Sumisu volcano, Izu-Bonin arc, Japan: site of a silicic caldera-forming eruption from a small open-ocean island

Abstract

Sumisu volcano was the site of an eruption during 30–60 ka that introduced ∼48–50 km³ of rhyolite tephra into the open-ocean environment at the front of the Izu-Bonin arc. The resulting caldera is 8 × 10 km in diameter, has steep inner walls 550–780 m high, and a floor averaging 900 m below sea level. In the course of five research cruises to the Sumisu area, a manned submersible, two ROVs, a Deep-Tow camera sled, and dredge samples were used to study the caldera and surrounding areas. These studies were augmented by newly acquired single-channel seismic profiles and multi-beam seafloor swath-mapping. Caldera-wall traverses show that pre-caldera eruptions built a complex of overlapping dacitic and basaltic edifices, that eventually grew above sea level to form an island about 200 m high. The caldera-forming eruption began on the island and probably produced a large eruption column. We interpret that prodigious rates of tephra fallback overwhelmed the Sumisu area, forming huge rafts of floating pumice, choking the nearby water column with hyperconcentrations of slowly settling tephra, and generating pyroclastic gravity currents of water-saturated pumice that traveled downslope along the sea floor. Thick, compositionally similar pumice deposits encountered in ODP Leg 126 cores 70 km to the south could have been deposited by these gravity currents. The caldera-rim, presently at ocean depths of 100–400 m, is mantled by an extensive layer of coarse dense lithic clasts, but syn-caldera pumice deposits are only thin and locally preserved. The paucity of syn-caldera pumice could be due to the combined effects of proximal non-deposition and later erosion by strong ocean currents. Post-caldera edifice instability resulted in the collapse of a 15° sector of the eastern caldera rim and the formation of bathymetrically conspicuous wavy slump structures that disturb much of the volcano’s surface.

Introduction

The Izu-Bonin arc extends from Hakone volcano on the Japanese island of Honshu to the Ogasawara Islands, 1,000 km to the south (Fig. 1). A chain of volcanic islands, extending from Izu-Oshima to Aogashima, is located along the northern half of the arc. Farther south, islands are less common and smaller, and the volcanic front is defined by a series of submarine caldera volcanoes.

Fig. 1

Map showing location of Sumisu volcano and other seafloor features south of Japan

We here describe and interpret the Sumisu volcano, the largest of these southern submarine caldera volcanoes and where ∼48–50 km³ of rhyolite tephra was erupted during 30–60 ka. We describe the geological development of the caldera area and interpret the caldera-forming eruption, particularly the ways in which large volumes of pumiceous tephra were dispersed in the open-ocean, arc-front environment. We examine the possibility that Sumisu eruptions may have been the source of the thick pumice deposits encountered during ODP Leg 126 seafloor drilling, 70 km to the south and southeast.

The present research, plus the small number of similar studies in today’s oceans (e.g. Carey et al. 2000; Fiske et al. 2001; Smith et al. 2003; Wright et al. 2003, 2006) are steps toward better understanding of large-scale, submarine silicic volcanism. Collectively, these studies complement those of similar volcanism on land, which has commanded much attention over the past several decades. In addition, studies of present-day seafloor silicic calderas will permit better interpretation of submarine caldera-forming eruptions and their products in ancient terrains, where deformation and metamorphism commonly mask original textures and structures. The association of economically important volcanic-hosted massive sulfide deposits with these volcanic centers (e.g. Gibson et al. 1999; Hudak et al. 2003; Mueller et al. 1994) underscores the need to know more about silicic submarine volcanism.

Previous Sumisu related studies

Recent bathymetric studies along the front of the Izu-Bonin arc have identified seven submarine caldera volcanoes between the islands of Izu-Oshima and Torishima (Murakami and Ishihara 1985; Nagaoka et al. 1991; Yuasa et al. 1991). Volcanological studies in the Sumisu area, using manned submersibles, remote operating vehicles (ROVs), camera sleds (Deep-Tow), and dredging, began in the 1980s in conjunction with the seafloor drilling projects of ODP Leg 126. (e.g. Taylor et al. 1990; Klaus et al. 1992). Iwabuchi (1999) carried out the first manned submersible investigations in the Sumisu area, and the present study builds on his pioneering work. Detailed petrological and geochemical studies of the Sumisu volcano have been presented by Tamura et al. (2005) and Shukuno et al. (2006).

The name Sumisu has interesting origins (S. Oshima, written communication, 2004). The first recorded Western contact with Sumisu took place in 1851, when the crew of an American sailing ship, captained by a Mr. Smith, discovered a 136-m-high sea stack and named it Smith Island (Fig. 2). The Japanese retained the name Smith but transliterated it to Sumisu, the name still in use today.

