Bull Volcanol (2002) 64:55–68 DOI 10.1007/s00445-001-0184-z

R E S E A R C H A RT I C L E

N. Geshi · T. Shimano · T. Chiba · S. Nakada

Caldera collapse during the 2000 eruption of Miyakejima Volcano, Japan

Received: 28 April 2001 / Accepted: 17 October 2001 / Published online: 20 November 2001 © Springer-Verlag 2001

Abstract A collapsed caldera, 1.6 km in diameter and 450 m in depth, was formed at the summit of Miyakejima Volcano during the 2000 eruption. The collapsed caldera appeared on 8 July, with a minor phreatic eruption, 12 days after seismic activity and magma intrusion occurred northwest of the volcano. Growth of the caldera took from 8 July to the middle of August, with seismic swarms associated with the continuous intrusion of magma northwest of the volcano. The growth rate of the caldera was about 1.4×107 m3/day, and the final volume of the collapsed caldera was about 6×108 m3. Major phreatomagmatic eruptions produced a total of about 1.6×1010 kg (1.1×107 m3) of volcanic ash after caldera growth. The caldera structure, and the nature of the eruptive materials of the first collapse on 8 July, suggest that the surface subsidence was caused by the upward migration of a steam-filled cavity, with stoping of the roof rock above the magma reservoir. The diameter of the stoping column was estimated to be 600–700 m from circumferential faults that developed in the caldera floor, and the collapse of the caldera wall enlarged the diameter of the caldera to 1.6 km. The total volume of the caldera and the horizontal diameter of the stoping column gave a subsidence of the caldera floor of 1.6–2.1 km. Keywords Collapsed caldera · Miyakejima Volcano · Phreatomagmatic eruption · Stoping · Underground cavity

Editorial responsibility: T.H. Druitt N. Geshi (✉) · T. Shimano · S. Nakada Volcano Research Center, Earthquake Research Institute, The University of Tokyo, 1-1-1, Yayoi, Hongo, Tokyo, 113-0032, Japan e-mail: [email protected] Tel.: +81-3-58415798, Fax: +81-3-38126979 T. Chiba Asia Air Survey Co. Ltd., 5-42-32, Asahi-machi, Atsugi, Kanagawa Pref., 243-0073, Japan

Introduction Circular collapse structures are common features on the summits of basaltic shield volcanoes and basaltic to intermediate stratovolcanoes (McBirney 1956; Nakamura 1964; Simkin and Howard 1970; Lockwood and Lipman 1987; Tilling and Dvorak 1993; Rymer et al. 1998). Such depressions are also known on silicic volcanoes (e.g., Hildreth 1987). These depressions are named pit craters, collapsed craters, subsided craters, or calderas. Although their diameters range from several meters to several tens of kilometers, their structural similarity may indicate a common mechanism of formation. From the viewpoint of the evolution of a volcano, flank eruption often precedes the formation of pit craters or calderas at basaltic shield volcanoes (Simkin and Howard 1970; Lockwood and Lipman 1987). Some volcanoes show changes in eruption style and petrological character of magmas after caldera formation, implying an intimate relationship between caldera formation and evolution of the shallow magma system, and a widely accepted idea is that the escape of magma from the shallow reservoir causes collapse of the summit of the volcano (Macdonald et al. 1970; Hirn et al. 1991). Therefore, detailed observation of the collapse of a summit caldera will provide us with information about the structural evolution of the shallow magmatic system of the volcano, such as the location, size, shape, and state of activity of the magma body. Few observations of caldera formation, however, have been reported (Galapagos Islands, Simkin and Howard 1970; Pit crater of Masaya Volcano, Rymer et al. 1998). This paper documents the evolution of the newly formed caldera of Miyakejima Volcano based mainly on aerial photos and surface surveys, and presents a model of the structural development of the caldera. This is the first detailed scientific observation of the formation of a collapsed caldera on an island-arc stratovolcano.

56

Fig. 1 Generalized map of Miyakejima Volcano. Distributions of lava and pyroclastics of 20th century eruptions are from Tsukui and Suzuki (1998). Outline of the summit caldera formed during the 2000 eruption, the location of submarine eruption on 27 June, and the approximate area of the earthquake swarm during July and August are shown

Geological setting of Miyakejima Volcano Miyakejima Volcano (814 m above sea level before the 2000 eruption) is an active basaltic andesite stratovolcano about 200 km south of Tokyo, in the Izu-Bonin volcanic chain (Fig. 1). The volcano has a sub-circular outline with a diameter of about 13 km and a height of about 1,000 m above the sea floor. The summit area is characterized by concentric double calderas: the Kuwakidaira Caldera (3.5 km across) and the Hatchodaira Caldera (1.5 km across), with the central stratocone of Mt. Oyama (Fig. 1). The main part of the volcanic edifice of Miyakejima was built more than 10,000 years ago. Kuwakidaira Caldera is considered to have formed during the building of the main part of the volcanic edifice, although the associated eruptive materials are not recognized. The main part of the volcanic edifice has been covered with lavas and pyroclastics within the last 10,000 years. The eruption at 2,500 years B.P. is the largest eruption of Miyakejima Volcano within the last 10,000 years. About 0.4 km3 dense rock equivalent (DRE) of eruptive materials (scoria, explosive breccia, and accretionary lapilli) were produced during the eruption, and the Hatchodaira Caldera was formed (Tsukui and Suzuki 1998; Tsukui et al. 2001). After the formation of Hatchodaira Caldera, eruption of basaltic an-

