Abstract
Ambrym is one of the largest volcanic islands of the Vanuatu arc. It has been the focus of exclusively mafic volcanism and has a structure dominated by a central 13 km-diameter caldera. Contained within the caldera are two major cone complexes, Marum and Benbow, which have been the locus of most historic eruptions. Vents within these are constantly in a state of strong degassing, with visible lava lakes periodically being observed in several subcraters. Vulcanian and strombolian explosive eruptions occur at least yearly, along with larger subplinian events every 20–30 years. The active vent systems are enclosed by several 100 m-deep vertical-walled pit craters that expose cross-sectional views through the transition zone between the conduit and the crater. Units include coherent magma bodies with interbedded pyroclastic successions. One of the Marum craters, Niri Taten, exposes portions of solidified lava lakes, magma pods that fed spatter cones, small shallow-level intrusions and larger sills that connect through a complex network of dykes to the surface and/or into the pyroclastic edifice. These features show that shallow-level infiltration of degassed and low-viscosity melts into pyroclastic-deposit-dominated volcanic sequences of Marum plays an important role in the growth of scoria and spatter cones. Once solidified, the large intrusive bodies apparently provide important buttressing of pyroclastic cones, but during emplacement they may also cause cone collapse and lateral escape of magma to form lava flows.
This paper describes the pit crater morphology and exposures in crater walls at Ambrym in Vanuatu (SW Pacific) in order to elucidate the role that shallow-level dyke and sill formation plays in the growth and structural stability of long-lived mafic volcanic cone complexes. Here the uppermost ‘subvolcanic’ intrusive processes are the focus. They occur within a critical transition zone where it is not well understand why some intrusions stall at very shallow levels and others go on to feed eruptions or, through other environmental events (e.g. cone-wall collapse), escape to the surface as lava flows. We highlight the need for better understanding of crater successions to develop improved models of the shallow-level plumbing systems of these types of volcanoes. This is needed to understand volcanic cone/structure growth and volcanic eruption mechanisms in mafic volcanoes.
Pit craters are typically regarded as the result of collapse of sections of a volcanic construct over a cavity in the shallow subsurface stratigraphy (Okubo & Martel 1998). They may form when lava flows run over faults, or any depression. They may also result from the sudden lateral drainage of lava from a shallow intrusive body or lava lake near the central magmatic feeding system, such as during lateral flank eruptions (Rymer et al. 1998; Roche et al. 2001). The volcanic conduit referred to here is the shallow-level feeding system of a volcano. The near-surface portion of the conduit is the volcanic vent, which often has an aperture to the surface defined as a crater. Explosion craters are connected to the deep magmatic feeding system of the volcano and commonly formed by erosion during explosive eruptions. Long-lived volcanic cone complexes commonly have wide and complex crater features that may represent a combination of formation processes. Some subcraters may be filled by lava lakes. This leads to additional complexity in that within some large coalesced crater-like structures, collapse of solidified lava lakes may lead to the development of individual and often nested subpit craters.
Large cone complexes with wide and deep craters are excellent sites to study near-vent pyroclastic deposits and/or extrusive and intrusive bodies, particularly because the geometry and facies of the units is fairly unequivocal. Sections within crater walls show insights into important emplacement processes in and near vents that can elucidate the internal mechanisms and evolution of long-lived volcanic systems. In spite of the obvious location and natural outcrop of such sequences, inherent dangers to access and work within active craters has meant that little attention has been given to describe and interpret the crater-wall units. When described, exposed coherent lava bodies are mostly interpreted to represent a layer-cake like stratigraphy of lava flows or solidified lava lakes.
The crater wall exposures are probably created through a combination of subsidence of the crater floor and/or eruption-induced explosive excavation. Pit crater-wall sequences inside of craters of complex volcanic cones appear to document apparent alternations of pyroclastic and effusive rock units, they may also show the geometry and nature of eruption-feeding systems.
