Introduction RESEARCH ARTICLE. Maya Yasui Takehiro Koyaguchi

Transcription

1 Bull Volcanol (2004) 66: DOI /s RESEARCH ARTICLE Maya Yasui Takehiro Koyaguchi Sequence and eruptive style of the 1783 eruption of Asama Volcano, central Japan: a case study of an andesitic explosive eruption generating fountain-fed lava flow, pumice fall, scoria flow and forming a cone Received: 25 August 2002 / Accepted: 16 June 2003 / Published online: 23 August 2003 Springer-Verlag 2003 Abstract The 3-month long eruption of Asama volcano in 1783 produced andesitic pumice falls, pyroclastic flows, lava flows, and constructed a cone. It is divided into six episodes on the basis of waxing and waning inferred from records made during the eruption. Episodes 1 to 4 were intermittent Vulcanian or Plinian eruptions, which generated several pumice fall deposits. The frequency and intensity of the eruption increased dramatically in episode 5, which started on 2 August, and culminated in a final phase that began on the night of 4 August, lasting for 15 h. This climactic phase is further divided into two subphases. The first subphase is characterized by generation of a pumice fall, whereas the second one is characterized by abundant pyroclastic flows. Stratigraphic relationships suggest that rapid growth of a cone and the generation of lava flows occurred simultaneously with the generation of both pumice falls and pyroclastic flows. The volumes of the ejecta during the first and second subphases are 0.21 km 3 (DRE) and 0.27 km 3 (DRE), respectively. The proportions of the different eruptive products are lava: cone: pumice fall=84:11:5 in the first subphase and lava: cone: pyroclastic flow=42:2:56 in the second subphase. The lava flows in this eruption consist of three flow units (L1, L2, and L3) and they characteristically possess abundant broken phenocrysts, and show extensive welding texture. These features, as well as ghost pyroclastic textures Editorial responsibility: T. Druitt M. Yasui () ) Department of Geosystem science, Nihon University, Sakura-josui, Setagaya-ku, Tokyo , Japan yasui@chs.nihon-u.ac.jp Tel.: ex Fax: T. Koyaguchi Earthquake Research Institute, University of Tokyo, Yayoi, Bunkyo-ku, Tokyo , Japan on the surface, indicate that the lava was a fountain-fed clastogenic lava. A high discharge rate for the lava flow (up to 10 6 kg/s) may also suggest that the lava was initially explosively ejected from the conduit. The petrology of the juvenile materials indicates binary mixing of an andesitic magma and a crystal-rich dacitic magma. The mixing ratio changed with time; the dacitic component is dominant in the pyroclasts of the first subphase of the climactic phase, while the proportion of the andesitic component increases in the pyroclasts of the second subphase. The compositions of the lava flows vary from one flow unit to another; L1 and L3 have almost identical compositions to those of pyroclasts of the first and second subphases, respectively, while L2 has an intermediate composition, suggesting that the pyroclasts of the first and second subphases were the source of the lava flows, and were partly homogenized during flow. The complex features of this eruption can be explained by rapid deposition of coarse pyroclasts near the vent and the subsequent flowage of clastogenic lavas which were accompanied by a high eruption plume generating pumice falls and/or pyroclastic flows. Keywords Asama volcano Clastogenic lava flow Explosive eruption Plinian eruption Pyroclastic cone Introduction The Asama 1783 eruption is the youngest sub-plinian to Plinian eruption of Asama volcano, central Japan (Fig. 1; see Appendix for general geology). This eruption generated pyroclastic flows and lavas and discharged abundant (0.5 km 3 dense rock equivalent: DRE) andesitic magma. A pioneering geological study of this eruption (Aramaki 1956, 1957) revealed that most of the eruptive products were generated in the climactic phase of the 3 months of activity, and that eruptive styles changed with time from pumice falls through pyroclastic flows to lava flows. This scenario of temporal variation in eruptive style was

