Today, we can observe the 1809 feeder system in four outcrops exposed along the walls of two adjacent craters (A and B in Figure 2), that were formed during the later stage of the eruption. Hereafter, we identify the outcrops of the feeder system as A–S at the southern wall of the crater A, A–N at the northern wall of the crater A, B–S at the southern wall of the crater B, and B–N at the northern wall of the crater B, respectively. The most important and clear outcrops are inside crater A and for this reason we focus particularly on the description of this crater. Only the uppermost part of the feeder system is exposed in crater B.

Craters A and B are located along the uppermost part of the eruptive Fissure 3 (Figures 1A, 2). These two craters and other small ones form a ~450 m long segment of the eruptive fissure that still crops out (although partially buried by younger volcanic products), between 2390 and 2340 m asl (Figure 2). The outcrop A–N is at the center of this segment and A–S is at the peripheral part, ~150 m from the center. The alignment of the craters indicates that the strike of the feeder dyke is 020°, parallel to the strike of the central portion of the NE Rift (Figures 1,2). We should define some key terms used in this paper to avoid any confusion. A “feeder system” connects to a vent and includes the entire magma plumbing system below the original ground surface. The term “dyke” is used only for a vertical intrusive body that fills a fracture in the host rock.

The cross section of the feeder system of the 1809 eruption, consisting of a feeder dyke and a sill in the lowest portion, the diatreme, crater and the scoria cone, is well–exposed only on the walls of crater A (Figure 3). The exposure of the feeder system in crater B is limited only to the diatreme part owing to the development of the talus deposit.

In the outcrop A–S, the feeder system up to 22 m deep from the existing, original ground surface before the 1809 eruption (bottom of Unit 3b) is exposed (Figure 5). The feeder system is divided in three portions: the deeper part consists of a dyke cropping out >16 m below the ground. Between 16 and 15 m in depth, the feeder forms a horizontal intrusion (hereafter “sill”) in Unit 6', immediately under the base of overlying Unit 5. The shallower part of the feeder system is made up of an upward–flaring diatreme. The feeder system is almost vertical at depths greater than 15 m from the original surface. It inclines~75° westward within the uppermost 15 m, and is perpendicular to the original ground surface (the base of Unit 3b inclined ~15°eastward; see Figures 3, 5).

In the outcrop A–N, the feeder system exposes its uppermost 15 m (Figure 6). The feeder system shifts from a dyke around 15 m deep to an upward–flaring diatreme. The feeder system inclines ~75 degrees westward, and is perpendicular to the original ground surface (the base of Unit 3b inclined ~15 degrees eastward).

In the outcrop B–S, the feeder system exposes its uppermost 7–8 m, in which the feeder system shows an upward–flaring diatreme. No feeder dyke is exposed. At the base of the outcrop, the width of the feeder is 16 m and increases to 30 m at the base of Unit 3.

The uppermost 15 m of the feeder system is exposed in the outcrop B–N. The whole feeder system in this outcrop is made of the diatreme and no feeder dyke is exposed. The width of the feeder system at the bottom of the outcrop is 3.6 m and the width increases to 22 m at the base of Unit 3.

The thickness of the feeder system depends on the host rock lithology (Figure 7). The feeder in the outcrop A–S has an almost constant thickness (0.2 m) in the central part of Unit 6. The thickness increases to 0.6 m in the volcanic ash layer at the top of Unit 6, and then converts to a 4 m width sill. The feeder intrudes in Unit 5 with almost constant thickness of 0.75 m. The feeder flares toward the original ground surface in Unit 4. The thickness of the feeder increases from 2 m at the base of Unit 4 to 13 m at the original ground surface.

The feeder system in the outcrop A–N increases its thickness from ~0.5 m at the deepest portion to ~25 m at the original ground surface (Figure 7). The dyke thickness gradually increases between 15 and 10 m deep, then remains at 4–5 m between 10 and 5 m deep. The thickness increases rapidly toward the surface in Unit 4, within 5 m from the original ground surface.