Volcanic craters may reach a few kilometres in diameter and the largest craters may closely resemble caldera features. In such cases, distinguishing wide craters from ‘true’ calderas can be ambiguous and has been the subject of controversy (Gudmundsson 1988; Gardner & Tait 2000; Lipman 2000; Browne & Gardner 2004; Kennedy et al. 2004). Interpretation of such wide craters and/or small calderas is complicated by erosional processes that may be involved in the widening of the original crater depression (Karátson et al. 1999). In large-volume andesitic stratovolcanoes, crater walls are predominantly composed of pyroclastic rock units interbedded with lava flows, as in the case of Vesuvius (Barberi et al. 1989). Such successions are typically exposed by explosive excavation of craters through plinian-style or vulcanian eruptions (Barberi et al. 1989; Bertagnini et al. 1991; Cioni et al. 1992). At mafic volcanoes, especially large-volume ocean island volcanoes such as Hawaii, larger, more coherent magmatic bodies are typically exposed in crater walls. Wide and deep craters can be excavated through phreatomagmatic explosions in the vent zone with associated phreatomagmatic tephra rings, such as the Deanakakoi or Uwekahuna Ash Member at Kilauea (McPhie et al. 1990; Dzurisin et al. 1995). Such disruption could section entire solidified lava lakes in half to be exposed in the newly formed crater walls. Magma–water interactions may also take place at deeper levels in the feeding system of a mafic volcano to form maar-like depressions in the summit zone of an otherwise dominantly effusive volcano (Kieffer & Creusoteon 1992). Aside from these explosive processes, crater-floor subsidence events appear to be the most common mechanism to form ‘pit-craters’ on mafic volcanoes. These are probably generated by the drainage of melt from the summit zone through lateral dykes to feed flank eruptions such as at Kilauea, Hawaii (Okubo & Martel 1998) or Masaya, Nicaragua (Rymer et al. 1998).
Here we demonstrate that crater walls of long-lived mafic explosive volcanoes commonly expose cross-sections of complex anastomosing networks of dyke and sill complexes, as well as lavas and pyroclastic deposits. Very shallow-level dyke and sill emplacement appears to correspond to feeder systems of lava flows. But, by contrast large ponded lava lakes in craters surrounded by pyroclastic walls may also be the source of lateral sills and dykes, as documented from Pliocene mafic volcanoes of Hungary (Martin & Németh 2004) or Peninsula tuff cone, California (Lavine & Aalto 2002).
The observations of crater-wall successions suggest that mafic volcanoes can store significant volumes of hot, degassed and low-viscosity melts in a highly porous pyroclastic rock–deposit-hosted confinement zone. Hence, the complex interactions of this intruded material with the host materials influence cone complex growth and perhaps short- and long-term stability of the cone. Such processes are increasingly being documented in association with scoria cones (Haller 2004; Németh & White 2006; Valentine et al. 2006) and inferred from larger stratocones (Tibaldi 2001, 2003).
Geological setting
Ambrym is located in the central part of the Vanuatu volcanic arc and comprises a triangular shaped, slightly east–west elongated island of approximately 45×30 km (Fig. 1). Ambrym eruptives are predominantly calc-alkaline basalts, including both high- and low-K variants (Gorton 1977; Picard et al. 1995). Few, if any, differentiated rocks occur. Analyses of more evolved rocks were reported by Picard et al. (1995), but revisiting of the sampling sites suggest that these samples were altered (Cronin & Németh 2005). A 13 km-diameter caldera forms the central part of the island and is believed to have formed about 2000 years ago (McCall et al. 1969). The formation of the caldera is currently under debate (Cronin & Németh 2005). It was initially considered to have formed during quiet and continuous subsidence (McCall et al. 1969) similar to many other mafic calderas (McCall 1963; Walker 1984, 1993; Lipman 2000). However, an enigmatic alternative was proposed by Robin et al. (1993), where huge phreatomagmatic explosions triggered caldera formation along with the development of a giant tuff cone and dacitic ignimbrites. Recent mapping and revisiting of key sites shows no evidence for the presence of deposits of the ‘Giant tuff cone’, no evidence of ignimbrites and no evidence whatsoever for large-volume tephra-producing plinian-style explosive eruptions in the geological record of the island (Cronin & Németh 2005). Hence, the initially proposed mechanism of effusive and relatively quiet caldera formation appears to be the most reasonable and the nature of volcanism since caldera inception has been consistent with this.
Overview map of the Vanuatu arc (a) and Ambrym Island (b). The names refer to the studied vents of the Marum volcanic complex.
Two large active cone complexes, Benbow and Marum (Fig. 2), are formed within the caldera, along with at least one other, now dormant, complex. In addition, a maar with c. 800 m-wide crater is located in the eastern side of the caldera. The bulk of the frequent historical activity has been centred at one or both of Benbow and Marum, although several flank eruptions have also occurred (Wiart 1995). Benbow (Figs 2c and 3) is a large, relatively simple cone structure, and only minor lava flows, sills and dykes are exposed in the pyroclastic deposit-dominated walls of the outermost crater. Smaller deeper-level craters may hold alternative successions but are not accessible. By contrast, Marum (Fig. 2b) is a somewhat more complex pyroclastic cone showing exposures of many thick successions of phreatomagmatic tephra and scoriaceous ash and lapilli beds with intercalated sills, dykes, lava sheets and ponds.