2 244 largely based on the simple model of Verhoogen (1951) that eruptive evolution can be accounted for by a vertical gradient in volatile contents of the pre-eruptive magma column. On the other hand, recent studies on the basis of the compilation of observations contained in old documents (e.g., Hagiwara 1986; Yasui et al. 1997; Yasui and Koyaguchi 1998b) and detailed field observations (e.g., Inoue 1998; Yasui and Koyaguchi 1998a, 1998b) have enabled us to reconstruct the sequence of this eruption with a time resolution of several hours or less. These studies have shown that some field observations and descriptions in old documents are not consistent with the above simple scenario. In this paper, we will reconstruct the sequence of diverse eruptive styles of the climactic phase of this eruption on the basis of the old documents, lithological features of the deposits, and petrological features of juvenile materials. We will briefly suggest more general implications for explosive eruption of andesitic magmas on the basis of this example. Outline of the 1783 eruption inferred from old documents The volcanic events of the 1783 eruption are recorded in various old documents (listed in Hagiwara 1986, 1987, 1988, 1993, 1995). They include descriptions of phenomena such as ash fall, presence of an eruption cloud and rumbling, the times when they occurred, and the localities where they were observed (Imai and Mikada 1982; Hino and Tsuji 1993; Yasui et al. 1997). Nearly 1,000 descriptions in 190 old documents and old drawings were analyzed in order to reconstruct the sequence of the eruption. Figure 2 shows in time series the frequency of descriptions of individual phenomena in the old docu- Fig. 1 Index map and simplified geological map of Asama volcano after Aramaki (1963), showing the major three volcanic edifices (Kurofu, Hotokeiwa, and Maekake volcanoes) and the distribution of the 1783 eruptive products. P1, P32, P54, P70, M6, and M69 are the localities that are described in the text and on Fig. 4. AVO Asama Volcano Observatory, University of Tokyo Fig. 2a e Frequency diagram showing the number of descriptions in the old documents. a Total number; b number of localities where ash fall was reported; c number of localities where rumbling was audible; d number of descriptions of witnessed eruption clouds; e number of descriptions that refer to an old Japanese term, Yake. Most of the descriptions of ash fall describe the size of fallen particles and thickness of the deposit as well as qualitative features (e.g., intensity) of ash fall. The term rumbling was used when low-frequency sounds were audible around the volcano, but it seems to have been used also for continuous ground vibration in some cases. The descriptions of eruption clouds include their dimensions, shapes (direction of cloud motion) and color. Yake means the occurrence of an eruption, rumbling, eruptive column, and/or eruption cloud in a more general sense. Although this term is ambiguous from the viewpoint of volcanology, the frequency of use of this term in the old documents also changes systematically

3 ments. The 3-month eruption of Asama volcano in 1783 started on 9 May The number of reports increases dramatically after the end of July, reaching a maximum on 4 August. The numbers of reports for each phenomenon synchronously change with time. The parallel trends in Fig. 2 show qualitatively waxing and waning of the volcanic activity. Based on the waxing and waning trends, the 1783 eruption is divided into six episodes. The term episode is used here for an active period that is separated from another episode by a clear repose period of more than a few days or a period of markedly weaker eruptions. The individual episodes are composed of single and/or multiple phases. The term phase is used for a single eruptive event described in the old documents. When one phase can be further subdivided for a reason such as change in eruptive style, we will use the term subphases. Although the definitions of these terms are rather arbitrary, they are useful in describing detailed features of the eruptive sequence. The eruptive sequence and representative events in each episode described in the old documents are summarized in Table 1. The errors of time in the old documents may be as much as a few hours judging by the accuracy of clocks at that time. In this study, we are concerned with the style and the products of the eruptions in episode 5. Episode 5 is considered to begin on the 2 August. During episode 5, the duration of individual eruptive phases and the mass of erupted magma increased exponentially. The activity before the evening of 4 August is divided into more than four phases, each of which was characterized by Plinian activity lasting for 4 5 h with or without generation of pyroclastic flows. The eruption intensity increased on the afternoon on 4 August. Some old documents and an old drawing describe the generation of a pyroclastic flow at about 4 p.m. The final phase of episode 5 started with a large scale Plinian eruption about 7 p.m. that lasted for about 15 h (Table 1). In Karuizawa town, which was nearly on the dispersal axis of the eruption cloud, intense fallout of