Bull Volcanol (2011) 73:1259–1277 DOI 10.1007/s00445-011-0477-9

RESEARCH ARTICLE

Mafic Plinian volcanism and ignimbrite emplacement at Tofua volcano, Tonga J. T. Caulfield & S. J. Cronin & S. P. Turner & L. B. Cooper

Received: 30 October 2009 / Accepted: 12 March 2011 / Published online: 27 April 2011 # Springer-Verlag 2011

Abstract Tofua Island is the largest emergent mafic volcano within the Tofua arc, Tonga, southwest Pacific. The volcano is dominated by a distinctive caldera averaging 4 km in diameter, containing a freshwater lake in the south and east. The latest paroxysmal (VEI 5–6) explosive volcanism includes two phases of activity, each emplacing a high-grade ignimbrite. The products are basaltic andesites with between 52 wt.% and 57 wt.% SiO2. The first and largest eruption caused the inward collapse of a stratovolcano and produced the ‘Tofua’ ignimbrite and a subcircular caldera located slightly northwest of the island’s centre. This ignimbrite was deposited in a radial fashion over the entire island, with associated Plinian fall deposits up to 0.5 m thick on islands >40 km away. Common subrounded and frequently cauliform scoria bombs throughout the ignimbrite attest to a small degree of marginal magma– water interaction. The common intense welding of the coarse-grained eruptive products, however, suggests that the majority of the erupted magma was hot, waterundersaturated and supplied at high rates with moderately low fragmentation efficiency and low levels of interaction with external water. We propose that the development of a

water-saturated dacite body at shallow (<6 km) depth resulted in failure of the chamber roof to cause sudden evacuation of material, producing a Plinian eruption column. Following a brief period of quiescence, largescale faulting in the southeast of the island produced a second explosive phase believed to result from recharge of a chemically distinct magma depleted in incompatible elements. This similar, but smaller eruption, emplaced the ‘Hokula’ Ignimbrite sheet in the northeast of the island. A maximum total volume of 8 km3 of juvenile material was erupted by these events. The main eruption column is estimated to have reached a height of ∼12 km, and to have produced a major atmospheric injection of gas, and tephra recorded in the widespread series of fall deposits found on coral islands 40–80 km to the east (in the direction of regional upper-tropospheric winds). Radiocarbon dating of charcoal below the Tofua ignimbrite and organic material below the related fall units imply this eruption sequence occurred post 1,000 years BP. We estimate an eruption magnitude of 2.24×1013 kg, sulphur release of 12 Tg and tentatively assign this eruption to the AD 1030 volcanic sulphate spike recorded in Antarctic ice sheet records.

Editorial responsibility: J.D.L. White

Keywords Tofua . Mafic . Plinian . Tonga . High-grade . Ignimbrite . Caldera

J. T. Caulfield (*) : S. P. Turner GEMOC National Key Centre, Dept. of Earth and Planetary Sciences, Macquarie University, 2109 Sydney, NSW, Australia e-mail: [email protected] S. J. Cronin Institute of Natural Resources, Massey University, Palmerston North, New Zealand L. B. Cooper Institute of Geochemistry and Petrology, ETH, Zurich, CH 8092, Switzerland

Introduction Recognition of the importance and large scale of explosive volcanism at mafic-intermediate centres has increased markedly in the last few decades (Williams 1983; Foden 1986; Sigurdsson and Carey 1989; Freundt and Schmincke 1995; Torres et al. 1995; Allen 2005; Palladino and Simei 2005). This is particularly because mafic eruptions are now

1260

recognised as extremely important climate forcing events because they inject large volumes of sulphur- and halogen-rich volcanic gases into the upper atmosphere (Thordarson and Self 2003; Witter and Self 2007). Reconstruction of explosive mafic eruptions from their deposits is often difficult: evidence of fragmentation style is poorly preserved when activity at this scale produces highly welded ignimbrite but is well preserved when it generates Plinian fallout. Triggering mechanisms hypothesised for highly explosive mafic eruptions commonly include water–magma interaction or basaltic intrusion into/replenishment of a storage chamber. Early successions produced by caldera-forming eruptions of this style are commonly characterised by an abundance of hydrothermally altered lithic/country rock clasts. Subsequent eruption progression and dynamics depend on the mechanism and pace of caldera collapse and resulting modification of vent geometry (Druitt and Sparks 1984; Rosi et al. 1996; Scandone 1996; Roche and Druitt 2001). Two main endmember modes of caldera collapse have been identified, associated with what have been referred to as overpressured and underpressured calderas (Marti et al. 2009). In the case of chamber overpressure, arcuate ring faults form prior to the onset of eruption in response to localised stresses, and early in the eruption a small volume of magma may be ejected through these restricted arcuate cracks, producing a firefountaining phase during which marginal, degassed magma is displaced by hydraulic action. Subsequent failure of the chamber roof overlying a shallow magma body rapidly decreases magma pressure, triggering paroxysmal volatile exsolution and expansion that drives a climatic pyroclasticflow forming stage via collapse of a high Plinian eruption column or spill-over of dense, low and hot plumes (Branney and Kokelaar 1992; Allen 2005). In contrast, underpressure calderas form by subsidence along ring faults during chamber recovery following an earlier eruptive phase (Marti et al. 2009). Studying the nature of volcanic deposits and their distribution within the associated volcanic successions provides a means of reconstructing eruption characteristics whilst distinguishing among products of discrete eruptive phases (e.g. Palladino and Simei, 2005). Examination of Tofua successions was carried out in the early 1970s as a series of exploratory studies (Bauer 1970; Baker et al. 1971; McReath 1972). Following subsequent advances in work on classification, nomenclature (Sparks et al. 1973; Druitt and Sparks 1984) and emplacement mechanisms (Branney and Kokelaar 1992; Calder et al. 2000) of explosive pyroclastic deposits, it is pertinent to reassess the mafic volcanic deposits of Tofua. Also, new information is needed to improve our understanding of local eruption hazards in Tonga, and potentially to identify events that have had a global impact on atmospheric chemistry. Accordingly, we present the results of field

Bull Volcanol (2011) 73:1259–1277

mapping, facies analysis, geochemistry and volatile budgets of Tofua volcano, Tonga, southwest Pacific. Here, we recognise two successions of variably welded ignimbrites representing paroxysmal VEI 5–6 eruptions, including a caldera-forming event within the last 1,000 years. The stratigraphic exposure afforded on Tofua offers a unique opportunity to constrain eruption triggers, dynamics and timing of these highly explosive mafic caldera-forming eruptions—events at large scales that, until now, have remained overlooked within the Tofua Arc.

Tofua Island Tofua Island is located mid-way along the active eastern Tofua arc (insert, Fig. 1). The island is weakly elliptical with a long axis of 8 km oriented northwest–southeast. It contains a 4-km diameter caldera containing in the south and east a freshwater lake (Fig. 1). The present-day caldera rim ranges between 426 and 500 m above sea level, whilst the lake is perched 25 m above sea level. Within the caldera, the northernmost ‘Lofia’ cone remains active. Exploratory geological work was carried out on Tofua in the early 1970s (Bauer 1970; Baker et al. 1971; McReath 1972). Bauer (1970) provided the most detailed and descriptive account of the stratigraphy and geologic history. Four major units were distinguished: (1) the Hamatua Formation comprising basaltic/andesitic/dacitic pyroxene bearing lavas of precaldera age; (2) the Hokula Froth Lava [andesitic]; (3) the Kolo Formation composed of lapilli fall and lapilli-tuffbreccia, tuff, unconsolidated ash and cinder, small basaltic andesite lava flows, and one thick pyroxene andesite lava flow; and (4) the Lofia Formation, 90% ash fall and tuff, 10% basaltic/andesitic lava flows. The accounts of the preand post-caldera deposits remain geologically sound, but much of the descriptive terminology and inferred depositional processes surrounding the explosive facies is outdated. Following Bauer (1970), the four main chronological subdivisions are retained with the Hokula Froth Lava, re-named as the vesicular platform lava and the Kolo formation now being interpreted to comprise the caldera-forming Tofua pyroclastic flow deposits and the younger Hokula flow interpreted as pyroclastic flow deposits from a later event associated with continued caldera deepening. Radiocarbon dating of a carbonised log recovered from the base of the Tofua ignimbrite sequence on the west coast, 200 m south of Hota’ane, gave an age of 970±50 year BP (Shotton and Williams 1971) corresponding to a calendrical age range of AD 1025–1210 (using OxCal: Bronk Ramsey 2008). Three post-caldera cones form an intra-caldera construct near its northern margin. Two extinct cones have been breached to the south by blocky lobate lava flows that terminate at the Lofia crater lake. Pyroclastic deposits from

Bull Volcanol (2011) 73:1259–1277

1261

vegetation

N

5

A’ 6

7

To eastern islands 30 km

Hokula

4

recent pyroclastic deposits pyroclastics post-caldera lavas

3 mantling breccia

caldera deepening Hokula ignimbrite mantling spatter intervening deposits

1

Tofua ignimbrite

caldera forming

2

Manaka

Lofia crater lake Hota'ane 178 W

southern fissure zone

174 W

Bas in sp read Tofu ing c a vol entr can e ic arc

Tafahi T57 dacite

Fonualei

18 S

Hamatua

Late

B’

tren

22 S

ch

Lau Rid

ge

Lau

Kao Tofua

22 S

1km field sections #1-7

178 W

slump scar

post-caldera cones

lava lobe front caldera rim

fissure cliff exposure

normal fault

coastal tuff cone

active 'Lofia' cone

settlement

174 W

Elevation (m)

