Phreatomagmatic eruptions associated with the caldera collapse during the Miyakejima 2000 eruption,Japan

The 2000 AD eruption of Miyakejima was characterized by a series of phreatomagmatic eruptions from the subsiding caldera. Six major eruptive events occurred, and they can be divided into the first and second periods separated by a 25-day hiatus. The phreatomagmatic eruptions produced a total of ~ 2 × 10¹⁰ kg of tephra, which mainly comprised fine-grained volcanic ash. The tephra layers could be divided into six fall units corresponding to the six major eruptive events.     

Submarine flank eruption preceding caldera subsidence during the 2000 eruption of Miyakejima Volcano, Japan

During the early part of a seismic swarm preceding eruption and caldera formation at Miyakejima Volcano, discoloured sea surfaces were observed 1.5 km off the western coast of Miyakejima on 27 June 2000. A later survey of the area using a multi-beam side scan sonar and a remotely operated small submarine revealed four craters of 20–30 m diameter aligned east-west in a 100×10–30 m area on the seafloor, with hot water at 140°C being released from one of the centres. Each crater consists of submarine spatter overlain in part by scoria lapilli. Dredged spatter from the craters was fresh, and there was no evidence of activity of marine organisms on the spatter surface, indicating that the discoloured sea surface resulted from magmatic eruption on the seafloor. This eruption occurred when a westward-propagating seismic swarm, initiated beneath Miyakejimas summit, passed through the area. Finding new magma on the seafloor demonstrates that this seismic swarm was associated with intruding magma, moving outward from beneath Miyakejima. Submarine spatter shows flattened shapes with a brittle crust formed by cooling in water, and its composition is aphyric andesite of 54 wt% SiO₂. The spatter is similar in whole rock and mineral composition to spatter erupted in 1983. However, the wide range of Cl in melt inclusions in plagioclase of the 27 June submarine spatter shows that it is not simply a remnant of the 1983 magma, which has only high Cl melt inclusions in plagioclase. The mixed character of melt inclusions suggests involvement of a magma with low Cl melt inclusions. The magma erupted explosively on 18 August from Miyakejimas summit, considered as the second juvenile magma in this eruption, contains low Cl melt inclusions in plagioclase. Based on these observations and the eruption sequence, we present the following model: (1) A shallow magma chamber was filled with a remnant of 1983 magma that had evolved to a composition of 54–55 wt% SiO₂. (2) Injection of the 18 August magma into this chamber generated a mixed magma having a wide range of Cl in melt inclusions contained plagioclase. The magma mixing might have occurred shortly before the submarine eruption and could have been a trigger for the initiation of the removal of magma from the chamber as an extensive dyke, which eventually led to caldera subsidence.Editorial responsibility: S Nakada, T Druitt     

Chronology and products of the 2000 eruption of Miyakejima Volcano, Japan

Lateral migration of magma away from Miyakejima volcanic island, Japan, generated summit subsidence, associated with summit explosions in the summer of 2000. An earthquake swarm beneath Miyakejima began on the evening of 26 June 2000, followed by a submarine eruption the next morning. Strong seismic activity continued under the sea from beneath the coast of Miyakejima to a few tens of kilometers northwest of the island. Summit eruptive event began with subsidence of the summit on 8 July and both explosions and subsidence continued intermittently through July and August. The most intense eruptive event occurred on 18 August and was vulcanian to subplinian in type. Ash lofted into the stratosphere fell over the entire island, and abundant volcanic bombs were erupted at this time. Another large explosion took place on 29 August. This generated a low-temperature pyroclastic surge, which covered a residential area on the northern coast of the island. The total volume of tephra erupted was 9.3×10⁶ m³ (DRE), much smaller than the volume of the resulting caldera (6×10⁸ m³). Migration of magma away from Miyakejima was associated with crustal extension northwest of Miyakejima and coincident shrinkage of Miyakejima Island itself during July–August 2000. This magma migration probably caused stoping of roof rock into the magma reservoir, generating subsurface cavities filled with hydrothermal fluid and/or magmatic foam and formation of a caldera (Oyama Caldera) at the summit. Interaction of hydrothermal fluid with ascending magma drove a series of phreatic to phreatomagmatic eruptions. It is likely that new magma was supplied to the reservoir from the bottom during waning stage of magmas migration, resulting in explosive discharge on 18 August. The 18 August event and phreatic explosions on 29 August produced a conduit system that allowed abundant SO₂ emission (as high as 460 kg s^–1) after the major eruptive events were over. At the time of writing, inhabitants of the island (about 3,000) have been evacuated from Miyakejima for more than 3 years.     

