and references therein). At Etna, similar textural features were observed in the deposits of the 11 February 2014 PDC37. Therefore, we interpret the second particle population (reddish-gray clasts) as representative of the lithic fraction of the deposit, deriving from the scoria cone that collapsed and triggered the 10 February 2022 PDC.

Fig. 6: BSE images of the fine components of the PDC deposits collected by the SEM.
figure 6

a Reddish/bright-orange clasts (0.5–1 mm). b Black, glassy clasts (0.5−1 mm). c Microlite-rich and microcrystalline clasts, indicated by the red arrows, and juvenile clasts, indicated by the white arrows. d, e Symplectite exsolution in olivine crystals found in clasts with microcrystalline groundmass. f Trellis-type lamellae textures in Fe-Ti oxides.

Full size image

Numerical simulation of the PDC

Numerical simulations of the PDC propagation have been performed using the open-source code IMEX-SfloW2D16,47 to retrieve a set of initial conditions and rheological parameters compatible with the observed scenario. In this regard, we adopted the Voellmy-Salm rheology, which depends on the dry-friction coefficient µ and viscous-turbulent friction coefficient ξ, to model the motion of a mixture of gas and solid particles. In addition, an initial velocity u0 was prescribed to the PDC. These unknown parameters (µ, ξ, u0) are the target of a calibration procedure, for which a workflow has been developed by Zuccarello et al.43 in order to run an ensemble of simulations (from hundreds to thousands of runs) and find the combinations of the parameters minimizing the misfit between the simulated and the observed PDC deposit43. The result of the calibration of the input parameters was used to develop a probabilistic hazard map associated with the pyroclastic avalanche inundation on the summit area of Mt. Etna43. In this work, we take into account the best-fit simulation, identified using the Jaccard Index (J) which is based on the overlap between the simulated and observed PDC deposit areas48, corresponding to µ = 0.33, ξ = 741 m/s², and u0 = 80 m/s (Fig. 7) and with a total propagation time of 50 s, which is close to the time taken by the avalanche to propagate (~ 1 minute). The reproduced thicknesses were then compared to the available measurements on the deposit, providing a good match between simulated and ground data (Fig. 8).

Fig. 7: Results of the numerical simulation of the 10 February 2022 PDC.
figure 7

a Simulated PDC computed by using as input parameters µ = 0.20, ξ = 300 m/s2 and u0 = 0 m/s (J = 0.375). b Simulated PDC computed by using as input parameters µ = 0.33, ξ = 741 m/s2 and u0 = 0 m/s (J = 0.271). c Simulated PDC (best fit) computed by using as input parameters µ = 0.33, ξ = 741 m/s2 and u0 = 80 m/s (J = 0.691). The total time for all simulations has been fixed at 240 s: the simulated deposit in (a) was reproduced within the total simulation time, while the best fit simulation (c) required only 50 s to fit the actual deposit, and then its motion ended, consistent with the observation from the thermal camera. The black line represents the area covered by the observed PDC deposit.

Full size image
Fig. 8: Comparison between simulated and measured thickness of the PDC deposit.
figure 8

The continuous black line indicates a perfect match between simulated and observed values (1:1), while dashed and dotted lines indicate differences of factors 2:1(1:2) and 5:1(1:5), respectively. For some samples in FS1 the measured thickness is underestimated because it was not possible to reach the bottom, while one sample from FS2 shows a much greater simulated thickness than the measured one on the deposit (> 5:1), likely due to the local effect of the topography on the deposition of the actual avalanche. The error bar represents 5% of the measured thicknesses on the deposit.

