Spatial vent opening probability map of Etna volcano (Sicily, Italy)
We produce a spatial probability map of vent opening (susceptibility map) at Etna, using a statistical analysis of structural features of flank eruptions of the last 2 ky. We exploit a detailed knowledge of the volcano structures, including the modalities of shallow magma transfer deriving from dike and dike-fed fissure eruptions analysis on historical eruptions. Assuming the location of future vents will have the same causal factors as the past eruptions, we converted the geological and structural data in distinct and weighted probability density functions, which were included in a non-homogeneous Poisson process to obtain the susceptibility map. The highest probability of new eruptive vents opening falls within a N-S aligned area passing through the Summit Craters down to about 2,000 m a.s.l. on the southern flank. Other zones of high probability follow the North-East, East-North-East, West, and South Rifts, the latter reaching low altitudes (∼400 m). Less susceptible areas are found around the faults cutting the upper portions of Etna, including the western portion of the Pernicana fault and the northern extent of the Ragalna fault. This structural-based susceptibility map is a crucial step in forecasting lava flow hazards at Etna, providing a support tool for decision makers.
KeywordsFlank eruption Dike Volcano structure Susceptibility map Spatial clustering Back analysis
Volcanoes usually erupt from the summit area, fed by the central conduit. However, many eruptions occur from vents along the volcano flanks. In some active volcanoes with open central conduit, including Kilauea, Stromboli, Vesuvio (between 1631 and 1944), and Etna, the occurrence of flank eruptions is similar or even more frequent than that of summit eruptions (Holcomb 1987; Neri and Acocella 2006; Acocella et al. 2006, 2009; Neri et al. 2011a). These flank eruptions are usually responsible for effusive activity, which may feed lava flows capable of flowing for long distances over the volcano’s slopes and invading vulnerable areas. Forecasting the possible location of flank eruptions, in addition to any summit activity, is an important challenge to improve our understanding of volcanic processes so as to achieve successful assessments of lava flow hazards (Newhall 2000; Sparks 2003; Behncke et al. 2005; Crisci et al. 2010; Cappello et al. 2011a; Bonaccorso et al. 2011; Ganci et al. 2012b).
Various approaches, based on statistical or probabilistic analysis of the geology and structure of a volcano and the processes driving magma transfer, were proposed to forecast the location of future eruptive vents. Wadge et al. (1994) evaluated a probability surface of future vent locations on Etna, starting from a classification of its flank eruptions from 1763 to 1989, and applying a non-homogeneous Poisson process. Connor and Hill (1995) proposed three non-homogeneous Poisson processes to estimate the recurrence rates and the associated probability of volcanic eruption in the Yucca Mountain (Nevada). These models were extended by Martin et al. (2004) in a Bayesian framework applied to the Tohoku volcanic arc (Japan) to combine different geological and geophysical information assuming that new volcanoes will not form far from existing ones. Recently, Marti and Felpeto (2010) proposed a numerical multi-criteria evaluation of the spatial probability of hosting a new crater in Tenerife (Canary Island) assigning weights (chosen through the expert judgment elicitation) to different layers of information obtained with structural criteria (for long-term analysis) and monitoring data (for short-term analysis).
Some of these studies have a common limitation due to the not well-detailed knowledge of the volcano structure, as well as its mechanisms of shallow magma transfer, preventing the formulation of a solid theoretical base for the statistical analysis. Recently, we completed a study on the modalities of flank eruptions at Etna of the last ∼2 ky, for which a detailed structural database was collected (Neri et al. 2011b). These geological records allowed new and broader insights into the dynamics of summit and flank eruptions, defining the main structural parameters and, possibly, the role of the magmatic intrusions in the framework of the volcano flank instability. During the last hundreds of years the central conduit of Etna is always open and active; in addition, lava flows from the summit vents are incapable of threatening the villages located at medium-low elevations on the flanks of Etna or reaching the Ionian coast where the city of Catania is located (Behncke et al. 2005). For these reasons, we analyze only flank events to identify areas where new eruptive vents may open at Etna. Our approach is based on the possibility of accounting for the available geological and structural data, highlighting the main pathways of magma transfer, as well as an understanding of their mechanisms to derive a susceptibility map containing spatial estimates suggesting the most likely future emission zones. The operative procedure consists in converting different structural datasets (faults, dikes, and eruptive fractures) in separate probability density functions (PDFs), giving each of them a relevance value (weight) with respect to the evaluation of susceptibility, and finally combining all the PDFs and weights in a non-homogeneous Poisson process.