pyroclasts caused the inhabitants to evacuate towards the SW. The final phase of episode 5 yielded diverse deposits with contrasting modes of emplacement, such as pyroclastic flows, lava flows, and constructed a cone as well as the pumice falls. Because it is considered that most of the magma of the 1783 eruption was erupted during this phase, we define this final phase of episode 5 as the climactic phase, hereafter. The climactic eruption ceased in the morning of 5 August. After an interval of a few hours, a large explosion with a loud booming sound occurred at 10 a.m. of 5 August and a pyroclastic flow and debris avalanche was generated that moved toward the northern foot of the volcano, devastating a village on the way (Aramaki 1956). We define this event as episode 6. The flow deposit from this event has unique lithological features and it is called the Kambara flow deposit ; it contains large angular juvenile blocks up to 60 m in diameter, while the matrix is mostly composed of accessory materials. There are at least three different hypotheses on the origin of the Kambara flow deposit: it was the result of (1) an explosion at the summit (Aramaki 1956), (2) a flank eruption (Inoue et al. 1994), or (3) a secondary explosion disrupting a lava flow (Hayakawa 1995). Based on the fact that the Kambara flow deposit is traced back to Onioshidashi lava flow, and that the angular juvenile blocks show a texture of quenched lava, we endorse the third hypothesis. The presence of abundant accessory materials indicates that the flow accompanied an erosion of the old volcanic edifice. This pyroclastic flow/debris avalanche flow entered the Agatsuma River and transformed into a water-saturated mudflow and flood. The mudflow and flood was a serious disaster, killing more than 1,000 people even in areas up to several hundred kilometers from the volcano. The origin of the Kambara flow is still open to question and study of this problem is in progress. Pyroclastic fall deposits 245 The main part of the 1783 pyroclastic fall deposits is dispersed toward the ESE, whereas thin (up to 16 cm) pale gray pumice fall layers are found in the NE and NNW (Fig. 3a, b). These deposits can be correlated with episodes on the basis of the dispersal of eruption clouds described in the old documents (Imai and Mikada 1982; Hino and Tsuji 1993; Yasui et al. 1997). The NE and NNW pyroclastic fall deposits were most likely deposited during episodes 4 and 3, respectively (see Table 1). Most of the ESE pyroclastic fall deposits correspond to episode 5, and the lower most part to episode 4. The ESE pyroclastic fall deposits consist of 22 fall units of pumice fall and ash fall. A representative stratigraphic section of the ESE fall deposits taken at the locality of M6 (5.6 km, E15S from the summit crater) is shown in Fig. 4. It is characterized by a stratified lower half and an massive upper half at every locality. The lower half consists of many thin (up to 10 cm) pumice fall units with finer grain-size distributions; these are layers, 2p-4p, 6p-9p, 11p, 13p, 15p-16p, and 17p. Layer 19p in the middle level is coarse-grained compared with the underlying units. The upper half (layer 21p) is composed of a single massive layer, although it rarely shows weak stratification. Layer 21p is the thickest and coarsestgrained; it represents nearly a half of the total volume (Fig. 5, Table 2). Distribution and grain-size distribution of the individual fall units are described in detail by Yasui et al. (1997). The isopach maps for representative units are given in Fig. 3c f. According to recent tephra dispersal models (e.g., Koyaguchi 1994), change in grain-size distribution as a function of distance from vent is a sensitive indicator of dynamics of spreading eruption clouds. The upper layers (layers 19p and 21p) tend to be coarse grained, even in the distal area, whereas the grain size of the lower layers (e.g., layer 17p) rapidly decreases with distance from the vent. Such a contrast in grain-size distribution implies that the eruption clouds of the upper layers (layers 19p and 21p) traveled faster than those of the lower layers