A’ 500 400 300 200 100 0

B’ Lofia crater section bend Lofia crater lake

Fig. 1 Tofua Island map detailing the major volcanic units and their distribution. 1–7 Field sections referred to in the text. Cross-section A′–B’ represents the island cross-section depicted in insert. Insert shows location

of Tofua Island (grey star) within the Tofua volcanic arc. Note section 8 is located offshore on Ha’afeva island 30 km to the east. Vertical exaggeration on cross-section is ×2

the currently active Lofia cone clearly form a mantle of tephra over the northwest caldera, reflecting the direction of

prevailing low-level (<3 km) winds (Fig. 1). This sequence is up to 3 m thick along the main dispersal axis and consists

1262

of sets of accretionary lapilli beds in thinly bedded surge and fall deposits, punctuated by frequent bomb-sag structures and lithic blocks up to 15 cm in diameter. SiO2 in whole-rock samples associated with Tofua's explosive caldera eruptions ranges between 52.45 wt.% and 56.37 wt.% (Caulfield et al. 2008). A single lava flow of pyroxene-bearing dacite (64.97 wt.% SiO2), in the Hamatau Formation of Bauer (1970), intercalated with pre-caldera lavas is the only evidence of the development of significantly evolved magmas on Tofua.

Stratigraphy of the caldera pyroclastic sequence Field survey The most important new first order interpretation is redefinition and reinterpretation of the lower ‘Tofua’ (island wide) and upper ‘Hokula’ (localised) informal units as ignimbrites, formed predominantly of vesicular juvenile material (pumice and shards), with features indicating a pyroclastic flow origin (after Sparks et al. 1973). The Kolo Formation can be separated into three sets of units: (1) a lower syn-caldera collapse ‘Tofua’ ignimbrite; (2) intervening fall and surge deposits, and; (3) an upper ‘Hokula’ ignimbrite related to caldera deepening. Sections were logged from three distinct zones (Fig. 1): proximal (caldera rim, sections 1–3); medial (northern coast, sections 4–7), and; distal (east islands, section 8). Representative stratigraphic logs of sections 1, 2, 4, 6, 7 and 8 are given on Fig. 2. The northern fissure zone (section 3, Fig. 1) was mapped across multiple sections and its composite stratigraphy illustrated (Fig. 3). The Tofua ignimbrite is by far the most voluminous unit: offshore observation of outcrops along the island’s perimeter show it is distributed in a radial fashion around the majority of the island with an unknown volume deposited offshore. Accessible, representative sections are located in the intra-caldera summit cliffs (sections 1 and 2) and on the northwest coast (‘Rope’ section 4). The deposit shows gradational lithofacies, both vertically and laterally, expressed by variations in degree of welding, colour and clast size/type. This dominantly high-grade (heavily sintered) deposit is characterised by common cauliform light-coloured pumice and darker scoria clasts. Fluctuation in fragmentation related external cooling may be responsible for both pumice and scoria examples. Alternatively, variable porosity of clasts may result from different degrees of welding compaction—early development of chilled margins on cauliform pumices conceivably reduced further compaction. Intervening deposits consist of a coarse, reversely graded lapilli tuff, along with distinctive bedded lapilli fall and surge units up to 0.5 m thick

Bull Volcanol (2011) 73:1259–1277

preserved in the northern caldera rim (section 3) and within 20 m long lenses (section 4) in parts of the northern coast. The upper Hokula ignimbrite unit is up to 14 m thick and best exposed in the northern fissure zone (section 3) and on the north coast (sections 4, 6, 7). This unit forms an almost continuous sheet-like outcrop from Hokula for over 2.5 km along shore to the west. East of Hokula the deposit is unwelded. From base to top, the unit is internally stratified showing a gradation in colour, degree of welding and clast concentration/type with no evidence for unconformities. It contains up to 25% lithics and displays a much greater lateral variation in degree of welding than the underlying Tofua ignimbrite. This lateral (and vertical) variability in welding of the pyroclastic deposits is common throughout all of the examined sections. High-grade welding is typically associated with the sites of thickest deposition. The degree of welding, however, varies considerably between sites and with stratigraphic height along any one continuous section over lateral distances of 20–50 m. Typically, juvenile matrix material is ash in both ignimbrites. Loss of primary pore space is common and greatest in heavily sintered zones. Essential lapilli and bombs are moderately to highly vesicular in the Tofua ignimbrite with diffuse dense margins containing a high density of fine cracks, suggestive of rapid exterior cooling of ductile juvenile material (Cole et al. 2005). In the upper Hokula ignimbrite lapilli and bomb-sized pyroclasts are rare (<5%), finely vesicular and contain abundant angular lithic fragments. Lithic clasts are angular to sub-rounded aphyric blue-grey lava fragments 4–8 cm in length. Spatter clasts are found within ≤100 m of the caldera rim, immediately beneath the proximal Hokula deposits. Clasts are elongate and >30 cm long, with increasing vesicularity towards clast centres. Proximal facies (caldera rim) Sections 1 and 2 are located along high and vertical red cliffs of the northeast and east of the inner caldera rim. Here, competent exposures of welded ignimbrite display distinct lateral and vertical lithofacies variation over tens of metres, distinguished by the degree of welding and clast abundance/type. In section 1, the exposed Tofua ignimbrite faces show irregular columnar cooling joints spaced 2–3 m apart. The lower 6 m are composed of red pyroclastic flow units distinguished by different porosities. Dark red/purple juvenile scoria blocks are clearly visible in most beds. A 2.5-m zone rich in scoria lapilli and bombs (50% scoria clasts) is found mid-unit. The degree of welding and flattening of scoria is greatest at the base where a distinctive deposit with a eutaxitic fabric is developed, with individual fiamme visible in outcrop. The top 6 m of Tofua ignimbrite is variably welded both vertically and laterally along

Bull Volcanol (2011) 73:1259–1277

1263

Fig. 2 Stratigraphy (sections 1, 2, 4, 6 and 7) of major ignimbrite outcrops found on Tofua and a correlative offshore locality, Finemui swamp (section 8) on Ha’afeva island, Kotu group. Correlations shown represent the island wide caldera-forming ‘Tofua’ ignimbrite,

the later ‘Hokula’ ignimbrite found only in the north of the island, and post ignimbrite lava and spatter deposits. Bold italics denote plates shown in Figs. 4 and 5. Drafted sections show the maximum clast size, after Cas and Wright (1987)

outcrop. Zones of equal welding intensity cross-flow unit boundaries, i.e. welded zones can include more than one flow unit. Weakly welded zones grade laterally on a 5–10 m scale into extremely high grade, lava-like lithofacies. On the northern caldera rim, the Tofua unit is mantled by a volcanic breccia. The surface of this can be traced down the upper flanks and appears to be continuous with the Hokula ignimbrite outcropping on the north coast. At section 2, a 13-m sequence of red/purple Tofua ignimbrite outcrops in which the degree of welding varies laterally on a scale of 10’s of metres. The lower 3 m of the cliff face is covered by a ramp of the poorly sorted, lithic rich, unwelded Hokula deposit that pinches out against the normally faulted southeast block (Fig. 4a). The unwelded unit is white/orange in colour and poorly sorted with a coarse lithic fraction (up to 60 cm) reflecting the proximal within-caldera location. As shown in Figs. 1 and 4a, between sections 1 and 2 this unwelded Hokula deposit ramps up against the pre-existing Tofua ignimbrite resulting in a subdued vegetated slope, relative to the steep scarps of the southeast faulted block (Fig. 4a).

Three sections were measured in the northern fissure zone (Fig. 3). Here, deposits overlie an unknown thickness of intra-caldera breccia, and thus post-date the Tofua ignimbrite. The breccia is composed of 40% each of variably coloured hydrothermally altered lithic clasts and poorly vesicular scoria lapilli, both set in a partially lithified orange ash matrix. Clasts are angular, up to 6 cm diameter, and display roughly the full range of silica content known from the island. The central and eastern sections contain an approximately 30-cm thick bed of unconsolidated tephra comprising juvenile scoria lapilli with similar vesicularity to those found in the underlying breccia. Additionally, the western section includes 1.5 m of alternating surge and ash fall beds forming an intervening deposit between the intra-caldera lithic breccia and the continuous scoria lapilli bed. Rootless lava flows and welded agglutinates (samples TF522, TF521, Fig. 3) within the proximal spatter deposits are common and show highly opposing dip directions. Within the upper 1–2 m, individual ropey spatter clasts are visible. In contrast, deposit/layer interiors are welded with thicker sections showing the highest