The feeder system cropping out along the A–S section can be divided into marginal parts (rim) and inner parts (core) (Figure 8). The rim develops below 7.5 m from the original topographic surface, i.e., the ground level at the beginning of the 1809 eruption (see Figure 5). It develops continuously below 14 m from the original surface, whereas it is truncated in several parts between 7.5 and 14 m. The rim consists of coherent intrusive rock, without any significant evidence of fragmentation (Figures 8A,B). The vesicularity in the rim is 32–43%. The typical size of bubble ranges from 0.5 to 4 mm. The shape of bubbles is flat and elongated along the nearby dyke wall. On the dyke wall, the bubbles form sub–horizontal lineaments (~10°dipping to South).

The inner parts are filled with pyroclastic materials (basaltic bombs, lithic blocks and finer volcanic sand and ash). Interstices of the blocks are partially filled with lapilli and sands, which show sub–horizontal bedding (Figure 8C). Some hollows and empty spaces are also found in the inner part (Figure 8D).

The feeder dyke at the base of the outcrop A–N consists of coherent intrusive rock, without any clear evidence of fragmentation. The upward–flaring part within 15 m from the original ground surface is filled with pyroclastics.

In outcrops B–N and B–S, all the exposed parts of the feeder consist of diatreme filled with pyroclastic materials. The transition between dyke and diatreme is not exposed.

The sill developing in Unit 6′ crops out only on the walls of crater A–S. It exhibits a lens–like shape (4.5 m in length along the horizontal section) and a maximum thickness of 0.6 m (Figure 9A). Both edges of the sill show rounded shape inflating toward the surrounding volcanic ash layer. The sill also consists of a marginal and inner part. The marginal part consists of coherent intrusive rock with relatively high vesicularity (49–53%). The typical size of bubble ranges from 0.5 to 6 mm. The shape of bubbles is flat and elongated sub–parallel to the outline of the sill. The bubbles also show the oblique alignment against the nearby wall of the sill (Figure 9B).

The horizontal width of the feeder system is larger at the original topographic surface, due to its upward–flaring shape in each cross section (Figures 3, 5, 7). The width of the feeder is larger (~30 m) at the outcrop B–S (Crater B, see Figure 2), located at the center of the segment of the eruptive fissure here examined (~450 m in length, between 2440 and 2260 m asl), and decreases toward both ends of the structural segment (Figure 10A). At the base of each outcrop the dyke thickness is almost constant, varying from 0.4 m (A–N and B) and 0.5 m (A–S).

The thickness of Unit 3b and 3c, which represent the magmatic products of the fissure in the first half of the eruption, is also greater at the outcrop B–S and then decreases toward A–S and B–N in the margins of the eruptive fissure. The width of the feeder correlates with the thickness of the scoriaceous tephra (Unit 3b and c) at the rim of the feeder (Figure 10B).

The propagation of the eruption fissure during the 1809 eruption indicates the dyke was fed from the shallow portion of the central conduit beneath the summit Central Crater (Figure 1A) and intruded horizontally along the northern rift zone, as also occurred during the recent flank eruptions in the NE Rift (Garduño et al., 1997; Andronico et al., 2005; Neri et al., 2011). The strike (toward NNE) and dip (subvertical) of the feeder observed in the outcrops of the craters A and B reflect the entire structure of the feeding system in the NE Rift zone. The preferred orientation of the bubbles in the feeder dyke (Figure 9B) indicates the sub–horizontal movement of the magma in the shallow portion, which is consistent with the outward and centrifugal propagation of the eruptive fissure.

During the approach to the topographic surface, the feeder dyke was temporally halted and deflected at the base of the lava layer (Unit 5; see stage 2 in Figure 11). As the overlying Unit 5 consisting of massive lava has greater stiffness than the underlying scoriaceous deposit, this mechanical discontinuity acts as a temporary mechanical barrier for the dyke propagation (Gudmundsson, 2006, 2011). On increasing the internal magmatic pressure, the blocked dyke intruded along the base of Unit 5 and formed a small sill (Figures 5, 9A, 11).

