drivers of explosivity and elevated hazard in basaltic fissure eruptions: the 1913 eruption of...

Journal of Volcanology and Geothermal Research 201 (2011) 194–209

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research

j ourna l homepage: www.e lsev ie r.com/ locate / jvo lgeores

Drivers of explosivity and elevated hazard in basaltic fissure eruptions: The 1913eruption of Ambrym Volcano, Vanuatu (SW-Pacific)

Károly Németh ⁎, Shane J. CroninMassey University, Volcanic Risk Solutions, Institute of Natural Resources, P.O. Box 11 222, Palmerston North, New Zealand

⁎ Corresponding author. Tel.: +64 6 3569099 fax: +E-mail address: [email protected] (K. Németh

0377-0273/$ – see front matter © 2010 Elsevier B.V. Adoi:10.1016/j.jvolgeores.2010.12.007

a b s t r a c t
a r t i c l e i n f o
Article history:Received 19 July 2010Accepted 9 December 2010Available online 22 December 2010

Keywords:tuff ringphreatomagmaticmaarscoriapyroclastic density currentfissure

Fissure-eruptions along linear structures can extend for several tens of kilometres with distinct separatemanifestations of volcanism along their length. They typically involve low-viscosity mafic magmas forminglong lava flows and cinder cones. Eruptions in 1894 and 1913 on Ambrym volcano, Vanuatu, showed howthese mildly explosive eruptions can rapidly transform into violent explosive events, producing significanthazard and widespread volcanic ash clouds. During the 1913 episode, a fissure began in the central calderaand basaltic magmas broke out in a series of locations down the island's western flank. In all sites over 100 min elevation, fissure outbreaks produced vigorous lava fountains and highly fluid lava flows that travelledrapidly to the shoreline. When the outbreaks propagated along the island's axis into coastal plain areas, aclimactic series of explosive eruptions occurred, producing a 1.2 km long by 600 mwidemaar and tephra ring.A further small tuff ring was formed later, creating a temporary island 400 m offshore. The onshore tephraring destroyed a hospital and associated buildings. Its last evacuating occupants were close witnesses to theeruption processes. Deposits exposed in the lower portion of the tephra ring show that this part of theeruption began with a mild phreatomagmatic explosive eruption from a narrow vent, followed by amagmaticscoria-producing phase. Subsequently a complex sequence of highly explosive phreatomagmatic eruptionsoccurred, producing pyroclastic surges, along with repeated distinctive breccia-horizons, rich in coral and lavacountry rock. These features tally with eye-witness accounts to indicate that the main eruption phase wasproduced by a periodically shifting locus of phreatomagmatic fragmentation and eruption along a single E-Wfissure. The glassy and vesicle-poor pyroclasts produced during this eruption phase were dominantlyfragmented in a brittle manner by magma water interaction. Low volatile content of the magma uponfragmentation is confirmed by FTIR analysis showing b0.5% H2O in chilled glass. These findings highlight thata degassed, mafic, fissure-fed eruption can under certain circumstances pose a major volcanic hazard if dykesintersect substrates with abundant available water.

64 6 3505632.).

ll rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Arguably the greatest hazard to life on basaltic volcanic islands isfrom explosive phreatomagmatic eruptions (Németh and Cronin,2009a). While rare, large-scale caldera-centred eruptions may occur(Robin et al., 1993), more common large-scale events are fissure-eruptions. Fissure eruptions are known in associationwith wide rangeof composition from rhyolite (Sutton et al., 2000; Nairn et al., 2001;Speed et al., 2002; Spinks et al., 2005; Wilson et al., 2006; Darragh etal., 2006; Gravley et al., 2007) to basaltic (Stothers et al., 1986;Thordarson and Self, 1993; Thordarson et al., 1996; Thordarson andLarsen, 2007). Many such fissure eruptions, involve some phreato-

magmatic explosive phases (e.g. Laki fissure eruption in Iceland:Thordarson and Self, 1993), but due to their remote locations theirimmediate hazard has not been well studied.

Many basaltic volcanic islands display strong axial rift structures,e.g., Lanzarote (Carracedo et al., 1992) or Taveuni (Cronin and Neall,2001), where magmas have risen in long dykes to focus belownumerous vents along an alignment, or en echelon set of fissures thatcan be tens of kilometres long. While this differs in principle fromdeep, mantle-derived magmas rising at isolated monogenetic volca-noes (Connor and Conway, 2000), at each eruption site along a fissuresystem, the volcanic styles and scales can be highly comparable. Withlow-viscosity arc basalts, such as those on several of the Vanuatuislands (Robin et al., 1993; Picard et al., 1995; Peate et al., 1997;Turner et al., 1999; Raos and Crawford, 2004), the hazard is linkedprimarily to explosivity, and depends mainly on environmentalconditions encountered by vertically and horizontally propagatingdykes and fissures and the degree to which point-focussing of magmadischarge occurs.

http://dx.doi.org/10.1016/j.jvolgeores.2010.12.007

mailto:[email protected]

http://dx.doi.org/10.1016/j.jvolgeores.2010.12.007

http://www.sciencedirect.com/science/journal/03770273

195K. Németh, S.J. Cronin / Journal of Volcanology and Geothermal Research 201 (2011) 194–209

Phreatomagmatism along fissure systems appears to be moststrongly controlled by the availability and recharge of external water,and the competence of the country rock, in a similar way to a diatremefed maar and tuff ring eruptions (Lorenz, 1986). On the islands ofAmbrym, Ambae, and Lopevi in central Vanuatu (Fig. 1A), the largestof historical eruptions have involved the rise of mafic magmas alongfissure systems of up to 20 km in length. Highly fluid lava flows andlava-fountains coexisted with sustained phreatomagmatic explosiveeruptions along the same alignment. This implies very low-internalgas contents and little magmatic-gas control on eruption explosivityand fragmentation. Here we investigate the conditions controlling themost explosive styles exhibited along basaltic fissure eruptions onvolcanic islands, by reanalysing the eyewitness accounts andexplosively emplaced deposits of the largest known recent explosiveeruption on Ambrym Island, Vanuatu in 1913.

The western flanks of Ambrym (Figs. 1B, C, D and E) haveexperienced several generations of fissure eruptions, including twomajor eruption sequences in 1894 (Purey-Cust, 1896) and 1913(Gregory, 1917; McCall et al., 1969). The December 1913 eruptionoccurred along a rift system that extended for over 19 km, formingtwo major lava sheets, at least five long lava flows into the sea,numerous spatter cones and at least one elongated tuff ring (Gregory,1917) (Fig. 1B). This phreatomagmatic eruption completely destroyeda large hospital/missionary headquarters within minutes of its onset,with the inhabitants narrowly escaping by sea. The crater was initiallyexcavated over 16 m below sea level and formed a lagoon within its c.1.2 km length and 0.6 kmwidth. Subsequent coastal modification anderosion of the tuff has cut off access to the sea to form a closeddepression filled by a brackish lake. An eruption a further 400 m-offshore occurred soon after this vent inception (Gregory, 1917),although it appears that the landform and major part of depositsassociated with this have since been eroded by wave action.

Erosion of the tephra ring during and since the eruption formed abroad, flat-lying apron of volcaniclastic deposits surrounding themainstructure (Németh and Cronin, 2007). An exceptionally strongtropical cyclone season in AD 2000 caused coastal erosion to clear anear-continuous 1.5 km section of the elongated tuff-ring, allowingclose examination of the proximal to distal pyroclastic facies of thedeposits contained within (Fig. 1C). These, in turn, provide an insightto the nature of explosive eruption processes on such fissure-formedphreatomagmatic eruptions, along with understanding gained fromstudies of other historic maar eruptions, e.g. of the AD 1977 Ukinrekmaar in Alaska (Kienle et al., 1980; Self et al., 1980; Büchel and Lorenz,1993; Ort et al., 2000; Pirrung et al., 2008).

Rift-edge volcanism is well-known on ocean islands in manytectonic settings, such as Hawaii (USA), Ambae (Vanuatu), Izu-Oshima (Japan), Tenerife (Spain) and Jeju (Korea). Detailed study ofthe volcanic structures and the processes that generated them are,however, comparatively rare (Sohn et al., 2008; Németh and Cronin,2009a). The volcanism is commonly phreatomagmatic and thedeposits form the foundation for the lateral growth of volcanicislands (Cole et al., 2001; Németh and Cronin, 2009a). Being at theinterface of land and sea means its activity poses volcanic hazards toadjacent coastal communities. Such volcanism in many parts of theSW Pacific is recorded in oral tradition (Nunn, 2003; Németh andCronin, 2009b).

2. Geological setting and morphology of the tephra ting

Ambrym has one of the highest magma-production rates along theVanuatu arc (Fig. 1A) and consists of a 35 by 50 km slightly east–westelongated triangular island with a central caldera system (McCallet al., 1969; Carney et al., 1985; Robin et al., 1993). Extending east andwest from the caldera, the island is cross-cut by an active fissure zonecontaining many scoria cones and fissure-fed lava flows. The northernpart of the volcano is considered the oldest and is inferred to be part of

an old shield volcano including old fissure vents (McCall et al, 1969).Within the central 12 km-wide caldera at about 800 m above sealevel, two active volcanoes are located, Marum and Benbow. Theseboth have multiple pit craters that are frequently active, producingsmall-volume Strombolian, phreatomagmatic and sub-Plinian erup-tions feeding extensive sedimentary systems within and outside thecaldera (Németh et al., 2009). It has been proposed that about2200 years ago a cataclysmic phreatomagmatic eruption lead to theformation of a giant tuff ring and subsequently to the formation of amafic caldera (Robin et al., 1993). The formation process of the calderaremains ambiguous, due to the poor exposure of deposits definitivelylinked to a giant tuff cone or a caldera-forming eruption.

