25 years recording of the long-base fluid tiltmeter installed at 2.850 m a.s.l. observatory of Pizzi Deneri on Etna volcano

Abstract

Etna is a highly active volcano with frequent eruptions and is closely observed with different monitoring techniques. One of the monitoring methods is a network of short-base tiltmeters that provides continuous control of ground deformation. Since 1996, this network has been integrated with a long-base fluid tiltmeter installed in two 80 m long connected tunnels at the high-altitude (2850 m a.s.l.) volcanological observatory of Pizzi Deneri (PDN) located on the volcano’s summit. The instrumentation was devised with an innovative configuration composed of mercury, free to move along the entire length of the tiltmeter in response to the ground tilt, and laser sensors to measure the changes in the mercury levels. In this study, for the first time, we present the entire 25-year series of data recorded by this instrumentation. During this long interval, the Etna volcano was characterized by numerous major eruptions due to dyke intrusions. The tilt variations recorded in the short-term during the rapid intrusive phases are presented and discussed. These signals contributed both to the modeling of eruptive processes, and in particular in the case of the 2002–2003 eruption, to the real-time understanding of the ongoing eruptive phenomenon, supporting the correct hazard assessment. In the medium-long term (years to decades), we show and discuss that the PDN signal is mainly related to the marked sliding of the entire eastern sector of the volcano.

Graphical Abstract

Introduction

A tiltmeter is a device that measures changes in the local surface inclination with a couple of sensors, oriented usually along two orthogonal components to obtain the tilt modulus and its direction. Instruments usually installed on volcanic areas to obtain accurate tilt measurements may be grouped in two main classes: short and long-baseline (Agnew 1986 and references there in).

Short-baseline devices measure ground tilt at single installations, generally by using bubble or pendulum based sensors and made to be installed inside boreholes or caves. They represent the most common tiltmeter type used on volcanoes as they are easy to manage and not too expensive (Dzurisin 2007).

Long-baseline tiltmeters (LBT) are fluid tube instruments measuring tilt across tens-to-hundreds of meters long-bases by the relative height difference of the fluid free surface at the ends of the tube. The long-base instruments have been developed by using water or other liquids, and may use different fluid level detectors (micrometers, pressure transducers or laser interferometers). LBT may be half-filled or completely filled; the first type was devised by Michelson (1914) and generally uses a water pipe, 10–20 cm in diameter, installed at the bottom of a 1–1.5 m deep trench, and accurately leveled.

The system acts like an underground pond where pressure and temperature variations occur slowly at depths of 1–2 m. Under such conditions, if the Earth’s surface tilts, water will flow so as to maintain an equipotential surface in the pipe. Half-filled devices show long-term stability and are unaffected by fluid thermal variations; however, due to the difficulty of laying out a straight and level tube, there are only a few examples of these instruments on volcanoes (e.g. Long Valley Caldera, Behr et al. 1992; Campi Flegrei, Amoruso et al. 2015).

Instead, several completely liquid filled tubes are present on volcanoes, the first, emplaced in 1958 at Kilauea Caldera, consisted of a tilt base made up of three brass hubs set in concrete piers and a portable water-tube leveling system with two micrometer-bearing waterpots connected by tubes (Eaton 1959). This system was successively installed on other volcanoes such as Soufrière Hills, Montserrat, during the 1966–67 seismic crisis (Shepherd et al. 1971), at Manam Island and on the floor of Rabaul Caldera in New Guinea (Mori et al. 1987; Scott and McKee 1982) and in 1976 at the Krafla Caldera, North Iceland (Tryggvason 1980).