desitic magma formed the stratocone of Mt. Oyama within the caldera. No summit collapse event has been recognized since the formation of Hatchodaira Caldera. The volcanic activity in the last 600 years has been characterized by flank fissure eruptions that have produced scoria and lava flows at 21- to 69-year intervals (Miyazaki 1984; Tsukui and Suzuki 1998). The volume of eruptive material of each eruption was ~1010 kg (~107 m3). Phreatomagmatic eruptions often occurred when the fissure reached coastal areas to generate explosive craters and tuff rings (Fig. 1). In the 20th century, eruptions occurred in 1940, 1962, and 1983 (Miyazaki 1984). The 1940 eruption was characterized by a fissure eruption on the northeastern flank with a minor summit eruption. About 4×1010 kg of lava and scoriae were produced (Tsuya 1941). The 1962 eruption was characterized by eruption from a 2.5-km-long fissure on the northeastern flank, producing ~2×1010 kg of lava and scoria (Suwa 1963). The 1983 eruption was also characterized by eruption from a 4.5-km fissure on the southwestern flank, which produced ~2×1010 kg of scoriae and lava flows (Aramaki and Hayakawa 1984). Phreatomagmatic eruption formed explosive craters and tuff cones around the southern coast during the 1983 eruption.

Summary of the Miyakejima 2000 eruption The Miyakejima eruption in 2000 was characterized by the formation of a collapsed caldera, which was ~1.6 km across and 6×108 m3 in volume in the summit area. Intermittent phreatic and phreatomagmatic eruptions from the

57 Fig. 2 A distribution of epicenters from 26 June to 31 December 2000 with the focal mechanisms of five major earthquakes >M 6 (Sakai et al. 2001). Arrows show displacement between June 2000 and September 2000 determined by GPS analysis of Geographical Institute of Japan. B Distribution of hypocenters from 8 July to 2 August in the area a–a′–b′–b in A projected to the vertical plane a–b (Sakai et al. 2001). C Distribution of hypocenters in Miyakejima volcano (area is shown with a square of dashed line in A) from 26 June to 21 July 2000, projected to the vertical plane (by the Earthquake Observation Center, ERI, University of Tokyo). D Temporal distribution of the depth of hypocenters from 26 June to 21 July in Miyakejima Volcano (by Earthquake Observation Center, ERI, University of Tokyo). Broken line shows upper limit of the distribution of earthquakes before the appearance of the collapsed caldera

caldera contrasted with historical volcanic activity. The total mass of eruptive materials was 1.7×1010 kg and mostly consisted of accessory and accidental materials. Nakada et al. (2001) divided the eruption into four stages: (1) intrusion stage, (2) summit subsiding stage, (3) explosive stage, and (4) degassing stage. Intrusion stage (26 June–8 July) In this stage, magma intruded from the reservoir beneath the summit of Miyakejima towards the northwest, with numerous earthquakes. Seismic activity began on the evening of 26 June beneath the volcano, and migrated northwestward from 26 June to 1 July at about 5 km/day (Fujita et al. 2001; Sakai et al. 2001). The epicenters show a tabular distribution to the northwest of Miyakejima Volcano (Fig. 2A, B). GPS data showed the deflation of the summit area of Miyakejima Island and inflation of the western part of the island in the vicinity of the earthquake swarm, which was migrating westward during this stage. A series of open cracks and normal faults showing a N–S extension formed in the western coastal part of Miyakejima Island. A minor submarine eruption, producing about 1–3×103 m3 of aphyric basaltic andesite spatter, occurred on the morning of 27 June about 1 km off the west coast of Miyakejima Volcano

(Fig. 1; Nakada et al. 2001). These observations suggest the intrusion of a dike to the northwest of the volcano. Seismic activity and ground deformation to the northwest of Miyakejima Volcano continued until the middle of August, suggesting the continuous migration of magma from Miyakejima Volcano. There is no evidence for submarine eruption in the vicinity of the earthquake swarm, except for the submarine eruption of 27 June. After the westward migration of the earthquake swarm, earthquake activity remained at a low level in Miyakejima Volcano. Earthquake activity beneath the summit area was reactivated on 4 July. The earthquakes showed a columnar distribution of about 2 km in diameter, and the top of the seismic swarm became shallower prior to the 8 July eruption (Fig. 2C, D). Summit subsiding stage (8 July–middle of August) A collapsed caldera grew in the summit area of the Miyakejima Volcano between the 8 July and the middle of August. The first summit eruption began at around 18: 40 h on 8 July in Hatchodaira Caldera. The eruption continued for 10 min with the generation of whitish eruptive clouds about 800 m above the summit. About 1×108 kg of volcanic ash and lapilli covered the eastern part of the island. No juvenile material was found in the eruptive

58

Fig. 3 Growth of the collapse caldera on the summit of Miyakejima Volcano. A Northern view of the summit area of the Miyakejima Volcano before the 2000 eruption (taken by O. Oshima in January 1983). Broken line “a” shows the outline of the collapsed caldera on 8 July 2000, and line “b” shows the outline of the collapsed caldera in September 2000. B Northern view on 9 July about 18 h after the appearance of the caldera (Asia Air Survey); C southeastern view on 22 July at 14 days after the start of the caldera growth (Asia Air Survey); D northeastern view on 3 September, about 2 months after the start of the caldera growth (S. Nakada)

products. Ballistics were scattered within 1 km of the crater. A collapsed caldera, 1 km across, appeared after the eruption (Fig. 3B). The second summit eruption occurred in the collapsed crater during the morning of 14 July and continued intermittently till noon of 15 July. Dark-colored eruption clouds rose higher than 1 km above the summit. Volcanic ash (~3×109 kg) was erupted and covered the northeastern part of the island. The volcanic ash contained about 25% fresh basaltic scoria, which is interpreted to be juvenile material. The collapsed caldera enlarged in depth and diameter continuously after 8 July, with seismic activity offshore to the NW of Miyakejima (Kaidzu et al. 2000; Sakai et al. 2001; Fig. 2A, B). By the middle of August, the caldera was 1.6 km in diameter and 450 m in depth (Fig. 3C, D). The enlargement in diameter and depth of