1264

Bull Volcanol (2011) 73:1259–1277

Fig. 3 Northern fissure zone, section 3. Composite section of the northern fissure zone detailing the proximal stratigraphy in the north of the island. Section correlates stratigraphy with eruptive events discussed in the text. Bold italics denote plate shown in Fig. 4b

grades due to increased loading and heat retention, producing highly domainal deposits (cf. Carey et al. 2008). Clast cores are composed of coalesced and elongate vesicles with iridescent interiors. The degree of vesicularity decreases towards the outer margins, which comprise glassy rims. Angular fragments of aphanitic lava and dense non-juvenile scoria (up to 12 mm) are included in the spatter. Nonjuvenile fragments are identified by their lower vesicularity and blocky shapes (e.g. Heiken 1972). They are very similar to clasts observed in spatter deposits from the mafic Siwi pyroclastic sequence, Tanna, Vanuatu (Allen 2005). Large lithic blocks (up to 2 m in diameter) are found lodged within the spatter deposits. This lithic-rich zone comprises a series of spatter mounds forming a distinctive hummocky topography. Capping the succession, and ubiquitous in the north of the island are a pair of mantling lithic breccias and underlying ash beds (samples TF1277, TF1276, Figs. 3 and 4b), similar to the breccia found at the top of section 1. Clasts are angular, average 6–8 cm in long axis and range from uniform aphyric blue grey lava to orange/red/white

coarse crystalline hydrothermally altered rock. Figure 3 is a composite section representing all the deposits of the northern fissure zone. A south facing slump scar, identified by a marked change in slope angle, is located north of the Lofia cone (Fig. 1). This structure may be the footwall of a buried fault. The fault scar spans almost 800 m east of the Lofia cone; however, exposures are best at its western extent where steep sections lack spatter cover. Numerous breadcrusted bombs litter the ground surface with its ubiquitous cover of medium scoria lapilli. The spatter encrusted Lofia cone is the most likely source of these blanketing deposits. The underlying section is composed of two fine ash/lithic breccia pairs, separated by up to 4 m of inter-bedded, poorly sorted ash and scoria. The lithic clasts are of the same hydrothermally altered types found in the caldera rim volcanic breccias in section 3; here, the maximum clast size reaches up to 50 cm. The section is capped by 2 m of agglutinated spatter, within which individual clast margins are still visible.

Bull Volcanol (2011) 73:1259–1277

1265

a

b section 2

umb l

T

nf UWH

fb

lmb

section 1 100m

c

d cc

T

iu

T

s

Fig. 4 a View 100 m south of section 1 taking in the eastern termination of the normally faulted southeast slump block (fb). Dashed line represents fault plane (nf), the sense of movement is given by the displacement of the Tofua ignimbrite scarp (T) that can be traced the full width of view. Left of the fault line the unwelded Hokula ignimbrite (UWH) forms a ramp up against the caldera wall. To the right, the caldera wall steepens markedly, scarps comprising older shield lavas (l) and the Tofua ignimbrite. b Upper (umb) and

lower (lmb) mantling lithic breccias and underlying ash units. Combined thickness of breccias and ash units is approximately 1 m. c Cauliform juvenile clasts (cc) from the upper reaches of the Tofua ignimbrite unit. d Striations (s) preserved on the upper surface of the Tofua ignimbrite unit (T). Intervening units (iu) comprise fall and surge beds deposited prior to the onset of the second Hokula explosive phase related to caldera deepening

Medial facies (northern coast)

These shear structures record a top-to-the-right sense of motion (as indicated by arrows, Fig. 5a). Above the shear zone, the Hokula unit is visible, distinguished from the Tofua ignimbrite by a much greater vertical and lateral variation in the development of welding and a marked increase in lithic clast content. The basal 2 m is composed of fine-grained ash-rich pyroclastic deposits with common fine (2–4 cm) scoria and lithic lapilli distributed throughout. Within this portion of the unit, a localised clast-rich lens composed of accessory blocks outcrops at a high angle to the surrounding horizontal flow fabric (Fig. 5b). This is fines-poor and contains rounded scoria blocks up to 30 cm in diameter. Above 2 m a series of fine-grained, undulose ignimbrite flow units are distinguished by sharp depositional boundaries, topped by a fines-poor, block and lapilli-rich horizon. The upper flow unit consists of 4 m of red/brown intensely welded ignimbrite, containing grey scoria clasts and a zone enriched in sub-rounded lithic clasts at its base (Fig. 2). Here, the depositional facies again is inferred to comprise a series of overlapping lobes, represented by intercalated lenses each up to 2.5 m thick (Fig. 5b). A well-developed

In section 4, the base of the Tofua ignimbrite is exposed and separated from the underlying vesicular platform lavas by an erosional unconformity. Here, the Tofua deposit is 7.5 m thick. The basal 2 m is very intensely welded, containing almost no visible clast remnants or fiamme; away from the measured outcrop, this unit continues to an unknown depth below sea level. Between 2 and 5 m the volumetric proportion of larger clasts (8–10 cm) reaches 30% comprising non-juvenile scoria and <5% well-rounded lithic clasts. The upper 5–7.5 m is dominated by creamcoloured cauliform pumice/scoria bombs (up to 35 cm) and coarse lapilli set in a pale grey heavily sintered ash matrix (Fig. 4c). Striations scored into the highly welded upper surface of the lower Tofua deposit are visible in the ‘Rope section’ outcrop (Fig. 4d). Atop the Tofua unit an up to 0.5 m thick succession comprising a coarse, reversely graded lapilli tuff and overlying well-bedded lapilli fall and surge units is pervasively fractured. Within these partially welded intervening units, welldeveloped irregular tension gashes are present (Fig. 5a).

1266

Bull Volcanol (2011) 73:1259–1277

a

b

tg

H

H iu a contact

T

T

c

d efb T l EHG

Fig. 5 Volcanic features found within the coastal ‘Rope Section’. a Contact between the Tofua (T) and Hokula (H) ignimbrites showing tension gashes (tg) within intervening fall and surge units (iu). These shear structures are believed to have formed in response to collapse of the underlying Tofua ignimbrite and/or slipping of the overlying Hokula unit in response to post-Hokula chamber recovery. b Intercalated ignimbrite flows (dashed lines) resulting from the pulsatory, intermittent collapse of the Hokula eruption column. A clast rich lens (solid line)

outcrops at a high angle to the adjacent flow, thickening away from the contact and is, therefore, of uncertain azimuth. Note insert marking the location of plate a. c Fallen block from the top of the Hokula unit displaying a strong eutaxitic flow band (efb). The irregular margin to the left of the hammer is composed of a broken block of the black flank lava flow (l) deposited above the ignimbrite in situ. d Extremely high grade (EHG) welded horizon found in the mid-section of the Tofua ignimbrite, west of the measured Rope section

eutaxitic flow band is located approximately 60 cm from the top of the unit (Fig. 5c). Imbricated elongate fiamme (up to 20 cm long) are visible, displaying feathered terminations. The unit is capped by black vesicular lava originating from the lowermost flow front found on the northwest flank (Fig. 1). An eroded tuff cone (>200 m in diameter) underlies the coastal pyroclastic sequence described above at section 5 (Fig. 1). The vent site for the cone was located a short distance off the present day coastline, and preserved in exposure is a cross-section through the cone, passing from the outer wall (left side Fig. 6), through the rim and into a relatively fresh internal crater wall dip slope (right side, Fig. 6). The tuff cone comprises several different facies of finely bedded pyroclastic deposits, typically ash-rich and very poorly sorted. Ashes are frequently highly angular and poorly vesicular. Common exotic lithic fragments occur and in the lower units, coral fragments are present. A coral sample from the base of this cone has been dated at 1,236± 31 years BP (Wk25502), providing a maximum age for

both the tuff and the overlying ignimbrites, consistent with the date from below the ignimbrite on the west coast. The Hokula and Tofua ignimbrite units are here draped over, and lie alongside, the eroded structure, respectively. The onsite observations show that the Hokula units on the eastern side (Hokula) are thin (<2 m) and typically poorly welded; whereas, the Tofua units on the western side have ponded against an eroded contact and the inner wall of the former crater of the tuff cone, are 7 m thick and show highgrade welding. The pyroclastic flow units thin out completely by the summit of the underlying remnant tuff cone. At Hokula locality, the uppermost deposit forms a competent sea cliff (section 6). The lower 8 m is composed of red, fine-grained ignimbrite with poorly sorted small diameter (up to 4 cm) dark red scoria and lithic lapilli. This unit grades upward into a 3 m thick dark red layer which is more strongly welded and rich in large clasts (60% total volume). Irregular jointing on a 2–3 m scale is most pronounced in the uppermost part of this unit, and reflects the very high degree of welding (Fig. 7a). This black

Bull Volcanol (2011) 73:1259–1277

1267

deposit shows coarse-tail grading and is white/orange in colour. This lower section contains the largest sub-rounded lithic blocks (up to 50 cm) and most abundant rounded scoria clasts (up to 30 cm). Immediately above, the deposit grades into a poorly sorted unwelded deposit with a clearly visible scoria-rich zone. The upper 5 m grades into red and finally black immediately beneath the vegetation cover (Fig. 7b). Along this section of coastline, the Tofua ignimbrite does not outcrop and there is also a lack of corresponding deposits within the caldera rim above (rim area west of section 1, see Fig. 1).