Improved level measurement techniques and more accurate installation in tunnels free of temperature fluctuations have led to a substantial improvement in the tilt signals that have reached a very low seasonal noise level (less than 50 nrad) and less than 5 nrad per day. Short and long-base fluid tiltmeters may in fact have similar precision but LBT are preferable since they are less susceptible to small-scale tilt fluctuations, such as local ground movement for differential settling, hole relaxation, snow loading, freeze/thaw cycles and rain effects (Dzurisin 2007). This advantage is particularly evident in long-term volcanic unrest (several years) with small expected tilt. Some more recent tunnel installations have been set up at the Geodynamic Observatory Cueva de los Verdes, Lanzarote, (Canary Islands, Kalinina et al. 2004) and inside Arimura tunnel at Sakurajima (Japan, Hotta et al. 2016).

Over the last decades, a tilt network made up of 16 short-base bi-axial instruments placed in shallow boreholes has been installed on Etna volcano (for a review, see Bonaccorso et al. 2004; Gambino et al. 2014). In 1996, inside the EC Tekvolc research project aimed at constructing innovative devices, in order to integrate this short-base network a long-base instrument was installed at the high-altitude volcanological observatory of Pizzi Deneri (PDN), located 2850 m a.s.l. on Mount Etna’s northeastern flank (Bonaccorso et al. 1998). In this work, we report on the technical solutions adopted to install this device, and present 25 years of tilt data recorded at one of the most active volcanoes in the world, which was characterized by various eruptive phases during this period.

The long-base fluid tiltmeter at PDN observatory

The site: PDN volcanological observatory

The high-altitude volcanological observatory of PDN is located at 2850 m a.s.l. on the NE flank of Etna volcano about 2 km away from the summit craters (Fig. 1). The observatory is composed of two very thin domes (Fig. 2) and its construction lasted about three years, from the 1975 to 1978, following the innovative and challenging project of D. Bini (Imperadori 2023) that was financed by Italian National Council of Research (CNR).

Fig. 1

Map of the permanent tilt network operating on Mt. Etna reporting surface faults and the sliding sector. The dashed line represents the boundary of the eastern sliding sector, while the magenta arrow indicates the directions of movement. Inset map shows the location of the Pizzi Deneri (PDN) Observatory, site of the long-base fluid tiltmeter. The purple lines highlight the crater rims (NSEC, SEC, VOR, BN, NEC)

Fig. 2

Google Earth satellite image of the PDN observatory. The position and geometry of the measurement points (white circles) is indicated. The traces of the two tunnels (blue dashed lines) and of the EW component (white dashed line) are also indicated. The tunnel along SW direction is entirely underground, the tunnel along SE direction is partially underground and the eastern portion is close to the ground surface

Two artificial underground orthogonal tunnels, 80 m long with a central common point, were also built simultaneously near the observatory. The directions of the two tunnels are respectively SW, i.e. in direction of the summit craters, and SE (Fig. 2). The first tunnel is almost entirely built a few meters under the ground surface, while the second one is partially under the ground and its northeastern side has the external wall in contact with the ground surface. Furthermore, the two ends of the two tunnels (SW point and SE point) and the central common point are free to move independently.

The long-base instrumentation described in this work has been positioned inside these two tunnels and thus measures the tilt variation along the same two directions. The two tilt components are then referred to as the Radial (SW) Component and the Tangential (SE) Component. We have also considered a third component connecting the two points SE and SW, called EW Component (Fig. 2).

The high-altitude of the PDN observatory places the instrument in a very extreme environment, characterized by snow cover of up to several meters in winter that often interrupts the energy supply (provided by solar panels) and makes it unreachable, hence limiting its regular maintenance.

Instruments

The tilt measuring system was built by using a transparent polyethylene tube (diameter 4 mm), filled with mercury and positioned along the two tunnels of the observatory (Fig. 3). At the two ends and in the central part, the tube is connected to beakers with a laser sensor fixed near the top, whose emitted ray is reflected by a floating target inside the beaker on the surface of the mercury.