Full size image

Textural modification of ultra-proximal deposits

Our integrated approach provides relevant insights on the pre-collapse conditions of the source which generated the 10 February 2022 PDC at Etna and on its flow propagation dynamics. Componentry analysis on the class size particles 0.5–1 mm indicates that within the pyroclastic deposit there are no significant variations of the components in the different stratigraphic units, where the juvenile fraction is significantly lower (<30% in volume) compared to the lithic class. The low percentage of juvenile fraction is interpreted to be related to the fresh magma erupted during the lava fountaining activity and recycled in the avalanche. Textural observation of ash matrix particles by SEM shows that the PDC is mainly composed of the cone-forming deposits before the collapse. In fact, the near-complete crystallization of the groundmass and the subsolidus transformation of olivine and oxides in the analyzed clasts have often been reported in the literature and are interpreted as related to the alteration (recycling) of pyroclastic products at high residual temperatures (>700 °C) and oxidizing conditions that induced the post-depositional modification of primary textures of pyroclasts in ultra-proximal environments49,50. Alternatively, these have also been explained by processes occurring within the magmatic column and are mainly related to rapid changes in oxygen fugacity conditions51.

We propose that these high thermal conditions may have been enhanced by abundant deposition and high accumulation rates of coarse tephra that fell on the middle-upper flanks of the SEC during the very frequent lava fountain episodes occurring before 10 February 2022, particularly between December 2020 and October 2021. A further source of heat comes from the circulation of volcanic fluids through the fracture-fields that generally develop at the summit of the SEC, also contributing to the hydrothermal alteration of the rocks constituting the scoria cone34,52. The burned wooden shelter, located at 2798 m and only partially covered by incandescent volcanic products, is evidence of a very hot and still glowing granular flow during the PDC emplacement (minimum temperature >450 °C)53.

Cascade trigger mechanisms for PDCs at Etna

The above interpretations are also consistent with the topographic monitoring based on satellite data. The DSM difference between August 2020 and July 2021 shows an accumulation of at least ∼110 m of pyroclastic deposits mostly located on the southern flank, which are associated to the rapid deposition of emitted volcanic products during the paroxysmal activity at SEC41. The very short time intervals between the lava fountaining episodes in February-April and May-June 2021 did indeed promote the rapid growth of the south flank of the SEC, preventing the heat dispersion of the buried pyroclastic deposit accumulated at the beginning of the sequences. The increase of the slope angle due to the fast accumulation of ultra-proximal products, combined with the high residual temperatures, favored the instability of the southern flank of the SEC. In addition, the hydrothermal alteration caused by cracks and fractures filled by volcanic fluids further weakened the scoria cone, finally resulting in the reduction of the rock mechanical properties54,55.

These conditions were preparatory for the gravitational collapse which generated the 10 February 2022 PDC. The opening of an eruptive vent feeding a lava flow on the south flank of the SEC may have led to the total failure of the ultra-proximal pyroclastic deposits, generating the collapse. This mechanism differs somewhat from the 2014 PDC event, where the magma below the SEC edifice moved toward its weakened southern flank until it was pushed, then causing the collapse37. Compared to the other minor events occurring in 2011–2013 and 2020–2022, the PDC produced by the 2022 collapse was characterized by a higher mobility. The granular flow was in fact able to overcome the upper scoria cone of the 2002–2003 craters, despite the flow propagating over a low slope (∼12–14°, much lower than the natural angle of repose of basaltic material14) and in absence of a confined valley. Indeed, a confined valley might favor longer travel distances12,56, as observed for Stromboli9 and for Fuego, where the topography predominantly controlled the propagation dynamics and distribution of the relatively “cold” granular flows (<200 °C)13.

The distance reached by the avalanche, the deposit thickness, and the emplacement time can only be reproduced by numerical simulations by imposing an initial velocity of several tens of meters per second43. This is consistent with observations from thermal cameras which show an impulsive motion of the hot avalanche, with an estimated average velocity of ∼30 m/s. The fast initial motion of the avalanche was then followed by a sudden slowdown due to the topographic effect of the 2002–2003 scoria cones that induced the deposition of the most concentrated part of the flow (Unit B and C; Supplementary Fig. 1) between FS1 and FS2. Then, the granular flow became less concentrated along FS3, where Unit B is observed only locally.