the caldera had mostly stopped by the middle of August, although intermittent small collapses of the caldera wall were observed. The structural evolution of the caldera is described in the next section. Step-like tilt changes were observed at this stage (Kumagai et al. 2001; Ukawa et al. 2000; Yamamoto et al. 2001). The step-like tilt change resulted in rapid inflation of the volcanic body with seismic swarms during continuous deflation. Each step-like tilt change event continued for as much as 2 min. A very-long-period seismic signal was accompanied by step-like tilt changes (Kumagai et al. 2001). From 8 July to 18 August, 46 events of tilt change were observed, at frequencies of one to three events for each 1–2-day period (Yamamoto et al. 2001). Explosive stage (middle of August–September) Phreatic and phreatomagmatic eruptions with ash emission restarted on 10 August and continued till late September to produce about 1.7×1010 kg of eruptive materials. The total mass of juvenile materials was about 4×109 kg, and the remainder of the eruptive materials consisted of fragments of lavas and pyroclastics from the old volcanic body and hydrothermally altered rocks. A relatively large eruption during the afternoon of 18 August produced ~8×109 kg of volcanic ash and

59 Table 1 Evolution of the collapsed caldera Date

9 July

14 July

22 July

28 July

2 August

4 August

26 August

3 September

18 September

Max. width (m) Deptha (m) Area (km3) Volume (108m3)

980 140 0.6 0.56b

1,100 300 0.8 2

1,200 450 1 2.65b

1,300

1,400 450 1.4 4.26b

1,400 450

1,600

1,600 450 1.9 6

1600

a Depth

1.25

4.5

1.9 6

1.9 6

is taken from the summit car park, 700 m in altitude; b Data from Geographical Survey Institute webpage

lapilli, which covered the whole area of the island. The eruption column rose higher than 10 km and drifted southward. Some of the ash cloud reached Hachijojima Island, located 110 km south of Miyakejima Volcano. Ballistic fragments larger than 1 m across reached the northwestern coast of Miyakejima Island, about 3 km from the summit. Basaltic volcanic bombs were issued towards the end of the eruption. Eruptive materials contained about 40% fresh basaltic fragments and bombs, which are considered to be juvenile material (Uto et al. 2001). Degassing stage (since September) During this stage, emission of volcanic gas from the collapsed caldera was observed. A steam plume rose continuously from the crater inside the collapsed caldera, and the active emission of more than 1×107 kg of SO2 gas per day has been observed from the middle of August until the time of writing (September 2001). No remarkable ash emission has been observed since 26 September, although dust clouds formed from the landslides of the caldera wall have often been observed. Seismic activity and ground deformation in the volcano has remained low.

Structural evolution of the collapsed caldera The collapsed caldera was first recognized in the early morning of 9 July, about 12 h after the first eruption at 18:40 h on 8 July. At noon on 9 July, the caldera was 985 m across, 140 m deep, and 5.6×107 m3 in volume (Table 1, Fig. 4). The collapsed caldera grew in diameter as a result of the continuous subsidence of the floor and landslides of the caldera wall until the middle of August: it was 1.22 km across, 450 m deep, and 2.7×108 m3 in volume on 22 July; 1,390 m across, 450 m deep, and 4.3×108 m3 in volume on 2 August; and 1,640 m across, 450 m deep, and 6×108 m3 in volume on 28 September. The structural evolution of the caldera was reconstructed based mainly on aerial photos taken on 9 and 22 July, and 2 August. 9–11 July (1–3 days after the start of caldera growth) Figure 5 shows the structure of the collapsed caldera at noon of 9 July, about 18 h after the first summit eruption. The collapsed caldera consisted of a caldera floor sur-

Fig. 4 Development of the collapsed caldera from July to September 2000. A Maximum diameter; B depth from the summit car park at ~700 m elevation; C area, and D volume of the collapsed caldera; E cumulative mass of erupted materials. Total mass of eruptive materials and mass of juvenile materials are shown. F Number of the earthquakes per day in the Miyakejima Volcano (data from the Japan Meteorological Agency)

60 Fig. 5 A Surface structure of the collapse caldera just after the first summit eruption (8 July), based on aerial photographs taken at 12:00 h on 9 July by Asia Air Survey Co. Ltd. Arrows show the horizontal displacement of the landmarks. B Structural classification of the caldera

rounded by a steep caldera wall. The caldera floor can be subdivided into an undeformed central zone and an extended marginal zone bordered by concentric faults (Fig. 5B). The central zone consisted of a coherently subsided block, which was oval in shape, about 300×500 m across, and elongated in a SW–NE direction. The original ground surface, which was covered by a small amount of the ejecta from 8 July, remained undeformed in its central zone. At the center of the block, two or three craters, 50–100 m in diameter, were lined up in WNW–ESE direction and eruptive materials were scattered around these craters. No steam emission was observed from these craters during the morning of 9 July. The marginal zone consisted of blocks that tilted inward towards the caldera. Some extension cracks and normal

faults subparallel to the outline of the caldera developed in this zone. The marginal zone was wide in the northwestern area (250 m) and absent in the southeastern area. A circumferential slope, 50 m high, divided the central and marginal zones. There was little talus deposit at the foot of the slope. The caldera wall surrounding the floor was 50–100 m high. Talus developed from the caldera wall on the marginal extended zone of the caldera floor. Horizontal displacements of the subsided caldera floor were estimated by using landmarks on aerial photos. The central zone subsided almost vertically with a small horizontal displacement, whereas the marginal zone moved a maximum of 100 m towards the center of the caldera with extensional deformation (arrows on Fig. 5A).