Fig. 3

PDN Observatory with schemes, images and specifications of the underground tunnels in which instrumentation is installed and the laser sensor used to measure the mercury level changes. Temperatures are recorded inside the tunnels at each of the three laser measurement points

The horizontal tilt of the ground is determined through the variation of the fluid level in the tube, measured at both extremities and at the central corner point. This type of laser sensor can detect variations in the order of µm (2σ = 5 µm) which, for the 80 m length of the tube, corresponds to a tilt resolution of 2σ/L ~ 0.05 µrad. The laser sensors used in the mercury tube are based on the optical triangulation method. The light beam emitted by the LED focuses on the target. A part of this light is reflected on the photosensitive detection element that produces an electrical signal proportional to the distance between the sensor and the measured surface (Fig. 3). More details on the sensor characteristics are reported in Bonaccorso et al. (1998).

Mercury was chosen as the liquid due to its various thermal and physical properties, making it particularly suitable for our purposes. It is characterized by:

• a very low freezing point (about − 30 °C), which favors its use in extreme thermal conditions such as those at the PDN observatory, especially during the winter months;

• a very low, almost nil, degree of evaporation at our temperatures of use (between − 10 and 10 °C), thus not needing to add further liquid over time;

• it also has the property of not "wetting" the walls and therefore not leaving residues of fluid materials on the sides of the beaker that may induce spurious level variations, thus attenuating this potential noise;

• high density (13.7 gr/cm³), allowing to use floating reflective targets without sinking problems;

• a low and stable coefficient of thermal expansion that for temperatures between − 20 and 20 °C is 1.8 × 10^–4 /°C.

Temperature monitoring is ensured by sensors positioned at the two ends and at the central common point that record daily changes within a few degrees and seasonal changes around 20 °C. Due to the diverse positions of tunnels with respect to the surface, seasonal temperature trends at the three sites are phase-shifted and show different amplitudes with a higher variation (20–25 °C) at the SE tunnel compared to the Central and SW ones (Fig. 4a).

Fig. 4

a Temperatures recorded at the three sensors (SE, Centre and SW); b comparison between the tangential (SE) component (in green) and the temperature measured at the SE sensor (in purple)

As highlighted in previous studies on the shallow depth installations (i.e. Bonaccorso et al. 1999, 2004), also at PDN station the seasonal temperature cycle is the main contribution to the annual oscillations of the recorded tilt signals (Fig. 4b).

The innovative idea of the instrumentation was therefore to exploit the properties of mercury and use laser instruments placed at the ends of the mercury tubes as measuring sensors, whose high-precision measurements could be entirely managed by an A/D converter and data logger. Data (vertical displacements and temperatures at the ends of the pipes) are acquired and controlled by a data logger (CR10 Campbell model), stored in a solid-state memory and also transmitted via RF modem for continuous real-time monitoring.

Currently, the PDN long-base instrument records a signal every 10 min, but, in the past, data were acquired at different sampling times (Table 1). Specifically, in the first two years of operation, from 1997 to 1998, the signal was sampled every 4 h; from January 1, 1999 to January 11, 2001 one data item per hour was acquired; from May 19, 2001 onward, the sampling rate changed to one data item every 10 min, which has been maintained until today, with the exception of summer 2000 (Bonaccorso 2006) and autumn 2011 when the acquisition was set to one data item per minute.

Table 1 Sampling times are shown on the left, periods for which there is no data on the right

Considering that we are reconstructing the 25-year history of the instrument located at 2850 m a.s.l., it is normal to have several periods where we have no data (the longest periods have been briefly summarized in Table 1). In most cases, the absence of data occurred during the late autumn–winter periods, when adverse weather conditions cause interruptions in power supply and data transmission (e.g. solar panels are often covered by snow). The copious amount of snow covering the volcano during these months, also makes it difficult to reach the site for regular maintenance. However, there are interruptions in data acquisition and transmission due to other reasons: between November 7, 2002 and October 23, 2003 the signal went off the measurement range due to the large-scale ground deformation caused by the violent 2002–2003 flank eruption that affected the northeastern sector of the volcano. In this case, the long-base tiltmeter continued to work, but recorded anomalous values indicating that the instrument’s range had been exceeded. The absence of data is sometimes due to technical interventions requiring an interruption in acquisition for longer or shorter periods. In particular, between December 3, 2016 and June 4, 2018, floor and wall breaking works were done for a complete separation of tunnel structure segments and a mercury tube wrapping with glass wool to increase thermal isolation.