Earthquake swarms preceding the 2000 eruption of Miyakejima volcano, Japan

The first sign of magma accumulating beneath Miyakejima, an island volcano in the northern Izu islands, Japan, came at around 18:00 on 26 June 2000, when a swarm of earthquakes was detected by a volcano seismic network on the island. Earthquakes occurred initially beneath the southwest flank near the summit and gradually migrated west of the island, where a submarine eruption occurred the next morning. Earthquakes then migrated further to the northwest between Miyakejima and Kozushima, another volcanic island and developed to the most intense earthquake swarm ever observed in and around Japanese archipelago. To better image how the initial magma intrusion occurred, we relocated hypocenters by using a station-correction method and a double-difference method. The relocated epicenters are generally concentrated near the upper bound of dyke intrusions inferred from geodetic studies throughout the initial stages of the 2000 eruption at Miyakejima from 26 to 27 June 2000. As for seismic activity westward off Miyakejima in the morning on 27 June, hypocenters from both a nationwide seismic network that were relocated by the double-difference method, and those from the volcano seismic network relocated by the station-correction method, formed a very shallow cluster that ascended slowly with time as it propagated northwestward from Miyakejima. This suggests that the dykes have both a radial and upward component of movement.Editorial responsibility: S. Nakada, T. Druitt     

Magma plumbing system of the 2000 eruption of Miyakejima Volcano, Japan

During the 2000 eruption at Miyakejima Volcano, two magmas with different compositions erupted successively from different craters. Magma erupted as spatter from the submarine craters on 27 June is aphyric basaltic andesite (<5 vol% phenocrysts, 51.4–52.2 wt% SiO₂), whereas magma issued as volcanic bombs from the summit caldera on 18 August is plagioclase-phyric basalt (20 vol% phenocrysts, 50.8–51.3 wt% SiO₂). The submarine spatter contains two types of crystal-clots, A-type and A-type (andesitic type). The phenocryst assemblages (plagioclase, pyroxenes and magnetite) and compositions of clinopyroxene in these clots are nearly the same, but only A-type clots contain Ca-poor plagioclase (An < 70). We consider that the A-type clots could have crystallized from a more differentiated andesitic magma than the A-type clots, because FeO*/MgO is not strongly influenced during shallow andesitic differentiation. The summit bombs contain only B-type (basaltic type) crystal-clots of Ca-rich plagioclase, olivine and clinopyroxene. The A-type and B-type clots have often coexisted in Miyakejima lavas of the period 1469–1983, suggesting that the magma storage system consists of independent batches of andesitic and basaltic magmas. According to the temporal variations of mineral compositions in crystal-clots, the andesitic magma became less evolved, and the basaltic magma more evolved, over the past 500 years. We conclude that gradually differentiating basaltic magma has been repeatedly injected into the shallower andesitic magma over this period, causing the andesitic magma to become less evolved with time. The mineral chemistries in crystal-clots of the submarine spatter and 18 August summit bombs of the 2000 eruption fall on the evolution trends of the A-type and B-type clots respectively, suggesting that the shallow andesitic and deeper basaltic magmas existing since 1469 had successively erupted from different craters. The 2000 summit collapse occurred due to drainage of the andesitic magma from the shallower chamber; as the collapse occurred, it may have caused disruption of crustal cumulates which then contaminated the ascending, deeper basalt. Thus, porphyritic basaltic magma could erupt alone without mixing with the andesitic magma from the summit caldera. The historical magma plumbing system of Miyakejima was probably destroyed during the 2000 eruption, and a new one may now form.Editorial responsibility: S Nakada, T Druitt     