61

period. The northeastern part of the marginal zone also subsided with an increase of inward tilting and extension. Talus from the caldera wall covered the original ground surface of the subsided caldera floor. 22 July and 2 August (14 and 25 days after the start of caldera growth) The diameter, depth, and volume of the collapsed caldera increased during the 14–25 days after the commencement of caldera growth (Figs. 4 and 7). No fumarole activity was observed in the caldera. Debris avalanche deposits and talus deposits from the caldera wall covered the caldera floor, and the original surface structure was completely buried (Fig. 3C). Debris avalanche deposits from the caldera wall covered the caldera floor, and formed lobes and hummocky rises. Pyroclastic cones and craters, which were recognized during the eruption of 14–15 July on the southern edge of the caldera floor, were buried on 22 July (Fig. 3C). Some concentric faults with displacements as high as several tens of meters developed in the marginal zone of the floor (Fig. 8). These faults surrounded an oval area about 600–700 m diameter at the center of the caldera floor. The rate of outward migration of the caldera wall as a result of continuous landslides occurred at ~10 m/day (Figs. 4A and 9). The diameter of the caldera was about 1.0 km on 22 July and 1.4 km on 2 August. The increase in the diameter of the flat floor with minor changes in depth reflect the shift of cross section of the caldera from a V-shape on 22 July to a U-shape on 2 August (Fig. 9B). Middle of August to September (40–60 days after the start of caldera growth)

Fig. 6 Growth of the collapsed caldera in the early stage. Arrows show a block on the northwestern part of the caldera floor that tilts and slides inwards. A Noon of 9 July at 18 h after the appearance of the caldera, taken by Asia Air Survey; B morning of 10 July at about 1.5 days after the appearance of the caldera, from the Chunichi Shimbun Co. Ltd.; C afternoon of 11 July at about 3 days after the appearance of the caldera, from the Mainichi Newspaper Co. Ltd. Whitish fog covered the bottom of the caldera

After the summit eruption of 8 July, the collapsed caldera enlarged continuously. Oblique aerial images taken on 10 and 11 July show continuous subsidence of the caldera floor without eruption (Fig. 6).The depth of the caldera floor exceeded 300 m on 11 July: 3 days after the beginning of the caldera collapse. The central zone subsided vertically with only minor deformation during this

Minor changes in the outline of the caldera were observed after the middle of August, although minor collapses and rock falls from the wall occurred intermittently. The diameter of the caldera reached 1.6 km by the end of August and then stopped increasing. Intermittent phreatic and phreatomagmatic eruptions occurred along the southern margin of the caldera floor from the middle of August, and a pyroclastic cone formed in the southern part of the caldera (Figs. 3D and 7). A steam plume rose continuously from the craters of the pyroclastic cone from the middle of August. Enlargement of the collapsed caldera Collapse of the caldera wall enlarged the diameter of the caldera. Figure 9 shows the change in caldera outline between July and September 2000. The caldera enlarged mainly towards the northeast and southwest in July, and towards the southeast in August. By the middle of August, the outline of the collapsed caldera had reached the rim of Hatchodaira Caldera, which was formed 2,500 years B.P. (Fig. 9A). The center of the final outline

62

Fig. 7 Structural evolution of the collapsed caldera on 9 July, 22 July, 2 August, and 18 September, based on aerial photos by Asia Air Survey Co. Ltd. and Japan Air Self Defense Force. Talus has a slope of about 30–45° and consists of collapsed materials from the caldera wall. Subsided blocks are recognized only on 9 July

of the caldera is located at the point where the craters were first observed on 8 July. The southern rim of the caldera floor, which was nested by craters and fumaroles, corresponds to the caldera rim on 9 July. The diameter, area, and volume of the collapsed caldera increased linearly from July to the middle of August (Table 1 and Fig. 4). The rates of increase in diameter, area, and volume were 20 m/day, 3.5×10–2 km2/day, and 1.4×107 m3/day, respectively from 9 July until the middle of August. The depth increased rapidly in the middle of July and remained at almost the same level (~450 m).

Discussion Cause of caldera collapse The total volume of eruptive materials during the Miyakejima eruption in 2000 was about 6.8×106 m3 in DRE, and this value is only 1.1% of the final volume of the collapsed caldera (6×108 m3). More than 75% of the eruptive materials consisted of rock fragments derived from the former volcanic body. Moreover, it was after the main phase of the caldera growth that major eruptions occurred (Fig. 4). These facts show that eruption of magmatic material was not the trigger for the formation of the collapsed caldera. In the early stages of the eruption in 2000, in the intrusion stage, swarms of earthquakes migrated to the northwest of the volcano, indicating the continuous es-

63

cape of magma through a newly formed dike. Ground deformation shows deflation of the Miyakejima Volcano and inflation of the seismic area northwest of the volcano, which suggests the growth of an intrusive body (Fig. 2). Some magma reached the sea floor to the west off the volcano and resulted in a minor submarine eruption on 27 June (Fig. 1). Because no evidence for the submarine eruption was found in the northwest of the volcano, this suggests that most of the magma migrated from the reservoir beneath the volcano to form a shallow intrusive body. The distribution of the seismic swarms (Fig. 2A) suggests the formation of a dike with a NW–SE strike (Kaidzu et al. 2000; Furuya et al. 2001),

Fig. 8 Circumferential faults that developed on the caldera floor. This shows new debris from the avalanche lobe that covers the faults. Photo taken by S. Nakada on 4 August from the top of the western caldera wall