Fig. 6 Coastal tuff cone, section 5. Outcrop provides important chronological and physical constraints on eruptive events and conditions. Inner slope of tuff cone dips steeply away from shore and camera. The Hokula (H) and Tofua (T) ignimbrite units can be seen to drape over and beside the eroded structure, respectively. The onsite observations show that the units on the eastern side are thin (<2 m) and poorly welded, draped across the outer flank of the tuff cone; whereas those on the western side have ponded against an eroded contact and the inner wall of the former crater of the tuff cone are 7 m thick and show high-grade welding. The pyroclastic flows thin out completely by the summit of the underlying remnant tuff cone. This evidence suggests the pyroclastic flows were largely hot and mobile, resulting in elongate ignimbrite deposits

fine-grained layer contains 20% vesicular light grey pumice enclaves, sub-circular in cross-section and up to 12 cm diameter. East of Hokula (section 7), the pyroclastic flow unit is unwelded and lies atop the vesicular platform lavas. At the contact, the upper 2 m of lava comprises a ‘rubbly’ zone of vesicular blocks (up to 40 cm across), inferred to be an autobrecciated ‘aa’ cooling crust. The base of the Hokula

Fig. 7 The Hokula ignimbrite. a Hokula welded ignimbrite (section 6) detailing the variation in welding from moderate grade (MG) to extremely high grade (EHG). Irregularly spaced joints (ij) distinguish this pyroclastic deposit from a lava flow. Scarp is approximately 18 m in height. b Coastal section of the unwelded Hokula ignimbrite (section 7), deposited on top of the autobrecciated vesicular coastal platform lavas (ab). Overall the deposit is poorly sorted, except for the strongly coarse tail graded (ctg) base and scoria-rich zone (srz) mid-unit

Distal facies (eastern islands) Between 2 and 20 m of shower-bedded tephra (planar-bedded scoria lapilli showing alternation in grainsize in 3–15 cm beds) and weathered volcanic soils cap the reef limestone or/ and coralline sands of many islands of the Ha’apai group east of Tofua, especially the Kotu subgroup islands closest (30 km) to Tofua and Kao. Within this succession, a thickness of up to 80 cm of bedded scoria lapilli is present. To obtain ages for these scoria lapilli and ash fall units, piston cores (Fig. 2) were taken from swamps in the central portions of Ha’afeva island (c. 45 km east of the centre of Tofua). A radiocarbon age of 970 ± 50 years BP (Rafter Radiocarbon Laboratory, New Zealand) for a pollen separate from below the thickest unit in this sequence (Fig. 2), is consistent with the reported radiocarbon age of 798–983 years BP for the carbonised log beneath the Tofua Ignimbrite (Shotton and Williams 1971; Bronk Ramsey 2008) and also compatible with the chemical correlation observed between bulk samples of Tofua ignimbrite and glass shards from the

a

b

srz

EHG ij

MG ctg ab

1268

Bull Volcanol (2011) 73:1259–1277

66

post-caldera volcanics Hokula ignimbrite intervening spatter base Hokula ignimbrite Tofua ignimbrite Ha'afeva island fall ignimbrite lithics

Mg#

61

56 4

51

1

2

3

46 52.5

53.5

54.5

55.5

56.5

SiO2 wt. %

Fig. 8 Plot of Mg# vs. SiO2 for samples labelled in Figs. 2 and 3. Major element XRF data were determined at Macquarie University and the University of Waikato (NZ). Mg# was calculated using a Fe2O3/FeO ratio of 0.24, the average of Tofua wet chemistry titrations performed by George Jenner (personal communication). This diagram shows that the resurgent volcanics and the ignimbrite clasts clearly form separate parallel differentiation trends. Fields 1–4 represent separate magma batches and order of eruption. The Tofua and Hokula ignimbrites are clearly compositionally distinct, occupying fields 1 and 4, respectively. The most mafic explosive juvenile material is from the base Hokula ignimbrite (field 2), whilst the intervening spatter deposits (field 3) are believed to be derived from an evolved Tofua-like cap. Significantly, the offshore deposits recovered from Ha’afeva island plot only within the Tofua field

Ha’afeva sequence (Fig. 8). The nearly aphyric nature (≤5% phenocrysts) of bulk rock samples makes this comparison possible. In addition, another sequence of 0.3 m fine lapilli fall deposits, showing the same chemical composition, on Fotoha’a island (north of Ha’afeva, also c. 40 km from Tofua) yielded a highly consistent age from dating of an immediately underlying anthropogenic deposit including coral fragments (914 ± 54 years BP; Wk25499).

Structural evidence Faulting has been the major structural control on the evolution of Tofua. Arcuate caldera-bounding faults define the caldera margins in the north and northeast (section 3) and south of the island. The southern fissure zone largely occupies the southeast quadrant of the caldera (Fig. 1). The fissure zone is a vast expanse of spatter deposits comprising four individual spatter cones and agglutinate dikes arranged in a northwest–southeast en-echelon pattern. A ubiquitous layer of scoria lapilli forms the uppermost deposit on the hangingwall block. Spindle bombs up to 25 cm long litter the scoria cover. A steep cliff runs the length of the fissure zone forming a footwall, with the fissure zone itself located atop the hanging wall block. The degree of displacement increases to the south, as evidenced by the thickening of the cliff fault-scarp exposure in the same direction. This

arrangement suggests scissor-like normal faulting, hinged at the most northerly extent of the fissure zone. Smaller scale structures run parallel to the main fault in the north and northeast, forming outside and concentric to the caldera faults (northern fissure zone). These structures were identified by Bauer (1970) as conduits for magma eruption. The proximal accumulation of a large volume of spatter has produced a hummocky topography topped by the pair of mantling lithic breccias. The style and stratigraphic level of deposits suggests that the spatter of the southern fissure zone and the rootless lava/agglutinate of the northern spatter zone are coeval. Magma displacement due to slumping/caldera deepening in the south produced an extensive spatter field site and likely coincided with eruption of a similar, but smaller volume of magma in the north through pre-existing caldera bounding faults.

Chemical composition Elsewhere on Tofua Island, the distinction and correlation of the two ignimbrite units is possible using field observations and geochemistry (Fig. 8). A full set of geochemical analyses from Tofua will be presented elsewhere. As shown in Fig. 8, the Tofua and Hokula data plot in distinct major element compositional fields (fields 1 and 4, respectively). A bulk Hokula ignimbrite sample TF746M from locality 6 has almost identical geochemistry (field 4) to the capping ignimbrite unit found above the Tofua ignimbrite section sampled in the northeast of the caldera wall, 06TF69 (section 1). A deposit with the same composition is not found in section 2; here, the caldera walls are composed entirely of the Tofua ignimbrite. However, the "ramp" of unwelded ignimbrite ramped up against the eastern caldera margin is the lithic rich, proximal equivalent of the unwelded coastal Hokula ignimbrite found east of Hokula (section 7, Fig. 7b). Section 2 marks the easterly extent of the mantling breccia deposits, confirming that the Hokula ignimbrite was cooler and less widespread in the east. The remainder of the caldera wall ignimbrite is correlative with the Tofua unit of the Rope section (Figs. 2 and 8). Importantly, at the contact between the Tofua and Hokula units, geochemical reversals are apparent. In proximal sections, the base of the Hokula unit comprises mafic laminated ash fall (base Hokula, field 2, samples TF1277, TF1276) overlain by relatively evolved basaltic-andesite spatter deposits (intervening spatter, field 3, samples TF522, TF521), before returning to more mafic compositions typical of this unit (field 4). This evidence suggests eruption of chemically distinct magma batches in close succession during the early stages of the second (Hokula) event. Sample TF1063 is the ash matrix of the lithic breccia found in the slump scar (south of section 3, see Fig. 1) and is

Bull Volcanol (2011) 73:1259–1277

geochemically identical to that found in the equivalent deposit from the summit sequence (section 3, Fig. 3). Although texturally diverse, the vast majority of juvenile clasts are chemically identical to their host matrix, e.g. the light-coloured cauliform clasts (06TF56) seen in the dark, intensely welded ignimbrite (06TF57) of the ‘Rope’ section 4 (Fig. 2). This is interpreted to reflect the mingling of a small volume of magma previously quenched (via an external water source) with hot juvenile pyroclasts. Indeed their cauliform shape, the development of internal vesicle gradients, and a lack of compaction even in a high-grade matrix of identical chemical composition, all favour quenching in the shallow conduit. These features may record the interaction of spatially distinct magmatic regimes during formation of pyroclastic flows that produced the ignimbrite deposits. Lithics in the ignimbrite, believed to be fragments of pre-caldera lava, form a linear compositional trend spanning ∼4 wt.% SiO2 (Fig. 8). The pre-caldera dacite plots on the same trend (not shown in Fig. 8), extending the range in SiO2 of pre-caldera lavas to 65 wt.%. Such a range indicates an extended period of differentiation, inferred to have occurred during the cone-building phase. Post-caldera deposits plot at the primitive end of the Tofua array, forming a parallel tend, but over a much smaller range of SiO2. The limited spread in SiO2 may reflect the time elapsed since caldera formation. Samples Ha3-3a and Ha3-3b, from different levels in the same major shower bedded unit on Ha’afeva island (section 8) are chemically indistinguishable from the Tofua (caldera-forming) ignimbrite (field 1, Fig. 8). This unit correlation is supported by the overlapping age ranges obtained from pollen separate/14C dates for organic material recovered from the base of each unit. Furthermore, the tephras are typically shower-bedded in cores and exposures, showing either variations in syn-eruptive wind conditions or, more likely, variations in eruption intensity, consistent with a collapsing and reforming column (Fierstein and Hildreth 1992). There is no evidence in the distal tephra for significant intra-eruption pauses. The thickness (0.5 m) and relatively coarse grain size of the fall units east of the volcano imply eruption columns of sufficient height to enter the higher-level westerly wind streams, i.e. above 3 km in altitude (Reid and Penney 1982). Current modal high-level wind speeds are found to be ∼24 m/s (Reid and Penney 1982); sieving of well-sorted distal fall deposits reveals a median clast size of 4 mm (−2 phi). The 1886 basaltic Plinian fissure eruption of Tarawera provides an analogue model for release heights calculated using horizontal displacement distances, average wind speed and pyroclast fall times (Walker et al. 1984). Applying this model to the deposits of Tofua, we infer a minimum release height of ∼12 km for the calderaforming column.