(a) Active vents in the Ambrym caldera. The picture was taken in July 2005 when a mild subplinian eruption took place from the Mbwelesu crater of the Marum volcanic complex. (b) Overview of the Marum volcanic complex. The broad tephra cone stands about 200 m above the ash plain of the Ambrym caldera. (c) Overview of the Benbow volcano, which is slightly smaller and has less complex architecture than the Marum volcanic complex. (d) Complex subsidence and explosion crater structure within the crater zone of Benbow. Note the tephra cover in the inner flat crater floor of the crater complex.
Aerial photograph of the Benbow crater system. Note the bench-like subsidence features forming a complex crater zone. The craters are surrounded by tephra rings covering the flat crater floor of Benbow.
Intense ‘passive’ degassing characterizes Ambrym activity during intereruptive periods. Measured SO2 fluxes of 180–270 kg s−1 in January 2005 constitute one of the world's highest rates of passive degassing, comparable with Mt Etna (Bani et al. in review). Normally, such high rates of degassing are associated with large basaltic eruptions or large lava lakes; Nyiragongo (Congo) had mean SO2 emission rates of 185 kg s−1 associated with its lava lake, calculated on the active period of the volcano between 7 October 2002 and 17 November 2003 (Carn 2004). Nyiragongo volcano is, however, on an intraplate rift zone, in comparison to a more common arc-basaltic system such as Ambrym. Ambrym has always been known as a large producer of gas, through its consistent production of acid rain that affects the crops of the local inhabitants. The ongoing gas loss occurs almost exclusively from the intra-caldera cone complexes of Marum and Benbow, indicating a steady locus for rise of magma at these locations.
Active cone complexes in the Ambrym caldera
Located near the centre of the caldera, the Marum cone complex stands about 400 m above the ash plain, reaching 1000–1100 m above sea level. Marum is comprised of an asymmetric outer tuff cone, elongated to the NW owing to the prevailing low-level trade winds (Figs 2a & b). Its outer flanks are mostly unvegetated, and are hence strongly eroded with a deep gully network. The inner flanks of the tuff cone are, by contrast, smooth, being draped by young pyroclastic deposit layers. The crater floor of the westernmost part of Marum is flat and also blanketed by young tephra that is commonly remobilized during heavy rain falls into the active, deeper easternmost craters. The currently active crater of the Marum cone complex, Mbwelesu (meaning ‘wild pig’), lies c. 700 m east of the highest rim of the tuff cone (Fig. 2a). This crater commonly hosts an active lava lake, and the latest Vulcanian and subplinian eruptions took place from Mbwelesu in July 2005. Immediately south of Mbwelesu, but still within the overall tuff cone, lies a younger crater Niri Mbwelesu (Tamo; elder son of Mbwelesu; Fig. 2a). This structure is surrounded by its own tephra ring; the inner walls of which expose a great succession of pyroclastic deposits, predominantly fine-grained fall and surge units, with rare interbedded spatter horizons. Niri Mbwelesu crater also commonly hosts a lava lake (Fig. 4a). On the outer flank of Marum, another new crater opened in 1988–1989, called Niri (Mbwelesu) Taten (younger son of Mbwelesu) (Fig. 4b). Niri Taten is a 300-m wide and c. 200 m-deep, vertical-walled pit crater. It is surrounded by a relatively coarse-grained, spatter cone that is slightly elongated to the west. This crater occasionally hosts a lava lake, and continuous degassing is common. Owing to the high and steep pit-crater walls around this vent, eruptive products are rarely able to leave the vent zone, and common recycling of pyroclasts occurs. More energetic explosive events, predominantly triggered by magma–water interaction, were able to drive out pyroclasts to form a low apron of ejecta a few hundred metres from the rim. Each crater of the Marum cone complex reaches up to 200 m in depth, and are typically floored by active lava lakes or open vents and, at times, colluvial fills of muddy breccia. On the wall of each crater and conduit in and around Marum are exposed a great variety of half-sections of lava lakes, lava ponds and shallow-level intrusions.
(a) Active lava lake in Mbwelesu crater in March 2002. Note the fresh tephra ring around the lava lake (arrow). (b) Oblique view of the Niri Taten crater. The crater is about 100 m deep and almost perpendicular.