Fig. 9 Development of the caldera from July to September 2000. A Plan view. B Cross section. Line A–B shows the location of the cross section. Solid lines are the caldera outline determined by vertical aerial photos and a map; broken lines are oblique aerial photos. The outlines were determined from aerial images taken on the specified dates: the vertical aerial photos of 9 July and 22 are from Asia Air Survey Co. Ltd., the oblique image of 14 July is from TV-Asahi; the oblique aerial photos of 2 August are from Ashia Air Survey Co. Ltd., 26 August. The outline on 16 January is from the eruption map (1: 5,000) taken with GSI. Stars show the location of the craters on 8 July and open circles show the active craters and fumaroles on 18 September. Shaded area in A shows the central coherent area recognized on the caldera floor on 9 July

64

which corresponds to the direction of the regional maximum horizontal stress in the northeastern margin of the Philippine-Sea plate, as indicated by dike swarms in the neighboring volcanoes (Nakamura 1977). The intrusion of the dike might have triggered the opening of a fracture northwest of the volcano, and the consequent fractures withdrew magma from the summit reservoir through the dike. Reduction in pressure inside the magma reservoir caused a gravity-driven downward displacement of the reservoir roof into the reservoir, and brittle deformation of the roof rock began when the gravity stress of the roof rock exceeded the strength of the roof rock. Upward migration of seismic activity, which began beneath the summit area just before the caldera collapse (Fig. 2D), suggests stoping of the roof rock of the reservoir. There was a 12-day delay from the start of the evacuation of magma from the reservoir (26 June) to the start of surface collapse (8 July). During this delay, underground stoping migrated upward from the roof of the magma reservoir to about 3 km from the surface. Assuming that the evacuation rate of magma from the reservoir was constant before and after the caldera collapse, about 1.8×108 m3 of magma had already evacuated from the reservoir before the start of caldera collapse. Formation of the collapsed caldera on 8 July The collapsed caldera appeared on 8 July with a phreatic eruption when the upward migrating seismic activity beneath the summit reached the surface (Fig. 2D). One of the remarkable structural characteristics of the collapsed caldera on 9 July was its caldera floor, which consisted of a coherent central zone and an extended and inwardtilted marginal zone (Fig. 5). Inward movement of the marginal zone (Fig. 5) meant shortening occurred between the central and the marginal zones, which suggests the existence of an outward-dipping circular reverse fault between them. Subsidence with the inward movement of the marginal zone also meant that an inward-dipping circular normal fault surrounded the outer rim of the marginal zone (Fig. 10A). These structural characteristics are very similar to those when the shape of the roof had a narrow thickness compared with the width when formed in analogue experiments of Roche et al. (2000, 2001). Deformation of the caldera floor was limited to the margin of the subsided zone, and the central zone did not suffer from deformation, which means that the concentration of gravitational stress at the roof margin was the result of an underground cavity. Some geophysical observations suggest the existence of a cavity just before the appearance of the collapsed caldera. Furuya et al. (2001) reported that the gravity changes exceeded –1 mgal around the summit area just before the appearance of the collapsed caldera (from June 1998 to 1–6 July 2001). They concluded that the gravity changes were caused by the generation of a vacant space, with a volume of 9.5×107 m3 and a depth

Fig. 10 Schematic illustration of the deformation of the caldera floor in the early stages of subsidence. Inner reverse fault borders of the central zone and marginal zone, and outer normal fault limits show the outline of the marginal zone on 9 July. There was continuous subsidence of the central zone from 9–11 July, and the marginal zone was tilted and extended inwards. By the middle of July, 1 week after the beginning of surface collapse, the caldera floor was buried by the collapsed materials from the caldera wall

of 1.7 km beneath the summit. They also concluded that a dike with a volume of about 2.5×108 m3 intruded to the west of the volcano during this period, based on the gravity changes in the island. Sasai et al. (2001) report a slow decrease in geomagnetic total intensity around the summit area from 1–7 July and then rapidly on 8 July, and they conclude that the geomagnetic changes were caused by the generation of a vacant space with a volume of 3.3–5.8×107 m3, which occurred at about 2 km depth from the summit and then migrated to the surface.

65

Seismic and geomagnetic observations also suggest a rapid collapse of the caldera block during the eruption on 8 July. A long-period seismic pulse (Kikuchi et al. 2001) was observed during the eruption on 8 July. Kikuchi et al. (2001) show that this seismic event was characterized by a single force directed initially upwards and later downwards during 12 s, and they interpret the focal mechanism to be an abrupt collapse of massive rock with a mass of about 5×1010 kg that fell about 200 m. Changes in geomagnetic total intensity around the summit area during the eruption on 8 July suggests the formation of a collapsed caldera of 93 m depth within 4 min (Sasai et al. 2001). These observations show that the rapid subsidence of blocks during the eruption on 8 July caused the formation of the collapsed caldera, which suggests the collapse of an underground cavity at shallow depth. The estimated depth of the magma reservoir from geophysical observations is, however, much deeper than the depth of the cavity collapse on 8 July. The inflation source before the 2000 eruption was at about 5 km beneath the summit, and the lower limit of the earthquake swarms beneath the summit area was about 3 km below sea level (Fig. 2C, D). GPS data show that the deflation source during the eruption was at about 4–5 km below sea level (Nakada et al. 2001). Temporal gravity changes suggest a deflation of ~7.6×107 m3 at 5 km depth below sea level (Furuya et al. 2001). This suggests that the collapsed cavity that occurred during the eruption of 8 July was not the magma reservoir itself, but the secondary cavity formed by the underground stoping process. Formation of some explosive craters at the center of the collapsed caldera, and the distribution of ballistic fragments as far as ~1 km from the new craters, bear witness to the explosive release of pressurized steam-rich gas. The cavity might have been filled with steam-rich gas, judging from the steam-rich eruption cloud and wet volcanic ash; and the collapse of the underground cavity would have cause the rapid release of this gas. The absence of juvenile materials in the eruptive material of the 8 July eruption also suggests that the cavity was filled with the steam-rich gas. Gypsum within the eruptive materials of the 8 July eruption has the lowest sulfur isotope ratio of the 2,000 eruptions (δ34S=4.9–7.1; Imai et al. 2001), indicating the emission of juvenile volcanic gas. This suggests that the steam ejected during the first summit eruption was derived from the magma reservoir by upward migration of the cavity. Continuous growth of the caldera Sudden collapse during the 8 July eruption was followed by continuous subsidence. During the 8 July eruption, about 5.6×107 m3 of rock collapsed into the underground cavity within a few minutes, and continuous subsidence of 1.7×107 m3/day continued until the middle of August without any eruption (Fig. 4). This change can be ex-