The vertical dyke re-injected from the sill into the overlying Unit 5 when the internal magmatic pressure became high enough to break the overlying Unit 5 (stage 3 in Figure 11). The re-propagation of the dyke into the overlying layer caused a drop in the internal excess pressure of the dyke and, consequently may have stopped the horizontal growth of the sill. The development of the feeder parallel to the joints system developing in Unit 5 (Figure 5) suggests that the dyke intruded along a pre-existing fracture in Unit 5. The inclined propagation of the uppermost part of the dyke caused the oblique eruption that formed an asymmetric scoria cone around the crater (Unit 3b, c, and d).

The texture of the components of the feeder system and the ejecta indicates the fragmentation of magma started the moment the vent opened. The feeder dyke was temporarily halted and formed a sill at the base of Unit 5 before reaching the surface (Figure 12, stage 2). Therefore, the cooled margin at the tip of the sill is a trapped magma which preserves the texture of magma before the eruption. The cooled margin consists of coherent intrusive rock without any evidence of fragmentation (Figure 8B). This suggests that the magma intruded the very shallow part without fragmentation. High vesicularity in the cooled margin indicates that the ascending magma reached high bubble–content before the vent opening. The existence of the coherent dyke margin up to 7.5 m from the original topographic surface indicates that the magma reached the surface within 10 m in fluid form (non-fragmented) before the opening of the vent. On the contrary, all the products cropping out at the rim of the crater (Unit 3b) are pyroclasts (basaltic spatters and welded breccia) and no lava flow was formed (Figure 4). This means that magma reaching the ground surface was already fragmented and erupted as a spatter from the onset of the eruption. This also implies that the fragmentation in the conduit started the moment the vent opened at the surface.

The fragmentation of the magma can be caused by rapid decompression in the conduit by the opening of vent at the surface (e.g., Alidibirov and Dingwell, 1996). Rapid decompression accelerates magma ascent and may cause strain–induced fragmentation (e.g., Papale, 1998). Formation of the sill at the base of the stiff layer (lavas of Unit 5) indicates that magma had a certain excess pressure before the arrival of the dyke at the surface (Gudmundsson, 2011). Release of the pressurized magma at the ground surface caused an explosive emission of magma.

The ejection speed of the pyroclastic clasts represents the rising speed of magma at the vent. This ejection speed can be evaluated from the incline angle of the conduit and from the distribution of the ejecta, assuming the ballistic flight of the volcanic bombs. The relationship between the inclination of the conduit (75° toward East) and the maximum thickness of the ejecta (Unit 4b) at ~80 m from the eruption site indicates the mean ejection speed at ~40 ms⁻¹, neglecting the air resistance and slope of the ground. The ascent speed before the vent opening is typically several ms⁻¹ judging from the typical propagation speed of the eruption fissure at the NE Rift (Neri et al., 2011, and references therein). This implies that the magma ascent rate accelerated at the opening of the vent and this acceleration triggered the explosive emission of fragmented magma.

The upward–flaring shape of the feeder observed in the eruptive system of the 1809 eruption is commonly observed in explosive feeder systems (e.g., Keating et al., 2008; Geshi and Oikawa, 2014). The width of the feeder and the size of the eruptive crater at the surface are determined by the shape of the uppermost part of the feeder. Several mechanisms are proposed for the formation of upward flaring feeder systems (e.g., Valentine and Groves, 1996). Among them, erosion from particle collision and conduit wall collapse due to variations in magma pressure and/or shock waves are possible prime mechanisms for the feeders of the 1809 eruption, judging from the variation of width of the feeder reflected in the host rock lithofacies (Figure 8). In the outcrop A–S, the feeder begins increasing in width ~15 m deep from the original ground surface. The rapid increase of the width from the boundary between Unit 5 (lava) and overlying Unit 4 (pyroclasts) suggests that the feeder was enlarged particularly in the less consolidated host rock, since the mechanical erosion processes, such as particle collision and wall collapse, are more effective in the less consolidated host materials.