Situated within the western side of the caldera, Benbow volcanic cone complex stands about 400 m above the surrounding ash plain (Fig. 2a and c). Benbow also comprises a large asymmetric tuff cone that is higher in the west (Fig. 2c). The cone complex encloses a 1500 m-wide crater, but its walls expose only pyroclastic deposits, with rare lava units (Fig. 3). Inside this ring, at approximatley, 100 m depth, is a bench-like complex crater zone (Fig. 3), within which two pit craters are inset by more than 200 m, hosting the currently active vents. These pit craters may have also been deepened by explosive processes, as attested by ash rings surrounding them. The vents produce continuous degassing, lava fountaining and mild Strombolian-style eruptions to form small intra-crater ash cones (Fig. 2d), as well as blanketing tephra layers deposited in the flat crater floor (Figs 2d and 3). In the innermost pit crater walls tabular, relatively thin lava flows are exposed that have no obvious connections to any lava outflows in the outer tuff cone flank (Fig. 3). This suggests that these flows are intra-crater lava flows that may have been connected to thicker, active lava lakes (Fig. 2d).
Other recent intra-caldera vents include outer cone flank fissures that have produced ‘aa’ lava flows, spreading up to 5 km within the caldera in 1986 and 1988–1989 (Monzier & Charley 1989). The 1988 flow lasted only 24 h and involved outpouring of approximately 106 m3 of lava. It occurred following the build-up of a large lava lake within Mbwelesu and Niri Bwelesu craters, and was also coincident with the inception of Niri Taten. The lava flow broke out from the southern flank of the Marum cone complex at an elevation equivalent to around half the depth of the Niri Mbwelesu crater (Fig. 2A and B). It appears that the magma escaped from the Niri Mbwelesu lava lake through the porous pyroclastic deposits of the outer walls of the crater.
Major effusive or pyroclastic eruptions have occurred on at least 20 occasions since 1820 (Wiart 1995). Whilst some of these have been confined to the caldera (e.g. 1986, 1988 and 1989), highly destructive flank lava flows or phreatomagmatic eruptions have also occurred in tandem with eruptions from the intra-caldera vents (e.g. in 1894, 1915, 1929 and 1951). These latter eruptions were apparently fed by lateral flow of degassed magma along dyke systems down the major east–west axis of the island.
Lava-lake and shallow-level lava-pond cross-sections
As described earlier, there has been a combination of explosive and effusive basaltic activity from the major pit-crater-hosted volcanic cone complexes, including an ephemeral presence of lava lakes. This alone should create a complex network of feeding and draining systems, as well as variable natured explosive deposits. In addition to this, however, the siphoning of lava from lakes to local flank vents (e.g. in 1988) or over longer distances down lateral flank rifts or dykes (e.g. in 1915) is likely to produce an additional range of overlapping deposits and structures. These are particularly evident in the more readily accessible sections of the Marum cone complex crater walls.
The inner crater-wall sections of Mbwelesu, and Marum, contain several stacked, more than 20 m-thick, columnar jointed, tabular lava bodies (Fig. 5). The lava bodies are massive, with 0.5–1 m-wide columns that are uniformly vertically jointed. No brecciated upper or lower layers were noted (Figs 5 and 6) and the overall bodies, where fully exposed, form a slightly dish-like architecture. All of these features indicate that these units derive from ponded lava lakes. In the western Marum crater wall the southern margins of these bodies show angular contacts with the former confining walls of the lava lakes, and minor brecciated zones a few metres thick occur (Fig. 5). The brecciated zones are better developed in the thinner extremities of the lava bodies and these have 0.1–1 m-diameter voids that are loosely filled with coarse, clast-supported breccias. Similar architecture is exposed on the northern Marum crater wall (Fig. 6), where the original angular lateral contacts of at least three lava-lake bodies are exposed against pyroclastic deposits. These wedge-shaped bodies taper towards the northwestern side of the exposure (left), but at their deepest on the east (right) lava has intruded in some places from the lakes into the pyroclastic deposits to form thin tapering sills (Fig. 6). The tapering sills have at times an irregular form that parallels the enclosing pyroclastic bedding. Of three successive lake-sill deposit sets (one with more than one thin sill leading from it across a faulted contact; Fig. 6), it appears that pyroclastic overburdens were, at most, 10–20 m thick at the time of their formation and that these may have been in a loose, compressible state. The sills are typically only 10% of the preserved lake-body thickness and they extend up to 50 m from the lake margins. These units are particularly well developed in the lowermost part of unit 1. All of the lake bodies show signs of drainage, excavation or gentle subsidence in their central portions. The margins of these lake bodies also show evidence of subsidence in relation to the former walls (of pyroclastics; Fig. 6). The tabular parts of the lava bodies (flows across adjacent crater floor) on the northern side of the crater are at a higher level than the central portions (Fig. 6), and units 2 and 3 shows signs of having drained from their original depths, with a former tide level on their eastern margins (Fig. 6). It is possible that the lake-drainage process and/or continuing crater-collapse processes caused tension on the marginal pyroclastic sequences, and thus allowed lateral flow of magma to form the thin tapering sills. Between each of these lake units, which were emplaced in successively broader craters, thin pyroclastic deposits occur, as well as crater-marginal colluvial deposits.