Fig. 2

Eastern wall of Sumisu Island; its summit is 136 m high

Field and analytical methods

Fieldwork was carried out during five cruises operated by the Japan Agency for Marine-Earth Science and Technology (JAMSTEC). Locations of our surveys, the equipment used, and shipboard operations are presented in Fig. 3 and Tables 1 and 2. Fieldwork utilizing a manned submersible, ROVs and a Deep-Tow camera sled differs in important ways from studies on land. The illumination for reliable observation is limited, and dives last for only a few hours. In addition, it is difficult to linger at an outcrop for detailed observations, because ROVs and Deep-Tows are tethered to moving ships. Instability of submersibles and ROVs can make detailed, in situ sampling difficult, especially on steep cliffs. On the contrary, these limitations are offset by the ability to investigate caldera walls from floor to rim. Many modern subaerial calderas are filled with lakes, limiting access to complete caldera-wall sections and post-caldera features.

Fig. 3

Bathymetry of Sumisu volcano, contour interval 100 m. a Proximal areas dominated by the caldera and the wavy slump structures that disturb the volcano’s surface. Blue arrows point to post-caldera cones. Dashed squares (A) and (B) outline areas shown in Fig. 10. Bathymetric color-key shown in Fig. 4. b Sampling and dive-track locations

Table 1 Cruise data in the Sumisu area

Table 2 Data relating to submersible/ROV dives, dredge hauls, Deep-Tow camera traverses

Single-channel seismic (SCS) surveys employed a 350 in.³ GI-gun with shot intervals of 6.9 and 8.0 s. Onboard data processing was done using PRO-MAX and SPW (PARALLEL GEOSCIENCES, Inc., U.S.A.). Bathymetric data were obtained with the multi-beam seafloor swath-mapping systems (SEABEAM2112/SEABAT8160) aboard JAMSTEC’s research vessels.

X-ray fluorescence (XRF) analyses of samples were performed at Institute for Frontier Research on Earth Evolution (IFREE), JAMSTEC, using RIGAKU SIMLTEX 12 for major elements and RIGAKU RIX3000 for trace elements. Analytical methods and conditions are described in Tani et al. (2005). Brine contamination was removed by repeatedly boiling samples in distilled water. Major elements were analyzed on 11-times diluted glass beads, and trace elements were analyzed on pressed pellets. The relative standard deviation (RSD) of the repeated analyses is <0.71% for major elements and <7.2% for trace elements.

Carbon and oxygen isotopes were measured to interpret the origin of carbonate that locally cements caldera-rim pumice and lithic clasts. The analyses were carried out at the Department of Earth and Planetary Science, University of Tokyo. The sample preparation methods and isotope measurement techniques are based on Morishita and Matsuhisa (1984). The carbonate samples were rinsed with distilled water and acetone and later dissolved in phosphate to produce CO₂ gas. The CO₂ gas was then analyzed by a Finnigan MAT252 mass spectrometer to determine ¹³C/¹²C and ¹⁸O/¹⁶O ratios. The data were standardized by PDB and VSMOW. The analytical precision of the delta values is less than ±0.13‰ for both carbon and oxygen isotopes.

Geology of Sumisu volcano

We define Sumisu volcano as the crudely elliptical cluster of submarine volcanoes centered at 31°30′N on the front of the Izu-Bonin arc (Fig. 4). Using the 1,000 m bathymetric contour to define its outline, the volcano measures about 40 × 25 km along E–W and N–S directions, respectively. A nearly unbroken slope to the east extends to the floor of the Izu-Ogasawara trench at ocean depths >9,000 m; the North Sumisu rift basin cuts deeply into the western flank of Sumisu volcano.

Fig. 4

Bathymetric map of the Sumisu–Torishima area; contour interval 100 m

Bathymetric knolls, representing the summits of volcanoes that coalesced to form the complex, dot Sumisu volcano’s surface. The most conspicuous of these, Daini-Sumisu knoll, lies about 6 km west of Sumisu caldera (Fig. 3), and others are roughly clustered along E–W directions. The clustered growth of these small- and medium-size basaltic to silicic volcanoes could have been the precursor to larger-scale magmatism that led to caldera-forming eruptions, as has been interpreted at numerous other volcanic complexes, both subaerial and submarine (e.g. the Crater Lake caldera in Oregon, U.S.A.; Bacon 1983; Bacon and Lanphere 2006, and the submarine Myojin Knoll volcano, 40 km north of Sumisu; Fiske et al. 2001). This interpretation is strengthened by ROV and Deep-Tow observations of Sumisu’s caldera walls, where cross sections of coalesced volcanoes are exposed.

Sumisu Island

The 136-m-high sea stack of Sumisu Island (Fig. 2) is made of reddened and oxidized basalt lavas, scoria, and associated intrusions. Beds dip irregularly a few degrees to the south, supporting the notion that the present-day feature is a remnant of a larger, pre-caldera island, the summit of which once lay somewhere to the north. Extrapolation of Sumisu’s present-day submarine slopes suggests that the summit of this island was about 200 m above sea level.