4 246 Table 1 Eruption episodes and activity during the Asama 1783 eruption based on the old documents. A small number (less than four) of descriptions were reported on the events of 8 May, 24, June and 15, 16, July; however, the accuracy of those reports is not well defined Episode Month Date Approximate time Remarks 1 May 9? Beginning of eruption. Rumbling was felt in several villages. Thirteen old documents mention that the first eruption occurred on 9 May; however, no detailed description of this activity was found About 46 days interval to the next records 2 June 25 Around 10 a.m. to the noon Rumbling. Vertical eruption column with black color. Ash fall in the ESE to SE. A vertically developed directed eruption column is depicted in some old drawings on from that day. Rumblings (both sound and ground vibration) were Night also reported from seven localities Rumbling. Ash fall in the E to ESE About 3 weeks interval to the next records 3 July 17 Around 8 p.m. Rumbling. Fire lightning in the night sky was seen from the ground 13 km SW of the volcano. Ash fall to the N and NNW Interval for several days Intermittent activities of rumbling associated with small eruptions followed by ash falls July 27 Afternoon (1) Eruption began after midday. Thick eruption cloud directed to the east and pyroclastic fall in the NE district Late evening (2) Pyroclastic fall in the NE district continued until night. A distant ash fall (176 km NE) was recorded. Rumbling was felt more than 300 km in the SW Afternoon Rumblings were felt in many places including the distant places more than 200 km. Ash fall in the NE to the ESE 29 After 2 p.m. Eruption for about 5 h, within which a few pulses were recognized. Eruption got intense abut 4 p.m. and ceased around sunset time, 7 p.m. Ash fall was recorded 400 km to the NE 30 Afternoon (1) Eruption got intense around 2 p.m. and ceased at sunset. Ash fall to the E and ESE July August 31 1 Late evening to the midnight (2) Ash fall to the N and E Interval for a half day August 2 Afternoon Eruption for several hours. 2 3 Night to early morning 3 4 Afternoon of 3 to the next morning 4 From noon to past 4 p.m. Weaker eruption than last 4 days. Ash fall to the NE and SE. Most intense eruption to date. Continuous eruption having two cycles of waxing and waning. Intense ash fall in the ESE including Edo (present Tokyo) almost continuous through the night. Rumbling was felt in many places including distant ones Continuous eruption with fluctuation in intensity. Intense pyroclasts fall in the ESE. Detailed waxing and waning of rumblings were documented at Kaga (present Kanazawa), to the 167 km WNW Intense eruption and heavy ash fall in the east Rumbling was felt up to 200 km from the volcano 5 Around 5 p.m. The sky was clear of ash fall for the first time Pyroclastic flow generated toward the NE-ENE flank. A low eruption cloud trailing to the east in an old drawing suggests generation of a pyroclastic flow. Some old documents describe the generation of a pyroclastic flow. 4 5 Night to the next morning Climactic phase Rising pillar of fire and falling pyroclasts including a fireball were seen from many places A famous old drawing depicted the details of the eruption column, eruptive cloud extending to the east, falling pyroclasts and lightning. Intense fallout of pyroclasts from the rising column onto the summit area is remarkable Most intense fallout of pyroclasts in the ESE like that caused by the falling of a heavy shower or hail Inhabitants in Karuizawa town nearly below the dispersal axis of the eruption cloud, were in panic due to the heavy pyroclastic fallout. Evacuation of people began at 9 p.m., triggered by the instant death of a man hit by a large pyroclast Reddish sky were seen at places more than 200 km from the volcano. Interval probably lasting several hours August 5 Morning Sunrise was seen from the west at the southern foot of Asama Volcano due to the heavy eruption cloud trailing towards the ESE. It was getting light slightly at dawn and got dark again around 7 a.m. 6 Ash fall continued until about 8 a.m. in many places in the ESE 10 a.m. A loud detonation sound was heard over a wide area Some collapse occurred on the northern flank of the volcano Mud or muddy rain began to fall about noon in the E-ESE

5 247 Fig. 3a f Representative isopach maps of the 1783 pyroclastic fall deposits (modified from Yasui et al. 1997). Values are in cm. Tr Trace. a NE and NNW pumice fall deposits; b totalized map for ESE pyroclastic fall deposit; c layer 17p, d layer 19p; e layer 21p; f layer 22a

