One of the key issues in forecasting volcanic eruptions is to detect signals that can track the propagation of dykes towards the surface. Continuous monitoring of active volcanoes helps

significantly in achieving this goal. The seismic data presented here are unique, as they document surface faulting processes close (tens to a few hundred meters) to their source, namely the

dyke tip. They originated nearby - and under - a seismic station that was subsequently destroyed by lava flows during eruptive activity at Etna volcano, Italy, in 2013. On February 20, a

~600 m-long and ~120 m wide NW-SE fracture field opened at an altitude between 2750 and 2900 m. The consequent rock dislocation caused the station to tilt and offset the seismic signal

data are interpreted as the seismic footprints of a magmatic dyke intrusion that moved at speed ~0.02 m/s (first stage) and 0.46 m/s (second stage).

Etna, located in southern Italy, is one of the most active volcanoes in the world in historical time. It is also one of the most highly monitored, with state-of-the-art networks of

geophysical and geochemical sensors operating continuously and providing surveillance over a territory of more than 1180 km2 prone to seismic and volcanic activity1.

The seismic station EBEL was set up on Etna in late 2004, at Belvedere at an altitude of 2899 m (Fig. 1). It was one of the 45 seismic stations on the volcano edifice belonging to the

permanent monitoring network run by Istituto Nazionale di Geofisica e Vulcanologia (INGV). It was equipped with Nanometrics 3-component Trillium (40 s cutoff period) seismometers. The

signals were sampled at a frequency of 100 Hz and were transmitted to the data acquisition centre in Catania. The other seismic stations used in our analyses had the same technical

characteristics; they were within a maximum distance of ~8 km from the summit craters and at altitudes between 1600 m (ESPC) and 3050 m (ECPN) (Fig. 1).

a) Fracture field (black lines), location of the active vents (Vent “a”, Vent “b”, Vent “c”) and centroid of volcanic tremor (solid grey circles) between 18 and 28 February 2013. The panel

is a zoom of the area indicated by the dashed grey square in Fig. 1c. The NW-SE dashed grey line indicates the trace of the cross section in Fig. 4. The light grey lava flows in a) formed in

2013. The snapshot b) was taken by Michele Mammino on 22 February 2013, six days before the seismic station breakdown (EBEL is visible in the foreground). c) Map of Etna and location of the

volcanic rifts (NE, W and S, respectively) and the seismic stations considered here (EBEL, ECPN, ESLN and ESPC). The map in a) was generated using a DEM owned by INGV; geographical

coordinates are expressed in UTM projection, zone 33N.

During the more than 8-year operational period of EBEL, numerous effusive (2004-2005, 2006, 2008-2009, 2013) and explosive (2006, 2007, 2008, 2011, 2012, 2013) eruptions affected the summit

of the volcano and the Valle del Bove (Fig. 1). EBEL was close to the active craters of Etna (~800 m from the New South East Crater, Fig. 1a) and, therefore, highly prone to breakdown in

case of paroxysmal/effusive activity. By February 28 2013, a lava flow had finally covered the station during a lava fountain episode, the 6th from the beginning of the year.

The last 11 operational days of EBEL offered us a wealth of field observations on ongoing eruptive and fracturing processes, giving us the rare opportunity to document these processes close

to their source. We use the seismic signal recorded as these processes were underway to piece together what happened at this time.

There is continuous activity at the summit craters of Etna. There is a constant release of gas and vapour, occasional Strombolian activity and impressive episodic high-rate paroxysmal

eruptions. At times, Etna also erupts through fissures in its flanks that are mainly clustered along three rift zones, the so-called NE-, S- and W-rift, respectively2. Dykes feeding eruptive

fissures stem radially from the central conduit or, more rarely, originate from independent batches of magma3,4,5.

The development of extensional fractures, faults and grabens accompanied magma intrusions and the subsequent volcanic activity during numerous historical flank eruptions of Etna (e.g., 1809,

1989, 1991–1993, 2001, 2002-2003, 2004-2005, 2009). This especially happened when the eruptive systems were along the NE- and S-Rift3,5,6,7,8. In the years from 1809 to 2010, the length of

the eruptive fractures varied from a few tens of meters up to ~9 km (∼2,3 km on average)5,8. The complete extension of these eruptive fractures covered time spans from tens to a few hundreds

of hours5; the mean velocity of propagation was 0.05 m/s, encompassed between the minimum value of 0.003 m/s (in 2004) and the maximum of 0.21 m/s (in 2008)5,7. However, the velocity was up

to ~1 m/s during the initial phase of the fracturing process, similar to that inferred at Krafla, Iceland9.