Caldera-wall geology

ROV observations (Fig. 5) show that dacitic and basaltic coherent facies (lava flows and intrusions) and associated breccias make up the lower 30–40% of caldera-wall sections. These units are overlain by a wide variety of gently dipping volcaniclastic deposits. Steeply dipping dikes cut all of these units. The dikes are distinguished from the lava flows by the presence of chilled margins and/or zones of sub-horizontal cooling joints. Overall, caldera-wall physiography consists of a series of steep cliffs separated by talus-covered shelves. Selected major and trace element data of caldera-wall samples are presented in Table 3.

Fig. 5

Sumisu caldera-wall sections from ROV traverses; lithologies in covered intervals are interpreted. Numbered dashed lines and number on arrows showing sampling point indicate ocean depth in meters below sea level (mbsl). a Southern wall. b Southeastern wall

Table 3 Major and trace-element data for selected samples from the Sumisu–Torishima area

Massive dacitic coherent units (Fig. 6a) and coarse monomictic breccias make up the bulk of the lower 100–200 m of the southern caldera wall (Fig. 5a, Dive 3K572 and 3K573). The traverse up the southeastern caldera wall encountered a small basaltic volcano, made up of lavas and dikes that is nested between the dacitic volcanoes that make up most of the pre-caldera edifice (Fig. 5b, Dive 3K575). The bathymetric knoll, seen on the southeastern caldera rim and rising to a depth of 200 m, may be the summit of this basaltic volcano. Two clasts of orthopyroxene-bearing basaltic andesite were collected from talus at the base of the southeastern wall, but outcrops of these rocks were not found.

Fig. 6

Facies exposed in Sumisu caldera wall. a Massive dacitic coherent unit; southern caldera wall, 840 mbsl. b Thinly bedded volcaniclastic deposits; southern caldera wall, 440 mbsl. c Fines-poor syn-caldera pumice deposit at top of western caldera wall, 330 mbsl. Widths of each view is uncertain but in the range 2 to 4 m

Upper parts of the southern caldera wall are dominated by buff colored, gently dipping pumiceous volcaniclastic deposits that in general are finer grained than the monomictic breccias below. These units first appear at depths of about 650–550 m and continue intermittently to the top of the caldera wall (Fig. 6b). Bed thicknesses range from a few cm to >5 m. The thick units (>5 m) are massive and consist mostly of fine breccia; these units contain scarce angular to subangular dark lithic clasts, most <20 cm in diameter. The thinner beds range in thickness from about a few centimeters to about 1 m and locally display low-angle truncations and channels. Cross beds were not observed. Several bedded intervals contain local concentrations of dark 10–20 cm subangular lithic clasts. We infer that these upper-caldera-wall deposits consist mostly of resedimented volcaniclastic debris, although some may have been direct products of eruptions.

Syn-caldera pumice and associated deposits

Perhaps the most noteworthy feature of Sumisu volcano’s geology is the paucity of proximal pumice genetically related to its caldera-forming eruption. Thin (2–5 m) deposits of angular, fresh, and pale-buff rhyolite pumice occur only locally at the top of the caldera wall, but these are rare. Shukuno et al. (2006) found the caldera wall to be composed of rocks ranging from basalt to rhyolite (<70.0 wt% SiO₂). In contrast, rhyolitic pumice samples collected from the few syn-caldera deposits found differ from those in the wall and have SiO₂ >71.0 wt% (Table 3). These deposits are unconsolidated or only slightly consolidated, and none are welded or sintered. The pumice clasts, most <25 cm in diameter, are highly vesicular and phenocryst-poor (<1 vol.%). The scarce phenocrysts are plagioclase, clinopyroxene, orthopyroxene, and titanomagnetite (Table 3).

The syn-caldera pumice deposits at Sumisu were found in situ at the following two places only:

• 1) ROV Dive 3K574 encountered a 3-4-m-thick massive, fines-poor pumice deposit capping the western caldera wall (Fig. 6c), but steep cliffs prevented sampling.

• 2) Deep-Tow camera traverse DT-03C found a similar, 2-m-thick fines-poor pumice deposit at the top of the northern caldera wall.

Syn-caldera pumice mantles an extensive area at and near the northeastern caldera rim, but we have no evidence that these accumulations are primary. Dive 3K576 encountered extensive pumice talus mantling the gentle outer slope of the Sumisu edifice from depths of 650 to 320 m. Dive H348 revisited the upper part of this same traverse and documented vaguely bedded, weakly cemented, pumice outcrops preserved in dozens of low, subparallel ribs (Fig. 7a). Accumulations of dense lithic blocks and lapilli were seen between some of the ribs. The long axes of these ribs are aligned along an azimuth of ∼120°, the average orientation of the southeast-directed ocean currents that we infer to have carried away most of the pumice, leaving the subparallel ribs as erosional remnants. The vague internal beds (<30 cm) are parallel from rib to rib, suggesting that continuous pumice lapilli layers once capped the northeastern caldera rim.