The first data were acquired in July 1997, and the complete tilt data set, updated in December 2022, is shown in Fig. 5, where the tilt recorded along the radial (SW) component, the tangential (SE) component and the EW component, respectively, is plotted.

Fig. 5

1997–2022 PDN recorded signals: the orange rectangles highlight the major intrusions (I); the gray and white rectangles ideally separate the different periods of Etna activity (F), described in detail in Table 2 and reported in Chapter 3; (1) indicates the period of no data due to the out-of-scale following the 2002–2003 eruption; (2) indicates the interruption of the signal due to technical problems; (3) indicates the insertion of new sensors and restoration of the signal; (4) indicates the period of no data due to major interventions on the instrument

Main phases of volcanic and eruptive activity over the last 3 decades

In recent decades, the eruptive activity of Etna volcano has been characterized by different types of phenomena. After the major effusive eruption of 1991–1993, the following period 1993–2001 (F1 in Fig. 5) saw a period of recharging the volcanic edifice that led to the two main effusive-explosive eruptions of 2001 (I1 in Fig. 5) and 2002–2003 (I2 in Fig. 5). These eruptions were activated by the vertical ascent of powerful dykes. The 2002–2003 eruption also triggered the sliding of the eastern flank of the volcano, which in the following years shifted eastward with metric displacements. During this phase in which flank sliding dominated, short recharge phases occurred with subsequent effusive eruptions on the eastern flank in 2004 and 2006 (F2 in Fig. 5). Another dyke ascent fed the 2008 eruption (I3 in Fig. 5). In 2011–2013, after a short recharge (F3 in Fig. 5), 64 lava fountains were erupted from the “South-East” (SEC) summit crater, and another 4 powerful lava fountains in 2015 from the “Voragine” (VOR) summit crater (F4 in Fig. 5). After a following short recharge phase (F5a in Fig. 5), another feeding dyke characterized the short eruption of 2018 (I4 in Fig. 5). From late 2020 to February 2022, the eruptive activity of the SEC crater featured a further sequence of 64 lava fountains (F6 in Fig. 5).

A detailed description of all these multiple eruptive phases is beyond the scope of this work, however Table 2 reports a summary describing the different phases of activity of the volcano and noting the main references, in which exhaustive details can be found.

Table 2 Summary table of the main phases of volcanic and eruptive activity from 1993 to 2022

As recorded by the permanent GPS and borehole tiltmeter networks, the main and predominant phenomena that induced deformations to the volcanic edifice were the four vertical feeding dykes of 2001, 2002–2003, 2008 and 2018, and the marked sliding of the eastern flank in 2002–2008. Moreover, as a primary signal change, during 2001–2008, the PDN long-base tiltmeter recorded a long-term marked variation associated with the marked sliding that affected the volcano’s eastern flank.

Tilt changes during the fast dyke intrusions

In general, the tiltmeters are effective instruments to very clearly detect the deformation associated with the rapid dyke intrusions. Lateral flank eruptions along fracture systems are generally preceded by a dyke emplacement, lasting several hours to a few days and revealed by ground deformation and seismicity (e.g. Cervelli et al. 2002; Battaglia and Bachelery 2003). At Etna volcano, since the 1981 eruption, tilt changes have accompanied the main eruptive phases (Bonaccorso 1999) allowing the identification of eruptive precursors related to intrusive processes and eruptive fracture opening. Tilt data have also made an important contribution to modeling the geometry and position of the intrusive sources (e.g. Aloisi et al. 2003, 2006, 2009, 2020). PDN long-base tiltmeter recorded very significant data during the intrusive phases of 2001, 2002–2003, 2008 and 2018 eruptions (Fig. 6).