plained by the changes in the strength of the roof rock by sectorial stoping as follows. Upward migration of stoping will form a brecciated column as it moves, and the brecciated roof rock will lose the strength to support itself within the cavity. This underground stoping process before the surface collapse on 8 July would be sufficient to form the brecciated column between the caldera floor and roof of the reservoir. No major cavity was, in contrast, formed within the brecciated column, and the deflation of the magma reservoir would have been related directly to the surface subsidence during the continuous subsidence of the caldera floor because no sudden collapse and ejection of steam on 8 July was observed. In the early stages of summit subsidence, the central part of the caldera floor subsided vertically with minor deformation, and the northwestern part of the floor tilted and extended inwards towards the caldera from 9–11 July (Fig. 6). Successive deformation of the caldera floor during this period shows that the subsidence of the caldera floor progressed by the displacement of concentric faults formed during the 8 July eruption (Fig. 10). Continuous subsidence of the central zone by displacement on the inner reverse fault caused inward extension and tilting of the marginal zone. After the eruption between 14–15 July, eruptive and collapsed materials from the caldera wall covered the caldera floor, and the original surface of the subsided block was unrecognizable. Some concentric faults developed on the caldera floor, which suggests the subsidence of the cylindrical stoping column (Fig. 8). The diameter of the caldera increased because of rock falls and avalanches from the caldera wall, yet the depth of the caldera was nearly constant (Fig. 4A, B) because collapse of the caldera wall was balance by subsidence of the caldera floor. The final outline of the caldera is much larger than the area of the stoping column indicated by the circumferential faults on the caldera floor. The oval area of about 600–700 m across (2.8–3.8×105 m2) surrounded by circumferential faults on the caldera floor (Fig. 8) may represent the diameter of the stoping column. Subsidence of the stoping column, which was about 600–700 m in diameter, must account for the whole volume of the caldera, which was 1.6 km in diameter. This means that the collapsed materials from the caldera wall covered the top of the stoping column, and the subsided depth of the stoping column was larger than the topographic depth of the caldera (about 450 m). The relationship between the area of the stoping column (2.8–3.8×105 m2) and the final volume of the caldera (6×108 m3) shows that the stoping column may have subsided 1.6–2.1 km into the reservoir. The average rate of subsidence of the stoping column was ~40–53 m/day, assuming 40 days of subsidence from July to August. The estimated value of subsidence gives information about the scale of the magma reservoir: the vertical extent of the reservoir should be greater than, or equal to, the subsidence value given above, otherwise, the floor of the reservoir would have supported the stoping column. This estimation, however, may give a minimum value for subsidence because chaotic

66 Fig. 11 Schematic illustration of the development of the collapsed caldera. A Intrusion stage before the surface collapse. Evacuation of magma from the reservoir caused stoping of the roof rock of the reservoir. Underground stoping formed a cavity at the top of the stoping column. B The early stage of the summit subsidence. Surface subsidence occurred on 8 July when the cavity reached a critical depth of less than 1 km. The roof rocks of the cavity could not carry their own weight and collapsed into the cavity. Release of steamy gas filled the cavity and caused a phreatic eruption. C The late stage of the summit subsidence. Continuous evacuation of magma from the reservoir caused the subsidence of the roof of the reservoir. The top of the stoping column was filled with the collapsed materials from the outward migrating caldera wall. D Explosive stage. Invasion of magma to the stoping column caused phreatomagmatic eruption after the westward intrusion of magma ceased

collapse may lead to a substantial dilation of the subsided column. Volcanic activity changed in early–middle August. Growth of the caldera seemed to stop as seismic activity northwest of the volcano and intermittent phreatomagmatic eruptions and continuous steam emission began from the caldera floor. Eruptive materials from August contain fresh basaltic fragments and this suggests that some of the magma intruded into the stoping column above the reservoir. An equilibrium degassing pressure of about 20–40 MPa was suggested from the content of volatiles found in the juvenile materials from the eruption in 18 August, which suggests the ascent of magma 1–2 km upwards through the stoping column (Saito et al. 2001). The development mechanism for the collapsed caldera is schematically illustrated in Fig. 11. In the early stages of volcanic activity, evacuation of magma from the reservoir caused the deflation of the reservoir (Fig. 11A). The roof of the reservoir lost its magmatic support and began to collapse. Underground stoping of the roof rock of the reservoir formed a cavity at the top of the stoping column. Surface collapse occurred on 8 July when the cavity reached a critical depth: when the strength of the roof rock could not support its own

weight (Fig. 11B). After the surface collapse, the stoping column in the reservoir subsided with the deflation of the reservoir (Fig. 11C). The diameter of the caldera was enlarged by rock avalanches and rock falls of the caldera wall. The top of the stoping column consisted of collapsed materials from the caldera wall. By the middle of August, the growth of the caldera had stopped because the westward migration of magma from the reservoir had ceased (Fig. 11D). Some of the magma intruded the stoping column and caused a series of phreatomagmatic eruptions from August to September. Absence of explosive eruption during the caldera collapse Pit craters or collapsed calderas commonly develop at the summits of shield volcanoes, and some observations of collapse events show that flank eruption or intrusion can channel magmas from the summit reservoir to trigger the subsidence of the volcanic center (the 1968 eruption of Fernandina, Galapagos; Simkin and Howard 1970). This is basically similar to the magmatic process suggested here for the caldera collapse of the Miyakejima 2000 eruption.