The range of rheological parameters of the Voellmy-Salm model allowed us to explore different flow regimes, from frictional (i.e., dominated by the Coulomb friction term) to turbulent. With the best-fit parameters (µ = 0.33, ξ = 741 m/s², and u0 = 80 m/s) the turbulent term is largely prevailing. The best-fit also indicates that frictional weakening (i.e., the reduction of the Coulomb friction) is not the dominating mechanism to explain the flow mobility. Some processes, not included in our model, might have contributed to the enhanced mobility of the avalanche. During the transport, in fact, an increased abundance of fine particles is observed both vertically, due to segregation processes57, and along the distance from the vent, as shown by the GSD pattern from proximal to distal sectors. The increase of finest particles may lead to the increase of the runout distance of the PDC58 through the build-up of the pore pressure and the consequent self-fluidization19,20,21,22. Vaporization of snow59 could have further enhanced the propagation of the avalanche. Such processes would mostly act to reduce the Coulomb friction term and, potentially, the required initial velocity. These effects could be investigated in the future using different rheological models. However, as explained above, the Coulomb friction term is not the dominant one in the present case, so that an initial velocity (consistent with impulsive motion shown by the thermal cameras) would in any case be required.

Such an initial impulse might have been driven by the contribution of an explosive process or by the rapid expansion of overheated fluids trapped within the cone43. While video-recordings enable excluding a lateral blast, even seismic and infrasound recordings cannot provide any clue on the possible occurrence of an explosion in the generation and propagation of the granular flow, since these signals were saturated by the lava fountaining activity60. In addition, as stated above, the analysis of the components of the PDC deposit pointed out that the granular flow derived mostly from the ultra-proximal pyroclastic deposits forming the SEC scoria cone before the collapse, leading us to exclude the contribution of a magmatic fragmentation which would have produced a greater amount of fresh juvenile clasts in the emplaced PDC deposit. Therefore, we hypothesize that the sudden unloading induced by the gravitational failure may have led to the expansion of the hydrothermal fluids overheated by the magma feeding the effusive activity before the collapse, thus providing the initial rapid acceleration to the PDC.

Implications for hazard assessment

This study highlights the importance of assessing the source conditions that may generate PDCs from the collapse of hot and altered portions of the edifice and remobilization of the deposits on basaltic volcanoes. These phenomena can pose a potential hazard due to their unexpected occurrence, sudden nature, and rapid kinetics of emplacement even on limited shallow slopes. The 10 February 2022 PDC at Mt. Etna raises the issue that relatively small PDCs may also occur at basaltic volcanoes, commonly characterized by small-moderate explosive eruptions (e.g. Bonadonna et al.61 and references therein), and usually more approachable in the summit areas. The combination of the eruptive activity with the morpho-structural features of the SEC was crucial in generating a highly-mobile PDC, without the possibility of issuing a prompt alert. It is worth noting that the three PDCs emplaced during the 13–14 December 2020 paroxysmal event were generated from the same area of the 10 February 2022 PDC. Recent studies have highlighted that hydrothermally weakened hidden rocks composing lava-domes, which have undergone previous collapses, are more susceptible to future collapses due to the load of new extruded material on the same site54. Such conditions might also be effective for a basaltic scoria cone and should be seen as precursory signals for careful risk management. In fact, the south flank of the SEC is expected to undergo further collapses in the future if the residual scar, left after the 10 February 2022 PDC, is gradually filled by an accumulation of pyroclastic products. At the time of writing (April 2025), a total of 10 paroxysmal episodes have occurred since 2023 (4 at the SEC and 6 at the near VOR crater), which already induced a partial filling of the detachment niche, with accumulation of unwelded tephra up to several meters (Fig. 9). These dynamics could re-establish pre-eruptive conditions at the SEC similar to those present before the 10 February 2022 eruption. In this regard, we suggest that periodic topographic monitoring through satellite data should be implemented to identify changes in morphology and the growth of relatively young and structurally simpler volcanic areas that are particularly prone to collapse due to excess slope or deformations caused by eruptive activity. The synoptic view provided by satellite imagery could provide great benefits also for the hazard monitoring of poorly known and remote volcanoes, especially those characterized by frequent activity with rapid growth of the summit portion of the edifice (e.g. Fuego and Pacaya volcanoes in Guatemala, Sangay in Ecuador), which may potentially experience collapse-PDC induced due to the accumulation of pyroclastic deposits.