Fig. 6

Tilt recorded by the PDN long-base tiltmeter during the dyke emplacement preceding the eruption onsets of the 2001, 2002–2003, 2008 and 2018 intrusive phases. Horizontal labels are UTC time. Dashed lines represent no-data periods. At that time, the PDN long-base had power problems and did not work at night

In 2001, the eruption was heralded by a strong seismic swarm and ground deformation, which started on July 12 at 22:00 UT evidencing a dyke emplacement that preceded the onset of the July 17 lateral eruption on the southern flank by just over 4 days (Bonaccorso et al. 2002). PDN long-base and borehole tiltmeters on the southern flank recorded sharp variations during July 12–14 (about 20 µrad) that continued more slowly until the onset of the eruption (Fig. 6). Tilt and GPS changes inferred a nearly N-S oriented vertical tensile dislocation, 2 km long and 2.5 km high, with an opening of ca 3.5 m, which crossed the volcano edifice on the upper southern flank (Bonaccorso et al. 2002).

The 2002–2003 eruption was characterized by both a vertical uprising dyke in the upper southern flank, and a second intrusion propagating radially in the NE flank, along the NE Rift. The onset of the eruption occurred on the night of October 26, 2002 and was marked by an increase in the seismicity and tilt changes (20:25–21:00 UT) that, during the first four hours, took place in the southern-upper part of the volcano; subsequently, the signals evidenced a clear processes shift from the summit craters towards the NE Rift that accompanied the NE dyke intrusion for about 20 h (Aloisi et al. 2003). During the first phase (from 21:00 to 24:00 UT on October 26), the long-base recorded tilt changed by about 5–6 µrad (Fig. 6), also detected by borehole stations of the southern flank. Starting from 00:10 UT on October 27, exceptional tilt variations (150 µrad) were observed (Fig. 7) as the fracture propagated along the NE Rift together with the northeastern stations, revealing the propagation of the intrusive process for about 4 km. Aloisi et al. (2006) performed a dynamic inversion to model the temporal evolution of the continuously recorded tilt data, both for the vertical dyke on the high south flank and during the radial propagation along the northeast flank (Fig. 7).

Fig. 7

2002 eruptive tilt signals and NE radial propagating intrusion (from A to C). D1 (white diamonds) and D2 (white dots) represent the two parts of the dyke intrusion modelled by Aloisi et al. (2006). Point A represents the starting point of the model, point B the orthogonal projection to PDN radial component, and point C represents the end of the intrusion. The upper part of fissure system (D1) was modeled using a tabular dislocation propagating for ca. 3.0 km, with incremental steps of 235 m (distance between the diamonds). The lower part (D2) was modeled using a tabular dislocation propagating for ca. 3.2 km with incremental steps of 76 m (distance between circles). Redrawn from Aloisi et al. (2016)

The 2008 eruption was characterized by the formation of a dry fracture field on the northern flank, and a lateral eruptive fissure on the upper eastern flank. It was accompanied by a strong seismic swarm which started on May 13 at 08:40 UT. Aloisi et al. (2009) inverted tilt and GPS data, dividing the intrusion process into two phases: the first highlights an ascending vertical dyke-like intrusion; the second the attempt of the magma to penetrate laterally northwards. The first dyke intruded in the high southern flank until the eruptive fissure, while the lateral intrusion stopped with no lava emission, but induced a dry fracture field in the northern flank. Also in the 2008 case, the long-base tiltmeter recorded strong variations of about 110 µrad that started at 08:40 UT, about 4 h before the eruptive fracture opening (Fig. 6).