67

Some caldera collapse events at basaltic volcanoes are accompanied by an explosive phreatomagmatic eruption (Kilauea caldera, Izu-Oshima caldera, Nakamura 1964; Swanson and Christiansen 1973; Decker and Christiansen 1984). The explosive phreatomagmatic eruption during caldera formation was triggered by the draining of magma by a large flank eruption and invasion of groundwater into the conduit (Decker and Christiansen 1984). In this case, collapse of the conduit caused the mechanical mixing of hot rock and water, and triggered a steam explosion in the conduit. In contrast, there was no major explosive eruption with base surge observed during the formation of the collapsed caldera in the Miyakejima 2000 eruption, and major ash-emission events occurred after subsidence of the caldera. The structural development of the collapsed caldera in the Miyakejima 2000 eruption, and the nature of its eruptive materials suggests that the collapsed block consisted of a former volcanic body with a small amount of magmatic material. Magma–water interaction did not occur, and the caldera floor subsided without extensive explosive activity during caldera growth. This is in contrast with the behavior of Hawaiian calderas, and suggests that the existence of magma in the collapsed block is one of the controls of explosive eruption during collapse of a caldera at a basaltic volcano.

Conclusions A collapsed caldera developed during the Miyakejima 2000 eruption. A collapsed caldera that was 1 km across and 150 m deep appeared during a minor phreatic eruption on 8 July after the occurrence of seismic activity beneath the summit. The structure of the caldera floor suggests that the surface block collapsed into an underground cavity at shallow depth. Some geophysical observations suggest the generation of an underground cavity just before the appearance of the collapsed caldera. After the sudden collapse during the 8 July eruption, the caldera floor subsided continuously until the middle of August, along with seismic activity northwest of the volcano. The dimensions of the caldera reached 1.6 km in diameter, 450 m in depth, and 6×108 m3 in volume by the middle of August. Growth of the caldera continued for about 1 month at about 1.4×107 m3/day. Landslides from the caldera wall enlarged the outline of the caldera at the same time as floor subsidence. The final outline of the caldera wall was much larger than the subsidence area indicated by the circumferential faults. The circumferential faults on the caldera floor suggest the existence of a stoping column that was 600–700 m in diameter. The relationship between the size of the stoping area and total volume of the caldera shows that the displacement of the subsided roof of the magma reservoir was 1.6–2.1 km, and this may indicate the vertical extent of the magma reservoir. Phreatomagmatic eruptions occurred after the growth of the caldera. The eruptive materials consisted mainly

of volcanic ash and contained less than 30% juvenile materials. The total volume of eruptive materials was about 6.8×106 m3 in DRE, but this value is only 1.1% of the collapsed volume. Acknowledgements We would like to thank the members of the Joint Universities Research Group with whom most of the observations and surveys of the eruption were carried out. Discussions with T. Fujii, K. Fujita, T. Kaneko, D. Miura, I. Miyagi, M. Nagai, M. Ohno, T. Ominato, A. Terada, A. Tomiya, and K. Uto were very helpful in understanding the eruptive phenomena. Critical reviews by O. Roche and T. Druitt greatly improved the manuscript. Aerial photos, offered by Asia Air Survey Co Ltd., Chunichi Shinbun Co. Ltd., Mainichi Newspaper Co. Ltd., Japan Air SelfDefense Force, and O. Oshima, were critical to our understanding of the evolution of the collapsed caldera. We are deeply grateful to the Japanese Meteorological Agency, the Tokyo Metropolitan Government, the Miyake Village, Metropolitan Police Department of Tokyo, and Japanese Self-Defense Force for assisting in our aerial and ground surveys during the eruption.

References Aramaki S, Hayakawa Y (1984) Sequence and mode of eruption of the October 3–4, 1983 eruption of Miyakejima (in Japanese with English Abstract). Bull Volcanol Soc Jpn Spec Issue 29:S24–S35 Decker RW, Christiansen RL (1984) Explosive eruption of Kilauea Volcano, Hawaii. In: National Research Council (ed) Explosive volcanism: inception, evolution, and hazards (studies in geophysics). National Academy Press, Washington, DC, pp 122–132 Fujita E, Ukawa M, Yamamoto E, Okada Y, Kikuchi M (2001) Volcanic earthquakes and tremors associated with the 2000 Miyakejima volcanic eruption (in Japanese with English Abstract). J Geogr 156:191–-203 Furuya M, Okubo S, Tanaka Y, Sun W, Watanabe H, Oikawa J, Maekawa T (2001) Caldera formation process at the 2000 Miyakejima volcano detected by spatio-temporal gravity change (in Japanese with English Abstract). J Geogr 110:217– 225 Hildreth W (1987) New perspectives on the eruption of 1912 in the Valley of Ten Thousand Smokes, Katomai National Park, Alaska. Bull Volcanol 49:680–693 Hirn A, Lépine J-C, Sapin M, Delorme H (1991) Episodes of pit crater collapse documented by seismology at Piton de la Fournaise. J Volcanol Geotherm Res 47:89–104 Imai A, Geshi N, Shimano T, Nakada S (2001) Sulfur isotope study of 2000 eruption of Miyake-jima volcano, Japan. Abstract 2001 Japanese Earth and Planetary Science Joint Meeting Kaidzu M, Nishimura T, Murakami M, Ozawa S, Sagiya T, Yarai H, Imakiire T (2000) Crustal deformation associated with crustal activities in the northern Izu-islands area during the summer, 2000. Earth Planet Space 52(8):ix–xviii Kikuchi M, Yamanaka Y, Koketsu K (2001) Source process of the long-period seismic pulses associated with the 2000 eruption of Miyake-jima volcano, and its implications (in Japanese with English Abstract). J Geogr 110:204–216 Kumagai H, Ohminato T, Nakano M, Ooi M, Kubo A, Inoue H, Oikawa J (2001) Very-long-period seismic signals and caldera formation at Miyake Island, Japan. Science 293:687– 690 Lockwood JP, Lipman PW (1987) Holocene eruptive history of Mauna Loa volcano. In Decker RW, Write TL, Stauffer PH (eds) Volcanism in Hawaii. US Geol Surv Prof Pap 1350: 509–535 Macdonald GA, Abbott AT, Peterson FL (1970) Volcanoes in the sea: the geology of Hawaii. Univ Hawaii Press