6 248 Fig. 4 Representative columnar sections showing the stratigraphic relationships between deposits extending in different directions. Pie graphs showing the proportion of glass shard with different colors (clear, brown, and dark brown) in the ash fall deposits (5a, 18a, 20a, and 22a; see the columnar section at M6 for their stratigraphic relationships) are also shown. Asterisk shows the data from an ash fall layer that overly the pyroclastic flow deposit of the first member at the locality of P1 (see Fig. 1 for the locality of P1) Fig. 5 Plot of thickness (log scale) versus square root of isopach area for the pumice fall layers of the Asama 1783 eruption. The volume estimates on the basis of these relationships using Fierstein and Nathenson s (1992) method are given in Table 2. ESE total* is based on the isopach map obtained by this study (Fig. 3b). ESE total** is based on the isopach map by Minakami (1942). A hypothetical segment with 10 km of the thickness half-distance is also shown on the diagram (e.g., layer 17p). From the typical vertical structure of wind speed in summer, it is inferred that column heights of layers 19p and 21p were higher than that of layer 17p, extending to more than 10 km above the volcano. From the large volume and the column height estimates, Yasui et al. (1997) concluded that layer 21p is the product of Plinian eruption in the climactic phase (i.e., the final phase of episode 5). It is also suggested that most of the lower half of the ESE pyroclastic fall deposits correspond to the intermittent Plinian eruptions from 2 to 4 August (i.e., the earlier phases of episode 5). The ESE pyroclastic fall deposits contain several fineash fall layers: layers 1a, 5a, 10a, 12a, 14a, 18a, 20a, and 22a (Fig. 4). Layer 1a is characterized by abundant lithic fragments as well as vesicular glass shards, and is considered to be the product of discrete explosions of episode 2. The ash layers, except for layer 1a, consist of glass shards, crystals, and small amounts of lithic fragments, and show various colors: pinkish gray, light

7 249 Table 2 Summarized table of volume estimations for the 1783 eruptive products Volume a Density DRE volume b (km 3 ) (kg/m 3 ) (km 3 ) Unit Layer bt c (km) Pyroclastic fall deposits NNW NNW NE NE ESE 1a p p p p d ESE Total Total Subunit Thickness e (m) Cone Unit B Ba , Bb d , Bc f , Total Member Area (km 2 ) Thickness g (m) Pyroclastic flow deposits First d Second f 21.1 P5/M30/D , Third f , Total Thickness h (m) Lava flows L1 d 4.5 U25/D , L2 f 4.4 U13/D , L3 f 0.8 U15/D , Total Total volume Climactic phase i Subphase Lava Cone Fall/flow First d 0.18 (84) 0.02 (11) 0.01 (5) 0.21 Second f 0.11 (42) (2) 0.15 (56) 0.27 a Volume estimations of tephra were based on Fierstein and Nathenson (1992; Fig. 5) b DRE (dense rock equivalent) volumes were calculated assuming that magma density is 2,500 kg/m 3 c bt: Thickness half-distance (Pyle 1989) d Deposits of the first subphase of the climactic phase e Thicknesses at the crater wall (Fig. 7b) f Deposits of the second subphase of the climactic phase g P Proximal area; M the eastern part of medial area; D distal area and the western part of the medial area h U Upper stream; D mid to down stream. Note that L1 is overlain by L2 in the mid to upper stream i Numbers in parentheses indicate weight percent of each mode of emplacement brown, and light purple. The color of glass shards varies from clear to dark brown under the microscope. Clear glass shards are silicic (SiO 2 around 73 wt%) and rarely contain microlites. Brown glass shards contain a small amount of microlites, and have relatively mafic (SiO 2 around 67 wt%) compositions. Dark brown glass shards contain numerous microlites in the groundmass, and also have relatively mafic compositions (SiO 2 around 64 wt%); however, its accurate composition cannot be determined because of the abundant microlites. The proportion of the brown glass shards tends to increase upward in the section (Fig. 4). Each ash layer tends to thicken towards the area where the pyroclastic flow deposits are concentrated (i.e., the ENE flank of the volcano, Fig. 3f) rather than towards the vent. One of the fine-ash fall layers can be traced to an ash layer that directly covers the pyroclastic flow deposits (see P1 and M6 in Fig. 4). Although we cannot eliminate the possibility that the fine-ash layers deposited during weak (or lulls in) Plinian activity, the above two observations would suggest that at least some of the ash fall layers are derived from the ash clouds of pyroclastic flows. The volume of these deposits has been crudely estimated using the method of Fierstein and Nathenson (1992) (Fig. 5, Table 2). The total volume of the ESE pyroclastic fall deposits is 0.12 km 3, whereas the NE and NNW pumice fall deposits are at most and km 3, respectively. Assuming that average density of deposits is 600 kg/m 3, dense rock equivalent (DRE) volume of the 1783 pyroclastic fall deposit is estimated to be 0.03 km 3 (see Fig. 5, Table 2 for volumes of individual fall units). It should be noted that these values provide a lower bound of volume estimation. Volume estimate on the basis of an isopach map using all the available data, including previous measurements by Minakami (1942), provides a slightly greater value (0.15 km 3, see Fig. 5).