The 2018 eruption was characterized by marked seismicity and ground deformation events and yet accompanied by effusive activity of short duration and a moderate amount of emitted lava. By using multidisciplinary deformation networks (GNSS, borehole strainmeters and tiltmeters), Aloisi et al. (2020) inferred a dyke propagating in the first kilometer a.s.l. followed by the eruptive fracture opening, and a successive elongated intrusion in the southern flank, at a depth of about -3 km b.s.l. that did not reach the ground surface. The initial seismicity (December 24 at 8:30 UT) was located beneath the summit craters and the long-base recorded tilt changes from 8:20–8:30 UT, as did borehole tiltmeters and strainmeters, showing a variation over 50 µrad (Fig. 6), thus preceding the onset of the eruptive activity (11:10 UT) by almost 3 h, 1 h and 30 min before changes were recorded at GNSS stations (Aloisi et al. 2020).

Long-term tilt variation

Thanks to data acquired over 25 years, we can also evaluate the long-term behavior recorded at the PDN site. The signal is clearly characterized by the presence of fast and significant variations associated with dyke intrusions (Fig. 5), that preceded flank eruptions. To better observe the LBT behavior over time, we removed dyke-related variations, in order to highlight the long-term trends in signals that are normally hidden by the rapid (usually lasting a few hours) and marked changes that accompany the dyke intrusion process (Fig. 8a, magenta arrows).

Fig. 8

a Signals without dyke-related variations, highlighting the sliding effect triggered by the 2001–2002 flank eruptions (blue arrows) and long-term trend along the three components (magenta arrows); b zoom on the period 2011–2016 showing thermo-elastic oscillations

To obtain this result, we used a simple Matlab routine, applying to a moving average of the signal the “filloutliers” function that detects and removes outliers and jumps once the difference between adjacent elements of the array is calculated (in our case, the array consists of a moving average of the 90-step tilt data). We finally applied the “cumsum” function to reconstruct the signal, thus making a cumulative sum of the individual array elements obtained after the previous filtering.

PDN tilt is modulated by seasonal oscillations (about 5 µrad) observed until December 2016 (Fig. 8b) and linked to the annual temperature cycles (Fig. 4). Temperature variations induce different expansion of the mercury, i.e. 1.8 × 10^–4/°C at the ends of the tube, thus causing a thermic tilt noise that Bonaccorso et al. (1998) estimated as 2 µrad/°C. Moreover, the external temperature changes induce distortion effects of the topography, therefore producing real thermo-elastic tilts (Berger 1975; Harrison and Herbst 1977). These can be amplified by near-surface material heterogeneity (Ben-Zion and Leary 1986) and by the surface topography, which has a major effect on tilt (Harrison 1976) and can amplify the thermo-elastic tilts (Harrison and Herbst 1977; Meertens and Wahr 1986).

Generally, induced seasonal tilt variations can be filtered out with a linear regression with temperature (e.g. Bonaccorso et al. 2004); in the case of PDN signals, temperature trends at the three sites are quite different and this makes the filtering more complex. In any case, after the interventions during Dec 2016–Jun 2018 data interruption period, noise linked to temperature effects has been significantly reduced thanks to adjustments that improved the degree of interdependence between the different segments of tunnel structure and to a better thermal isolation made around the mercury tubes by using a cover of fiber glass. Moreover, on observing Fig. 8b, it is evident how the seasonal tilt variations have little influence on the detection of long-term trends on PDN signals.