In the western and eastern edges of Ambrym, many tuff rings havedeveloped at or near sea level, apparently as a result of interaction offissure-fed magma and near-surface water and/or water saturatedsediments. These typically display 1000 m-diameter craters with low(b100 m) rims. Many of these phreatomagmatic volcanoes erupted inrecent pre-historic and historic times. Among the eruption records, themost recent and violent eruptive tephra ring eruptions took place in theyears of 1896 and 1913 at the western edge of Ambrym (Purey-Cust,1896; Frater, 1917;Gregory, 1917). InwestAmbrym, a cliff section at sealevel exposes phreatomagmatic tephra units up to ~15 m thick (Fig. 1B).The tephra ring surrounds an oval depression, a shallow water filledmaar (Fig. 1B). The highest point of the tephra ring is ~85 m above sealevel at its northern margin, and its form is best preserved along thewestern and northern flanks (Fig. 1B and C). The outer flanks slopegently (b10°) outward, with units to the east mantling lava flows andolder scoria cones (Figs. 1C, D and E). The tephra ring is partially open tothe north and temporary re-connection to the sea occurs after strongwave action during cyclones (Fig. 1B). The tephra ring passes graduallywestward into a volcaniclastic debris fan, thinning out over severalhundred metres. The tephra ring was formed in a low-lying plain thatwas possibly the floor of an older volcanic crater (Frater, 1917). Inaddition, a concurrent eruption, centred c. 400 m offshore may havecontributed to the deposits. From both sources, shallow beds of coralwere disrupted and coral fragments are dispersed throughout parts ofthe sequence. The entire structure is nowcoveredby tropical shrubs andtrees. The wide crater and low surrounding rim are similar to youngmonogenetic tephra/tuff rings worldwide. Tephra is preserved up toapproximately 1 km outward from the crater rim.

Within amonth of the eruptions water depth in the deepest part ofthe lagoon in the centre of the tuff ring was measured at 15 m,indicating between 20 and 30 m of crater floor subsidence/excavationthrough the eruption, which is similar to many maar volcanoes(Lorenz, 1986). Surveys several months later revealed the crater floorwas up to 37 m deep, indicating ongoing post-eruptive subsidence ofthe maar crater.

3. AD 1894 and 1913 Ambrym eruptions

3.1. October 1894 rift-edge eruptions

At 0600 hrs on 16 October 1894 an ash cloud erupted from thecentral vent of Benbow and covered much of the western flanks of theisland. Soon afterward, lava flows commenced toward the sea. A“dense blackwall” of ashwas observed from the top of an old tuff ring,(Dip Point) in the western edge of Ambrym (Purey-Cust, 1896).Scoriaceous coarse ash was deposited across most of westernAmbrym. The eruption was accompanied by strong volcanotectonicearthquakes with associated landslides, including the western shorecliffs. Lava entered the sea in many locations along the NW coast.Initial contact of lava with the sea produced no explosion, although afew seconds later, littoral explosions began, generating extensivesteam-charged jets of lava fragments, radiating ocean surface waves,and over-riding steam clouds. The scale of the central-vent activityappears to have been similar to that observed during a series of central

196 K. Németh, S.J. Cronin / Journal of Volcanology and Geothermal Research 201 (2011) 194–209


vent eruptions on Ambrym during 2005 (Németh and Cronin, 2008).The 1894 eruption differed from this most recent activity, byproducing lava flows along the rift axis, which ponded in severallocations (Purey-Cust, 1896).

3.2. The December 1913 eruption

At around 1700 hrs on the 6th of December 1913, an eruptionbegan in the central caldera area in a similar manner to the 1894events (Marshall, 1915; Gregory, 1917). This eruption was observedfrom a number of perspectives on Ambrym and from neighbouringislands. A very close-observation of the inception of the climacticexplosive phase was made as a new vent opened up within thegrounds of a Mission Station and Hospital at the westernmost edge ofthe island (Fig. 1B, D and E). The eruption changed the coastline ofwest Ambrym, creating a new lagoon and extending the coast by up to800 m surrounding this (Gregory, 1917). The Mission Station andHospital was built within an older tuff ring and many other older tuffrings are located nearby, indicating that the 1913 events represent anexample of a process common at the margins of the rift-like structurecrossing through Ambrym.

A very large earthquake one month before the eruption wasfollowed by increasingly frequent precursory seismicity that wasstrongly felt across the entire western flanks of the island prior tomagma outbreak. The initial manifestation of this eruption was in thecentral caldera area, focused on Benbow (Fig. 1E) (Frater, 1917). At1700 hrs the first explosions were seen from the central caldera,generating a high eruption column of steam and ash. Sporadic newexplosions with dark ash clouds burst through the steam on intervalsdropping from 3 to 1 min. Within an hour of the eruption onset, apersistent sub-Plinian eruption column had developed (Marshall,1915). This cloaked thewestern part of the island in a dense ash cloud.Within the next few hours a series of satellite lava and spatter vents,opened progressively westward of Benbow along the upper flanks(Fig. 1E). Lava streamed down toward Port Vatu and Baulap, with thelatter flow reaching the sea at 2100 hrs and flowing for another dayafterwards (Marshall, 1915). From the neighbouring island of Paama,it was estimated that a new vent appeared every few kilometres alongthe western ridge of Ambrym (Fig. 1E).

The greatest magma output was concentrated at six closely-locatedvents immediately outside thewestern edge of the caldera (Figs. 1D andE). Another area of intense fissure eruptions occurred in the Meltunganarea, forming large lava streams (Fig. 1E). Lava from these craters andfissures along the rift fed eight lava flows of 1.6–16 km-length and 90 to1200mwidth,many reaching the sea during the first night. A flow fromthe Meltungan area began at 2300 hrs and reached the sea at KrongPoint at 0200 hrs on 7 December, with another following shortlyafterwards fartherwestward (Fig. 1E). These lavaflows carried rock, soiland large uprooted trees on their surfaces, with one c. 180 m-wide flowon flat ground estimated to be bulldozing through tall forest at c. 6.5 to8 km/h (Marshall, 1915; Frater, 1917). Upon entry to the sea, substantiallittoral explosions occurred, generating high steam columns, withincandescent pyroclastic explosions seen at night. Lava flows were ofboth pahoehoe and aa type, many also forming narrow lava channelsand tubes. In a major area of lava fissure formation at Meltungan, lavafountains were active for almost a week, with lava flows being mostvigorous for the first few days of the eruption.

Fig. 1. Overviewmaps of Ambrym Island, Vanuatu. A) Location of Ambrym in relation to otheto the western edge of Ambrym. Arrow points to the Hospital-tuff ring. Letters correspondcorrespond to the locations of the stratigraphic logs. Thin arrow points to Dip Point, while tcrater and the sea, D) Contour map of Ambrym with the location of its main central vents anorth Ambrym eruption sites, E) Google Earthmap view of the area effected by the 1913 erupto “g” represent the time sequence of volcanic events after Marshall (1915). Small black stayoung judging from their vegetation cover andmorphology, while light (yellow on the onlineinferred to be slightly older then the SW vents due to their denser vegetation cover and mHospital-tuff ring, while large star with “g” is the likely position of a Surtseyan tuff cone form

Around 12 hours after the initial outbreak, in the early morning ofthe 7th of December, new vents opened near the western coast of theisland, immediately inland of the hospital site (Fig. 1E). A few hourslater, and during evacuation of the hospital, explosions beganexcavating the side of a hill c. 400 m inland of the buildings and by1000 hrs, the hill had all but disappeared and a lava stream began toflow from the hill and through the hospital grounds (Marshall, 1915).Within 30 minutes, evacuees on the sea witnessed the suddenformation of a new vent within the centre of the hospital grounds(Fig. 2A), as “electric flame flashed from the ground, and immediatelyfollowing was a mighty roar and a crash” (Marshall, 1915). In theintense heat, building materials disintegrated rapidly. All buildingswere incorporated into a vertical jet of steam and pyroclasts that roseN6 km within 1 minute (Fig. 2B and C) (Frater, 1917). A major phaseof ash production began at this point (Fig. 2C), with wet ash plasteringsurrounding islands and even a ship at sea several hundred kilometresaway on its passage between Sydney and Fiji. For four days theeruption continued with short pauses occurring sporadically (Mar-shall, 1915). The elongated form of the resultant tuff ring indicatesthat a migration of the eruption focus occurred along a westward-propagating fissure system (Fig. 2D).

After the Hospital eruption, at around 1500 hrs on the 7thDecember (Marshall, 1915), a vent opened beneath the sea, c.1.6 km from the hospital in ca. 45 m of water, building a tuff ringthat within 16 hours bridged the 400 m to the coast of Ambrym. Theseexplosive eruptions continued for another four days (Marshall, 1915)and hampered the operation of radios from a ship anchored nearby(Frater, 1917). A major lahar also occurred, down one of the lava flowpaths near Craig Cove, over a 6 hour period on 9th December. Thereare no eye-witness accounts for the generation mechanism for thelahar, but it is inferred to have been triggered by the destabilisation ofloose ash caused by heavy rain on the upper flanks of Ambrym.