Types and frequency of Etna eruptions
Etna’s eastern and southern flanks are characterized by slow but continuous seaward displacement (Fig. 1; Solaro et al. 2010). Northward, the unstable area is delimited by the Pernicana fault system (PFS; Tibaldi and Groppelli 2002) and to the South-West by the Ragalna fault system (RFS; Neri et al. 2007). Recent InSAR observations reveal differential slip along faults inside the unstable area, delimiting blocks characterized by different velocity of deformation (Solaro et al. 2010). A close relationship between flank deformation and eruptive activity at Etna is described in Bonaccorso et al. (2006).
Distinct eruptive cycles have been observed since 1865, with flank eruptions occurring in clusters, separated by periods of quiescence and summit activity (Behncke and Neri 2003). Flank eruptions seem to occur more frequently during accelerations of flank movements, facilitated by the opening of fractures close to the summit of the volcano and by increased activity of the main fault systems (Neri et al. 2009).
Geological and structural data
Based on field measurements, satellite data, and geophysical observations, Neri et al. (2011a) have completed a detailed study of the main structural features of lateral eruptions at Etna. Here, we carried out a complete revision of the location and opening dynamics of all eruptive vents occurring since 1900, one of the most active periods in the documented eruptive history of Etna, comprising 35 summit and 33 flank events. Moreover, we collected frequency and strike of outcropping and buried eruptive fractures active in the past ∼2 ky, of the dikes cropping out in the Valle del Bove (VdB), and of the main faults that can potentially be used as pathways for intruding magma. All geological and structural data were used to construct a georeferenced database containing the mapping of: (1) the outcropping and buried eruptive fissures, (2) the dikes cropping out in the VdB, and (3) the main faults.
At higher elevations (>2,000 m), almost all eruptive fissures are radial with respect to the volcano summit. However, these radial fissures cluster along an overall ∼N-S trend, that is perpendicular to the direction of regional extension in the area.
At lower elevations (<2,000 m) a few of the longest fissures (>5 km) propagating downslope are rotated by 10–40° with regard to the radial path. The vast majority (∼96 %) of the entire fissure population is radial, with a negligible population (∼4 %) slightly deviating from the radial path. These latter fissures, located at the lower extremities of the North-East and South Rifts, mainly show NE and SE directions, respectively (Fig. 2). These directions are consistent with those of the nearby structures controlling the slip of the unstable eastern flank of the volcano (Solaro et al. 2010), so that a relation between eruptive activity and flank instability may be expected at these lower elevations. Fissures older than 1600 are uncertain in age and were inserted into the same georeferenced layer (Fig. 2a). This group includes 200 fissures, largely with a S-ward (43 % of the cases), NE-ward (23 %), and W-ward (20.5 %) direction.
We identified 68 flank eruptions and 130 fissure systems emplaced between 1600 and 2011. Nineteen of these eruptions occurred during the seventeenth and eighteenth centuries, generating 22 fissures (Fig. 2b; in some cases, the same eruption produced multiple eruptive systems with different orientations). Thirteen flank eruptions formed 27 fissures during the nineteenth century (Fig. 2c). Twenty-seven and six events in the twentieth and twenty-first centuries, respectively, generated 81 fissures (Fig. 2d). The total of fissures is oriented with a S-ward (39 % of the cases), NE-ward (22 %), E-ward (18 %), and W-ward (11 %) direction, and a minor percentage with a N-S orientation (8 %). The frequency distribution of the post-1600 eruptive systems shows that ∼60 % are located above 2,400 m, 30 % crops out in between 1,800 m and 2,400 m, and the remnant 10 % are below 1,800 m a.s.l.
Faults related to eruptive fissures
Georeferenced mapping of Etna faults has been carried out, merging field, seismic, and magnetotelluric data (Cardaci et al. 1993; Monaco et al. 2008; Bonforte et al. 2009; Falsaperla et al. 2010; Siniscalchi et al. 2012) with structural data acquired through satellites (Solaro et al. 2010) and geochemical mapping (Siniscalchi et al. 2010; Neri et al. 2011b).