Fig. 9: Panoramic views of the SEC from the South.
figure 9

a A few months after the generation and propagation of the 10 February 2022 PDCs, showing the deep scar formed in the southern flank of the cone. b The SEC a few weeks before the start of the paroxysmal activity at the VOR crater, showing the scar only partially filled mainly by lavas and minor tephra deposits. c The scar cloaked and hidden by the accumulation of tephra fallout after only four lava fountain episodes at the VOR crater. It is also evident that both the 2022 PDC and the following lava were also entirely mantled and obscured by thick tephra deposits. Photos by D. Andronico.

Full size image

Finally, the hazard associated with the hot, diluted, over-riding ash cloud that formed in the summit areas and moved independently of the PDC should not be underestimated, as observed during the Etna 2014 event (see Fig. 4 in ref. 37). Indeed, this type of cloud typically travels significantly longer distances than the concentrated part of the PDC62, covering larger areas and capable of having a serious impact on people in multiple ways (from lethal suffocation, skin burning to eye injuries63).

This study demonstrates how basaltic volcanoes like Etna can experience repeated phases of lateral collapses due to cyclic, paroxysmal eruptive behavior. This type of activity increases the instability of volcanic edifices and causes a prolonged hydrothermal alteration of deposits and products that rapidly accumulate above the growing scoria cone. In order to assess the associated volcanic hazards and avoid serious risks to people and infrastructure, we conclude that a multidisciplinary approach, combining field, laboratory, and satellite measurements, is required to predict in advance the re-establishment of the conditions that may trigger such type of PDCs.

Methods

Satellite imagery acquisition and DSM difference

In order to map and quantify the volumes of the PDC deposit, we exploited high spatial resolution tri-stereo Pléiades acquisitions recorded on 13 and 25 July 2021, the WorldView-1 images of 27 July 2021, and the tri-stereo Pléiades NEO acquired on 15 July 2022. The 2021 images were processed to obtain a DSM by applying the fusion technique presented in Ganci et al.44. The Pléiades Neo imagery was processed using the free and open source MicMac photogrammetric library (http://micmac.ensg.eu) developed by the French Institut Géographique National, which consists of three main steps: (a) tie points recognition and matching between images; (b) calibration and orientation, assessing relationships between viewpoints and objects; (c) correlation, producing dense matching for 3D scene reconstruction. The resulting DSMs have a horizontal resolution of 1 m or 0.5 m (for Pléiades NEO3) and an average vertical accuracy, estimated by using GPS Ground Control Points64 outside the area covered by the volcanic deposits, between 1.2 and 1.4 m. The thickness distribution of height variations was computed by DSMs difference after the co-registration performed through the Nuth and Kääb algorithm, which finds and potentially removes the horizontal and vertical shifts between the DSMs using the slope-aspect method65.

Field survey

Detailed surveys were possible only when the disappearance and melting of ice and snow and the rise in low temperatures generally present above 2700 m, made it possible to excavate the tephra deposit, which was otherwise too hardened and difficult to study. Five campaigns were then carried out to map the PDC deposit and select the outcrops to study from 28 April 2022 to 26 October 2022. A number of trenches were dug along the eastern lobe of the PDC deposit, not affected by the syn-eruptive lava flows, from which we described 16 stratigraphic sections. On the basis of the recognized stratigraphic units, 26 samples composed of ash particles to decimetric clasts were collected. Only one sample has been collected in the distal sections characterized by a thin, decimetric thickness. One more outcrop in the west part of the deposit has been qualitatively and quantitatively described, but no samples were collected.