Following a three week hiatus, activity was renewed on 1 January1914 with fissure eruptions south of Marum in the central caldera(Fig. 1D). These eventually produced another tube-fed lava flow thatdescended the slopes to Port Vatu (Fig. 1D). Several months after therift-edge eruptions, surveyors were surprised to find their soundingweights melted and recorded small submarine explosions sporadi-cally offshore of the Hospital tephra ring.

The AD 1913 eruption appears to have been typical of many of theprehistoric eruptions on the island. Its initiation at the central vents inthe summit caldera of the island was followed by fissure eruptionswith lava flows and spatter cones forming along the axis of the E-Wtrending rift, and concluded with phreatomagmatic explosive erup-tions at the coastal margins, along with some littoral explosionswhere lava entered the sea. The tuff-ring and a small island wereproduced during this event, although the latter has since erodedaway. It may have been eroded in only months after the eruption,judging from maps made soon thereafter (Marshall, 1915).

4. Tephra ring succession

The northern margin of the Hospital tuff ring at west Ambrym iscut to expose a succession of phreatomagmatic tephra units up to15 m thick. The succession can be subdivided into 4 majorstratigraphic units (Figs. 3A–C and D).

r islands in the Vanuatu archipelago, B) Google Earth oblique view looking from the NWto the locations of the stratigraphic logs, C) Topography map of west Ambrym. Lettershick arrow marks a shore platform, which was a former connecting point between thend the eruptive products of the 1894, 1913 vest Ambrym, 1889 east Ambrym and 1910tion in west Ambrym. Lava flow directions aremarked by white arrows. Letters from “a”rs in the SW represent vent locations of phreatomagmatic volcanoes that are relativelyversion) stars indicate other young phreatomagmatic vents in the NW side of the islandature morphology. Large star marked by “e” is the location of vent site referred to theed in the final stage of the eruptions but has completely been shoaled away by today.

Fig. 2. Photographs of the 1913 west Ambrym eruptions. A) A view taken in the on-going phase of the eruption record a tuff ring (arrow) already developed by the time thephotograph was taken. This tuff ring is inferred to be the basal tuff ring sequence marked as stratigraphic unit M1, B) Eruption cloud emerges in a well-defined vent zone (arrow), C)Complex eruption scenario. Continuous up-rush and cock's tail jetting (circle) have been captured on the photograph in the left side (dashed arrow) of the newly formed tuff ring.Note the simultaneously active vent in the right hand side (white rectangle) erupting over the initial tuff ring (stratigraphic unit M1, white arrow), D) Devastated zone of the“Hospital-tuff ring”, days after the eruption.

Fig. 3.Near-continuous coastal sections of the Hospital-tuff ring. A) east to west coastal section photo mosaic starting from the shore platform. Black arrows point to the stratigraphymarker horizon of M2 unit. Black rectangle marks the field of view on “C”, B) connecting view of the coastal section. Black arrows point to the marker horizon M2 unit. C) close-upview of the region near the former crater-lagoon opening. Note the Unconformity surfaces marked by arrows. Person in circle is for scale, D) Close-up view of the contact between thePH1, M2 and PH2 stratigraphy unit. Hammer is for scale in circle. Star marks the sampling location of AR9b sample.


image of Fig.�2

image of Fig.�3


4.1. M1 basal tephra ring

A basal tephra ring structure is partly exposed as the stratigraphi-cally lowest unit of the coastal succession (Figs. 3B and 4A) and itappears to continue below sea level. The M1 unit consists predom-inantly of 10–30 cm-thick, clast-supported coarse ash and lapilli bedsdominated by scoriaceous particles (Fig. 4A). M1 passes gradationallyinto the overlying and more widely-distributed PH1 (Fig. 4B). It hasalso a gradational contact upward to a phreatomagmatic tephra unitPH2 (Fig. 4C). M1 contains abundant coral blocks up to 40 cm,commonly enclosed within fluidal shaped lava bombs (Fig. 5A and B).The clast supported scoriaceous beds with laterally continuousbedding characteristics imply an air fall origin, while the matrix-supported and stratified interbeds are more consistent with depositsof pyroclastic density currents (Fig. 4A).

Stratigraphic unit M1 appears to be part of a different structure tothe overlying stratigraphic units. This is supported by the historicphotographs that show a distinct tuff ring already developed along theshore line at of the onset of the main phase of the coastalphreatomagmatic eruptions (Fig. 2A, B and C). M1 either representsan earlier phase of the eruption or an earlier tephra/tuff ring. On thebasis of the absence of the gradational contact, along with no evidencefor erosion or soil development features, we infer that M1 was formedduring the initial phase of eruptive events that took place in thewestern extremity of a line of vents.

4.2. PH1 — phreatomagmatic tephra succession

PH1 comprises weakly to well-bedded, poorly sorted, cross-bedded to dune-bedded lapilli and ash layers, up to 10 m thick(Fig. 3A and C). It contains abundant coral fragments up to 30 cm indiameter, indicating that the fragmentation level of themagma–waterinteraction was in strata of coral and/or coralline sand–gravel. Finecoral particles and carbonate silt form coatings on juvenile glassy

Fig. 4. Stratigraphic units around the preserved older tuff ring segment referred toM1 stratigseem to have a gradual contact. The same contact relationship can be seen on “B”. The right hgrained base-surge deposit dominated beds, part of the PH2 unit (on “C”). PH1 is here abse

particles. The succession is dominated by typical proximal pyroclasticdensity current deposits. It has interbeds of coarse pyroclastic brecciaand sporadically distributed large ballistic blocks. The coarse beds aremost likely associated with vent-clearing (e.g. Houghton andSchmincke, 1989; Houghton et al., 1996), which in this case mayrepresent lateral vent growth, or migration of the explosion locusalong a propagating fissure. The breccias are rich in large fluidalshaped bombs, angular, chilled glassy bombs and large coralfragments. PH1 has a lensoid 3D geometry and is capped by ametre-thick scoriaceous fall deposit (M2; Fig. 3C and D).

4.3. M2 scoria-fall succession

M2 is a predominantly clast-supported ~1.5 m thick lapilli fall unitthat gradually thins toward the west (Fig. 3A and B). The unit mantlesthe PH1–M1 units across the entire exposed cliff face. The clasts areplaty, angular, and highly vesicular with irregularly shaped vesicles,all suggesting a magmatic gas-expansion driven fragmentation of therising magma (Houghton et al., 1996; 1999). The unit can besubdivided into at least five (10–30 cm-thick) coarse bed sets eachoverlain by cm-thick beds of laminated fine ash (Fig. 3D). The beds areungraded, moderately sorted with weak stratification.

4.4. PH2 — capping phreatomagmatic tephra

The capping PH2 tephra unit totals up to 10 metres in thickness(Fig. 2A and B), and can be subdivided into three major repeated bedsets. Each set is represented by poorly sorted ash and lapilli-dominated beds, peppered with common coarse accidental lithicblocks (Fig. 6A and B). The units show cross bedding and areconsistent with deposition from pyroclastic density currents (Watersand Fisher, 1971; Chough and Sohn, 1990; Bull and Cas, 2000;Vazquez and Ort, 2006). These sub-units are separated by laterallycontinuous pyroclastic breccia horizons containing abundant, large

raphy unit (A). Dashed linemarks the scoriaceous lapilli bed (M2). PH1 andM1 howeverand side (western) of the basal tuff ring (M1 stratigraphy unit) however covered by finent or indistinguishable due to lack of any characteristic sedimentary indicators.

image of Fig.�4

Fig. 5. Coral fragments are abundant in M1 unit. Coral (in circle) have been identified inlarge scoriaceous blocks (“A”) and as individual accidental clasts (white clasts) in thelapilli tuff sequences (“B”). Black arrow points to a cauliflower shape bomb, while whitearrow marks a rounded, recycled tuff picked up from the underlying rock units.

Fig. 6. A) Overview section of the PH2 stratigraphy unit exhibits distinct pyroclasticbreccia (pb) horizons indicating significant vent clearing events in the course of theeruption. B) Abundant ballistic bombs and blocks (white arrow) and thick packages(double arrows) of pyroclastic breccia horizons (pb) indicate violent and changeablevent conditions in this stage of the eruption. M2 unit is marked in both views.


and angular lithic blocks and juvenile bombs, along with commoncoral fragments. The accidental lithic fragments are predominantlycrystalline basalt with a range in textures, as well as fragments ofindurated phreatomagmatic lapilli-tuff. Ballistic blocks up to 1.5 m indiameter commonly show impact sags in the bedding below. Towardthe outer margins of the tuff ring exposures, the PH2 unit thins to c.4 m of fine, dune-bedded, cross-laminated ash (Fig. 7A, B and C).These units eventually thinned to become 5–10 cm, laterally contin-uous, accretionary lapilli-bearing beds. The pyroclastic brecciahorizons show a gradual transformation into single bomb andblock-rich horizons, showing common impact sags in the bedsbelow (Fig. 7A).