8 250 Furthermore, because of poor exposure, we could not fully determine the thickness distance relationship of the deposits in the distal area, which is considered to be characterized by a longer thickness half-distance (e.g., Bonadonna et al. 1998). If a hypothetical distal segment (say, 10 km of the thickness half-distance) is assumed (dotted line in Fig. 5), the estimate of the total volume will increase up to 0.05 km 3 (DRE). In a later section all these uncertainties in volume estimation will be taken into consideration. Cone The edifice named Kama-yama is a pyroclastic cone on the summit of Asama volcano. It occupies a saucer-like, shallow depression on the somma, Maekake-yama, and rises from the northern slope of Maekake-yama (Fig. 6). The center of Kama-yama is located about 150 m NE of that of Maekake-yama. The shape of Kama-yama is nearly axisymmetrical except its northern slope (Fig. 6). The northern outer slope of the cone is truncated by a U- shaped, small depression, and it is steeper (up to 30) than the other sectors (~20). The altitude of the summit of Kama-yama (2,568 m) is higher than that of Maekakeyama (2,524 m). Old documents and drawings described the rapid growth of the cone during the eruption. Several old drawings, in which abundant incandescent spatter is falling onto the vent area coeval with a Plinian eruption column, suggest that contemporaneous fountains formed a cone around the vent during the Plinian eruption (see Fig. 15 in Sigurdsson 2000). These records, as well as the following geological evidence, suggest that Kama-yama formed during the 1783 eruption (Yasui and Koyaguchi 1998b). The surface of Kama-yama is covered with deposits of loose lapilli and ash including volcanic bombs from the Vulcanian explosions after the 1783 eruption (hereafter the recent Vulcanian deposits ). There is a crater with a diameter of m and 200 m deep on the top of Kama-yama (Fig. 7a). A stratified section of the cone (>100 m thick) is exposed on the inner crater wall. The crater wall sequence is divided into three parts based on unconformities and differences in lithology: units A, B, and C in ascending order (Fig. 7b). These units are deposits of pyroclastic fall or lava fountains, judging from the fact that these layers mantle underlying topographic undulation with uniform thickness. Units A and B are alternations of welded and non-welded pyroclastic deposits, and they are separated by a distinct erosion surface. Unit C consists of the recent Vulcanian deposits. Unit B is stratigraphically correlative with the deposits of the 1783 eruption (Yasui and Koyaguchi 1998b). Unit B is further subdivided into three units: subunits Ba, Bb, and Bc (Fig. 7b). Thickness of subunit Ba is uniform around the crater rim (20 m), whereas those of subunits Bb and Bc decrease at the northern crater wall from 50 to 15 m and from 10 to 3 m, respectively. Subunit Ba is composed of several pyroclastic fall cooling units. Fig. 6 Map showing topography and geology around the summit region of Asama volcano (modified from Yasui and Koyaguchi 1998b). Values on the contours show the altitude in meters. P46 and P60 are representative outcrops that are described in the text. Cross section (A A 0 ) is also shown Each cooling unit, recognized by columnar joints of a distinct width, is several meters in thickness. Subunit Bb is a massive, densely welded pyroclastic fall deposit. Although a weak stratification is observed in the upper part of subunit Bb, wide columnar joints crossing the entire thickness of subunit Bb indicate that subunit Bb is a single cooling-unit. Subunit Bc is composed of stratified layers of oxidized pyroclasts with variable degrees of welding. This subunit also outcrops on the northern outer slope at altitudes between 2,275 and 2,400 m (e.g., P46 and P60 in Fig. 6), just below the recent Vulcanian deposits. The non-welded parts of subunit Bc consist of strongly oxidized, reddish brown blocks and small amount of lapilli and ash, and they are moderately well sorted (s=1.2 and Md=Ÿ5.4). The blocks are slightly flattened and some of them show eutaxitic texture. The dips of the elongated fiamme in densely-welded parts are parallel to that of the outer slope of Kama-yama. The