Fig. 7

Post-caldera features of Sumisu caldera. a Slightly cemented “rib-like” exposures of resedimented syn-caldera pumice deposit; northeastern caldera rim, 375 mbsl. View is ∼4 m across. b Lithic-rich deposit mantles the southern caldera rim, 380 mbsl. View is ∼3 m across. c Vertical view of dacitic lava dome; southern caldera floor, 895 mbsl; absence of sediment cover suggests young age. View is ∼3 m across

A clast-supported veneer of angular coarse lithic clasts (most <50 cm in diameter) was observed to overlie the syn-caldera pumice deposits described at sites (1) and (2) above, as well as elsewhere over wide areas of the caldera rim (Fig. 7b). This veneer is the youngest caldera-wall unit found and consists of a wide variety of igneous rocks, including aphanitic to porphyritic basalt and dacite. We could estimate the thickness of this veneer only along the two traverses noted above, where it is 1–2 m thick. Elsewhere, its thickness is unknown, because it was observed from above (from ROVs and the Deep-Tow camera sled), and cross sections were not visible.

In places on the western, southern and southeastern caldera wall, this veneer of lithic clasts is cemented by calcite, forming a conspicuously rough-surfaced 1.5–2 m layer of what we informally call gasa-gasa [Japanese for rough]. Thin sections of the gasa-gasa show crystals of calcite cementing the subangular to angular lithic and pumice fragments that are <5 cm in diameter. About 80–90% of the fragments are angular basalt and dacite and the remainder is subangular rhyolitic pumice that may have been derived from the caldera-forming eruption.

The calcite cement was separated for detailed carbon and oxygen isotope study (Table 4). δ¹⁸O_PDB of the calcite ranges from +3.19 to +3.81‰, and δ¹³C_PDB ranges from +2.91 to +3.88‰. The δ¹⁸O values fall in the range of +3.20 to +4.92 reported for the carbonate precipitated from seawater circulating in fissures and fractures within ultramafic rocks at ambient ocean-floor temperatures (Bonatti et al. 1980). We suggest that calcite cement is the product of non-hydrothermal seawater circulation that affected parts of the caldera rim.

Table 4 Carbon and oxygen isotope compositions of calcite that cement lithic clasts mantling the caldera rim (gasa-gasa)

Talus and caldera floor

Talus and tumbled blocks were observed on all caldera wall traverses and conceal 25–40% of caldera-wall sections (Fig. 5). More extensive talus occurs at the base of the wall, and we infer that such deposits surround the entire caldera floor. It is likely that the caldera walls were originally steeper than they are at present, but ongoing wasting is diminishing these slopes and increasing the rim-to-rim diameter of the caldera.

Most of the caldera floor is covered with pumiceous sand and pebbles, although lobes of coarser debris, likely the product of avalanches and grain flows from nearby caldera walls, were commonly observed. Dive 3K574 encountered subangular lava boulders >1 m in diameter that had rolled or skidded onto the floor and traveled for distances of a few hundred meters.

Caldera-floor volcanism

The floor of Sumisu caldera is dotted with post-caldera cones, providing evidence for post-caldera volcanism (Fig. 3a). The 800-m-high cone of Shirane is the youngest and most conspicuous, and its summit lies just 8 m below sea level. Dive 2K1392 visited the western base of Shirane and collected one subangular fragment of dacite (69.2 wt% SiO₂). Since Shirane is a relatively large structure (about 3 km in diameter), it is not certain whether this dacitic composition is representative of the entire Shirane composition. Given its substantial size, however, it is likely to be a polygenetic volcano.

Two cones lie along the southern and southeastern margins of the caldera floor (Fig. 3a). Dive 2K1394 traversed these cones and revealed that they are small-volume extrusive lava domes the surfaces of which are free of sediment cover (Fig. 7c). These domes are mostly composed of dacites having two distinct compositions, 68.1–68.5 wt% SiO₂ (southern dome) and 69.1–69.2 wt% SiO₂ (southeastern dome).

Older knolls, 100–250 m high and completely mantled by pumiceous sand and pebbles, are also present and imply that caldera-floor eruptions have spanned a significant part of post-caldera time. Dive 2K1392 encountered one of these knolls near the center of the caldera. Others lie along the northwestern and northern caldera floor margins (Fig. 3a).

Seismic reflection profiles

Processed reflection profiles and the interpreted geology from two single-channel seismic (SCS) transects are shown in Fig. 8.