4.5. Distal reworked tephra succession

In the distal sections, the fine grained, dune- and cross-beddedlapilli tuff and tuff beds of PH2 are gradually replaced by a successionof poorly sorted, 0.1–0.4 m-thick beds with a dominantly fine-ashmatrix containing lapilli and blocks (Fig. 7D). These appear to relate tosyn- and immediately post-eruptive rainfall-induced reworking of theinner tephra ring, forming debris and hyperconcentrated flows

(Chough and Sohn, 1990; Lajoie et al., 1992). These fine ash matrixbeds are laterally continuous and gradually pass into beach sanddeposits on the northernmost side of the tuff ring around 1 km fromthe tephra ring interior.

4.6. Syn-eruptive reworking/erosion

Fine mud laminae between coarse-grained tabular lapilli andlapilli tuff beds indicate drapes that formed by water loss from thesettling deposits. This may also indicate brief intervals betweendepositional events and eyewitness reports show that these pausesranged between a few minutes and a few hours. The inter-layering ofprimary deposits with reworked volcaniclastic sediments indicates anongoing remobilisation of tephra deposited on the ring during theeruption. This was probably due to high water availability from steamcondensed from the passing base surges. The matrix-supported,tabular-bedded and heterogeneous volcaniclastic beds are inferred tobe deposited from hyperconcentrated flows and debris flows in thedistal areas, significantly reshaping the tephra ring and forming anapron of reworked tephra. Gully formation in the upper tuff ringappears also to have started during the eruption (Németh and Cronin,2007).

image of Fig.�5

image of Fig.�6

Fig. 7. A) Distal sections of the PH2 unit primarily composed of base surge deposited fine grained tuff beds. Stars represent the collected AR17a and b samples from a base surgemega-ripple. B) Indurated tuff bed in the distal section of the PH2 unit represents accretionary lapilli-rich horizons (star marks the AR21 sample collection site). C) In the most distalexposed sites the preserved tuff ring succession of the PH2 unit condensed to be a less than 5 m thick succession of fine-grained, bedded tuff. Star points to the collection location ofthe AR22 sample. D) The tuff ring succession grades laterally into a reworked tephra fan in the distal regions. Stars mark the AR23 and 24 sample collection locations.


5. Facies relationships and magma fragmentation

The presence of a laterally persistent scoria fall unit (M2) betweentwo major phreatomagmatic units without evidence of time breaksand the gradual transition from the M1 magmatic fragmentation-dominated unit into the phreatomagmatic PH1 unit and the lateralfacies variations indicate the operation of a continuous series of ventswith explosions involving various degrees of phreatomagmatic vs.magmatic fragmentation. The result of such eruption mechanismswas the repeated and alternating succession forming the Hospitaltephra ring phase of the eruption (Fig. 8). The alternating scoria falldominated bed with pyroclastic density current deposits indicaterepeated transitions between dry and wet eruption styles. Currentindicators, such as cross-bedding, dunes, lateral variations in grainsize and bed thickness tally with the eye-witness descriptions ofrepeated pyroclastic density currents (Fig. 2B and C). Thesepyroclastic density currents carried fine carbonate silt, along withcommon blocks of coral debris. Wind conditions and/or ventorientation and location during the early phases of the eruptionconcentrated M1 deposition toward the western side of the ring. Thelocal, crest-like-form of M1 shows a cross-cut through the earlieststages of the growing tephra ring rim (Fig. 4A, B and C). The ventappears to have shifted eastward by the time of deposition of M2, witha dryer fragmentation mechanism producing scoria lapilli fall thatthickens toward the east. The breccia horizons within PH1 and 2 andthe alternations between dryer and wetter phases of the eruption arereadily explained by a migrating locus of the fissure-fed eruptionthroughout this complete eruption sequence. Within the Hospitaltephra ring, the active locus of dominant explosive activity also shiftedduring its formation. Eyewitness descriptions are that violent (but

small compared to what was to come) surface explosions movedtoward the hospital, culminating in the sudden production of a hugesteam and ash plume (Fig. 2A, B and C). A similar progression occurredin the 1977 Ukinrek maar eruption (Kienle et al., 1980; Ort et al.,2000).

In many fissure eruptions, an initially spread eruption becomeslocalised to certain points of intense activity over time (e.g. Walker,1991; 1993; Walker et al., 1984; Valentine and Keating, 2007; Keatinget al., 2008; Valentine and Gregg, 2008). This occurred during thelava-fountain phases of the Ambrym eruptions in many places alongthe upper flanks. During eruption of the Hospital tuff ring (Figs. 1B, Eand 2D), progress of the explosion locus appears to have travelledwest to east. This is counter to the overall propagation direction of thefissure formation during the eruption, and indeed the next surfaceoutburst of activity took place offshore westward of the Hospitaltephra ring.

An abrupt eruption-style change recorded by the onset of the M2unit indicate that at least one phase was of dry, magmatic-driveneruptions, before resumption of typical phreatomagmatic explosions,generating PH2. Such abrupt changes in eruption style are well-documented in many successions and have commonly been inter-preted to represent intercalated deposits from simultaneous erup-tions through closely spaced vents (Houghton and Schmincke, 1989;Houghton et al., 1996; 1999). The capping PH2 succession thinstoward the west, implying that the eastern vent area stayed active thelongest. It contains three major pyroclastic breccia horizons that arelaterally continuous, which indicate repeated periods of either ventwidening, deepening, or migration reported commonly from variousphreatomagmatic volcanoes (Lorenz, 2003; Sohn and Park, 2005;Risso et al., 2008; Ort and Carrasco-Núñez, 2009). A step-wise

image of Fig.�7

Fig. 8. Simplified stratigraphy logs along the coastal sections. Letters mark the section locations introduced in the maps of Fig. 1.


migration of the vent along a lineation is consistent with the elongatefinal geometry of the tephra ring and the eyewitness reports.

6. Controls on explosivity — pyroclastic texture and volatilecontent constraints

The 1913 Ambrym eruption began with an explosive phase thatrapidly transformed into a fissure eruption of low-energy lava-fountains and highly fluid and rapid lava flows. This implies a lowinternal (magmatic) gas content — engendering inefficient magmafragmentation and gentle effusive, or lava fountaining activitydominating the eruptive phases occurring on the middle slopes ofthe island, i.e., producing the highly fluid lava flows and lava fountainsin the Meltungan area (Fig. 1E). A rapid and dramatic change tophreatomagmatic magma fragmentation caused extremely explosiveactivity at the Hospital tephra ring. This is most readily explained bymagma intersecting water or saturated rock as the fissure propagatedtoward the sea. To explore whether external water was the mostdominant forcing factor of this explosive eruption phase, moredetailed magma fragmentation and chemical studies were undertak-en. In addition, petrographic photomicrographs and Scanning Elec-tron Microscope images were analysed using ImageJ (http://rsbweb.nih.gov/ij/) and Benoit (http://www.trusoft-international.com/) soft-wares to characterise the shape parameters of the juvenile particles(Zimanowski et al., 1997; Büttner et al., 1999; 2002; 2006).

6.1. Grain size distribution

Grain size distribution analysis was performed by dry sieving onsamples from the proximal part of PH1 unit (AR9b) and the distalsection of PH2 unit (AR17a, b, AR21, AR22, AR23, AR24) (Fig. 9). The+5 Φ fraction represents a cumulative value for finer clasts andtherefore the diagrams are slightly distorted in the fine grain size

ranges. The overall distribution of the analysed samples is unimodal.Matrix-rich samples such as AR22, 23, AR17b, however, show abroadly dispersed grain size distribution pattern, common inpyroclastic surge deposits, while the other samples are morecharacteristic of fall deposits (Dellino and LaVolpe, 1996; 2000;Dellino and Liotino, 2002).

The analysed samples were dominantly juvenile pyroclast rich.The fine fractions however, contained abundant carbonate silt derivedfrom coral (b15%). Larger coral fragments and crystals occurred asrare constituents of coarser fractions.

6.2. EMP and FTIR analysis

Polished glass shards of a sample of a base surge bed from the PH2unit were analysed by energy dispersive (EDS) electron microprobe(Jeol JXA-840A) at the University of Auckland. The chemical data werecollected using a Princeton GammaTech Prism 2000 Si (Li) EDS X-raydetector, a 5 μm focused beam, accelerating voltage of 12.5 kV, beamcurrent of 600 pA and 100 second live-count time. Calibration of theanalyses used a suite of Astimex™ mineral standards.

The 1 to 2Φ fraction of particles from AR17b was analysed,representing the PH2 distal base surge unit (Table 1). The sampleswere fresh with no apparent surface alteration. The measured glassypyroclasts were vesicle- and microlite-poor, trachy-basalt to basaltictrachy-andesite in composition, containing around 53 wt.% SiO2

(Table 1). The total alkali content of the measured glasses wasbetween 5 and 6 wt.%, indicating a moderate alkaline compositionaltrend (Table 1), belonging to one of the magma series noted beforefrom Ambrym (Robin et al., 1993). The total values of the analysedvolcanic glass major elements ranged between 95.65 and 97.07%,reflecting a low original volatile content of the glass and some post-eruptive hydration and initial palagonitization.

http://rsbweb.nih.gov/ij/


http://www.trusoft-international.com/

image of Fig.�8

Fig. 9. Grain size distribution diagrams of sieved samples from the Hospital-tuff ring sequence.