View of the complex architecture of the Marum crater zone looking towards the west. Note the multiple solidified lava lakes. Along the lava-lake margins small veneers of lava represent the ‘lava lake shoreline’ draped over the host tephra. Note also the explosion pit cut through the solidified lava lakes.
Architecture of the Mbwelesu crater zone. Along the solidified lava lakes small intrusions have penetrated the tephra wall that formerly hosted the lava lakes.
Phreatomagmatic eruptions and crater-deepening processes
In the Mbwelesu crater, the NE crater wall (Fig. 7), the upper contact of the uppermost coherent lava-lake body, is highly irregular, with 10–100 m-wide depressions and intervening pinnacles or spurs of material that project up to 10 m into the overlying breccia. This unit is directly overlain by a more than 30 m-thick, massive body of scoriaceous pyroclastic deposits, predominantly lapilli and bomb breccias. The complex irregular upper contact of the coherent lava body appears to be generated by loading, which implies that the breccia was emplaced rapidly over the lava lake while it was still fluid. The fluid lava was able to squeeze up into the growing pyroclastic deposit pile above the lava, forming effectively giant flame structures (Fig. 7). These textural features indicate that the lava-lake formation in this crater was accompanied by lava fountaining and/or mild Strombolian-style eruptions in one of the neighbouring craters. In this same face of the Mbwelesu crater (Fig. 7), additional explosive processes can be recognized that have mined out parts of the uppermost lava-lake unit to form a shattered cone of crater-fill breccia with associated pyroclastic deposit beds laid above the lava body.
Section through the Mbwelesu crater. Along solidified lava lakes, this section exposes a complex pit crater and explosion crater architecture. The dashed line circle highlights a fragmented zone of volcanic rocks inferred to be a pyroclastic-filled conduit of a small explosion pit. The full white line circle shows a zone of an intrusion squeezed into the scoriaceous agglomerate-filled zone. The irregular upper surface of the solidified lava lakes indicates that the lava lakes were still relatively mobile during the deposition of the pit-crater-fill tephra and the lava was able to squeeze into the unconsolidated sediment.
Aside from processes of formation and deformation of still-fluid lava lakes, post-cooling modification of these lava bodies appears common. Large vertical-walled pit-crater depressions, such as the one in the northwestern crater wall of Marum cone complex (Fig. 5), appear to have been mined by explosive disruption of already solidified intra-crater lava flows and/or lava lakes. Deposits mantling the outer flanks of the Marum cone complex are of two main type: (1) 0.1–0.5 m-thick, brown, extremely poorly sorted, ash-dominated beds with common lithic lapilli, as well as large angular block-sized clasts of dense solidified lava. The matrix of these units has a high fine-ash component and they are often vesiculated, implying an initially high water content. The planar and cross-bedded characteristics of the deposits are typical of near-source pyroclastic surge deposits (base surges) and tend to thicken into present erosion valleys. Interbedded with these units are: (2) black, moderately well-sorted and mantling lapilli deposits, commonly 0.3–1 m thick and shower bedded. The large block components of the type (1) beds are derived from already solidified lava lakes, and the bed structures suggest that the craters are primarily reamed out by regular phreatic or phreatomagmatic eruptions (Fig. 8). These phreatomagmatic explosions appear to occur after lava lakes become solidified in the active craters and may be the precursor for following Strombolian activity (producing the lapilli fall units), before degassed magma rises to form a new lava lake in the newly deepened crater.
(a) Large angular blocks of lava disrupted from older lava units of Niri Taten by explosive eruption. These blocks are hosted in vesicular tephra that is characteristic for base surge and phreatomagmatic fall. (b) Large angular blocks derived from solidified lava lakes around the Mbwelesu crater. (c) Cauliflower bombs around the Niri Mbwelesu crater indicating a magma–water interaction.