Figure 5 shows that, except for the Timpe Fault System (to the east) and a few minor faults on the W flank, most of the active faults are located along the peripheral continuation of the North-East and South Rifts. Also, almost all the tectonic deformation at the end of the focused North-East Rift seems related to the activity of one main single fault (the Piano Provenzana fault, Fig. 5, becoming to the east the Pernicana Fault; Ruch et al. 2010); conversely, at the end of the wider South Rift, the deformation is partitioned among different fault systems. Therefore, a correlation appears between the end and width of a rift and its continuation as a broader or narrower zone of deformation.
Moreover, there is a strong link of these faults at the termination of a rift with the instability of the W and S flanks (Bonforte et al. 2011). In some cases, there is evidence that fissure eruptions in the summit area become major faults, characterizing the activity of the Etna unstable flank, on the lower slopes (i.e., the NW-SE 1989 fissure system, which connects the summit region of the volcano with the tectonic structures of the lower SE flank; Falsaperla et al. 2010). Again, the more focused deformation controlling flank instability on the N boundary (PFS) matches with the narrower North-East Rift and contrasts the more diffused deformation of the unstable flank to the south, matching with the wider South Rift. All these features suggest a link between the location and activity of the North-East and South Rifts and the instability of the E and S flanks of the volcano. It is expected that, on the upper slope, gravity and magma emplacement play a common role, feeding each other and justifying the preferred magmatic activity along the North-East and South Rifts.
The probabilistic model
Most of new fissures are expected to be radial and focus above 1,600 m.
The control of the regional tectonics suggests that numerous fissures (∼36 %) are expected to have an overall ∼N-S trend, therefore to develop along the South Rift.
A link between flank instability and the occurrence of fissure eruptions along the North-East and South Rifts is expected as result of the gravitational instability of the E and S flanks.
Moreover, some reasonable assumptions can be made about the contribution of each structural dataset to construct the susceptibility map. In principle, the most importance is given to recent (post-1600) eruptive fractures, considering them as the current, main expression of the volcano structure related to the eruptions. The contribution of the other datasets is considered less important (pre-1600 eruptive fractures), or considerably less important (faults and dikes), because of their “passive” role in the intrusion processes (faults) and their age (dikes).
To translate these qualitative assumptions into a mathematical language, we initially convert the spatial distribution of past volcanic structures, such as faults, dikes, and eruptive fractures divided by age, into separate PDFs. Successively, through a back analysis, we assign each PDF a weight, measuring the importance and quality of data. Finally, the PDFs and their relative weights are combined via a weighted summation and modeled in a non-homogeneous Poisson process to obtain the susceptibility map.
Probability estimations based on kernel technique
The structural georeferenced database of Etna includes four datasets: eruptive fissures of the past ∼2 ky separated per age into two datasets, pre-1600 and post-1600 fissures, dikes cropping out in the VdB, and main tectonic structures that can potentially be used as pathways for intruding magma and/or influence the superficial stress field of the volcano, namely faults.
The local intensity function λxy was computed using a kernel function, which is presently the most feasible tool for probabilistic modeling of the long-term future patterns of volcanic events. A kernel function is a PDF centered at each data sample location while exerting an influence in the surrounding region (Diggle 1985). It is employed to calculate the probability surface from the location of past, discrete events, in function of distance to nearest-neighbors and a smoothing factor.
Choice of the smoothing parameter
Back analysis for weights selection
To assign a weight to each PDF, we introduced an unbiased procedure for the back analysis, which aims to determine parameters and/or properties on the basis of a sample of known data. We employed it to identify how historical structural data have affected the most recent eruptions. Taking into account the age of the eruptions, we divided all the structural data in two parts, using the first one as learning set and the second one as testing set. The year 1981 was fixed as a time barrier, since it determines two statistically meaningful sets (49 fissures for learning and 10 for testing). Moreover, starting from 1981, the volcano activity shows an increasing temporal trend (Salvi et al. 2006; Bebbington 2007; Cappello et al. 2011b).
The methodology adopted for elaborating the susceptibility map at Etna is entirely based on the characterization of past events through a mathematical model. Figure 8 shows how the expected future eruptive fractures are concentrated in the rift areas radiating from the summit of the volcano toward its periphery. Irrespective of their location, the available structural data and analog and theoretical models (Acocella et al. 2009) demonstrate the statistical predominance of the radial structures (fissures, dikes, and faults). For this reason it is legitimate to expect the next flank vent to open at Etna will belong to a radial eruptive fissure, with its upper extremity corresponding to or trending toward the summit crater area.