Basaltic glass chips were also doubly polished as parallel-sidedwafers around 90 μm thick and Fourier Transform Infrared (FTIR)analyses were carried out on them on a Thermo Electron NicoletContinuum Microscope and Nicolet 4700 Spectrometer at Massey

University. Samples were placed on a water-free NaCl support and 512scanswere collectedwith a 4 cm−1 resolution using a ceramic IR sourceand anMCTAdetector cooled by liquid nitrogen. A positive dry air purgein themicroscope and spectrometer was used to minimise interference

image of Fig.�9

Table 1Electron microprobe chemical composition of volcanic glass shards from a fine grained base surge bed of the Hospital-tuff ring of the Ambrym 1913 AD eruption site.

Sample SiO2 TiO2 Al2O3 FeO MnO MgO CaO Na2O K2O P2O5 SO3 Cl Cr2O3 NiO Total

AK17B1–2 51.26 0.91 15.34 10.65 0.27 4.96 7.38 3.16 2.26 0.49 −0.06 0.06 0.05 0.05 96.79AK17B1–2 51.81 0.89 17.98 8.44 0.14 3.02 9.12 3.24 1.86 0.44 −0.02 0.13 0.03 −0.02 97.07AK17B1–2 50.70 0.94 14.91 10.63 0.22 4.11 8.26 2.93 2.29 0.66 0.09 0.05 −0.07 −0.09 95.65AK17B1–2 50.65 1.04 14.97 10.59 0.27 4.00 8.23 3.19 2.43 0.51 0.02 0.09 0.07 0.11 96.17AK17B1–2 51.04 1.02 15.11 10.82 0.20 4.03 8.19 3.19 2.45 0.36 −0.09 0.10 0.11 −0.09 96.44AK17B1–2 50.90 0.90 15.04 10.58 0.18 3.81 8.08 3.19 2.41 0.43 −0.10 0.11 0.09 0.02 95.64AK17B1–2 50.50 1.15 14.71 10.83 0.18 4.26 8.53 3.18 2.28 0.44 0.02 0.09 0.11 −0.07 96.21AK17B1–2 51.22 1.15 15.02 10.99 0.26 3.97 8.21 3.31 2.37 0.54 −0.01 0.04 −0.02 −0.03 97.03AK17B1–2 51.31 1.19 14.72 10.61 0.33 3.89 8.05 3.37 2.44 0.61 −0.03 0.05 0.03 −0.10 96.47AK17B1–2 51.11 1.08 14.82 10.77 0.13 4.04 8.16 3.30 2.33 0.46 0.06 0.01 0.02 0.02 96.31


by atmospheric water. The water and CO2 wavelength values wererecalculated to the nominated volatile content related to the totalvolume. None of the experiments showed peaks in the CO2wavelength,indicating CO2 values below the detection limit. Water peaks wereidentified and their intensity was recalculated to volume ratios(Table 2). The results suggest that the water content of the chilledglass shards ranges between 0.15 and 0.43%, (Table 2). These low valuesindicate that themagmawas degassed and volatile-poor upon eruptionandhence the explosivity is inferred to havebeenprimarily drivenby anexplosively efficient degree of magma–water interaction.

6.3. Optical microscopy

Samples were viewed in both untreated conditions and followingacid-washing with ultrasonic cleaning. In the untreated samples, thecoarse fractions were dominated by sub-rounded coarse ash particles(Fig. 10A), andwere coated by fine carbonate silt and fine ash that wasremoved by the washing (Fig. 10B and C). Treated particles wereangular glass-dominated fragments, along with up to 15% coralfragments as accessories.

Glassy fragments showed uniform texture across all grain sizefractions, with low vesicularity and microlite content. Vesicles weredominantly b100 μm across, spheroid and isolated, separated by thickglass walls; they rarely covered more than 15% of a thin sectionthrough a juvenile pyroclast. Microphenocrysts of olivine andpyroxenes occur but are not common (Fig. 10D). The glassy particleswere heavily fractured (Fig. 10E), with bubble-wall surfaces especiallyin the coarse ash fractions (Fig. 10F).

6.4. Scanning electron microscopy (SEM) and back-scattered electron(BSE) study

These measurements were carried out on b1 Φ glassy juvenileparticles at the Manawatu Microscopy Centre at the Massey

Table 2FTIR analysis data from volcanic glass shards from base surge beds of the Hospital tuffring of the Ambrym 1913 AD eruption site.

Spot# FTIR peak measurementH2Ot-3535 cm−1

H2Ot concentration (%)by Beer's Law

KN-1 0.398 0.43KN-2 0.31 0.34KN-3 0.205 0.22KN-4 0.159 0.17KN-5 0.158 0.17KN-6 0.275 0.3KN-7 0.283 0.31KN-8 0.221 0.24KN-9 0.136 0.15KN-10 0.34 0.37KN-11 0.233 0.25KN-12 0.245 0.27

Section thickness was 92 μm.

University on an FEI Quanta 200 SEM operated under 20 kVacceleration (with a dual backscattered electron detector for BSE).The juvenile particles are glassy, and heavily fractured even in thefinest ash ranges of only tens of microns in diameter. Larger particles(coarse ash), showed higher vesicularity, but with isolated bubbleshaving thick walls (Fig. 11A). Even in the most highly vesiculatedparticles, the outer margin of the particles was marked by heavilyfragmented outer rims (Fig. 11B and C). In most cases 10–20 micronthick peel-like layers were observed that gradually curl from the solidparticle core (Fig. 11B). This is inferred to be glassymaterial (i.e., not analteration or palagonitization rim) since it has the same texturalappearance as themain bodyof the particle. Bubblewalls are commonlytruncated by fracture surfaces (Fig. 11C). Coarse and fine ash particlesare dominantly blocky with sharp edges (Fig. 11D, E and F), showingsurface step-like fractures, as well as adhering fine carbonate silt(Fig. 11E and F).

BSE images were useful to characterise microlite-content, vesiclepatterns and hidden fracture networks in the interior of the juvenilepyroclasts. The blocky nature of the particles is very pronounced(Fig. 12A) and pyroclasts showed many hidden fractures and afractured glassy outer margin (Fig. 12B). Microlite contents were verylow, with occasional olivine and pyroxene microphenocrysts present(Fig. 12B–D and E). The vesicularity of the glassy pyroclasts was low(b20%), and vesicles were generally spherical (Fig. 12C), althoughoccasionally, elongated tube-like vesicles were identified in coarse ashparticles (Fig. 12F).

6.5. Morphological parameters of juvenile pyroclasts

During phreatomagmatic volcanism magma fragmentation iscaused by molten fuel coolant interaction (MFCI) to produce twodistinct groups of juvenile particles: active and passive particles (e.g.;Heiken, 1972; Lorenz, 1973; Wohletz, 1983;1986; Heiken andWohletz, 1986; Kurszlaukis et al., 1998; Morrissey et al., 2000;Büttner et al., 2002; 2006; Zimanowski et al., 2003). Active particlesare directly involved in the thermal to mechanical energy transferduring the MFCI process, while passive particles are blebs of melt thatare rapidly transported away in a ductile form from the fragmentationzone by kinetic energy release of theMFCI (Büttner et al., 2002; 2005),forming smooth droplet-like shapes (Zimanowski et al., 1997). Activeparticles are typically b130 μm in diameter and generated by brittlefragmentation, thus capturing the MFCI event (Dellino and LaVolpe,1996; Büttner et al., 1999; 2002; Morrissey et al., 2000; Dellino et al.,2001; Dellino and Liotino, 2002; Dellino and Kyriakopoulos, 2003;Zimanowski et al., 2003; Németh, 2010). Here we examined the ashfaction of selected pyroclastic density current beds for the ash fraction300–30 μm in diameter.

We used light microscopy, SEM and BSE images that weretransformed to black and white bitmaps, and used the “ImageJ”software package (http://rsbweb.nih.gov/ij/) to obtain shape para-meters such as particle perimeter, particle surface area, and Feretdiameter (e.g. the longest distance between any two points along the


Fig. 10. A) Spheroid shape ash particles with strong carbonate dust adhering. B) Angular shape ash fragments. C) Glassy pyroclasts with angular shape. D) Angular shapesideromelane glass shards in photomicrograph. Arrow pointing at larger crystal hosted in glass. Fractured glassy pyroclast (arrow) in photomicrograph. F) Bubble wall ofsideromelane glass shard (arrow) in photomicrograph.


particle boundary). From these parameters we calculated: 1)compactness (defined as the ratio of the particle area and the particlebreadth multiplied by the width, where the particle breadth is the

Fig. 11. A) Coarse ash SEM image reveal low vesicularity and thick glassy inner structure ofglass shard. C) Enhanced fractured outer margin of a sideromelane ash particle showing bvesicular sideromelane glass ash in SEM image. E) Step-like surface texture (arrow) of amodeadhering carbonate dust (arrow).

distance between the leftmost and rightmost pixels of an object, whilewidth is the distance between the uppermost and lowermost pixels ofan object), 2) elongation (defined as a ratio of the Feret diameter – the

the pyroclast. B) Thin glassy rim, that peels off from the inner part of the sideromelanelocky glass peels ready to detach from the main part of the particle. D) Angular, non-rately vesicular coarse sideromelane ash particle. F) Fine sideromelane ash particle with

image of Fig.�10

image of Fig.�11

Fig. 12. A) BSE image showing sideromelane glass shard that reveals the angular and low vesicularity nature of the majority of its particles. B) A coarse ash sideromelane glass shardshow low vesicularity and few microphenocryst (light zones) as well as heavily fractured outer rims. C) Sideromelane glass shards are low to non-vesicular. Occasional vesicles areseparated by thick glassy zones. D) The outline of the ash particles are angular, however, coarser particles can have bubble wall-defined margins. E) Microlite rich glassy pyroclastsare rare but can be identified. F) Tube-like vesicles can be identified in coarse ash fractions however, they are not common.


longest segment in an object – vs. the mean intercept to perpendic-ular, which is the mean length of the radius of an object perpendicularto the Feret diameter), 3) circularity (defined as a ratio of the particleperimeter vs. the perimeter of the circle with the same area of theparticle) and 4) rectangularity (defined by the ratio of the particlediameter vs. 2x the sum of the particle breadth and width) asdimensionless parameters following recommendations of previousauthors (Dellino and LaVolpe, 1996; Dellino et al., 2001; Büttner et al.,2002; Dellino and Liotino, 2002).