Dike intrusion associated with the 2000 eruption of Miyakejima Volcano, Japan

The model for the 2000 dike intrusion event between Kozushima and Miyakejima volcano, Japan, was reinvestigated. After the sudden earthquake swarm in Miyakejima volcano, a dike intrusion of large volume was detected by the nationwide GPS network (Geonet). The displacements detected with GPS stations over an area with a radius of about 200 km shows a distribution that is consistent with the dike source being located near Miyakejima volcano.The dike was intruded northwestwards between Miyakejima and the neighboring Kozushima volcano. We searched for the parameters in the models that reproduce the regional displacements due to dike intrusion between Miyakejima and Kozushiima islands. We tested three models, (1) the model with a single dike, (2) the model with a dike and a point dislocation source which represents a creep dislocation source and (3) the model with a dike and a deflation source which represents a magma reservoir. Though all three models can match the horizontal displacements near the source area, model 1 fails to reproduce the regional displacements in the central part of Japan. Both models 2 and 3 can reproduce the regional displacement for horizontal components. Model 3 produces slightly better results than model 2 for vertical components. The balance in the volume budget for models 2 and 3 is also consistent with the observations. These results show that we cannot distinguish between the two models using only GPS observation. As there is no direct evidence for such a large creep or ductile source (corresponds to M7 or more) as proposed in model 2 and the active seismic region migrated back and forth within the linear swarm region, the model with a dike and a deep magma source is preferable. For the deflation point source, we obtained a deflation volume of 1.5 km³ at the depth of 20 km below the dike. An additional ~0.95 km³ of volume loss through caldera collapse and edifice deflation took place at Miyakejima. We conclude that the magma that intruded the dike came in part from below Miyakejima and in part from below the sea floor between Miyakejima and Kozushima, perhaps from reservoirs at the Moho.Editorial responsibility: S Nakada, T Druitt     

Vesiculation path of ascending magma in the 1983 and the 2000 eruptions of Miyakejima volcano, Japan

The vesiculation of magma during the 1983 eruption of Miyakejima Volcano, Japan, is discussed based on systematic investigations of water content, vesicularity, and bubble size distribution for the products. The eruption is characterized by simultaneous lava effusion and explosive sub-plinian (‘dry’) eruptions with phreatomagmatic (‘wet’) explosions. The magmas are homogeneous in composition (basaltic andesite) and in initial water content (H₂O = 3.9±0.9 wt%), and residual groundmass water contents for all eruption styles are low (H₂O <0.4 wt%) suggestive of extensive dehydration of magma. For the scoria erupted during simultaneous ‘dry’ and ‘wet’ explosive eruptions, inverse correlation was observed between vesicularity and residual water content. This relation can be explained by equilibrium exsolution and expansion of ca. 0.3 wt% H₂O at shallow level with different times of quenching, and suggests that each scoria with different vesicularity, which was quenched at a different time, provides a snapshot of the vesiculation process near the point of fragmentation. The bubble size distribution (BSD) varies systematically with vesicularity, and total bubble number density reaches a maximum value at vesicularity Φ ∼ 0.5. At Φ  ∼ 0.5, a large number of bubbles are connected with each other, and the average thickness of bubble walls reaches the minimum value below which they would rupture. These facts suggest that vesiculation advanced by nucleation and growth of bubbles when Φ < 0.5, and then by expansion of large bubbles with coalescence of small ones for Φ > 0.5, when bubble connection becomes effective. Low vesicularity and low residual water content of lava and spatter (Φ  < 0.1, H₂O  < 0.1 wt%), and systematic decrease in bubble number density from scoria through spatter to lava with decrease in vesicularity suggest that effusive eruption is a consequence of complete degassing by bubble coalescence and separation from magma at shallow levels when magma ascent rate is slow.