Summit crater area
The zone with the highest probability of new vents opening is centered on the summit crater area and radially around it, to ∼2,900 m a.s.l. These summit eruptive fractures are related to the dynamic of the central crater area, which is characterized by (1) paroxysmal episodes lasting a few hours; (2) lava overflows of longer duration (days to months) from the summit craters; and (3) effusive events from fractures on the external flanks of the summit cones, most of which do not propagate at lower elevations, due to the superficial nature of the magmatic intrusions.
The distribution of the summit craters, four of them to date, seems to be controlled by two main structural trends, oriented SSW-NNE and NW-SE, respectively: NNE from the Voragine one encounters the North-East Crater (born in 1911), at the northern base of which begins the Northeast Rift, whereas going SE from the Voragine, one encounters the Southeast Crater (1971), which in turn overlooks the upper portion of the South Rift zone. The position of Bocca Nuova (1968) is less clear; the volcano-tectonic structures within the Bocca Nuova, often oriented N-S and NNW-SSE, seem to point to a dynamic link to the Central Crater, the huge, single summit crater that existed during the first half of the twentieth century, and within which the Voragine and Bocca Nuova have developed. The four summit craters have sometimes shown contemporaneous eruptive activity, but with different style and eruption rates. This implies a structural independence of the respective feeder conduits in their most superficial portion (<2–3 km from the surface).
Most recently, a new pyroclastic cone has started growing on the lower east flank of the South-East Crater, reaching a height of about 150 m during 25 paroxysmal episodes in 2011–2012 (Vicari et al. 2011; Ganci et al. 2012a).
Rift zones and other areas of lateral vents
Below 2,900 m a.s.l., Fig. 8 shows a N-S zone of elevated probability of new eruptive vents opening, ∼7 × 2.5 km wide, extending from the summit area to ∼1,700 m elevation on the south flank. This is the uppermost portion of the South Rift, which is the strip of terrain along the western rim of the VdB. At lower elevations, this rift area opens fanwise toward SSE and SE, down to rather low altitude (minimum elevation ∼400 m), coinciding with the most densely populated area of the volcano. In the SE area, several NW-SE pre-1600 fissures and NW-SE faults crop out influenced the extension of the susceptible zone at lower elevation, although the low weight assigned by the performance index elaboration (0.10 to faults, 0.15 to pre-1600 eruptive fissures, see paragraph 4.5).
A second area (∼8 × 2 km wide) with a high probability of new vents forming extends from the summit toward N-NE (downward to ∼1,000 m a.s.l.). It includes the North-East Rift, characterized by numerous eruptive fissures and pyroclastic cones. It has been active for at least 15–20 ky; in the recent past, the eruption frequency here was lower than on the S flank. Yet this rift plays a fundamental role in the structural setting of the volcano, because it represents the NW boundary of the sliding eastern flank (Ruch et al. 2010). The North-East Rift indeed shows faults and fractures characterized by predominantly extensional kinematics, which pass into left-lateral transtensive kinematics in its lower portion, where it is linked to the transtensive left-lateral Pernicana fault system, the northern margin of the zone affected by flank instability. Along the complex structural North-East Rift—PFS system, which is ∼3–4 km deep (Ruch et al. 2010), one can imagine the ascent of dikes and magmatic fluids (Siniscalchi et al. 2012), which use the tectonic discontinuities as pathways to the surface, while facilitating the movement of one block against the other due to a “lubrificating” action on the fault planes. Outcropping and buried dikes and pyroclastic cones, deformed by the tectonic activity, indeed occur along the PFS around 1,450 m altitude (Tibaldi and Groppelli 2002), testifying to the existence of this volcano-tectonic setting.
Structurally and morphologically, the North-East Rift is linked to the North-East Crater. During the 1960s to 1980s, this crater produced vast and thick lava fields, which bury much of the North-East Rift down to ∼2,500 m elevation. Yet another zone of elevated probability of new vents opening extends from the summit area toward ENE. This is the East-North-East Rift (Ruch et al. 2010), whose lower portion passes into the Ripa della Naca faults (∼1,000 m altitude), and includes the Valle del Leone and a portion of the N wall of the VdB, as well as a number of parasitic cones. Both the eruptive fissures and dikes outcropping in this area were conditioned, during their propagation, by the presence of the N wall of the VdB, due to their shallow paths (from a few tens to a few hundreds of meters of depth). In some cases, the intrusions halted at the base of the VdB wall (as in 1989 and 2001), whereas on other occasions the fractures cut through the wall, first changing direction and adapting the trend of the wall itself, and then resuming their original direction (toward NE and ENE), thus propagating outside the VdB (as in 1971 and 1979). Analyzing the geometric properties and structure of the dikes outcropping in the area, an identical evolution of the past eruptive activity can be inferred. The kinematic indicators present in the magmatic intrusive bodies (that is, the form and arrangement of bubbles, flow structures, morphology of the imprint of glassy margins on the host rocks) clearly indicate a subhorizontal (0–10°) or slightly inclined (10–30°) mode of propagation of magma within the fractures (Acocella et al. 2009).