68 McBirney AR (1956) The Nicaraguan volcano Masaya and its caldera. EOS Trans Am Geophys Union 37:83–96 Miyazaki T (1984) Features of historical eruptions at Miyake-jima volcano. The 1983 eruption of Miyake-jma (in Japanese with English Abstract). Bull Volcanol Soc Jpn Spec Issue 29:S1– S15 Nakada S, Nagai M, Yasuda A, Shimano T, Geshi N, Ohno M, Akimasa T, Kaneko T, Fujii T (2001) Chronology of the Miyakejima 2000 eruption: characteristics of the summit collapsed crater and eruption products (in Japanese with English abstract). J Geogr 110:168–180 Nakamura K (1964) Volcano-stratigraphic study of Oshima Volcano, Izu. Bull Earthqu Res Inst Univ Tokyo 42:649– 728 Nakamura K (1977) Volcanoes as possible indicators of tectonic stress orientation-principle and proposal. J Volcanol Geotherm Res 2:1–16 Roche O, Druitt TH, Merle O (2000) Experimental study of caldera formation. J Geophys Res 105:395–416 Roche O, Vries VWD, Druitt TH (2001) Sub-surface structures and collapse mechanisms of summit pit craters. J Volcanol Geotherm Res 105:1–18 Rymer H, van Wyk de Vries B, Stix J, Williams-Jones G (1998) Pit crater structure and processes governing persistent activity at Masaya volcano. Nicaragua. Bull Volcanol 59:345– 355 Saito G, Kawanabe Y, Uto K, Sato H, Takada A, Ito J (2001) Petrological characteristics and volatile content of magma of August 18, 2000 eruption of Miyakejima volcano. Abstracts of 2001 Japanese Earth and Planetary Science Joint Meeting Sakai S, Yamada T, Ide S, Mochizuki M, Shiobara H, Urabe T, Hirata N, Shinohara M, Kanazawa T, Nishizawa A, Fujie G, Mikada H (2001) Magma migration from the point of view of seismic activity in the volcanism of Miyake-jima Island in 2000 (in Japanese with English abstract). J Geogr 110:145– 155

Sasai Y, Uyeshima M, Utada H, Kaghiyama T, Zlotnicki J, Hashimoto T, Takahashi Y (2001) The 2000 activity of Miyake-jima volcano as inferred from electric and magnetic field observations (in Japanese with English abstract). J Geogr 110:226–244 Simkin T, Howard KA (1970) Caldera collapse in Galapagos Islands, 1968. Science 169:429–437 Suwa A (1963) Volcano Miyake-jima and the 1962 eruption (in Japanese). J Geogr 72:293–305 Swanson DA, Christiansen RL (1973) Tragic base surge in 1970 at Kilauea volcano. Geology 1:83–86 Tilling RI, Dvorak JJ (1993) Anatomy of a basaltic volcano. Nature 363:125–133 Tsukui M, Suzuki Y (1998) Eruptive history of Miyakejima Volcano during the last 7,000 years (in Japanese with English abstract). Bull Volcanol Soc Jpn 43:149–166 Tsukui M, Niihori K, Kawanabe Y, Suzuki Y (2001) Stratigraphy and formation of Miyakejima Volcano (in Japanese with English Abstract). J Geogr 156:156–167 Tsuya H (1941) The eruption of Miyake-sima, one of the seven Izu Islands, in 1940. Geological observations of the Miyakesima eruption of 1940 (I). Bull Earthqu Res Inst 19:263–294 Ukawa M, Fujita E, Yamamoto E Okada Y, Kikuchi M (2000) The 2000 Miyakejima eruption: crustal deformation and earthquakes observed by the NIED Miyakejima observation network. Earth Planet Space 52:xix–xxvi Uto K, Kazahaya K, Saito G, Ito J, Takada A, Kawanabe Y, Hoshizumi H, Yamamoto E, Ukawa M, Fujita E, Okada Y, Kikuchi M (2001) Step-like tilt change occurred during the caldera-forming stage of the 2000 Miyakejima volcanic activity (in Japanese with English Abstract). J Geogr 156:181–190 Yamamoto T, Miyagi I, Tomiya A, Satoh H, Hamazaki S, Shinohara H (2001) Magma ascending model of 2000 Miyakejima eruptions: evidence form pyroclastics of August 18 and SO2-rich volcanic gas (in Japanese with English Abstract). J Geogr 156:257–270