1269

The incompatible elements Ba and Ce are plotted versus Mg# in Fig. 9. Both plots show depletion in incompatible elements, with anomalous samples taken from laminated ash fall deposits believed to represent the onset of the Hokula phase (base of the Hokula ignimbrite, field 2, Fig. 8). With the convincing case for chemically distinct magma batches, it seems probable that an influx of fresh magma was responsible for triggering the subsequent caldera deepening ‘Hokula’ event.

Volatiles Glass melt inclusions hosted in phenocrysts from small diameter (<12 mm) basaltic andesite Tofua tephras were analysed for volatile contents using SIMS at the Dept. of Terrestrial Magnetism, Carnegie Institute of Washington and for major elements and volatiles using EMP at the Massachusetts Institute of Technology (see Cooper et al. (2010) for set-up and running conditions during analytical sessions). Inclusion hosts were olivine, clinopyroxene, orthopyroxene and plagioclase. The highest H2O and S contents were measured in the most magnesian clinopyroxenes (Table 1). The maximum H2O content was 4.16 wt. % and this inclusion has a S concentration of 869 ppm (Cooper 2009). CO2-H2O saturation curves were calculated for basaltic compositions using VOLATILECALC (Newman and Lowenstern 2002). The calculations indicate that the magma was stored at ∼5 km depth (<2 kbar). Inclusions with the highest H2O contents are hosted within the most primitive phenocrysts, and so the upper values are taken to represent the undegassed source magma. The close agreement between whole rock and inclusion chemistry is consistent with the measured volatile contents being representative of the host magma. The least squares fractionation trends shown in Fig. 9 were used to forward model the H2O content of the intermediate composition tephras to that of the dacite, using Ce as a proxy (see Caulfield (2010) for full modelling rationale). The models estimate a H2O concentration of 6.5 wt.% for the dacitic magma. Comparing these intensive parameters with experimentally determined equilibrium conditions for the Mount St. Helens dacite (Rutherford et al. 1985) and the empirical solubility curves of Moore et al. (1998) it is very likely that the dacite was water saturated at ≤2 kbar.

Interpretation Unit correlation The lowermost unit of the ‘Rope section’ (section 4) is the oldest and largest outcropping ignimbrite on the island, and

1270

Bull Volcanol (2011) 73:1259–1277

a

16

least-squares FC

b

Ce (ppm)

Ba (ppm)

250 200 150

12

8

4

100

2

2 0 30

50 30

40

50

60

70

40

50

60

70

Mg#

Mg# post-caldera volcanics Hokula ignimbrite intervening spatter base Hokula ignimbrite Tofua ignimbrite Ha'afeva island fall ignimbrite lithics

Fig. 9 Incompatible trace element vs. Mg# variation diagrams showing the highly incompatible element a Ba and incompatible element b Ce. Significantly, the depletion in incompatible elements shown by the base Hokula ignimbrite samples (field 2) provides

strong evidence for chamber recharge by a mafic magma genetically unrelated to the Tofua and Hokula compositions. Dashed line represents two-stage least-squares fractional crystallisation model used to estimate the water content of the dacitic magma

is probably the deposit of the caldera-forming ‘Tofua’ eruption. At this location, two distinct cooling packages are clearly distinguishable, separated by heavily sheared intervening units. These units comprise intervening fall and

surge deposits (Fig. 5a), giving evidence for variations in mass eruption rate and style and possibly also indication of a pause in the eruption sequence. Structural evidence suggests major slip motion between the ignimbrites

Table 1 Major element concentrations and volatile contents of melt inclusions, composition of melt inclusion hosts and wholerock major element data for Tofua (06TF70) and offshore (Ha3-3) tephras

Preparation and analysis of melt inclusions and inclusion hosts follows the methods in Cooper et al. (2010) a

Refers to EMP analyses at the American Museum of Natural History (Ha'afeva sample) and the Massachusetts Institute of Technology (Tofua sample) b Refers to SIMS analysis at the Carnegie Institution of Washington

Sample

06TF70

Ha3-3

06TF70

Ha3-3

Sample type Melt inclusion name Melt inclusion host Mg#host (near melt inclusion)a SiO2 (wt%) TiO2 Al2O3 FeO MnO MgO

Melt inclusion cpx4-MIa Clinopyroxene 90.3 52.83 0.30 14.43 9.07 0.12 7.33

Melt inclusion cpx2-MIa Clinopyroxene 84.6 50.49 0.48 17.54 9.74 0.15 5.41

Whole rock

Whole rock

53.08 53.08 53.08 9.67 0.17 5.87

55.30 0.70 16.09 9.75 0.18 4.09

CaO Na2O K2O P2O5 Total LOI H2O (wt%)b CO2 (ppm)b F (ppm)b S (ppm)b S (ppm)a Cl (ppm)b Cl (ppm)a

10.00 0.80 0.14 0.06 95.08

10.12 1.32 0.24 0.06 95.55

11.08 1.63 0.38 0.09 98.22 0.11

9.46 2.03 0.56 0.12 98.28 0.15

4.16 18 119 869 754 748 806

3.19 – 111 833 741 495 595

Bull Volcanol (2011) 73:1259–1277

occurred on the upper surface of the Tofua ignimbrite either during or immediately following the eruption of the second (Hokula) ignimbrite. Tofua has an average flank slope of 35°. Striations at field section 4 imply a SE directed sliding motion along a shallow dip angled in the same direction (Fig. 4d). Grooves preserved on the upper surface of the older ignimbrite imply that it had cooled to a state of near solidity. The cooling time calculated for the Tofua ignimbrite is one day using the AshPac computational model (Miller and Riehle 1994; Riehle et al. 1995). This result is based on an eruption temperature of 900°C and an average deposit thickness of 15 m. A lack of weathering or soil deposits suggests the interval was no longer than a few decades (cf. Chadwick et al. 1999). Fracturing of these intervening units may have occurred due to collapse of the underlying Tofua ignimbrite into the caldera and/or slippage of the overlying Hokula deposit in response to post-Hokula chamber recovery. The Hokula unit is made up of multiple intercalated lenses that are readily distinguishable in outcrop (Fig. 5b), separated by localised segregations of co-ignimbrite lithic breccia (e.g. Freundt and Schmincke 1985; Druitt and Bacon 1986). The intercalated lenses, each up to 2.5 m thick, show that discrete depositional units overlap and were deposited one at a time. We correlate the two mantling breccia/ash pairs on the present caldera rim (Fig. 4b) as the vent-proximal equivalents of the Hokula ignimbrite. These match geochemically, lithologically and geographically by being confined to the north and east caldera rim. The lithic clast fraction shows an average size decrease away from the vent whilst intercalated ignimbrite flow units are further evidence for pulsatory eruptions. Offshore tephras sharing the same geochemistry as the Hokula unit have not been identified, implying that it represents a smaller scale of activity than the caldera-forming Tofua event. Ignimbrite emplacement Both the Tofua and Hokula ignimbrites form single individual cooling units comprising several sub-units bounded by discordant contacts (Fig. 2) and lack evidence for significant intra-event time breaks. Layered, variably welded ignimbrite units are believed to form by progressive aggradation during sustained passage of a single particulate flow with properties varying over time (Branney and Kokelaar 1992). Here, however, we see evidence for multiple lobes and pulses. Changes in source dynamics through time may be reflected in the facies variation recorded within discrete ignimbrite units. For example, that the upper layer of the Tofua ignimbrite (section 4, Fig. 2) displays an enrichment in rounded scoria bombs and lapilli, may reflect an increase in abrasion in response to a rise in the suspended load. The influence of topography is clearly

1271

seen in the mid-level of the Hokula Ignimbrite where a localised clast rich lens outcrops at a high angle to a series of fine-grained discordant welded beds. In detail, the clastrich lens abuts and thickens away from the contact with the adjacent welded deposit documenting the variable, and in this case indeterminable, azimuth of flow (Fig. 5b). A thin bed of imbricated fiamme is located in the upper part of the Hokula deposit (Fig. 5c), providing evidence for the pervasive loss of primary vesicularity due to welding. Moreover, an extremely high-grade band of welded Tofua ignimbrite of variable thickness spans a 1.5 km stretch of the coastline west of section 4. This represents an extremely high-grade zone in which large amounts of internal heat were retained within the ignimbrite body, which thickened in response to the underlying topography (Fig. 5d). These features suggest the extent of welding was dependent upon local topography rather than being vent controlled; notably, the most densely welded zones are not in proximal sections. The depositional lithofacies of the coastal tuff cone buried by the ignimbrites suggests it was formed by pyroclastic density currents and fall deposits from explosive phreatomagmatic eruptions (cf. Chough and Sohn 1990; Sohn 1996). The Tofua and Hokula ignimbrites were deposited from either high energy, highly mobile flows that were unable to deposit on or/and surmount the obstacle of the tuff cone (Fisher et al. 1993), or were emplaced by low-energy flows pinching out by the summit of the underlying tuff cone (Valentine et al. 1992; Giordano et al. 2002). Associated offshore fall units suggest high column heights during at least part of this eruption sequence, supporting the highly mobile alternative. Highgrade welding and high mobility indicate rapid emplacement with little heat loss, implying a hot flow consistent with the mafic/intermediate pyroclast composition and high mass-ejection rates. Welding and compaction has resulted in a significant loss of primary vesicularity in juvenile clasts. For the Hokula unit, the absence of an associated fall unit on distal islands, its restricted distribution, high variability in degree of welding, and structure comprising multiple small lobes indicate that it may, in contrast, have been emplaced with relatively low energy from low, dense plumes spilling over the northern caldera rim. This extremely domainal welding is similar to the ‘local welding’ style described in the 1875 eruption deposits of Askja volcano (Carey et al. 2008). These authors identify the availability of supplied heat and insulation as the primary controls on local welding. This style of welding suggests that the Hokula phase was cooler than the Tofua phase with a lower degree of lateral heat transfer. The proximal mantling breccias probably formed from near vent explosive fall-out (e.g. Walker 1985), forming a coarse veneer of vent- and conduit-derived clasts ripped up during the early stages of the second explosive phase. The