The phreatomagmatic/phreatic explosive origin of crater-opening processes is also shown by an apron of deposit making up the surface around the 1988–1989 excavated Niri Taten crater (Fig. 8a). In this unit, common angular volcanic lithic blocks are enclosed within fine vesicular-textured ash. This eruption preceded with development of a lava lake and at least two phases of lava-flow effusions from a vent a few 100 m to the west of Niri Taten during 1988 and 1989 (Monzier & Charley 1989). Episodic deepening of the Niri Taten crater is reported during these two years of activity and a lava lake still periodically appears in this crater as observed by the authors in March 2002 and January 2005. During parts of this eruption sequence, and notably before a short period of intense lava flow generation in 1988, the Niri Bwelesu crater contained a high-level lava lake, which was not present immediately after the event (Monzier & Charley 1989).
The generation of phreatic and phreatomagmatic explosions during recently observed vents in these craters (e.g. July 2005) is preceded by a lowering of lava in the conduit, followed by colluvial infill of the crater base (and partially crater-wall collapse). The pyroclastic colluvial fill often traps rainfall in this humid tropical climate to develop muddy lakes in the crater base (Figs 5 and 6). In addition, the crater floors lie approximately 150 m below the surrounding level of caldera fill, which may also provide a regular source of groundwater inflow into the vents feeding the ephemeral muddy lakes. The potential presence of a shallow sill network below the crater systems may act as an inwards-dipping aquitard, thus favouring groundwater flow into the crater areas. The rapid accumulation of water in the crater floors, along with blanketing fine-grained sediment, sets up the environment for intense explosions once magma rises again within each crater. This leads to 2–6 h-long phreatomagmatic explosive eruptions, generating fine-ash fallout (e.g. July 2005; Fig. 2a) and excavating the crater fill, as well as the country rock base.
Magma pods and spatter feeding systems
Solidified magma pods are exposed in cross-section in the western Niri Taten crater – conduit wall (Fig. 9). The two 10 m-high lava pods are connected with the same set of columnar joints, showing that they cooled together. However, with each pod there is an associated, stratigraphically separate, spatter horizon, suggesting individual spatter eruptions fed from each of the lava pods (Fig. 9). Following eruption from the lowermost pod, the crater rim was added to with further pyroclastic deposits from a nearby vent, and the pod was reused as part of a conduit system to feed the pod above. A branch from system of relatively thick dykes (3–5 m wide) feeds this system (Fig. 9). Within this same exposure, the uppermost lava pod connects directly to a subhorizontal lava body that mirrors the stratigraphy of pyroclastic beds. This structure has coherently vertical jointing, like the lava pod, and has no basal or top brecciated zones. These features suggest that the unit is a very shallow sill structure, which was emplaced just a 1 m below the surface, probably during the spatter eruption.
Overview of a lava-pond and lava-spatter complex from the conduit wall of the Niri Taten vent. Note the fine-grained, probably phreatomagmatic, tephra exposed in the crater wall. An irregular marginal sill is fed from the lava pod and post-dates the eruption of the lava spatter.
Sills and dyke systems
The largest lava bodies exposed in these craters, as described earlier, have been lava lakes and flows. Some of these have small lateral sills, where lava has leaked into the surrounding pyroclastic deposit walls (see the previous discussion), but the feeding systems of these larger units are rarely exposed. In the upper pyroclastic deposit structures of the Marum and Benbow craters isolated thin irregular spiderweb-like dykes occur, typically less than 1 m in width. These structures do not show peperitic contact zones and do not typically lead to major surface eruption structures (Figs 6 and 7).
By contrast, deeper level dyke and sill systems of greater complexity are exposed in the ESE walls of the Niri Mbwelesu crater (Fig. 10). In this case, as in the Niri Taten walls, the major feeding dykes are about 1–3 m in width. Several sills are interconnected through oblique dykes with irregular, chilled margin. There are occasional peperitic contact zones, especially where dykes are in contact with fine-ash and lapilli successions (Fig. 10a). Many of the sills and the overall major sill complex in the Niri Mbwelesu crater are dish-like in cross-section. The largest sill complex is approximately 20 m thick and strongly columnar jointed. Many of the sills have joint patterns in radial, curving array. The thinner interconnecting sills form strongly U-shaped cross-sections, in some cases following the bedding of the pyroclastic deposits, in other cases suggesting a temporary inflation during their intrusion followed by settling during magma withdrawal (Fig. 10). The master sill complex has no obvious connection to any surface outflows in the eastern outer flanks of Niri Mbwelesu, and similarly no equivalent structures exposed in the opposite western crater wall. This appears to show that they were probably sourced from a point east of the present crater of Niri Mbwelesu, which lies effectively in the centre of the crater. The sill complex is about 50 m below the top of the crater wall, and covered in some cases by spatter, which may be related to this feeding system. Like the Niri Taten example, this complex was emplaced at very shallow levels of the volcanic edifice either by lateral intrusion from an extremely high lava-lake level, or more likely by intrusion subvertically from the vent area below. Similar, but less common individual U-shaped thin (m-scale) sills are also exposed in the southern side of the crater wall, which are independent of the major sill and dyke complex of Niri Mbwelesu and relating to spatter – agglutinate horizons (Fig. 10). These sills are platy textured, and strongly chilled.