Petrological characteristics and volatile content of magma from the 2000 eruption of Miyakejima Volcano, Japan

Among the series of eruptions at Miyakejima volcano in 2000, the largest summit explosion occurred on 18 August 2000. During this explosion, vesiculated bombs and lapilli having cauliflower-like shapes were ejected as essential products. Petrological observation and chemical analyses of the essential ejecta and melt inclusions were carried out in order to investigate magma ascent and eruption processes. SEM images indicate that the essential bombs and lapilli have similar textures, which have many tiny bubbles, crystal-rich and glass-poor groundmass and microphenocrysts of plagioclase, augite and olivine. Black ash particles, which compose 40% of the air-fall ash from the explosion, also have similar textures to the essential bombs. Whole-rock analyses show that the chemical composition of all essential ejecta is basaltic (SiO₂=51–52 wt%). Chemical analyses of melt inclusions in plagioclase and olivine phenocrysts indicate that melt in the magma had 0.9–1.9 wt% H₂O, <0.011 wt% CO₂, 0.04–0.17 wt% S and 0.06–0.1 wt% Cl. The variation in volatile content suggests degassing of the magma during ascent up to a depth of about 1 km. The ratio of H₂O and S content of melt inclusions is similar to that of volcanic gas, which has been intensely and continuously emitted from the summit since the end of August 2000, indicating that the 18 August magma is the source of the gas emission. Based on the volatile content of the melt inclusions and the volcanic gas composition, the initial bulk volatile content of the magma was estimated to be 1.6–1.9 wt% H₂O, 0.08–0.1 wt% CO₂, 0.11–0.17 wt% S and 0.06–0.07 wt% Cl. The basaltic magma ascended from a deeper chamber (10 km) due to decrease in magma density caused by volatile exsolution with pressure decrease. The highly vesiculated magma, which had at least 30 vol% bubbles, may have come into contact with ground water at sea level causing the large explosion of 18 August 2000.Editorial responsibility: S. Nakada, T. DuittAn erratum to this article can be found at     

Relationship between caldera collapse and magma chamber withdrawal: An experimental approach

Collapse calderas have received considerable attention due to their link to Earth's ore deposits and geothermal energy resources, but also because of their tremendous destructive potential. Although calderas have been investigated through fieldwork, numerical models and experimental studies, some important aspects on their formation still remain poorly understood. One key issue concerns the volume of magmas involved in caldera-forming eruptions. We perform analogue experiments to correlate the structural evolution of a collapse with the erupted magma chamber volume fraction. The experimental device consists of a transparent box (60 × 60 × 40 cm) filled with dry quartz sand and a water-filled latex balloon as a magma chamber analogue. Evacuation of water through a pipe causes a progressive deflation of the balloon that leads to a collapse of the overlying structure. The experimental design allows to record the temporal evolution of the collapse and to track the evolution of fractures and faults. We study the appearance and development of specific brittle structures, such as surface fractures or internal reverse faults, and correlate each different structure with the corresponding removed magma chamber volume fraction. We also determine the critical conditions for caldera onset. Experimental results show that, at any stage of caldera developments, the experimental relationship between volume fraction and chamber roof aspect ratio fits a logarithmic curve. It implies that volume fractions required to trigger caldera collapse are lower for chambers with low aspect ratios (shallow and wide) than for chambers with high aspect ratios (deep and small). These results are in agreement with natural examples and previous theoretical studies.     

Geochemical Changes in Spring Water Associated with the 2000 Eruption of the Miyakejima Volcano, Japan

Water was sampled from eight springs and a lake in volcanic Miyakejima Island of Japan after the 2000 eruption. Major chemical and isotopic compositions of the water were analyzed. Significant increases of sulfate ion are observed in several springs where the thickness of ejecta exceeds 32 mm. A good relationship of Cl/S mole ratios between spring water and leachate of the ejecta is observed. Sulfur isotopic compositions of the spring water become close to that of leachate of the ejecta as time elapses after the eruption. Consequently the sources of the added sulfate ion in the spring water after the eruption are interpreted to be anhydrite and adhered sulfur of the ejecta.     