Semi-automated thresholding methods were used to separateparticle outlines from backgrounds, instead of manual retracing of theparticle outlines. This is quicker, but challenging due to the unevengrey-scales resembling from variations in illumination, inhomoge-nities or particle surface cover (e.g., alteration and adheringmaterial).BSE images appeared most reliable for this method, although someinconsistencies occurred (Fig. 13A–C and D). On BSE images, a hiddenfracture network was well defined, suggesting a complex fragmen-tation history of the larger particles (100–300 μm) (Fig. 13A). Usingpixel erosion (i.e. causing a certain layer of pixels to be removed alongany edges in the image) the fracture network became even morepronounced (Fig. 13B), while an overall dilation of the image alongedges made the particle more homogeneous (i.e. the outline becamemore continuous) (Fig. 13C). BSE images were also the best images toidentify vesicles and determine their modal occurrence (Fig. 13D).Using automatic thresholding within ImageJ software, particlesappeared to become more complex, hence manual thresholding wascarried out toward the average greyscale value of a particle as well asto the value of its margin. This allowed tracing of slight changes in thecomplexity of the obtained bitmap particle images. Later each of thethresholded particle shapes was analysed and compared.

As an end-product, we applied the Büttner et al. (2002)discrimination diagram, plotting circularity multiplied by elongation

vs. rectangularity multiplied by compactness (Fig. 14). On thisdiscrimination diagram light microscopy and SEM imagery producedvalues that plotted slightly below the empirically defined boundarybetween brittle (phreatomagmatic) and ductile (magmatic) fragmen-tation fields. Finer grained particles, however, plotted dominantly onthe brittle field of the diagram, suggesting the pattern expected offragmentation during MFCI processes. By contrast, values calculatedusing BSE imagery clearly plotted within the brittle fragmentationfield (Fig. 14), especially those produced by automatic thresholding,which as stated above appeared to create artificial complexity in thesamples. Finer ash particles plotted deeper within the brittle field.

Overall, the larger (~200–300 μm) ash particles have a moreductile fragmentation affinity, while the finer grained particles areclearly the product of brittle fragmentation. This trend indicates thatthe rising magma went through MFCI triggered explosive fragmen-tation, and the generated kinetic energy dispersed larger, still fluiddroplets of melt (coarse ash). The presence of complex microfracturesin the interiors of many coarse ash particles may indicate that theywere derived from near the fragmentation front where the magmaand water interacted.

7. Discussion

The initial explosive activity of the 1913 Ambrym eruption rapidlygave way to effusive behaviour of a very fluid magma erupted alongan en echelon set of fractures and fissures. Lavas erupted during themiddle phases of the eruption advanced rapidly to the coast, withlittle violence in the lava-fountaining and effusion observed by near-by eye witnesses. The indications of low magmatic volatile contents,such as isolated immature vesicles showing little evidence formagmatic-gas driven fragmentation, also demonstrate that themagma was not likely to produce explosive activity from its internal

image of Fig.�12

Fig. 13. This diagram summarises the potential variations that automatic thresholding can cause in particle shapes. “A” shows an image that has been created by the automaticthresholding method. It is clear that the boundary of the particle is strongly influenced by its fractured nature; however, there are fragments that seem to be separated from themainbody of the particle rather artificially due to the thresholding method. Applying pixel erosion along the particle (white) regions causes the fracture network to become moreenhanced and the particle boundary to appear more complicated (“B”). This complexity is purely a result of the pixel erosion, and has little to do with the original shape of theparticle. Conversely, we can apply pixel dilation to create a homogeneous rim around the individual particle fragments; this creates an apparently more realistic form that is similarto the original shape of the particle (“C”). Because the outline of the particle can change dramatically upon application of all of these techniques, we have used an interactive method,and manually altered, where needed, the particle outline during the bitmap image creation process. For this reason particle shape analysis was performed on three different datasetsdefined as “automatic” (using automatic thresholding), “average” (applying pixel erosion and/or dilation and manual particle outline correction) and “full” (applying pixel erosion)particle segmentation techniques. On “D” an image is shown from one of the most vesicular coarse ash particles where automatic thresholding of the original bitmap image hasprovided good particle segmentation.


gas-content. This implies that the highly hazardous phase of thiseruption was produced primarily by magma–water interaction.Fissure-fed eruptions may thus pose significant volcanic hazards topopulations on mafic volcanic islands, and this should be accommo-dated in volcanic hazard assessments. The migration along the mainfissure during the 1913 Ambrym eruption was westward from thecentral vent toward the rift-edge. However, in detail, this trend mayhave locally been reversed; the geological record at the Hospital tuffring showed a local west to east migration over a b1 km lineation. Thislocalised vent migration may have been controlled rather by near-surface processes, and/or shallow geology such as an alternation ofolder hard lava sheets and volcaniclastic deposits of varying porewater content (e.g. Németh et al. 2009). Accumulation of loose,permeable volcaniclastic debris aprons surrounding the rift-edges ofsuch volcanic islands provides ideal conditions for promotingexplosivity upon contact with rising magma. Even strongly degassedmagma if it encounters confined aquifers that are pressurised mayengender strong explosivity. The newly developed crater floor in the1913 AD Hospital tephra ring gradually subsided following theeruption, possibly indicating compaction of the disrupted crater andupper conduit. Such processes have been reported to occur at maar–diatreme volcanoes (Suhr et al., 2006). Photographs soon after the1913 AD Hospital tephra ring eruption show minor (m-10 m) initialcrater floor subsidence (e.g. Fig. 2D), suggesting, that the explosiveeruptions caused significant mass deficit to trigger crater floorcollapse and the formation of a shallowmaar, that gradually deepenedsubsequently.

8. Conclusion

We have related historic eye-witness accounts to a completecoastal exposure of one of the largest rift-edge phreatomagmatictephra rings in West Ambrym, Vanuatu. This demonstrates evidencefor some of the highest levels of explosivity found on mafic island arcvolcanoes. It occurred during a dominantly effusive fissure formingeruption and demonstrates that when fissures propagate into thesaturated deposits fringing volcanic islands, highly hazardous erup-tions may be likely.

The stratigraphy of the oval 1913 Ambrym Hospital tephra ringdemonstrates that an initial explosive magma–water interactionoccurred in a coral fragment dominated substrate, probably only afew tens of metres in depth. The efficiency of magma fragmentationwas possibly increased by the geometry of contact between a thindyke penetrating through water-saturated sediments. The explosivetuff-cone forming eruption was produced from vents that opened in astep-wise fashion from the west to the east, locally opposing themigration trend of the main fissure eruption. This may have beencontrolled by the local exhaustion of water at successive eruption sitesalong the fissure, which may in-turn has been controlled by the localgeological structure. The crater floor of the tephra ring subsided about30 m following the eruption, suggesting that the newly formedvolcano is a shallow maar. In its paroxysmal stage the eruptionproduced a thick (over 20 m) phreatomagmatic pyroclastic succes-sion with interbedded tuff breccia horizons suggesting continualinstability of the conduit, probably resulting from a step-wise lateral

image of Fig.�13

Fig. 14. Discrimination diagram introduced by Büttner et al. (2002) to distinguish particle shape characteristics that can be associated with brittle (e.g. phreatomagmatic) to ductile(e.g. magmatic) fragmentation. The diagram separately shows image outlines analysed from particle outline creation from SEM, BSE Automatic, BSE Average, BSE Full and LMmethods. Thick line separates the theoretical field of particle shapes connected to brittle versus ductile fragmentation processes.


motion of the explosive-locus with vent clearing blasts associatedwith shifts along the fissure. The textural characteristics of thejuvenile pyroclasts and FTIR-measured water contents confirm thatthe magma was in a degassed, volatile-poor state, meaning thatexplosivity was controlled by interaction with external water. Thishighlights for volcanic hazard assessments to focus on the hydro-geology and sedimentary structure of areas that are potentiallysubject to fissure eruptions, such as margins of volcanic islands thatare dominated by saturated volcaniclastic deposits.

Acknowledgements

This work was supported by the Foundation for Research Scienceand Technology - IIOF project “Facing the Challenge of AucklandVolcanism”. Logistics during the field campaigns were possible by theChiefs of Ambrym. Special thanks due to Douglas Charley (DGMWR,Port Vila) for his constant and enthusiastic support of our projects inVanuatu. SEM and BSE study was conducted at the ManawatuMicroscopy Centre of the Massey University under the expertguidance of Doug Hopcroft. Constructive reviews by Peter Kokelaarand Steve Self and editorial work by Claus Siebe improved the qualityof the manuscript significantly, for which we are grateful.