(a) A complex network of shallow intrusions in the crater wall of Niri Mbwelesu. The sills(s) are dish-like in cross-section, and dykes commonly develop peperitic margins, especially in contact with fine-ash beds. Note the thick succession of fine grained, base surge and phreatomagmatic-fall-dominated succession intruded by the sill complex. (b) The western crater wall of the Niri Mwelesu exposes a great succession of pyroclastic deposits occasionally intruded by thin sills that are dish-like shape in cross-section.
The presence of, particularly, the large intrusive bodies within the walls of the Marum cone complex buttress the structure where they are present and contribute in part to its relative stability, compared to eroding parts of the cone comprised of pyroclastic deposits. Intrusion of these bodies, however, would have caused major changes to the internal friction angles and stability of the cone margins while the magma was still in a fluid state. In addition to lateral breakouts through sill structures to form flank lava flows (as interpreted for the 1988 events), wholesale collapse of cone sides is a possibility under these circumstances. This process could result in similar lateral spreading and breakouts of lava (causing partial collapse of the overriding cone) to that described from rapidly growing spatter cones during the 1986 eruption of the Izu-Oshima volcano (Sumner 1998).
Composite crater development
The craters of the active intra-caldera volcanic cone complexes of Ambrym have a composite origin that involves two main processes:
Magmatic drainage in shallow subsurface sill systems, either via flank eruptions of lava (e.g. the 1988–1989 Niri Taten-related lava flows) or the lowering of active magma levels in the shallow levels of the conduit system; and
Excavation through phreatic and phreatomagmatic explosions occurring in cycles associated with lava retreat and rise in conduits. Lava retreat is followed by colluvial infill and water accumulation. This confinement of rising magma in the conduit by wet sediment leads to explosive vent clearance and mining of country rock, progressively deepening crater levels.
Process (1) appears to be the most important for establishing the initial pit crater structure, while (2) contributes significantly to its ongoing evolution. The results of (2) are the excellent exposures through exhumed lava-lake and dyke – sill structures described in this paper. Step-like repeated drainage of magma from the central feeding system appears to create the main architecture of nested collapse features in the main craters of Marum and Benbow. The presence of tuff-rings and aprons surrounding these features show that phreatomagmatic explosions may be equally important in their development, although owing to the mechanism by which this process works a depression needs to be created first.
The dimensions of the pit craters of Ambrym are similar to those described from Masaya, Nicaragua (Rymer et al. 1998). Moreover, the style of volcanism and its frequency and magnitude are also similar (Rymer et al. 1998). Continuous SO2 and CO2 emission are also a feature of Masaya, although perhaps not to the same magnitude as Ambrym (Baxter et al. 1982; Allen et al. 2002; Duffell et al. 2003). These features at Masaya are interpreted to be the direct result of a very shallow magmatic feeding system that pulsates in height, and rapidly switches location in the upper volcanic edifice (Rymer et al. 1998). The geophysical model for Masaya suggested a complex network of very shallow (hundreds of metres deep) cavities, some filled by gas-charged foamy magma occupying the edifice summit (Rymer et al. 1998). On Ambrym, the field descriptions and observations, alongside Bani et al.'s (in review) documentation of high gas emission rates and frequent small volume mafic explosive eruptions, indicate a similar shallow feeding system to that at Masaya. The presence of sills and dykes at extremely shallow levels, especially at Niri Mbwelesu and Niri Taten, support the idea of high-level magma infiltration into the otherwise unconsolidated pyroclastic deposits of the Marum–Benbow complexes. As these are at a considerably greater height than the surrounding caldera area, it appears that the magma pathways are relatively stable, intruding melts to high levels in the pyroclastic cones, at times 200–300 m above the surrounding caldera surface.