9 251 Fig. 7a,b Photo and columnar section of Kama-yama crater. a The view from the east side of the crater. At the northern crater rim, the upper part of subunit Bb and the lower part of subunit Bc are eroded, and the upper part of subunit Bc mantles the undulating erosion surface (thick dashed line). b Schematic columnar section of the crater wall petrological features of juvenile materials in these deposits indicate that these are products of the 1783 eruption (Yasui and Koyaguchi 1998b). The sequences of unit B and the ESE pyroclastic fall deposits share a common feature: both the deposits are characterized by a stratified lower half and a massive upper half (Yasui and Koyaguchi 1998b). This pattern is consistent with the fact that the eruption started with intermittent explosions, followed by the continuous, Plinian eruption of the climactic phase. It is suggested that cone formation and Plinian eruptions occurred simultaneously, and that subunit Bb may be correlative with the Plinian deposits of the climactic phase (layer 21p). The upper part of unit B on the northern crater wall is directly continuous to the surface of the Onioshidashi lava flow on the northern outer slope of Kama-yama. The topographic depression on the northern outer slope, which forms the surface of the lava flow (see Fig. 6), can be traced to the point where unit B is thinnest. These features, as well as the fact that the lava flow carries abundant meso-blocks of bedded cone deposits (coherent sections of cone tens of meters in length or diameter; see Sumner 1998 for its definition) on its surface, suggest that the generation of the Onioshidashi lava flow accompanied the partial failure of the cone. At the northern crater rim, the thickness of subunits Bb and Bc decreases due to erosion of the upper part of subunit Bb and the lower part of subunit Bc and the upper part of subunit Bc mantles the undulating erosion surface (Fig. 7a). This feature implies that failure of the northern part of the cone occurred before and/or during the deposition of subunit Bc. Volume of the 1783 cone is approximated by a truncated cone with a slope of 20, a basal radius of 560 m, and a relative height of 80 m, and it is estimated to be 0.05 km 3 (Table 2). On the basis of the data from the outcrops on the outer slope (P46 and P60 in Fig. 6), we assume that average density of the welded cone deposits is 2,000 kg/m 3, from which DRE volume of the cone deposits is estimated to be 0.04 km 3 (see Table 2 for volume estimates for individual subunits). Pyroclastic flow deposits The 1783 pyroclastic flow deposits are widely distributed toward the northeast up to 8 km from the vent. Scoriaceous pyroclastic flow deposits also sporadically crop out overlying pumice fall deposits (layer 21p) on the eastern and southern slope of Maekake-yama just below the summit (see Fig. 6). Vertical sections of the pyroclastic flow deposits are exposed in many gully walls particularly within 4 km of the summit crater on the NE flank (Fig. 8). However, medial exposures of the pyroclastic flow deposits are limited because their strong welding has limited erosion and only few shallow gullies have formed. The 1783 pyroclastic flow deposits are composed of many lithological units. Flow units can be defined by sedimentary structures such as the grading of grain-size distribution, lithological features of constituent blocks, and topography. Because the original topography of the flow units (e.g., lobes bounded by levees) is wellpreserved, the boundaries of these flow units can be identified from aerial photographs (Fig. 8). In addition to flow units, we can define cooling units; typically a single cooling unit is defined by vertical variation in degree of welding, an upper oxidized zone, and/or through-going

10 252 Fig. 8 Map showing distributions of the flow units of the 1783 pyroclastic flow deposits (units A to U) and lava flow (Onioshidashi lava). L1 L3 shows the flow units of the lava flow defined by Inoue (1998). 1 Distal part of the 1783 pyroclastic flow deposits, where the boundaries of the flow units are not well-defined; 2 the first member of the 1783 pyroclastic flow deposit, which is interbedded with layer 21p; 3 secondary deposits derived from the 1783 pyroclastic flow deposits of the upper stream; 4 area where the 1783 pyroclastic fall deposits overlie the pre eruptive products; 5 the ejecta of the 1108 a.d. eruption. The 1783 pyroclastic flow deposits are thickest in the area enclosed by the dashed line development of columnar joints ( simple cooling unit in the sense of Smith 1960). A single cooling unit can be composed of a single flow unit or several flow units. Some cooling units have several zones of dense welding and partial welding as well as intercalated oxidization zones ( compound cooling unit in the sense of Smith 1960). The variation in pattern of the lithological units reflects various patterns for timing of deposition of different flow units. The 1783 pyroclastic flow deposits are divided into three members (Fig. 4). The major part of the pyroclastic flow deposits forms the second member. According to the stratigraphic relations on the ENE flank and the eastern slope (Figs. 8 and 9), most of the pyroclastic flows are considered to have followed deposition of pumice in the climactic Plinian eruption (i.e., layer 21p). On the other hand, minor pyroclastic flow deposits and fine-ash fall layers derived from pyroclastic flows are interbedded within pumice fall deposits (see Fig. 4). We define these minor pyroclastic flows during or before the climactic Plinian eruption as the first member. In the proximal region, some flow units with distinctive lithological features overlie the second member, these are defined as the third member. The second and third members were grouped together as Agatsuma pyroclastic flow in Fig. 9 Block diagram showing stratigraphic relationships between the flow units of the pyroclastic flow deposits (A to U) and other eruptive products (pumice fall and lava flows). For the distribution of individual flow units see Fig. 8