Finally, another area of elevated probability of new vents opening extends from the summit toward W and WSW, downward to ∼1,700–1,800 m a.s.l. This is the most elevated portion of the West Rift, where flank cones reach a minimum elevation of ∼1,300 m and most recent eruptive manifestation dates back to 1974. Compared to the other rift zones, some geological data (significant documented explosive activity) and instrumental evidence (seismic swarms preceding some eruptions) suggest that some West Rift eruptions might have been fed directly from a deep magma reservoir, bypassing the central conduit of the volcano.
The susceptibility map in Fig. 8 furnishes an essential element for volcanic (possible lava flow invasion) and seismotectonic (shallow earthquakes accompanying the emplacement of the feeder dike) hazard evaluation, helping the government organizations charged with land-use planning. This result has significance for tephra fallout models as well, because most tephra associated with these eruptions accumulates in proximal to medial distances from the vent.
Furthermore, the South Rift is characterized by eruptions of comparatively long duration (mean duration ∼30 days), and greater volumes (>24 × 106 m3 of lava per eruption; Neri et al. 2011b), increasing the exposure of urban infrastructures in this sector of Etna to damage.
The towns on the eastern flank are in part “protected” from lava flows by the VdB, a morphological depression (5 × 7 km wide) able to capture lava flows emitted from the eruptive fissures below the summit craters and along the upper portions of the South and East-North-East Rifts. Only eruptions of exceptionally long duration (months to years) and very high eruption rates (>30 m3/s) could form lava flows capable of traveling across the entire valley and seriously threaten the population centers of the eastern sector (Fornazzo in 1979, Zafferana Etnea in 1992).
The eruptions in the northern sector are generally short but violent, and originate from rather long fractures (>6 km in length; Neri et al. 2011b). The main population centers exposed to lava invasion are Randazzo and Linguaglossa, downslope of the North-East Rift, and Piedimonte and Sant’Alfio, which lie below the East-North-East Rift.
The western flank of Etna is much less urbanized. The only town lying within the ideal trajectory of lava flows emitted from the West Rift is Bronte, whereas Adrano and Paternò are more remote and thus at considerably lower hazard.
The map of the spatial probability of opening of future vents at Etna was based on the statistical analysis of spatial distribution of the available geological information. It is a susceptibility map, not a hazard map. Indeed, it does not consider the potential consequences of volcanic events, but identifies the areas susceptible to opening of new eruptive vents. Moreover, estimating the probable distribution of future vent locations is the first and perhaps most important step in forecasting lava flow hazards. The more rigorous the evaluation of the susceptibility map, the greater the accuracy of the lava flow invasion hazard map. We obtained a quantitative evaluation of the probable location of future eruptions analyzing the past geological and volcanological history of Etna, introducing a recurrence rate, which defines a possible spatial vent density. The methodology proposed exploits all available structural datasets through a non-homogenous Poisson process to obtain the intensity of volcanic events at sampling points representing potential future vents.
Our susceptibility map has great potential value for Etna because it would allow better management of its territory. Future work will include the statistical analysis of the temporal distribution of past lateral eruptions for constructing a spatiotemporal probability map of vent opening at Etna.
This study was performed with the financial support from the V3-LAVA project (DPC-INGV 2007–2009 contract). Comments by Editor Agust Gudmundsson and two anonymous reviewers greatly improved the manuscript.