Fig. 8

Single-channel seismic profiles; a–b, a’–b’ and c–d, c’–d’ show raw data and interpreted geology, respectively. The deeper, repetitive reflections on a–b and c–d are multiple echoes and should be disregarded

Poor reflections were obtained beneath Sumisu’s knolls and caldera wall, probably because of steep slopes and the absence of suitable reflectors at depth. In contrast, volcaniclastic aprons bordering these features contain good reflectors and permit the following interpretations:

• 1) Seismically defined strata making up the western pre-caldera edifice of Sumisu volcano abut Daini-Sumisu knoll. The shape of this knoll and the lack of strong seismic reflections within it suggest it is a lava dome. This is supported by the Deep-Tow camera observation of Daini-Sumisu’s western slope and its summit (Fig. 3b, DT-9C) that encountered abundant large (<3 m in diameter) angular to subangular pumice clasts mantling its surface.

• 2) Strata forming the apron east of Sumisu caldera (uncolored in A’–B’) thicken dramatically toward the west, suggesting extensive deposition of volcaniclastics prior to the growth of Sumisu volcano.

• 3) The southern apron of the Sumisu complex overlaps, and is therefore younger, than the northern apron of Minami Sumisu volcano.

Discussion

Distal transport of Sumisu tephra: evidence from ODP cores

Nishimura et al. (1991, 1992) described rhyolite pumice layers as much as 118 m thick from ODP Leg 126 cores at Sites 790/791 (in the Sumisu rift), Site 788 (at the crest of the arc front), and Site 793 (on the eastern forearc slope). These sites lie about 70 km south and southeast of Sumisu (Fig. 1). Although these workers noted the unrestricted route for bottom-hugging gravity-current transport from Torishima volcano to the Sumisu rift sites, they favored Minami Sumisu caldera (Fig. 4) as the more likely source, because geochemical data then available suggested similarities of Minami Sumisu pumice clasts with those from the cores. These workers eliminated Sumisu caldera from consideration because a ridge of small knolls separating the northern and southern basins of the Sumisu rift was thought capable of blocking pumiceous gravity currents originating in the north. Klaus et al. (1992), however, presented bathymetry showing that the blocking ridge identified by Nishimura et al. (1992) actually contains a narrow channel through which pyroclastic gravity currents from Sumisu could have traveled to Sites 790/791.

Major- and trace-element data from Sumisu caldera, Minami Sumisu caldera and Torishima obtained by us and others since the early 1990s permit a reevaluation of pumice provenance (Table 3). A plot of SiO₂ versus K₂O (Fig. 9a) shows that the data from all nearby sources lie in the low-K field of Gill (1981). Most pumice samples from Minami Sumisu plot slightly higher than those from Sumisu, Torishima and the Leg 126 samples, but these data are not useful discriminators.

Fig. 9

a K₂O versus SiO₂ (wt%) diagram for silicic rocks (SiO₂ >55 wt%) in the Sumisu–Torishima area. Open circles, Sumisu caldera (this study); solid squares, Minami Sumisu caldera (this study); open squares, pumice samples from Minami Sumisu caldera (Yuasa and Nohara 1992); solid circles, submarine pumice samples from Torishima (K. Tani, unpublished data). The numbered fields show the range of Pleistocene pumice compositions collected at ODP Leg 126 sites 788, 790 and 793 (Gill et al. 1992). Dashed line separates the low-K and middle-K fields of Gill (1981). b Zr/Y versus SiO₂ (wt%) diagram of the same data shown in a

More useful is the incompatible element ratio (Zr/Y; Fig. 9b) which clearly distinguishes Minami Sumisu pumice samples from those obtained at other nearby sites. Minami Sumisu caldera, even though situated at the volcanic front, has geochemical characteristics different from those at the adjacent volcanic-front volcanoes, which, as noted by Hochstaedter et al. (2000), have Zr/Y ratios <2.5. Minami Sumisu caldera is therefore an unlikely source of the thick pumice layers at the ODP sites. The subordinate role of the Minami Sumisu as a source of voluminous seafloor pumice deposits is also implied by the fact that its caldera volume (∼3.8 km³) is about an order of magnitude less than that of the Sumisu caldera (∼36 km³).

Discriminating between Sumisu and Torishima sources is more difficult. From available major and trace element data, pumice samples from both volcanoes are indistinguishable from each other, as well as from the Leg 126 pumice samples. Moreover, pumice samples from all of these sources have the identical phenocryst assemblage of clinopyroxene, orthopyroxene, plagioclase and titanomagnetite, the common assemblage in silicic rocks along the front of the Izu-Bonin arc (Tamura and Tatsumi 2002).