Implication of the temporal sulphur isotopic variation during the 2000 eruption of Miyakejima Volcano, Japan

Abstract   The variation of sulphur isotopic composition during the 2000 eruption of the Miyakejima Volcano was examined in order to monitor the temporal change of the volcanic activity. The δ³⁴S values of water-soluble sulphate leached from volcanic ash effused during intermittent eruptions from July to September 2005 range from +5 to +11‰ with a fluctuation of ca 3‰ within a single eruption. The δ³⁴S value of sulphuric acid mist collected with 'Cu-metal trap' placed on the flank of the volcano from December 2000 to January 2001 is +6.2‰. These sulphur isotopic compositions of sulphate, which were isotopically equilibrated in the subvolcanic hydrothermal system, indicate that the temperature of the hydrothermal system beneath the caldera increased after the period of intermittent phreatic and phreatomagmatic eruptions. Then, the δ³⁴S value of sulphuric acid trapped from January to March 2001 was +9.0‰ and the δ³⁴S value of water-soluble sulphate on volcanic ash emitted with minor eruption in May 2001 was +11.0‰, suggesting a decrease in temperature of the subvolcanic hydrothermal system.     

Interaction between caldera collapse and eruptive dynamics during the Campanian Ignimbrite eruption,Phlegraean Fields,Italy

The Campanian Ignimbrite (36000 years B.P.) was produced by the explosive eruption of at least 80 km³ DRE of trachytic ash and pumice which covered most of the southern Italian peninsula and the eastern Mediterranean region. The eruption has been related to the 12-x15-km-diameter caldera located in the Phlegraean Fields, west of Naples. Proximal deposits on the periphery of the Phlegraean Fields comprise the following pyroclastic sequence from base to top: densely welded ignimbrite and lithic-rich breccias (unit A); sintered ignimbrite, low-grade ignimbrite and lithic-rich breccia (unit B); lithic-rich breccia and spatter agglutinate (unit C); and low-grade ignimbrite (unit D). Stratigraphic and componentry data, as well as distribution of accidental lithic types and the composition of pumice clasts of different units, indicate that coarse, lithic-rich breccias were emplaced at different stages during the eruption. Lower breccias are associated with fines-rich ignimbrites and are interpreted as co-ignimbrite lag breccia deposits. The main breccia unit (C) does not grade into a fines-rich ignimbrite, and therefore is interpreted as formed from a distinct lithic-rich flow. Units A and B exhibit a similar pattern of accidental lithic types, indicating that they were erupted from the same area, probably in the E of the caldera. Units C and D display a distinct pattern of lithics indicating expulsion from vent(s) that cut different areas. We suggest that unit C was ejected from several vents during the main stage of caldera collapse. Field relationships between spatter agglutinate and the breccia support the possibility that these deposits were erupted contemporaneously from vents with different eruptive style. The breccia may have resulted from a combination of magmatic and hydrothermal explosive activity that accompanied extensive fracturing and subsidence of the magma-chamber roof. The spatter rags probably derived from sustained and vigorous pyroclastic fountains. We propose that the association lithic-rich breccia and spatter agglutinate records the occurrence of catastrophic piecemeal collapse.     