Grain-size and textural analyses

Grain-size analyses were performed at the Sedimentology Laboratory of INGV-OE. The samples were first dried in the oven at 40 °C for 24 h and then manually sieved at 1φ steps in the grain-size range from −6φ (63 mm) to +5φ (0.032 mm). Samples preparation for textural analyses was carried out at the laboratories of the Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Pisa. Samples were washed with deionized water in an ultrasonic bath to remove impurities, dried at 60 °C, mounted with epoxy resin in 1-inch stubs, polished, and prepared for textural and geochemical analyses. The texture and components of volcanic particles (analyzed in the 1-0φ portion, 0.5–1 mm) were studied with a petrography microscope and a Scanning Electron Microscope (SEM) Zeiss EVO MA; images have been collected in back-scattered electrons (BSE) mode.

Numerical modeling

The open-source code IMEX-SfloW2D was used to model the propagation of the PDC. The model solves numerically the depth-averaged balance equations for a granular flow in global Cartesian coordinates47, applying a finite-volume central-upwind scheme in space and an implicit–explicit Runge–Kutta scheme for the discretization in time. The topography is given by a 10 m resolution Digital Surface Model in ESRI ASCII format, and the simulations are run on a uniform grid of equally sized square pixels. The Voellmy-Salm rheological model adopted in this study is an empirical one, widely used in hazard assessment studies where its performance is still comparable with more sophisticated models (e.g. Gruber & Bartelt66, Gueugneau et al.67). In its simplicity, it has proven able to reproduce some of the main features (both qualitative and quantitative) of rapid granular avalanches. The implicit treatment of the friction terms in IMEX_SfloW2D would allow us to easily implement alternative rheological models, some of which have been proved effective in reproducing laboratory experiments20,68 and for the reconstruction of recent PDC events8. In this study, however, model-related uncertainty would not be reduced, because calibrating more sophisticated models would require accurate information on the flow stratification and porosity, total grain-size distribution, and temperature, which was not available. The use of the Voellmy-Salm model is, in any case, justified a posteriori by the results of the inversion (see Sect. Cascade trigger mechanisms for PDCs at Etna). The input parameters defining the avalanche scenarios are the volume and the initial location of the mobilized granular material, the physical properties of the solid and gas components, and the parameters characterizing the Voellmy-Salm model. Both the rheological parameter µ and ξ, together with the initial flow velocity u0, were varied in prescribed intervals consistent with previous studies of other geophysical granular flows69. In particular, the ranges explored for µ, ξ, and initial velocity u0 are 0.20–0.55, 300–900 m/s2, and 0–140 m/s, respectively. The mobilized material was obtained by excavating a volume with an ellipsoidal shape of 1 × 106 m3 on the Digital Surface Model (DSM) representative of the pre-PDC topography in July 202144, assuming that a fraction (up to ∼35%70) of the total material derived from the collapse (∼1.6 × 106 m3) has been dispersed in the over-riding ash cloud and has not been considered in the model. An example of the input file required to run the simulation is provided in the Supplementary material.

An ensemble of 300 simulations was defined to constrain the best conditions to fit the observed deposit. The total time of each simulation was fixed at 240 s in order to avoid bias in the results due to the possible greater expansion of the numerical deposit at a certain combination of input parameters for a time greater than 1 min (i.e., the time of PDC emplacement), not matching with the observed deposit. The Jaccard Index was used to define the best fit, computed as J = (A ∩ B)/(A∪B), where (A ∩ B) and (A∪B) are respectively the intersection and the union between the simulated (A) and observed (B) area of the deposit. A more detailed description of the calibration process is reported in Zuccarello et al.43.