- Acocella V, Neri M, Scarlato P (2006) Understanding shallow magma emplacement at volcanoes: orthogonal feeder dikes during the 2002–2003 Stromboli (Italy) eruption. Geophys Res Lett 33: L17310. doi:10.1029/2006GL026862
- Behncke B, Neri M, Nagay A (2005) Lava flow hazard at Mount Etna (Italy): new data from a GIS-based study. In: Manga M, Ventura G (ed) Kinematics and dynamics of lava flows, Spec Pap Geol Soc Am 396, pp 189–208Google Scholar
- Burton MR, Neri M, Andronico D, Branca S, Caltabiano T, Calvari S, Corsaro RA, Del Carlo P, Lanzafame G, Lodato L, Miraglia L, Salerno G, Spampinato L (2005) Etna 2004–2005: an archetype for geodynamically-controlled effusive eruptions. Geophys Res Lett 32:L09303. doi:10.1029/2005GL022527 CrossRefGoogle Scholar
- Cappello A, Vicari A, Del Negro C (2011a) Assessment and modeling of lava flow hazard on Etna volcano. Boll Geofis Teor Appl 52(2):299–308Google Scholar
- Cappello A, Vicari A, Del Negro C (2011b) Retrospective validation of a lava flow hazard map for Mount Etna volcano. Ann Geophys 54(5). doi:10.4401/ag-5345
- Connor CB, Hill BE (1995) Three nonhomogenous Poisson models for the probability of basaltic volcanism: application to the Yucca Mountain region, Nevada. J Geophys Res 100:10,107–10,125Google Scholar
- Corsaro R, Neri M, Pompilio M (2002) Paleo-environmental and volcano-tectonic evolution of the south-eastern flank of Mt. Etna during the last 225 ka inferred from volcanic succession of the «Timpe», Acireale, Sicily. J Volcanol Geotherm Res 113:289–306. doi:10.1016/S0377-0273(01)00262-1 CrossRefGoogle Scholar
- Diggle PJ (1985) A kernel method for smoothing point process data. Applied Statistics. J R Stat Soc Ser C 34:138–147Google Scholar
- Holcomb RT (1987) Eruptive history and long-term behavior of Kilauea Volcano. US Geological Survey, Professional Papers 1350:261–350Google Scholar
- Kiyosugi K, Connor CB, Zhao D, Connor LJ, Tanaka K (2010) Relationships between temporal-spatial distribution of monogenetic volcanoes, crustal structure, and mantle velocity anomalies: an example from the Abu Monogenetic Volcano Group, Southwest Japan. Bull Volcanol 72:331–340CrossRefGoogle Scholar
- Martin AJ, Umeda K, Connor CB, Weller JN, Zhao D, Takahashi M (2004) Modeling long-term volcanic hazards through Bayesian inference: an example from the Tohuku volcanic arc Japan. J Geophys Res 109. doi:10.1029/2004JB003201
- Monaco C, De Guidi G, Catalano S, Ferlito C, Tortorici G, Tortorici L (2008) Carta Morfotettonica del Monte Etna. Litografia Artistica Cartografica, FirenzeGoogle Scholar
- Neri M, Guglielmino F, Rust D (2007) Flank instability on Mount Etna: radon, radar interferometry and geodetic data from the southern boundary of the unstable sector. J Geophys Res 112: B04410. doi:10.1029/2006JB004756
- Neri M, Giammanco S, Ferrera E, Patanè G, Zanon V (2011b) Spatial distribution of soil radon as a tool to recognize active faulting on an active volcano: the example of Mt. Etna (Italy). J Environm Rad 102:863–870. doi:10.1016/j.jenvrad.2011.05.002
- Newhall CG (2000) Volcano warnings. In: Sigurdsson H (ed) Encyclopedia of volcanoes. Academic, San Diego, pp 1185–1197Google Scholar
- Siniscalchi A, Tripaldi S, Neri M, Giammanco S, Piscitelli S, Balasco M, Behncke B, Magrì C, Naudet V, Rizzo E (2010) Insights into fluid circulation across the Pernicana Fault (Mt. Etna, Italy) and implications for flank instability. J Volcanol Geotherm Res 193:137–142. doi:10.1016/j.jvolgeores.2010.03.013 CrossRefGoogle Scholar
- Siniscalchi A, Tripaldi S, Neri M, Balasco M, Romano G, Ruch J, Schiavone D (2012) Flank instability structure of Mt Etna inferred by a magnetotelluric survey. J Geophys Res 117:B03216. doi:10.1029/2011JB008657
- Weller JN, Martin AJ, Connor CB, Connor LJ, Karakhanian A (2006) Modelling the spatial distribution of volcanoes: an example from Armenia. In: Mader HM, Coles SG, Connor CB, Connor LJ (eds) Statistics in volcanology. Spec Pub IAVCEI, Geol Soc London, pp 77–88Google Scholar