Aeromagnetic surveys of Kuttyaro and Aso caldera regions,Japan

Detailed total-intensity aeromagnetic surveys of the Kuttyaro and Aso caldera regions, eastern Hokkaido and central Kyushu, were made during early 1964 under the auspices of the U.S.-Japan Co-operative Science Program in conjunction with a project for geophysical studies of calderas in Japan. Each caldera has a maximum diameter of about 22 km; the flights cover a 60 × 60 km rectangular area in each region. The Kuttyaro survey also encompasses the older caldera Akan, south-west of Kuttyaro, and the younger caldera Mashu to the east. All three lie within the Chīshīma (Kurile) volcanic zone. The isomagnetic contour map shows this zone as a belt of short wave-length anomaies which trends east-northeast across the region. Broad wavelength anomalies with trends intersecting the Chīshīma belt at an acute angle probably reflect structural relief on the Neogene volcanic basement concealed beneath Kuttyaro pyroclastic flows. The centre of Kuttyaro caldera coincides with the sharp southern termination of a strong basement high, whereas caldera faults and post-caldera domes have little magnetic expression. Mashu caldera is marked by a minimum in the position of the caldera lake; a symmetrical positive anomaly centering southeast of the caldera suggests either a buried older volcanic edifice or an intrusion. Akan caldera is represented by a magnetic depression encompassing a positive anomaly produced by its central post-caldera cone. The depression extends north of the geologically-deduced boundary of the caldera and may include an earlier collapse structure. Several volcanoes and lava sequences in the region produce negative anomalies due to inverse polarization. The most significant feature of the Aso isomagnetic map is a large, elongate positive anomaly that occupies the southern half of the caldera and extends about one caldera diameter to the south-west along the trend of the Median Tectonic Line of south-west Japan. Whether the anomaly represents the pre-Tertiary basement complex or a younger intrusion perhaps associated with Aso eruptive activity is uncertain. However, the causative body is abruptly truncated within the caldera by a major east-south-east structure passing through the eastern rim and coincident with the approximate locus of resurgent central vent eruptions. The structure may be a fault system that provided egress for the Aso pyroclastic flows. Superimposed on the basement anomaly are the effects of the topography of the caldera, the superficial caldera structure, and the post-caldera cones. An area of intense solfataric activity in the Kuju group of young volcanoes north of Aso has a pronounced negative anomaly. These two surveys illustrate the utility of the magnetic method for investigations of basement structure in caldera regions. They have served as a guide in interpreting reconnaissance aeromagnetic profiles flown concurrently for this project across some 14 other calderas or caldera-like structures in the Japanese islands.     

Subterranean structure of collapse caldera associated with andesitic and dacitic eruptions — Structural evolution of the Miocene Kakeya Cauldron,Southwest Japan

Three stages of collapse and doming of the inner subsided block are recognized in the Miocene Kakeya cauldron. The mechanism of the first collapse is not clear, but the second and third are volcanic in origin. The second collapse was triggered by eruptions of silicic andesite lava flows and pyroclastic ejecta. The boundary fault between the subsided block and its surroundings is nearly vertical. The subsided block formed a distinct basin structure, and its marginal part was intensely deformed by faulting. The third collapse took place cylindrically, accompanied by voluminous eruptions of dacitic pyroclastic materials. The collapsed block formed a basin structure with a gently dipping marginal part. The doming of the inner subsided block was due to increase of pressure in a magma chamber.The structure formed by the second collapse is not consistent with the concept of the subterranean structure of either the so-called «Krakatau»- (funnel-shaped) type or Valles-type calderas. The second collapse is transitional between «Krakatau»- and Valles-type calderas or a new type of volcanic depression. The features of the third collapse and the resurgent doming are similar to those of Valles-type calderas, except for the size of cauldrons and composition of magmas related to collapse. The similarities indicate that the Kakeya cauldron was formed in an extensional tectonic setting similar to that for Valles-type calderas.     

Variability in eruptive dynamics associated with caldera collapse: an example from two successive eruptions at Batur volcanic field,Bali, Indonesia

Batur volcanic field (BVF) in Bali, Indonesia, underwent two successive caldera-forming eruptions, CI and CII (29,300 and 20,150 years b.p., respectively) that resulted in the deposition of dacitic ignimbrites. The respective ignimbrites show contrasted stratigraphies, exemplify the variability of dynamics associated with caldera-forming eruptions and provide insights into the possible controls exerted by caldera collapse mechanisms. The Ubud Ignimbrite is widespread and covers most of southern Bali. The deposits consist dominantly of pyroclastic flow with minor pumice fall deposits. The intra-caldera succession comprises three distinct, partially to densely welded cooling units separated by non-welded pyroclastic flow and fall deposits. The three cooling units consist of pyroclastic flow deposits only and together represent up to 16 distinct flow units, each including a thin, basal, lithic-rich breccia. This eruption was related to a 13.5×10 km caldera (CI) with a minimum collapsed volume of 62 km³. The floor of caldera CI is inferred to have a piecemeal geometry. The Ubud Ignimbrite is interpreted as the product of a relatively long-lasting, pulsating, collapsing fountain that underwent at least two time breaks. A stable column developed during the second time break. Discharge rate was high overall, but oscillatory, and increased toward the end of the eruption. These dynamics are thought to reflect sequential collapse of the CI structure. The Gunungkawi Ignimbrite is of more limited extent outside the source caldera and occurs only in central southern Bali. The Gunungkawi Ignimbrite proximal deposits consist of interbedded accretionary lapilli-bearing ash surge, ash fall, pumice lapilli fall and thin pyroclastic flow deposits, overlain by a thick and massive pyroclastic flow deposit with a thick basal lag breccia. The caldera (CII) is 7.5×6 km in size, with a minimum collapsed volume of 9 km³. The CII eruption included two distinct phases. During the first, eruption intensity was low to moderate and an unstable, essentially phreatomagmatic column developed. During the second phase, the onset of caldera collapse drastically increased the eruption intensity, resulting in column collapse. The caldera floor is believed to have subsided rapidly, producing a single, short-lived burst of high eruption intensity that resulted in the deposition of the uppermost massive pyroclastic flow.Editorial responsibility: T. Druitt     