1272

slump scar breccia/ash pairs likely represent equivalent intra-caldera “ramp” deposits, which are inferred to be necessarily thicker because they must occupy the space created by the syn-eruptive subsidence of the caldera floor. Identification of two distinct stages to the eruption sequence makes a piecemeal collapse scenario for Tofua volcano highly improbable (cf. Branney and Kokelaar 1994). The two vent proximal mantling breccia/bedded ash successions, intercalated coastal flows and clast rich lenses forming the upper unit suggest that the Hokula ignimbrite was emplaced via intermittent collapse of an unstable column.

Comparison with other ignimbrites from caldera-forming eruptions Common characteristics of intermediate (broadly andesitic) ignimbrites include: a significant proportion of both lithic and juvenile scoria clasts, evidence for involvement of external water (e.g. chilled clast margins), spatter deposits and a variable degree of welding extending to extremely high grade facies (Mellors and Sparks 1991; Robin et al. 1994a). The mid-fifteenth century Kuwae caldera, Vanuatu, was produced by a series of submarine phreatomagmatic explosions. These phreatomagmatic phases are represented by dominantly basaltic and dacitic units composed of poorly vesicular blocky vitric lapilli and blocks with welldeveloped glassy chilled margins (Robin et al. 1994b). This feature, combined with only rare development of welding under very thick deposit piles (>30 m) may indicate that this pyroclastic regime was overall cooler and wetter than that on Tofua. In contrast, more recent work on proximal deposits finds no evidence for a high intensity (VEI 6) eruption, implying dominantly submarine activity (Nemeth et al. 2007). Preserved vesicular juvenile material within the Tofua deposits is restricted to rounded cauliform pumices, requiring more rapid cooling histories involving external water. Further diversity is seen in the pyroclastic deposits of Gran Canaria, Canary Islands (Ignimbrite D) that is inferred to be the deposit of a drier system (Kobberger and Schmincke 1999). The highly peralkaline rhyolite would have had a similar viscosity to the more mafic compositions considered here and so serves as a useful comparison. Pervasive rheomorphism within the unit recorded a large component of gravitational remobilisation (Kobberger and Schmincke 1999), a feature not seen in the Tofua deposits. The Siwi pyroclastic sequence, Tanna island, Vanuatu, is composed of basaltic andesites and andesites. It contains only isolated welded packages and a much greater volume of accidental material producing highly heterogenous deposits (Allen 2005). Both the Siwi and Tofua sequences contain vent proximal spatter agglomerate, interpreted as ‘fountain’ deposits.

Bull Volcanol (2011) 73:1259–1277

Eruption dynamics and trigger The development of a water-saturated dacite body at <5 km depth is inferred to have led to significant chamber overpressure (Fig. 10a). Associated magmatic upwelling induced roof failure and magma vesiculation. A paroxysmal eruption ensued, producing the main Tofua eruption column (Fig. 10b). Fall deposits on surrounding islands are a testament to the significant height (at least 12 km) attained by the column and require moderately high mass flux rates (Carey and Sparks 1986; Fig. 10b). A Plinianstyle and magnitude is supported by shower-bedded fallout on surrounding islands. Tofua itself is too small to preserve significant fallout, and any deposits present are likely to be hidden below ignimbrite deposits. Based on an estimated original summit elevation of 2,100 m (an average flank slope of 35° and island radius of 3 km), a maximum total volume of 13 km3 of pre-caldera lava was displaced during this eruption (area of volcanic cone and caldera cylinder). At least part of the lithic rich nature of the proximal facies will reflect transport-related sorting. The abundance of hydrothermally altered lithics found in the vent breccia implies that they originated at the border of the conduit: maximum friction at the column margins precluded their entrainment into the central portion of the eruptive column (Freundt and Schmincke 1985). Away from the vent, the abundance of cauliform clasts in the welded ignimbrite units documents an influence of water. A period of quiescence followed the Tofua Ignimbrite eruption. The duration of this period is unknown, but striation grooves on the upper surface of the Tofua Ignimbrite are proof that lithification had proceeded to the point where dragging associated with emplacement of the Hokula unit was preserved. Normal faulting in the southeast of the island produced a slumped ‘fault block’. It is apparent from the present day ∼4° southward slope of the caldera floor that much of the chamber stress inferred to result from faultblock displacement was transferred to and released within the northern fissure zone through pre-existing secondary faults identified by Bauer (1970) as magma conduits (Fig. 10c). At this point in the eruption sequence, the most mafic and incompatible element-depleted ash fall deposits were formed. Significantly, this depletion in incompatible elements shows that this magma is genetically unrelated to the Tofua and Hokula batches. In contrast, subsequent spatter deposits appear to have tapped residual Tofua magma that had undergone further differentiation; they constitute the most evolved juvenile explosive material found on the island. A capping of spatter on the southeast fault block suggests mobilisation of magma through similar extra-caldera faults related to slumping. It is possible, based on the thickness of spatter, the sense of tilt of the caldera floor and the capping of co-ignimbrite lithic breccia in the

Bull Volcanol (2011) 73:1259–1277 Fig. 10 Cross-section reconstruction of events leading to the present day island form depicted in Fig. 1, transect A′–B′. a Development of a water saturated dacite body at <5 km depth likely led to chamber overpressure. Associated magmatic upwelling induced roof failure. b With magma vesiculation a paroxysmal eruption ensued, producing the main Tofua eruption column, collapse of which emplaced the Tofua ignimbrite. c Normal faulting in the southeast of the island produced a slumped ‘fault block’ in response to underpressured conditions and mafic recharge. The base Hokula ignimbrite tapped a distinct, incompatible element depleted composition. Intervening spatter deposits followed, tapping what remained of the Tofua magma, producing the northern and southern fissure zones. d Finally, the Hokula magma was erupted and emplaced the Hokula ignimbrite in the north and east of the island

1273

a original stratovolcano propagation of ring faults

A’

B’ upwelling magma

5 km overpressured chamber

b ash cloud Tofua ignimbrite eruption column

marginal vent breccia

column collapse

A’

B’

overpressured eruption

c

southern fissure zone northern fissure zone

A’

B’

chamber roof block

underpressured chamber

normal faulting in SE quadrant

chamber recovery/mafic recharge

d

Hokula pyroclastic eruption column

A’

B’

tilting ro

of block

Tofua ignimbrite

northern section, that slumping of the SE block was immediately followed by the Hokula eruptive phase in the north of the island. There, rootless lavas are found with highly

fault block

pressure release along ring fault

opposing dip directions on the margins of the northern fissure zone. The outermost margins of the flows contain discernable spatter clasts where heat retention was insufficient to allow

1274

complete coalescence. Flow interiors appear massive and likely formed from agglutinated spatter clasts. Such characteristics are attributed to coalescence of hot, ductile pyroclasts during very rapid welding (Branney and Kokelaar 1992; Sumner and Branney 2002). Variation in vesicularity amongst and within clasts is a function of variable compaction, and degree of post-depositional vesiculation. Angular fragments of aphanitic lava and dense non-juvenile scoria (up to 12 mm) are included in the spatter deposits and likely represent recycled pyroclasts from the Tofua explosive phase. Outsized lava-like boulders lodged in the upper part of the spatter mounds represent eruption of collapsed preexisting bodies of unstable, heavily agglutinated spatter. The occurrence of localised geochemical reversals either side of the Tofua/Hokula contact suggests tapping of different levels of a stratified chamber and/or magmatic mingling as the major explosive phases proceeded. Indeed, Kennedy et al. (2008) have demonstrated that rotation and tilting of a caldera roof block as it sinks can produce a stirring effect, resulting in the eruption of compositionally and physically distinct magmas in close succession (Fig. 10c). A strong component of tilting is in agreement with the large volume of eroded roof block material found in the mantling lithic breccia deposits on the northern rim. Stratigraphic relationships of the Tofua/ Hokula contact are crucial in identifying a likely eruption trigger for the second explosive phase. The basal Hokula ignimbrite tapped a distinct, incompatible element depleted magma, then intervening spatter deposits tapped what remained of the Tofua magma. Finally, the Hokula magma was erupted (Fig. 10d). Eruption of Tofua-like spatter during block slumping, immediately followed by the Hokula phase requires that the distinct (more mafic) Hokula magma had already been emplaced beneath the caldera. Thus, it is entirely possible that the second, caldera deepening, event was triggered by mafic chamber recharge. The post-caldera volcanics extend to lower silica and higher magnesium than any of the juvenile material from preceding explosive phases (Fig. 8). This suggests post-caldera replenishment of mafic magma, which subsequently differentiated to produce a resurgent trend similar to that of the pre-caldera lavas.