Extensional fractures, faults and grabens accompany magma intrusions also in other volcanoes, such as Kilauea, Hawaii10 and Stromboli, Italy (in 2002 and 2007)11,12 . However, this is not a

rule. For example, the dyke-induced tensile stresses associated with the feeder dyke did not generate any new surface faults at Krafla, Iceland, in 198013.

There are five active craters at Etna’s summit: the Central Crater - CC (divided into Voragine – VOR and Bocca Nuova - BN), the North-East Crater (NEC), the South-East Crater (SEC) and,

since 2007, the youngest “New” South-East Crater14 (NSEC, Fig. 1a). The NSEC built up on the southeastern segment of a ~2-km-long fracture field formed since 1998 on the eastern edge of the

summit area7. Its cone formed on the lower eastern flank of the “old” SEC15 during 56 eruptions (mainly lava fountaining episodes) between 2007 and 2014, building up a cone ~300 m in

height16,7. At a larger scale, it is worth noting that SEC, NSEC and numerous eruptive fissures opened along the high SE flank of the volcano between 1985 and 2006 following the same NW-SE

structural trend. This system is the continuation in the summit area of a shallow (1–4 km) tectonic structure separating two sectors of Etna prone to flank instability and characterized by

different kinematics7,17.

Table 1 summarizes eruptive activity at the summit craters and seismic data recorded at EBEL in February 2013 (all times are UT). During the month, NSEC produced six short-lived (a few

hours-long) paroxysmal eruptions. The first four episodes occurred between February 19 and 21 with repose periods lasting a few hours; the last two paroxysms were on February 23 and 28,

respectively (Fig. 2). Two main stages of fracturing phenomena also took place at the southeastern base of NSEC on February 20 and 28.

a) Eruptive activity, b) spectrograms of the seismic signal at ESPC, ESLN, EBEL, c) amplitude of volcanic tremor at EBEL and d) latitude of the centroid of volcanic tremor from 18 to 28

February 2013. Coloured bands in d) mark the approximate latitude of the North-East Crater (NEC), Bocca Nuova (BN), Voragine (VOR) and South-East Crater (SEC) system.

A slow increase in the amplitude of background seismic radiation (so-called volcanic tremor) heralded the first paroxysmal eruptive activity, which started late at night on February 18

(Table 1). During this first paroxysm, weak Strombolian activity also occurred at BN. The tremor peaked at ~1.12 × 105 nm/s during the climax of the lava fountain around 04:05 the day after

(February 19, Fig. 2a,c). The amplitude of volcanic tremor was calculated from the root mean square (RMS) value of the seismic signal over consecutive 5-min intervals, considering the bottom

25% (25th percentile) of the RMS value in order to remove any undesired transient event (Fig. 2c).

The tremor gradually decreased in the morning of February 19, but increased again soon after 21:00 the same day, leading to another lava fountain within about 17 h. The second paroxysm began

around midnight between February 19 and 20. It culminated at 00:50 with the opening of a new 600 m long eruptive fissure at the base of the NSEC that extended down to 2850 m above sea level

(Vent “a” in Fig. 1). Vent “a” opened at 0.17 m/s in ~1 hour and fed a lava flow for almost 2½ half days.

The first evidence of rock dislocation at EBEL occurred on February 20, about 6 h after the end of the second lava fountain. Between 09:14 and 09:44, the seismic signal underwent repeated

low-frequency oscillations (Fig. 3a). The largest offset was of ~0.8 × 104 nm/s. Towards the end of this phenomenon, the amplitude of volcanic tremor increased, heralding the third eruptive,

paroxysmal activity shortly after (Table 1, Fig. 2). When the third paroxysm took place between 13:00 and 14:45, a dry (that is without magma emission) fracture field over a length of ~600 

m and a width up to ~120 m opened in the area of Belvedere, close to EBEL (see the graben structure in Fig. 1a, b, partially overlapped by Vent “c” that would open later, on February 28).

The fracture field was mainly characterized by NW-SE trending extension fractures with opening displacement (apertures) as great as 1.45 m and normal faults with moderate vertical

displacement (