Genesis of magmas producing pumice flow and fall deposits of towada caldera,Japan

Towada caldera, lying near the northern end of Honsyu, Japan was constructed by eruptions of lavas and pyroclastic materials in three separate periods. At the ends of the first and second periods, great amounts of pumice were erupted in the form of pumice flow and fall respectively. Each pumice cruption was followed by collapse of the center of the cones resulting in double calderas. The lavas of these three periods and the pumice of the first and second periods were chemically analysed. The result was plotted in several different types of variation diagrams. The points for the lavas and pumice lie generally on smooth curves, indicating that the magmas which caused the pumice cruptions belong to the same general differentiation series as do the lavas. If SiO₂/FeO+Fe₂O, is plotted against sodification index (MgO x 100/MgO+FeO+Fe₂O, +Na₂O+K₂O), points for the lavas lie on a straight line, whereas those for the pumice lie on another straight line branching from the former at some point in the middle stage of differentiation. The rate of increase of this ratio in the pumice is greater than in the lavas, implying that less SiO₂ and more iron were subtracted from the magmas producing the pumice than from those producing the lavas. This was probably caused by crystallization of a greater amount of magnetite in the former magmas possibly due to higher oxygen partial pressure which may be in turn related to higher water content. It is not necessary to postulate melting of the crust in order to generate magmas of the pumice eruptions of the central type.     

The subsurface structure of the AIRA caldera and its vicinity in Southern Kyushu, Japan

Southern Kyushu, Japan, includes a chain of large and small calderas and active volcanoes, and the greatest part of it is covered with thick pyroclastic ejecta. The regional and local structures of this area are discussed from the standpoint of physical volcanology, with consideration of all available data.The regional structure of this area is examined in the light of gravity and geomagnetic anomalies. Two layers of the earth's uppermost crust are defined by spectrum analysis of the gravity anomalies. These two layers are identical with the two identified by seismicwave velocities. The Bouguer gravity anomalies are relatively high and rather monotonous over outcrops of the Mesozoic basement and the granite, but are relatively low and perturbed over calderas and caldera-like structures. Two low-gravity anomalies in Kagoshima Bay are remarkable. One is circular, with its center on the Aira caldera. The other is elongated between the Satsuma and Oosumi peninsulas. The southern end of the latter anomaly is occupied by the Ata caldera. Discussion of the gravity anomalies of the Aira caldera suggests that the subsurface basement has a funnel shape and is overlain by ‘fallback’. The sub bottom geology of the caldera suggests that it is formed by a few smaller depressions, though the distribution of the overall gravity anomalies is parallel with its shape.The southern part of Kagoshima Bay is characterized by a graben-like topography and low-gravity anomalies and, moreover, by several calderas. The middle part, between the Aira and Ata calderas, may have a graben-like structure. A profile crossing the bay through Sakurajima volcano is modeled on the basis of results from drilling and gravity surveys. The basement has a graben-like structure and is filled with coarse and low-density deposits, and the structure continues northwards to the Aira caldera with a funnel shape.A comparison of this area with the Taupo-Rotorua depression in New Zealand and Lake Toba in Indonesia, leads the authors to the conclusion that such major volcanic depressions may have been formed by amalgamation of a series of caldera-like structures which were formed by multiple violent explosions accompanied by ejection of a tremendous amount of pyroclastic material.