Bull Volcanol (2011) 73:1259–1277

(column height) (cf. Parfitt and Wilson 2008). An estimate of volatile emission during an eruption can be made by comparing the volatile contents of melt inclusions with the volatile concentrations measured in matrix glasses of the erupted magma (Witter and Self 2007). A clinopyroxene hosted glass melt inclusion within sample Ha3-3 (distal caldera-forming tephra, section 8, Fig. 2) yielded the highest offshore H2O content of 3.19 wt.%. The same inclusion contains 833 ppm S (Table 1). Fresh matrix glass from a Tofua tephra is found to contain 30 ppm S (Keller et al. 2008). Using the method of Sharma et al. (2004), taking a vesicle-free crystal mass fraction of 0.35 and 2.24×1013 kg as the mass of erupted magma, we estimate a mass of 12 Tg of S. This is equivalent to 14 Tg SO2 loading taking an aerosol conversion efficiency of 60% (e.g. Bluth et al. 1992). Thus, the petrologic budget suggests that degassing of the silicate melt alone resulted in significant aerosol release. Exsolution of sulphur from a magma is however strongly influenced by the presence of CO2-H2O rich vapour in the melt, in which it is highly compatible (e.g. Keppler 1999). The column height and the mafic, volatile rich composition of this magma imply that the paroxysmal Tofua ignimbrite eruption was almost certainly a climate-forcing event, indicating that it should be recorded in ice-core records. We infer that this event may be the source of either the AD 1030 or AD 1257 volcanic sulphate spikes recorded in Antarctic ice sheet records (Budner and Cole-Dai 2003; Zou et al. 2006) on account of their associated southern hemisphere cooling signals (Oppenheimer 2003). Based on our estimates of erupted mass and sulphur release, the caldera forming eruption cloud of Tofua would have emitted 520 ppm S per unit mass of erupted magma. The AD 1257 eruption is believed to have produced 130±30 Tg of S (Oppenheimer 2003 and references therein). If the Tofua eruption was the source of the “great mid-13th century eruption”, its magnitude would have needed to have been 2.5×1014 kg, an order of magnitude greater than predicted. It seems more likely therefore that the AD 1030 date, previously correlated with the 5×1013 kg, 10 Tg eruption of Baitoushan volcano, Northern China (Zielinski 1995), may have originated in the southern hemisphere.

Eruption magnitude and sulphur release Conclusions Based on caldera dimensions (width 4 km, height 0.5 km) we estimate that the volume of juvenile magma erupted could have been as much as 8 km3. Despite the mafic composition of eruptive products, this is comparable to the volume of the 1883 Krakatau eruption of 10.1 km3 DRE (Mandeville et al. 1996) and more than an order of magnitude greater than the 0.3 km3 DRE ejected during the 1980 Mt. St. Helens eruption (Carey et al. 1995). A VEI value of 5–6 is obtained combining our estimates of magnitude (total volume) and intensity

In the last 1,000 years, Tofua Island has undergone two phases of highly explosive activity, emplacing two high-grade ignimbrites, each comprising a series of flow units. The thick, relatively coarse nature of fall units found on islands to the east of Tofua imply that the first of these eruptions produced a plume high enough to enter the westerly stratospheric wind streams at an altitude of ∼12 km. This explosive event is estimated to have had a VEI of 5–6, and given its mafic

Bull Volcanol (2011) 73:1259–1277

composition it likely had an important atmospheric impact. The first and largest event was of highest intensity, likely triggered by development of a H2O saturated dacitic magma. Subsequent interaction of the rising body of magma with external water may have enhanced explosivity. The second event accompanied caldera-deepening as a result of chamber instability, likely accentuated by further recharge of a chemically distinct magma depleted in incompatible elements. The caldera-forming Tofua ignimbrite is of evolved basaltic-andesite composition. The dominantly mafic Hokula ignimbrite displays chemical anomalies in basal parts of ventproximal northern sections, likely recording the concurrent eruption of a distinct primitive magma delivered by recharge, and one with more-evolved composition similar to that of the Tofua ignimbrite. It is possible therefore that a portion of the magma from the Tofua event was retained in the crust and was erupted during the onset of the second explosive phase. Evidence of direct magma–water interaction is recorded only by frequently cauliform clasts in the earliest part of the eruption sequence. The main explosive eruptive phases were moderately fragmental comprising pyroclasts that were largely hot and highly mobile, resulting in elongate ignimbrite deposits. The intensely welded basal section of ignimbrite overlying a coastal tuff cone shows that the Hokula ignimbrite was also hot and mobile. The lower content of dense juvenile clasts containing angular lithic fragments, and a volumetrically greater proportion of lithics in the Hokula unit, suggests eruption of magma through pre-existing conduits and recycling of pyroclasts. The degree of welding of both ignimbrite deposits was mostly dependent on depositional thickness, in turn controlled by local topography. The thickest sections are very strongly welded and preserve extremely little primary pore space. The recognition of these young events in the Tofua arc demonstrates a higher degree of hazard to local communities than previously considered. All of the age estimates for this major eruption (once calibrated), overlap with ages of several major explosive eruptions recorded as sulphate spikes in Antarctic ice cores. Based on evidence presented here, we propose the AD 1030 date as the most likely candidate. Tofua may, then, represent the “smoking gun” for one of the acid peaks in the ice-core records of this era. Acknowledgements JTC wishes to thank Kelepi Mafi and the Tongan Survey for sampling permission and logistical assistance, Graham Smith for invaluable help with fieldwork and the crew aboard the yacht Bliss for essential coastal sampling and photography during the field season. Oded Navon is thanked for helpful discussion and editorial comments. Detailed and constructive reviews provided by Armin Freundt, Jacopo Tacceucci and James White helped greatly in improving the clarity and scope of the manuscript. ST acknowledges the support of an ARC Federation Fellowship. SJC acknowledges the support of FRST Contract MAUX0401 “Learning to live with volcanic risk”. The analytical data were obtained using instrumenta-

1275 tion funded by DEST Systemic Infrastructure Grants, ARC LIEF, NCRIS, industry partners and Macquarie University. This is contribution 719 from the Australian Research Council National Key Centre for the Geochemical Evolution and Metallogeny of Continents (http://www.gemoc.mq.edu.au).

References Allen SR (2005) Complex spatter- and pumice-rich pyrooclastic deposits from an andesitic caldera-forming eruption: the Siwi pyroclastic sequence, Tanna, Vanuatu. Bull Volcanol 67:27–41 Baker PE, Harris PG, Reay A (1971) The geology of Tofua Island, Tonga. Roy Soc NZ Bull 8:67–79 Bauer GR (1970) The geology of Tofua Island, Tonga. Pac Sci 24:333–350 Bluth GJS, Doiron SD, Schnetzler CC, Krueger AJ, Walter LS (1992) Global tracking of the SO2 clouds from the June, 1991 Mount Pinatubo eruptions. Geophys Res Lett 19(2):151–154 Branney MJ, Kokelaar BP (1994) Volcanotectonic faulting, soft-state deformation, and rheomorphism of tuffs during development of a piecemeal caldera, English Lake district. Geol Soc Am Bull 106:507–530 Branney MJ, Kokelaar P (1992) A reappraisal of ignimbrite emplacement: progressive aggradation and change from particulate to non-particulate flow during emplacement of high-grade ignimbrite. Bull Volcanol 54:504–520 Bronk Ramsey C (2008) Deposition models for chronological records. Quat Sci Rev 27(1–2):42–60 Budner D, Cole-Dai JH (2003) The number and magnitude of large explosive volcanic eruptions between 904 and 1865 A.D.: quantitative evidence from a new South Pole ice core. In: Robock A, Oppenheimer C (eds) Volcanism in the Earth's atmosphere. AGU, Washington D.C., pp 165–176 Calder ES, Sparks RS, Gardeweg MC (2000) Erosion, transport and segregation of pumice and lithic clasts in pyroclastic flows inferred from ignimbrite at Lascar Volcano, Chile. J Volcanol Geotherm Res 104(1–4):201–235 Carey RJ, Houghton BF, Thordarson T (2008) Contrasting styles of welding observed in the proximal Askja 1875 eruption deposits II: local welding. J Volcanol Geotherm Res 171(1– 2):20–44 Carey S, Gardner J, Sigurdsson H (1995) The intensity and magnitude of Holocene plinian eruptions from Mount St. Helens volcano. J Volcanol Geotherm Res 66:185–202 Carey S, Sparks RSJ (1986) Quantitative models of the fallout and dispersal of tephra from volcanic eruption columns. Bull Volcanol 48:109–125 Cas RAF, Wright JV (1987) Volcanic successions. Allen and Unwin, New York Caulfield JT (2010) Nature and timing of magma genesis at Tofua Volcano, Tonga. Unpublished PhD thesis, Macquarie University Caulfield JT, Turner SP, Dosseto A, Pearson NJ, Beier C (2008) Source depletion and extent of melting in the Tongan sub-arc mantle. Earth Planet Sci Lett 273:279–288 Chadwick OA, Derry LA, Vitousek PM, Huebert BJ, Hedin LO (1999) Changing sources of nutrients during four million years of ecosystem development. Nature 397:491–497 Chough SK, Sohn YK (1990) Depositional mechanics and sequences of base surges, Songaksan tuffring, Cheju Island, Korea. Sedimentology 37:1115–1135 Cole PD, Fernandez E, Duarte E, Duncan AM (2005) Explosive activity and generation mechanisms of pyroclastic flows at Arenal volcano, Costa Rica between 1987 and 2001. Bull Volcanol 67:695–716