References

Büchel, G., Lorenz, V., 1993. Syn-and Post-Eruptive Mechanism of the Alaskan UkinrekMaars in 1977. In: Negendank, J.F.W., Zolitschka, B. (Eds.), Paleolimnology ofEuropean Maar Lakes. Springer-Verlag, Berlin, Heidelberg, pp. 15–60.

Bull, S.W., Cas, R.A.F., 2000. Distinguishing base-surge deposits and volcaniclasticfluviatile sediments: an ancient example from the Lower Devonian Snowy RiverVolcanics, south-eastern Australia. Sedimentology 47 (1), 87–98.

Büttner, R., Dellino, P., Zimanowski, B., 1999. Identifying magma–water interactionfrom the surface features of ash particles. Nature 401 (6754), 688–690.

Büttner, R., Dellino, P., La Volpe, L., Lorenz, V., Zimanowski, B., 2002. Thermohydraulicexplosions in phreatomagmatic eruptions as evidenced by the comparison between

pyroclasts and products fromMolten Fuel Coolant Interaction experiments. Journal ofGeophysical Research-Solid Earth 107 (B11) Article Number: 2277.

Büttner, R., Zimanowski, B., Mohrholz, C.O., Kummel, R., 2005. Analysis of thermo-hydraulic explosion energetics. Journal of Applied Physics 98 (4) Article Number:043524.

Büttner, R., Dellino, P., Raue, H., Sonder, I., Zimanowski, B., 2006. Stress-induced brittlefragmentation of magmatic melts: theory and experiments. Journal of GeophysicalResearch-Solid Earth 111 (B8) Article Number: B08204.

Carney, J.N., Macfarlane, A., Mallick, D.I.J., 1985. The Vanuatu Island-arc — an outline ofthe stratigraphy, structure, and petrology. Ocean Basins And Margins 7, 683–718.

Carracedo, J.C., Rodriguez Badiola, E., Soler, V., 1992. The 1730–1736 eruption ofLanzarote, Canary slands: a long, high-magnitude basaltic fissure eruption. Journalof Volcanology and Geothermal Research 53, 239–250.

Chough, S.K., Sohn, Y.K., 1990. Depositional mechanics and sequences of base surges,Songaksan tuff ring, Cheju Island, Korea. Sedimentology 37, 1115–1135.

Cole, P.D., Guest, J.E., Duncan, A.M., Pacheco, J.M., 2001. Capelinhos 1957–1958, Faial,Azores: deposits formed by an emergent surtseyan eruption. Bulletin ofVolcanology 63 (2–3), 204–220.

Connor, C.B., Conway, F.M., 2000. Basaltic Volcanic Fields. In: Sigurdsson, H., Houghton,B.F., McNutt, S.R., Rymer, H., Stix, J. (Eds.), Encyclopedia of Volcanoes. AcademicPress, San Diego, pp. 331–343.

Cronin, S.J., Neall, V.E., 2001. Holocene volcanic geology, volcanic hazard, and risk onTaveuni, Fiji. New Zealand Journal of Geology and Geophysics 44 (3), 417–437.

Darragh, M., Cole, J., Nairn, I., Shane, P., 2006. Pyroclastic stratigraphy and eruptiondynamics of the 21.9 ka Okareka and 17.6 ka Rerewhakaaitu eruption episodesfrom Tarawera Volcano, Okataina Volcanic Centre, New Zealand. New ZealandJournal of Geology and Geophysics 49 (3), 309–328.

Dellino, P., Kyriakopoulos, K., 2003. Phreatomagmatic ash from the ongoing eruption ofEtna reaching the Greek island of Cefalonia. Journal of Volcanology and GeothermalResearch 126 (3–4), 341–345.

Dellino, P., La Volpe, L., 2000. Structures and grain size distribution in surge deposits asa tool for modelling the dynamics of dilute pyroclastic density currents at La Fossadi Vulcano (Aeolian Islands, Italy). Journal of Volcanology and GeothermalResearch 96, 57–78.

Dellino, P., LaVolpe, L., 1996. Image processing analysis in reconstructing fragmentationand transportation mechanisms of pyroclastic deposits. The case of Monte Pilato-Rocche Rosse eruptions, Lipari (Aeolian Islands, Italy). Journal of Volcanology andGeothermal Research 71 (1), 13–29.

Dellino, P., Liotino, G., 2002. The fractal and multifractal dimension of volcanic ashparticles contour: a test study on the utility and volcanological relevance. Journal ofVolcanology and Geothermal Research 113 (1–2), 1–18.

Dellino, P., Isaia, R., La Volpe, L., Orsi, G., 2001. Statistical analysis of textural data fromcomplex pyroclastic sequences: implications for fragmentation processes of the

image of Fig.�14


Agnano-Monte Spina Tephra (4.1 ka), Phlegraean Fields, southern Italy. Bulletin ofVolcanology 63 (7), 443–461.

Frater, M., 1917. Volcanic eruption, Ambrym Island (1913). Geological Magazine 6 (4),496–503.

Gravley, D.M., Wilson, C.J.N., Leonard, G.S., Cole, J.W., 2007. Double trouble: pairedignimbrite eruptions and collateral subsidence in the Taupo Volcanic Zone, NewZealand. Geological Society of America Bulletin 119 (1–2), 18–30.

Gregory, J.W., 1917. The Ambrym eruptions of 1913–1914. Geological MagazineDecember 1917, 496–503.

Heiken, G., 1972. Morphology and petrography of volcanic ashes. Geological Society ofAmerica Bulletin 83 (7), 1961–1968.

Heiken, G.H., Wohletz, K.H., 1986. Volcanic Ash. University of California Press, Berkeley.246 pp.

Houghton, B.F., Schmincke, H.U., 1989. Rothenberg scoria cone, East Eifel — a complexstrombolian and phreatomagmatic volcano. Bulletin of Volcanology 52 (1), 28–48.

Houghton, B.F., Wilson, C.J.N., Rosenberg, M.D., Smith, I.E.M., Parker, R.J., 1996. Mixeddeposits of complex magmatic and phreatomagmatic volcanism: an example fromCrater Hill, Auckland, New Zealand. Bulletin of Volcanology 58 (1), 59–66.

Houghton, B.F., Wilson, C.J.N., Smith, I.E.M., 1999. Shallow-seated controls on styles ofexplosive basaltic volcanism: a case study from New Zealand. Journal ofVolcanology and Geothermal Research 91 (1), 97–120.

Keating, G.N., Valentine, G.A., Krier, D.J., Perry, F.V., 2008. Shallow plumbing systems forsmall-volume basaltic volcanoes. Bulletin of Volcanology 70 (5), 563–582.

Kienle, J., Kyle, P.R., Self, S., Motyka, R.J., Lorenz, V., 1980. Ukinrek Maars, Alaska .1. April1977 — eruption sequence, petrology and tectonic setting. Journal of Volcanologyand Geothermal Research 7 (1-2), 11–37.

Kurszlaukis, S., Büttner, R., Zimanowski, B., Lorenz, V., 1998. On the first experimentalphreatomagmatic explosion of a kimberlite melt. Journal of Volcanology andGeothermal Research 80 (3–4), 323–326.

Lajoie, J., Lanzafame, G., Rossi, P.L., Tranne, C.A., 1992. Lateral facies variations inhydromagmatic pyroclastic deposits at Linosa, Italy. Journal of Volcanology andGeothermal Research 54, 135–143.

Lorenz, V., 1973. On the formation of Maars. Bulletin of Volcanology 37 (2), 183–204.Lorenz, V., 1986. On the growth of maars and diatremes and its relevance to the

formation of tuff rings. Bulletin of Volcanology 48, 265–274.Lorenz, V., 2003. Maar – diatreme volcanoes, their formation, and their setting in hard-

rock or soft-rock environments. Geolines - Journal of the Geological Institute of ASCzech Republic 15, 72–83.

Marshall, P., 1915. The recent volcanic eruptions on Ambrym Island. Trans. NewZealand Inst. XLVII, 387–391.

McCall, G.J.H., LeMaitre, R.W., Malahoff, A., Robinson, G.P., Stephenson, P.J., 1969. Thegeology and geophysics of the Ambrym Caldera. New Hebrides, SymposiumVolcanoes and Their Roots, Oxford, England, pp. 682 – 696.

Morrissey, M.M., Zimanowski, B., Wohletz, K., Büttner, R., 2000. PhreatomagmaticFragmentation. In: Sigurdsson, H., Houghton, B.F., McNutt, S.R., Rymer, H., Stix, J.(Eds.), Encyclopedia of Volcanoes. Academic Press, New York, pp. 431–446.

Nairn, I.A., Self, S., Cole, J.W., Leonard, G.S., Scutter, C., 2001. Distribution, stratigraphy,and history of proximal deposits from the c. AD 1305 Kaharoa eruptive episode atTarawera Volcano, New Zealand. New Zealand Journal of Geology and Geophysics44 (3), 467–484.

Németh, K., 2010. Volcanic glass textures, shape characteristics and compositions fromphreatomagmatic rock units of the western Hungarian monogenetic volcanic fieldsand their implication to magma fragmentation. Central European Journal ofGeosciences 2 (3), 399–419.

Németh, K., Cronin, S.J., 2007. Syn- and post-eruptive erosion, gully formation, andmorphological evolution of a tephra ring in tropical climate erupted in 1913 inWest Ambrym, Vanuatu. Geomorphology 86, 115–130.