The positive correlation between the width of feeder system and the thickness of the tephra erupted (Figure 9B) also suggests the vigorous mechanical erosion by particle collision and wall collapse during the explosive emission of tephra. The greatest thickness of tephra at A–N and B–S outcrops indicates the larger magma flux at the center of the segment. The feeder was widened by the vigorous emission of pyroclastic materials during the fissure eruption. Conversely, the widening by the erosion was less effective in the peripheral zone (Figure 10A).

The existence of a partially–filled hollow at the center of the feeder (Figures 8C,D) and the change in eruption style recorded in chronicles and manifested in the erupted products (Figure 4), suggest the withdrawal of magma from the erupting conduit and the abrupt termination of the eruption in the zone of the craters A and B (Figure 12). The change of eruption style is testified by the erupted products exposed along the walls of the craters A (Figure 4) and B. The eruption style changed from the oblique lava fountaining, which produced Unit 3b and 3c, to the explosive activities and crater enlargements that formed the lithic–rich Unit 3d. The eruptive vent exposed at the outcrop A–S is covered with Unit 3d (Figure 5). This suggests that the eruption was terminated by the withdrawal of magma, as also testified by the feeder outcrop in A–S.

According to the historical chronicles (Gemmellaro, 1860), the propagation of the eruptive fracture system lasted about two days, from March 27 to 29, but the final step of this migration (from ~2400 to ~1400 m asl) occurred in a few hours, covering a distance of ~5500 m. Therefore, this migration lowered the eruptive vents by about 1000 meters with respect to the craters A and B (see Figure 12). This dramatic and sudden drop in the level of the magma would have brought the magma into contact with the main groundwater contained in the volcanic body, which is located just above the contact surface between the volcanic pile and its impermeable (clay–rich) basement. In the central zone of the NE Rift, the impermeable basement is at ~1200–1300 m asl (Branca and Ferrara, 2012; Siniscalchi et al., 2012). Therefore we assume the level of magma fragmentation may have interacted with the aquifer when the vents shifted from ~2400 to 1450–1325 m asl, generating phreatomagmatic (Surtseyan) activity along the higher portion of Fissure 3 (see Figure 12B). On the other hand, we cannot exclude that other shallow and discontinuous aquifers interacted with the magmatic dyke during its migration downward.

Similar magma–groundwater interactions were recently documented during the 2001 flank eruption on the southern slope of Mt. Etna (Behncke and Neri, 2003b; Calvari and Pinkerton, 2004), when the 2570 m eccentric vent was the site of powerful phreatomagmatic activity as the dyke cut through a shallow aquifer. This phenomenon did not occur during the 2002–2003 flank eruption, which affected the NE Rift (Andronico et al., 2005). In this case, the eruptive fractures propagated from the NE Crater area (~3300 m) downward to ~1800 asl, i.e., the magmatic dyke did not migrate deep enough to interact with the groundwater. Yet, eruptions of this type – rapid propagation of unusually long eruptive fissure systems – have occurred at Etna a few times per century, most recently in 1923, 1928, and 1981 (Acocella and Neri, 2003; Neri et al., 2011).

Scenarios where dykes show various types of interaction with aquifers can be envisaged in many basaltic volcanic settings worldwide, ranging from monogenic volcanic fields as Auckland (New Zealand) to large polygenic volcanoes, including the Hawaiian shield volcanoes. In particular, the rapid propagation and drainage of a dyke, as in the case of Etna's 1809 eruption, represents a potential hazard because of the enhanced possibility of a sudden and unexpected magma–groundwater interaction. Phreatomagmatic activity not only represents a threat to people within a certain distance from the eruptive vents, but it is also capable of producing abundant quantities of volcanic ash, which would affect much wider areas and potentially disrupt air traffic.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.