The onset of the large complex crater, pit-crater structures in Marum and Benbow on Ambrym are probably due to purely magmatic processes, similar to the example described above of the smaller scale Niri Taten crater formation during the 1988–1989 lava flows. A number of large-scale historical flank eruptions have occurred, associated initially with eruptions in intra-caldera craters (e.g. 1894, 1915: Purey-Cust 1896; Frater 1917; Gregory 1917). This pattern implies magma rise in the intra-caldera areas, followed by lateral drainage of degassed magma along the east–west rift system to the flank eruption sites, which are up to 20 km distant. These large flank eruptions may cause enough volume removal to each to establish a new pit crater or considerably increase the size of an existing one. Subsequent modification of these craters is probably through repeated phreatomagmatic and phreatic explosions as described above, although clearly these are most active once a major depression has already been formed.
Once the pit craters were formed in Marum and Benbow, they appear to be long-lived structures, maintained, and kept clear by ongoing phreatic and phreatomagmatic explosions. During the period of historical observations, there have been new craters develop (e.g. Niri Mbwelesu and Niri Taten), but the older ones have remained. At the Marum complex, there is also a clear progression of increasing size with increasing age from the Niri Taten through to the craters of the original Marum cone complex. This shows the effect of ongoing collapse (magmatic) and explosive (phreatomagmatic) processes on the landform once established.
The role of hydrogeology on the ongoing eruption mechanisms at Ambrym may be similar to that documented from Piton de la Fournaise, Reunion (Join et al. 2005). Piton de la Fournaise as a complex shield volcano, and groundwater flow is driven by the distribution of extensive lava layers that encapsulate tephra units (Join et al. 2005). Such architecture is expected in Ambrym, and could drive rainwater to the central part of the volcanic edifice where lenses of groundwater may form and facilitate magma–water interaction in a case where a dyke encounters such zones. Certainly, once the pit craters have excavated to a depth below the caldera surface at Ambrym, groundwater inflow probably adds significantly to the already high input of water through rainfall. Groundwater inflow may also be exacerbated by a basal aquatard layer that funnels flow towards the crater region, formed by cooled sills or buried lava flows, such as inferred from Masaya on the basis of geophysical measurements (Rymer et al. 1998).
Conclusion
Major intra-caldera edifices have developed on Ambrym volcano through several hundred years of effusive and explosive activity. The relatively large composite edifices of Marum and Benbow indicate stability in the source of magma supply. That large pit craters have also developed in these structures appears to be a function of large-scale drainage of magma from its initial intra-caldera central reservoir to feed major flank eruptions, such as in 1894 and 1915. Once these pit craters form, an interplay of processes between the cyclic rises and falls in magma levels in the upper conduits and influx of ground- and rainwater to the craters causes repeated phreatomagmatic and phreatic explosions. As magma levels in conduits fall, colluvium fills the crater base and becomes saturated to form muddy lakes. Magma rising again, encounters these fills and highly explosive phreatic and phreatomagmatic eruptions ensue, clearing the vent and excavating the base of the crater into country rock. Occasionally, magma levels rise to the degree that large lava lakes are formed. These either cool, and solidify, and partially drain through formation of lateral sills, or fully drain if the lava is able to leak laterally to form flank lava-flow outbursts (such as in 1988–1989).
A notable feature of the Marum system craters is the presence at an extremely shallow level of interconnected sill and dyke complexes and lava ponds associated with lava spatter vents. This suggests that shallow intrusive events are common during construction of edifices that outwardly appear to be pyroclastic-deposit-dominated. The presence of such shallow sills and dykes are important structural considerations in the growth and short- and long-term stability of mafic cone complexes. Once cooled, they perform an important buttressing function, but when emplaced these ‘hidden’ melt pockets may escape through the pyroclastic deposit wall to initiate unexpected lava flows. This process may also lead to destabilization of portions of the edifice. In addition, large ponded magmatic bodies can retain heat for long periods, and may provide preferential ‘pre-heated’ pathways for new, fresh melts to reach the surface.
Acknowledgments
This work has been supported by the New Zealand FRST Post-doctoral research grant (K. Németh) (MAUX0405) and FRST-PGST funding – MAUXO401 (S. J. Cronin). We thank D. Charley, E. Garae, and the people of Lalinda and Ranon villages for assistance in carrying out the fieldwork. Helpful reviews by A. Dini and S. Self are also thankfully acknowledged.
- © The Geological Society of London 2008
References
Volcanic craters, pit craters and high-level magma-feeding systems of a mafic island-arc volcano: Ambrym, Vanuatu, South Pacific
- Article
- Abstract
- Geological setting
- Active cone complexes in the Ambrym caldera
- Lava-lake and shallow-level lava-pond cross-sections
- Phreatomagmatic eruptions and crater-deepening processes
- Magma pods and spatter feeding systems
- Sills and dyke systems
- Composite crater development
- Conclusion
- Acknowledgments
- References
- Figures & Data
- Info & Metrics