Our findings in the Torishima area, however, help to narrow the uncertainty (JAMSTEC unpublished cruise reports; KR02-16 and KY03-10). Only 4 of 12 dredge hauls near Torishima recovered mixtures of basalt, andesite and rhyolite pumice; the remaining 8 contained no pumice. Ten Deep-Tow camera traverses showed the sea floor surrounding Torishima to be underlain mainly by thick deposits of basalt scoria and coarser volcaniclastic debris, and newly acquired bathymetry shows no convincing evidence for the existence of a submarine caldera. Finally, outcrops on Torishima Island expose rocks ranging from basalt to dacite but no rhyolite.

We therefore conclude that Torishima was an unlikely source of the thick seafloor rhyolitic pumice deposits described by Nishimura et al. (1991, 1992) and infer that much of the thick seafloor pumice encountered at ODP Leg 126 Sites 790, 791, 788, and 793 was erupted from the Sumisu volcano. If true, large volumes of tephra from Sumisu’s caldera-forming eruption traveled a minimum of 70 km to the south and southeast.

Caldera age

We found no direct evidence for the age of Sumisu caldera, but its youthfulness is suggested by the following:

• 1) The Sumisu magmatic system has fed recent eruptions and still emits heat. Iwabuchi (1999) discovered vent communities (including mussels and clams) clustered around warm (12.1°C) seeps near the summit of a caldera-floor cone. The two dacite domes discovered during this study (Dive 2K1394) are young features having no dusting of fine sediment (Fig. 7c). Small Sumisu-area eruptions are reported to have taken place in 1672, 1869, 1870, 1871, and 1873 (Siebert and Simkin 2002).

• 2) Caldera rim planation. Extensive parts of the northern and western caldera rim, lying at ocean depths of 140–300 m, have smooth, nearly-horizontal surfaces suggesting erosional planation during low stands of Pleistocene sea level. The maximum low sea-level stand occurred at about 18 ka (Chappell and Shackleton 1986), implying that the caldera rim (and thus the caldera itself) predates that event.

• 3) Correlations with dated ODP drill cores. Nishimura et al. (1991, 1992) used biostratigraphic- and sediment-accumulation-rate evidence from ODP drill cores in the Sumisu rift to suggest that thick pumiceous bed II and the combined III/IV beds at ODP sites 790/791 were deposited at 31 and 61 ka respectively. Based on the geochemical correlations discussed above, we suggest that Sumisu’s caldera-forming eruption took place during the late Pleistocene, sometime during 30–60 ka.

Edifice instability

Thin-skin slumping

Concentric, wavy slump structures, although superficial, are conspicuous features of Sumisu bathymetry (Figs. 4 and 10) and deserve special attention. Keating et al. (2000) described similar features on submarine aprons flanking the volcanoes in Samoa and called them slump sheets. We prefer the informal term wavy slump structures, because of their map-view resemblance to waves formed by dropping a rock into a pond.

Fig. 10

Wavy slump structures; profiles show stair-step-like morphology. See Fig. 3a for index map. a Enlargement of northeastern slope; contour offsets suggest 1–1.5 km of downslope motion on 1°–7° slopes. b Enlargement of southeastern slope; profile shows slumps moved on 1°–4° slopes

Detailed bathymetric maps of Sumisu’s northeastern and southeastern slopes (Figs. 3a and 10) show these slump structures have crest-to-crest wave lengths of 1–3 km and frontal scarps 50–220 m high. The accompanying profiles show they traveled along 1°–7° slopes (Fig. 10). Many of these slump structures have bathymetrically defined depressions on their upper surfaces, and single channel seismic profiles show that the structures are 70–100 m thick; some display only slightly deformed internal bedding (Fig. 8). We interpret that the slumps resulted from the simultaneous slippage and associated rumpling of the uppermost, unlithified part of the Sumisu edifice. Offsets of bathymetric contours on the northeastern slope of the volcano (Fig. 10a) show that single slumps traveled 1–1.5 km downslope. Manned submersible observations along the upper southeastern slope of the Sumisu edifice (Iwabuchi 1999) showed the surface of these slumps to consist mostly of matrix-free jumbles of lithic clasts ranging in diameter from 10 cm to about 5 m.

The triggering mechanism for the wavy slump structures is unknown. Similar slumps disturb the slopes of Minami Sumisu volcano and the Torishima volcano complex, 25 and 85 km south of Sumisu, respectively, suggesting that these slumps share a common origin. In contrast, single channel seismic profiles across the Myojin Knoll and Myojinsho submarine volcanoes, respectively 80 and 60 km to the north, show the surfaces of these edifices to be undisturbed. We favor the interpretation that one or more large tectonic earthquakes in the Sumisu–Torishima area triggered the wavy slump structures, one of the mechanisms suggested by Keating et al. (2000) for similar features in the Samoa area. Such earthquakes, common in the study area, are related to the nearby subduction beneath the Izu-Bonin trench. If true, the smooth-sloped Myojin Knoll and Myojinsho edifices to the north were sufficiently distant to have escaped the effects of these earthquakes.