Németh, K., Cronin, S.J., 2008. In: Thomson, K., Petford, N. (Eds.), Volcanic Craters, PitCraters and High-Level Magma-Feeding Systems of a Mafic Island-arc Volcano:Ambrym, Vanuatu, South Pacific. : Structure and Emplacement of High-LevelMagmatic Systems, 302. Geological Society of London Special Publications, pp.87–102.

Németh, K., Cronin, S.J., 2009a. Phreatomagmatic volcanic hazards where rift-systemsmeet the sea, a study from Ambae Island, Vanuatu. Journal of Volcanology andGeothermal Research 180 (2–4), 246–258.

Németh, K., Cronin, S.J., 2009b. Volcanic structures and oral traditions of volcanism ofWestern Samoa (SW Pacific) and their implications for hazard education. Journal ofVolcanology and Geothermal Research 186 (3 – 4), 223–237.

Németh, K., Cronin, S.J., Stewart, R.B., Charley, D., 2009. Intra- and extra-calderavolcaniclastic facies and geomorphic characteristics of a frequently active maficisland-arc volcano, Ambrym Island, Vanuatu. Sedimentary Geology 220 (3–4),256–270.

Nunn, P.D., 2003. Fished up or thrown down: the geography of Pacific Island originmyths. Annals of the Association of American Geographers 93 (2), 350–364.

Ort, M.H., Carrasco-Núñez, G., 2009. Lateral vent migration during phreatomagmaticand magmatic eruptions at Tecuitlapa Maar, east-central Mexico. Journal ofVolcanology and Geothermal Research 181 (1–2), 67–77.

Ort, M.H., Wohletz, K., Hooten, J.A., Neal, C.A., McConnel, V.S., 2000. The Ukinrek maarseruption, Alaska, 1977: a natural laboratory for the study of phreatomagmaticprocesses at maars. Terra Nostra 2000 (6), 396–400.

Peate, D.W., Pearce, J.A., Hawkesworth, C.J., Colley, H., Edwards, C.M.H., Hirose, K., 1997.Geochemical variations in Vanuatu arc lavas: the role of subducted material and avariable mantle wedge composition. Journal of Petrology 38 (10), 1331–1358.

Picard, C., Monzier, M., Eissen, J.-P., Robin, C., 1995. Concomitant Evolution of TectonicEnvironment andMagmaGeochemistry, Ambrym volcano (Vanuatu, NewHebridesarc). In: Smellie, J.L. (Ed.), Volcanism Associated with Extension at Consumed PlateMargins, 81. Geological Society of London Special Publications, pp. 135–154.

Pirrung, M., Büchel, G., Lorenz, V., Treutler, H.C., 2008. Post-eruptive development of theUkinrek East Maar since its eruption in 1977 AD in the periglacial area of south-west Alaska. Sedimentology 55 (2), 305–334.

Purey-Cust, H.E., 1896. The eruption of Ambrym Volcano, New Hebrides, South-WestPacific, 1894. Geographical Journal VIII, 585–602.

Raos, A.M., Crawford, A.J., 2004. Basalts from the Efate Island Group, central section ofthe Vanuatu arc, SW Pacific: geochemistry and petrogenesis. Journal of Volcanologyand Geothermal Research 134 (1–2), 35–56.

Risso, C., Németh, K., Combina, A.M., Nullo, F., Drosina, M., 2008. The role ofphreatomagmatism in a Plio-Pleistocene high-density scoria cone field: LlancaneloVolcanic Field (Mendoza), Argentina. Journal of Volcanology and GeothermalResearch 169 (1–2), 61–86.

Robin, C., Eissen, J.P., Monzier, M., 1993. Giant tuff cone and 12-km-wide associatedcaldera at Ambrym Volcano (Vanuatu, New-Hebrides-Arc). Journal of Volcanologyand Geothermal Research 55 (3–4), 225–238.

Self, S., Kienle, J., Huot, J.P., 1980. Ukinrek Maars, Alaska 2. Deposits and formation of the1977 craters. Journal of Volcanology and Geothermal Research 7 (1–2), 39–65.

Sohn, Y.K., Park, K.H., 2005. Composite tuff ring/cone complexes in Jeju Island, Korea:possible consequences of substrate collapse and vent migration. Journal ofVolcanology and Geothermal Research 141 (1–2), 157–175.

Sohn, Y.K., Park, K.H., Yoon, S.H., 2008. Primary versus secondary and subaerial versussubmarine hydrovolcanic deposits in the subsurface of Jeju Island, Korea.Sedimentology 55 (4), 899–924.

Speed, J., Shane, P., Nairn, I., 2002. Volcanic stratigraphy and phase chemistry of the 11900 yr BP Waiohau eruptive episode, Tarawera Volcanic Complex, New Zealand.New Zealand Journal of Geology and Geophysics 45 (3), 395–410.

Spinks, K.D., Acocella, V., Cole, J.W., Bassett, K.N., 2005. Structural control of volcanismand caldera development in the transtensional Taupo Volcanic Zone, New Zealand.Journal of Volcanology and Geothermal Research 144 (1–4), 7–22.

Stothers, R.B., Wolff, J.A., Self, S., Rampino, M.R., 1986. Basaltic fissure eruptions, plumeheights, and atmospheric aerosols. Geophysical Research Letters 13 (8), 725–728.

Suhr, P., Goth, K., Lorenz, V., Suhr, S., 2006. Long lasting subsidence and deformation inand above maar–diatreme volcanoes — a never ending story. Zeitschrift derDeutschen Gesellschaft für Geowissenschaften 157 (3), 491–511.

Sutton, A.N., Blake, S., Wilson, C.J.N., Charlier, B.L.A., 2000. Late Quaternary evolution ofa hyperactive rhyolite magmatic system: Taupo volcanic centre, New Zealand.Journal of the Geological Society 157, 537–552.

Thordarson, T., Larsen, G., 2007. Volcanism in Iceland in historical time: volcano types,eruption styles and eruptive history. Journal of Geodynamics 43 (1), 118–152.

Thordarson, T., Self, S., 1993. The Laki (Skaftar-Fires) and Grimsvotn Eruptions in 1783–1785. Bulletin of Volcanology 55 (4), 233–263.

Thordarson, T., Self, S., Oskarsson, N., Hulsebosch, T., 1996. Sulfur, chlorine, and fluorinedegassing and atmospheric loading by the 1783–1784 AD Laki (Skaftar Fires)eruption in Iceland. Bulletin of Volcanology 58, 205–225.

Turner, S.P., Peate, D.W., Hawkesworth, C.J., Eggins, S.M., Crawford, A.J., 1999. Twomantle domains and the time scales of fluid transfer beneath the Vanuatu arc.Geology 27 (11), 963–966.

Valentine, G.A., Gregg, T.K.P., 2008. Continental basaltic volcanoes — processes andproblems. Journal of Volcanology and Geothermal Research 177 (4), 857–873.

Valentine, G.A., Keating, G.N., 2007. Eruptive styles and inferences about plumbingsystems at Hidden Cone and Little Black Peak scoria cone volcanoes (Nevada, USA).Bulletin of Volcanology 70 (1), 105–113.

Vazquez, J.A., Ort, M.H., 2006. Facies variation of eruption units produced by the passageof single pyroclastic surge currents, Hopi Buttes volcanic field, USA. Journal ofVolcanology and Geothermal Research 154 (3–4), 222–236.

Walker, G.P.L., 1991. Structure, and origin by injection of lava under surface crust, oftumuli, lava rises, lava-rise pits, and lava-inflation clefts in Hawaii. Bulletin ofVolcanology 53 (7), 546–558.

Walker, G.P.L., 1993. Basaltic-volcano Systems. In: Prichard, H.M., Alabaster, T., Harris,N.B.W., Nearly, C.R. (Eds.), Magmatic Processes and Plate Tectonics, 76. GeologicalSociety of London Special Publications, pp. 3–38.

Walker, G.P.L., Self, S., Wilson, L., 1984. Tarawera 1886, New-Zealand — a basalticPlinian fissure eruption. Journal of Volcanology and Geothermal Research 21 (1–2),61–78.

Waters, A.C., Fisher, R.V., 1971. Base surges and its deposits: Capelinhos and Taalvolcanoes. Journal of Geophysical Research 76, 5596–5614.

Wilson, C.J.N., Blake, S., Charlier, B.L.A., Sutton, A.N., 2006. The 26.5 ka Oruanui eruption,Taupo volcano, New Zealand: development, characteristics and evacuation of alarge rhyolitic magma body. Journal of Petrology 47 (1), 34–69.

Wohletz, K.H., 1983. Mechanisms of hydrovolcanic pyroclast formation: grain-size,scanning electron microscopy, and experimental studies. Journal of Volcanologyand Geothermal Research 17, 31–63.

Wohletz, K.H., 1986. Explosive magma–water interactions: thermodynamics, explosionmechanisms, and field studies. Bulletin of Volcanology 48, 245–264.

Zimanowski, B., Büttner, R., Lorenz, V., Hafele, H.G., 1997. Fragmentation of basaltic meltin the course of explosive volcanism. Journal of Geophysical Research-Solid Earth102 (B1), 803–814.

Zimanowski, B., Wohletz, K., Dellino, P., Büttner, R., 2003. The volcanic ash problem.Journal of Volcanology and Geothermal Research 122 (1–2), 1–5.

drivers of explosivity and elevated hazard in basaltic fissure eruptions: the 1913 eruption of...

Documents