1276 Cooper LB (2009) Volatiles in Tonga Arc magmas and their role in unravelling subduction zone processes. Unpublished PhD thesis, Boston University Cooper LB, Plank T, Arculus RJ, Hauri EH, Hall PS, Parman SW (2010) High-Ca boninites from the active Tonga Arc. J Geophys Res 115:B10206. doi:10210.11029/12009JB006367 Druitt TH, Bacon CR (1986) Lithic breccia and ignimbrite erupted during the collapse of Crater Lake calders, Oregon. J Volcanol Geotherm Res 29:1–32 Druitt TH, Sparks RS (1984) On the formation of calderas during ignimbrite eruptions. Nature 310:679–681 Fierstein J, Hildreth W (1992) The plinian eruptions of 1912 at Novarupta, Katmai National Park, Alaska. Bull Volcanol 54 (8):646–684 Fisher RV, Orsi G, Ort M, Heiken G (1993) Mobility of a largevolume pyroclastic flow—emplacement of the Campanian ignimbrite, Italy. J Volcanol Geotherm Res 56(3):205–220 Foden J (1986) The petrology of Tambora Volcano, Indonesia: a model for the 1815 eruption. J Volcanol Geotherm Res 27:1– 41 Freundt A, Schmincke H-U (1985) Lithic-enriched segregation bodies in pyroclastic flow deposits of Laacher See volcano (East Eifel, Germany). J Volcanol Geotherm Res 25:193–224 Freundt A, Schmincke H-U (1995) Eruption and emplacement of a basaltic welded ignimbrite during caldera formation on Gran Canaria. Bull Volcanol 56:640–659 Giordano G, De Rita D, Cas RAF, Rodani S (2002) Valley pond and ignimbrite veneer deposits in the small-volume phreatomagmatic 'Peperino Albano' basic ignimbrite, Lago Albano maar, Colli Albani volcano, Italy: influence of topography. J Volcanol Geotherm Res 118(1–2):131–144 Heiken GH (1972) Morphology and petrology of volcanic ashes. Geol Soc Am Bull 83:1961–1988 Keller NS, Arculus RJ, Hermann J, Richards S (2008) Submarine back-arc lava with arc signature: Fonualei Spreading Centre, northeast Lau Basin. Tonga J Geophys Res. doi:10.1029/ 2007JB005451 Kennedy BM, Jellinek AM, Stix J (2008) Coupled caldera subsidence and stirring inferred from analogue models. Nature Geosci 1:385–389 Keppler H (1999) Experimental evidence for the source of excess sulfur in explosive volcanic eruptions. Science 284:1652–1654 Kobberger G, Schmincke H-U (1999) Deposition of rheomorphic ignimbrite D (Mogan Formation), Gran Canaria, Canary Islands, Spain. Bull Volcanol 60:465–485 Mandeville CW, Carey S, Sigurdsson H (1996) Sedimentology of the Krakatau 1883 submarine pyroclastic deposits. Bull Volcanol 57:512–529 Marti J, Geyer A, Folch A (2009) A genetic classification of collapse calderas based on field studies, and analogue and theoretical modelling. In: Thordarson T, Self S, Larsen G, Rowland SK, Hoskuldsson A (eds) Studies in volcanology: the legacy of George Walker Special Publications of IAVECI, 2. pp 249–266 McReath I (1972) Petrogenesis in Island Arcs with particular reference to the Tonga and South Sandwich arc and Deception Island, together with a study of minerals from plutonic blocks from the Carribbean. Unpublished PhD thesis, Leeds University Mellors RA, Sparks RS (1991) Spatter-rich pyroclastic flow deposits on Santorini, Greece. Bull Volcanol 53:327–342 Miller TF, Riehle JR (1994) A user's manual for ashpac: a program for predicting cooling, outgassing and compaction of pyroclastic deposits: Tech. Rep. TR94-14. Applied Research Lab Penn State Univ, State College:47pp Moore G, Vennemann T, Carmichael ISE (1998) An empirical model for the solubility of H2O in magmas to 3 kilobars. Am Miner 83:36–42

Bull Volcanol (2011) 73:1259–1277 Nemeth K, Cronin SJ, White JDL (2007) Kuwae caldera and climate confusion. The Open Geology J 1:7–11 Newman S, Lowenstern JB (2002) VolatileCalc: a silicate melt-H2OCO2 solution model wirtten in visual basic for excel. Comp Geosci 28:597–604 Oppenheimer C (2003) Ice core and Paleoclimate evisence for the timing and nature of the great mid-13th Century volcanic eruption. Int J Climatol 23:417–426 Palladino DM, Simei S (2005) Eruptive dynamics and caldera collapse during the Onano eruptionn, Vulsini, Italy. Bull Volcanol 67:423– 440 Parfitt EA, Wilson L (2008) Fundamentals of physical volcanology. Blackwell, London, p 230 Reid SJ, Penney AC (1982) Upper-level wind frequencies and mean speeds for New Zealand and Pacific island stations. NZ Meteorol Serv Misc Pub 174 Riehle JR, Miller TF, Bailey RA (1995) Cooling, degassing and compaction of rhyolitic ash flow tuffs: a computational model. Bull Volcanol 57:319–336 Robin C, Eissen J-P, Monzier M (1994a) Ignimbrites of basaltic andesite and andesite compositions from Tanna, New Hebrides Arc. Bull Volcanol 56:10–22 Robin C, Monzier M, Eissen J-P (1994b) Formation of the midfifteenth century Kuwae caldera (Vanuatu) by an initial hydroclastic and subsequent ignimbritic eruption. Bull Volcanol 56 (3):170–183 Roche O, Druitt TH (2001) Onset of caldera collapse during ignimbrite eruptions. Earth Planet Sci Lett 191:191–202 Rosi M, Vezzoli L, Aleotti P, De Censi M (1996) Interaction between caldera collapse and eruptive dynamics during the Campanian Ignimbrite eruption, Phlegraean Fields, Italy. Bull Volcanol 57:541–554 Rutherford MJ, Sigurdsson H, Carey S, Davis A (1985) The May 18, 1980, eruption of Mount St. Helens: melt composition and experimental phase equilibria. J Geophys Res 90:2929– 2947 Scandone R (1996) Factors controlling the temporal evolution of explosive eruptions. J Volcanol Geotherm Res 72:71–83 Sharma K, Blake S, Self S, Krueger AJ (2004) SO2 emissions from basaltic eruptions and the excess sulfur issue. Geophys Res Lett 31:L13612. doi:11029/12004GL019688 Shotton FW, Williams REG (1971) Birmingham University radiocarbon dates. Radiocarbon 13:141–156 Sigurdsson H, Carey S (1989) Plinian and co-ignimbrite tephra fall from the 1815 eruption of Tambora volcano. Bull Volcanol 51:243–270 Sohn YK (1996) Hydrovolcanic processes forming basaltic tuff rings and cones on Cheju Island, Korea. Geol Soc Am Bull 108 (10):1199–1211 Sparks RS, Self S, Walker GPL (1973) Products of ignimbrite eruptions. Geology 1:115–118 Sumner JM, Branney MJ (2002) The emplacement history of a remarkable heterogenous, chemically zoned, rheomorphic and locally lava like ignimbrite: 'TL' on Gran Canaria. J Volcanol Geotherm Res 115:109–138 Thordarson T, Self S (2003) Atmospheric and environmental effects of the 1783–1784 Laki eruption: a review and reassessment. J Geophys Res 108(D1):4011. doi:4010.1029/ 2001JD002042 Torres RC, Self S, Punongbayan RS (1995) Attention focuses on Taal: decade volcano of the Philippines. EOS 76(24):241–248 Valentine GA, Wohletz K, Kieffer SW (1992) Effects of topography on facies and compositional variation in caldera related ignimbrites. Geol Soc Am Bull 104:154–165 Walker GPL (1985) Origin of coarse lithic breccias near ignimbrite source vents. J Volcanol Geotherm Res 25:157–171

Bull Volcanol (2011) 73:1259–1277 Walker GPL, Self S, Wilson L (1984) Tarawera 1886, New Zealand—a basaltic Plinian fissure eruption. J Volcanol Geotherm Res 21:61–78 Williams SN (1983) Plinian airfall deposits of basaltic composition. Geology 11:211–214 Witter JB, Self S (2007) The Kuwae (Vanuatu) eruption of AD 1452: potential magnitude and volatile release. Bull Volcanol 69:301– 318

1277 Zielinski GA (1995) Stratospheric loading and optical depth estimates of explosive volcanism over the last 2,100 years derived from the Greenland Ice Sheet Project 2 ice core. J Geophys Res 100:20937–20955 Zou L, Li Y, Cole-Dai J, Tan D, Sun B, Ren J, Wei L, Wang H (2006) A 780-year record of explosive volcanism from DT263 ice core in east Antarctica. Chin Sci Bull 51:2771–2780