The most important long-term variation began during the 2001 flank eruption, accelerating after the 2002–2003 intrusion and continuing until about 2008 (Fig. 8a, blue arrows). This major variation occurred during a strong increase of the eastern flank sliding seaward, moving up to tens of centimeters per year (i.e. Bonaccorso et al. 2006, 2015a, b; Bonforte et al. 2007; Palano et al. 2017). After the eruption of 2002–2003, the PDN tiltmeter continued to measure very strong variations related to the acceleration and marked sliding of the eastern flank (phase F2 in Table 2, Bonaccorso et al. 2006). The instrumentation stopped measuring correctly when it reached the limit of the laser measurement range. The system was adjusted in 2004 and in the following years showed cumulative signal trends of about 80 µrad till the end of 2007. The 2002–2004 period with no signal was probably characterized by larger variations with similar trends. Thus, it is not possible to accurately quantify the total tilt variation at PDN in those years. During 2001–2008, a large long-term variation was also recorded by the CDV borehole tiltmeter located inside the unstable eastern flank (Bonaccorso et al. 2015a, b). At CDV station, the negative trend is similar to the one recorded at PDN (Fig. 9), and both stations are affected by the overall sliding of the eastern flank, accelerated by the intrusions associated with the 2001 and 2002–2003 eruptive episodes.

Fig. 9

Radial tilt components recorded at the two stations detecting the instability of the eastern flank, namely the CDV borehole tiltmeter and the PDN long-base tiltmeter, installed in the middle eastern sector and in the summit crater area, respectively

Discussion and conclusion

In this work, 25 years of data (1997–2022) from the long-base tiltmeter installed along the two shallow underground tunnels at the PDN volcanological observatory were presented. The installation represented an innovative technological application (laser sensors for measuring changes in the level of mercury used as a fluid) and, although the high-altitude PDN site has extreme meteorological conditions and consequent difficulties in performing regular maintenance, the instrument has provided valuable information from the long-term (decades) to the short-term (hours-days).

In the short-term, the PDN long-base tiltmeter has proven to be an effective support in understanding eruptive events. This instrument was able to provide a valuable contribution in detecting dyke intrusions that preceded and accompanied the main flank eruptions of recent decades. Together with the other borehole-type tiltmeters, it was able to indicate signal variations before and during emplacement, allowing both the warning of the processes in progress and the modeling of the sources, as occurred for the major eruptions of 2001, 2002–2003, 2008 and 2018.

In particular, as the most significant example, we underline the importance of the PDN long-base tiltmeter during the 2002 dyke propagation along the NE Rift, in which the data proved invaluable for hazard mitigation and crisis management (Bonaccorso et al. 2015a, b). The main critical aspects were on the evaluation of the extension of the propagating eruptive fissures and on their stopping. The strategic position of PDN, located in the summit area and close to the NE Rift, has made it possible to follow and constrain the propagation of the dyke's position over time (Fig. 6) since the beginning of its propagation on the night of October 26 and to establish its halting on late October 27, thus enabling the INGV to inform in near real-time the Civil Protection, who had to respond to the pressing public concern on the extent of the lava field and the potential risks for the villages on the mid-NE flank of the volcano.

In the medium-long term (years to decades), the signal is influenced by seasonal variations essentially related to thermo-elastic effects caused by the marked temperature changes that affect the tunnels due to their near surface position. In the long-term, however, the signal maintained a trend that cumulated approximately 200 µrad on all three components (radial, tangential and EW Component) during the entire observation period. This continuous tilt signal is connected to the instability and lateral sliding of the eastern flank of the volcano, which represents the sector with the greatest movement and an amplification of the movements towards the East (e.g. Palano 2016 and references therein). During this trend, the tilt recorded a marked acceleration during 2002–2008, namely the period when the eastern flank recorded the greatest movements after the intrusions of 2001 and 2002–2003 (i.e. Bonaccorso et al. 2006; Bonforte et al. 2008). This behavior was also recorded in the borehole tilt station of CDV, which is located inside the eastern flank. In general, the long-term pattern of the tilt, characterized by monotonic growth with a marked acceleration in 2002–2008, has a similar trend to that of the historical series of the GNSS signal recorded on the eastern flank (Palano et al. 2023). Therefore, in the long-term the PDN tilt is primarily affected by the dynamics of flank sliding, cumulating tilt during this continuous and dominant process.