Sector collapse

A ∼15° sector of the eastern caldera rim is missing as a result of small-scale collapse (Fig. 3). The post-caldera cone of Shirane partly fills the gap in the caldera rim that resulted from this collapse event.

Paucity of caldera-rim pumice deposits and abundance of lithic clasts

As noted above, only minor syn-caldera pumice deposits were found at or near Sumisu’s caldera rim, whereas the same area is littered with lithic clasts. Assuming the pumice and the lithic clasts were both products of the caldera-forming eruption, how, and when, were these two fractions partitioned?

Proximal, syn-eruption partitioning

Freundt (2003) conducted tank experiments in which heated mixtures of pumice and lithic clasts were allowed to flow into water. Although orders of magnitude smaller than the events during Sumisu’s caldera-forming eruption, these experiments showed that dense clasts were deposited first and that less dense clasts traveled for greater distances. On a much larger scale, proximal deposition of dense lithic clasts also occurs during subaerial caldera-forming eruptions (e.g. Druitt and Sparks 1982; Walker 1985; Druitt and Bacon 1986).

Post-eruption erosion

Strong ocean currents, common in the Sumisu area, could have preferentially eroded pumice clasts from the syn-caldera deposit, leaving dense lithic clasts behind. Dive H348 on the northeastern caldera rim encountered erosional remnants of a thin-bedded pumiceous sequence (Fig. 6d) that originally capped the northeastern caldera rim. Some of the pumice scoured from this area may be contained in the pumiceous scree that mantles the outer slopes of the Sumisu edifice just to the northeast (Dive 3K576).

We conclude that the scarcity of pumice and the high abundance of dense lithic clasts in proximal areas are likely to be the result of both syn-eruption preferential deposition of dense lithic clasts and post-eruption erosion of the more mobile pumice clasts, especially during the low stands of late Pleistocene sea level.

Volume of the caldera-forming eruption

Tephra deposits resulting from Sumisu’s caldera-forming eruption are located entirely below sea level, so it is difficult to quantify their volume. However, by employing a template method to sum the volumes of the present-day caldera, the now-vanished pre-caldera island, and a 200-m-thick funnel-shaped mass to represent the volume of caldera fill, we estimate that about 36 km³ of Sumisu’s pre-caldera edifice was lost as a result of the caldera-forming eruption. At Crater Lake caldera, which has a slightly larger diameter than Sumisu caldera, Bacon (1983) estimated that ∼9–12% of the pre-caldera edifice lost during the caldera-forming eruption has been enclosed as non-juvenile lithic clasts in the syn-caldera deposits. Assuming a similar percentage of non-juvenile lithic clasts in Sumisu, we therefore estimate that a volume of ∼32–33 km³ of Sumisu’s pre-caldera edifice is missing. If the bulk density of the erupted juvenile tephra is two thirds that of the missing edifice, a tephra volume of ∼48–50 km³ is implied. As noted above, the multiple pumice layers encountered during nearby ODP drilling suggest that Sumisu may have been the site of repeated large-volume eruptions, but we here focus attention on the most recent event that we infer to be genetically related to the present caldera.

Caldera-forming eruption

Considering its large volume, Sumisu’s caldera-forming eruption must have overwhelmed the proximal marine environment. Dispersal mechanisms were probably complex, and most details are unknown. With these caveats in mind, we discuss what we consider the three most important dispersal mechanisms (Fig. 11).

Fig. 11

Cartoon showing four stages of tephra dispersal from the Sumisu caldera-forming eruption, proximal details omitted. Vertical exaggeration X2; sea level (SL). a Eruption begins on island. b Eruption climax. (1) Proximal fallback mass, bounded by a seawater exclusion front, temporarily displaces the sea. (2) Hot pyroclastic flow travels over raft of floating pumice. c Eruption wanes, caldera formed. (3) Submarine pumice quenched as seawater gains access to submarine parts of hot, proximal mass (shown schematically by blotchy stipple pattern). (4) Pyroclastic density current travels into deeper water. (5) Water-column tephra hyperconcentration. (6) Floating pumice. d Eruption ends. (7) Strong ocean currents erode proximal pumice, leaving veneer of dense lithic clasts. (8) Floating pumice widely dispersed. (9) Deposits of pyroclastic gravity currents mantle distal sea floor

Rafts of floating pumice

Tangible evidence for pumice rafts, such as accumulations of abraded, rounded clasts of the type illustrated by Manville et al. (1998) and Bryan et al. (2004), was not observed. We nevertheless suggest that such rafts, highly mobile in the open-ocean environment, played an important role in tephra dispersal. As noted earlier, the upper part of the Sumisu edifice was probably an island prior to its caldera-forming eruption, and the eruption column rose into the air. In a process described by Fiske et al. (2001), pumice clasts cooling in the atmosphere would have ingested air as magmatic steam condensed, producing light-weight clasts that floated when they fell into the sea. Hot pyroclastic flows may have traveled over the temporary dry land provided by thick proximal pumice rafts.

Using the August 1883 eruption of Krakatau as an example, the existence of vast fields of floating pumice at Sumisu is a high probability. Floating pumice blanketed wide parts of the Indian Ocean following the Krakatau eruption (Simkin and Fiske 1983). Three months after the eruption, for example, a merchant vessel 3,000 km west of Krakatau encountered a pumice raft 1,800 km across that locally was sufficiently thick and buoyant to support sailors who jumped onto its surface. More remarkable, 11 months after the eruption (July 1884) masses of pumice washed up on the coast of Zanzibar, >6,000 km from Krakatau. Sumisu’s caldera-forming eruption, perhaps two to three times larger in volume than that at Krakatau, probably blanketed widespread parts of the western Pacific with floating pumice.

Water-column hyperconcentrations

As a process transitional between pumice rafts and pyroclastic gravity currents, we suggest that a large part of the water column, if sufficiently choked with slowly settling tephra, can spread laterally to become an important dispersal mechanism. In Fig. 11c, such a hyperconcentration moves toward the right side of the panel.

Such a scenario could explain the characteristics of the 30-m-thick Quaternary pumice layers cored at ODP Site 788, at the crest of the arc-front ridge 70 km south of Sumisu (Fig. 4), an unlikely place for bottom-hugging gravity currents to be deposited. Nishimura et al. (1991) showed cores from this site to be well sorted, matrix-free, and to contain <1% dense lithic clasts, consistent with an origin involving tephra fallout from water-column hyperconcentrations.

Pyroclastic gravity currents

Sumisu’s caldera-forming eruption, which began as a subaerial event on a small island, sent an eruption column into the air and delivered hot fallback debris to the proximal marine environment at prodigious rates. Large volumes of this debris entered the sea and remained under water to travel downslope to ODP sites, 70 km distant. But how did this subaerial-submarine transition take place? We suggest that the high density of seawater, as compared with air, temporarily constrained the lateral dispersal of tephra. This resulted in the proximal growth of a thick and rapidly growing mass of hot tephra, a transient feature the upper part of which lay above sea level, and the base of which was depressed to ocean depths of perhaps a few hundred meters. Hot tephra in the submerged part of the mass was initially filled with steam and other magmatic gases (Fig. 11b), an environment that temporarily excluded surrounding seawater. In time, however, the steam condensed (Fig. 11c), seawater entered the mass, and while still below sea level, hot pumices quenched, ingested water, and were transformed from floaters to sinkers. The resulting sludge of water-saturated pumice was then available for downslope transport in pyroclastic gravity currents. Pumice clasts that failed to saturate during this process may have risen toward sea level, accumulating along the base of the pumice raft.

After the eruption (Fig. 11d), Sumisu’s caldera rim may have lain at ocean depths of 100 m or less, and parts of it were likely emergent during times of late Pleistocene low stands of sea-level. Strong ocean currents scoured this rim, resedimenting, or removing altogether, much of the syn-caldera pumice deposits originally present. As a result, the thin proximal carpet of dense lithic clasts is the most conspicuous deposit remaining from the caldera-forming eruption.

Conclusions

Sumisu volcano, a large submarine volcano at the front of the Izu-Bonin arc, was the site of a late Pleistocene caldera-forming eruption that introduced ∼48–50 km³ of rhyolite tephra and non-juvenile lithic clasts into the open-ocean environment. Our chief conclusions are:

• 1) A cluster of small seafloor dacitic and basaltic volcanoes formed at ocean depths of ∼1,000 m; these coalesced and grew above sea level to form an island ∼200 m high.

• 2) The caldera-forming eruption resulted in the destruction of most of the island and the surrounding shallow-water parts of the Sumisu edifice. The 8 × 10 km rim of the present-day caldera lies at ocean depths of 100–400 m.

• 3) Syn-caldera tephra was dispersed mostly by a combination of sea-surface pumice rafts, hyperconcentrations of slowly settling tephra that drifted laterally in the water column, and pyroclastic gravity currents that traveled tens of kilometers downslope into ocean depths >2 km.

• 4) Geochemical similarities suggest that the thick pumice deposits encountered at ODP Leg 126 localities, 70–90 km distant, originated at Sumisu volcano. It is likely that Sumisu’s caldera-forming eruption was the source of one of these deposits.

• 5) Post-caldera edifice instability resulted in the failure of a ca 15° sector of Sumisu’s eastern caldera rim and the formation of wavy slump structures that disturbed widespread parts of the volcano’s surface.