11 253 Fig. 10 Schematic sketches of blocks contained in the pyroclastic flow deposits of the second and third members previous literature (e.g., Aramaki 1956). Representative features of the individual members are described below. One pyroclastic flow deposit of the first member crops out along the gully wall of the ENE flank (no. 2 in Fig. 8). It is a small-scale, non-welded, scoriaceous flow deposit, which is lenticularly interbedded with layer 21p. The upper surface of the deposit undulates with a wavelength of several meters. Welding and oxidation are rarely observed. The unit consists of rounded, vesicular, darkbrown blocks, and matrix ash. Large pumiceous, rounded blocks are concentrated in the upper part. It is directly mantled by a 3-cm fine-ash fall layer. This is mainly composed of brown and clear, vesicular glass shards, and can be traced to a thin fine-ash layer in layer 21p (Fig. 4). The fact that the pyroclastic flow deposit is interbedded with the pumice fall deposits (layer 21p) suggests that this small-scale pyroclastic flow was generated by partial collapse of the Plinian column. The rest of the first member can be recognized only from ash fall layers derived from pyroclastic flows (Fig. 4), and detailed features are not well-defined. The second member is composed of many flow units (e.g., units G-U in Figs. 8 and 9). Individual flow units consist of dark brown scoriaceous blocks and matrix ash. The deposits of the second member are characterized by strong welding. An oxidized zone of up to 30 cm thick develops in the upper surface of some welded layers. Generally, individual welded layers can be traced horizontally along the gully wall in the ENE flank. The thickness of individual welded layers vary from 3 m in the valley to 30 cm toward the ridge. Blocks are not flattened even in the strongly-welded part; the aspect ratio of 201 blocks ranges from 0.3 to 1.0. Generally speaking, welding is caused by compaction because of loading and/or agglutination due to high temperature. The presence of thin welded layers and undeformed blocks in the welded layer may indicate that these deposits were emplaced at high temperatures and immediately agglutinated without post-depositional compaction. Some upper flow units (units A F), are characterized by strongly developed reverse grading; they consist of an upper clast-supported, blocks rich zone, and a lower matrix-supported zone (Fig. 4). We define these flow units as the third member. The second and third members are separated by a distinctive oxidized zone, so that they belong to different cooling units; however, these two members are penetrated by common cooling joints at some localities. The upper zone commonly contains blocks of up to 3 m in diameter. Small amounts of oxidized ash are contained interstitially between the blocks. The lower half is mostly composed of dark brown ash and is depleted in large blocks. Some flow units of this member preserve a remarkable lobe-like topography and/or levees (e.g., unit D). The 1783 pyroclastic flow deposits contain blocks of diverse morphology including composite blocks. At least three types have been identified (Fig. 10): isolated blocks (type 1), composite blocks of type-1 material (type 2), and composite blocks showing various degrees of welding (type 3). Size of an individual block of type 1 is typically less than 1 m, whereas composite blocks (i.e., types 2 and 3) are up to 9 m (long axis). The internal features of type- 3 blocks, such as interlayering of welded and non-welded material, are similar to those of the pyroclastic deposits of the cone. Blocks of this type may have originated from collapsed parts of the cone. The total amount of large

12 254 thickness of the eastern part of the medial area is estimated to be 30 m. From available field data, average thickness of the deposits in the proximal area, including the flank slope of Maekake-yama (one third of the total area), is estimated to be 5 m and that in the distal area and the western part of the medial area is estimated to be 10 m. From these values, total volume is calculated to be 0.25 km 3 (0.15 km 3 DRE assuming that average density of the deposits is 1,500 kg/m 3 ; see Table 2). Lava flows Fig. 11 Diagrams showing masses per unit area for types 1, 2, and 3 blocks with different sizes (Ÿ7f to 12f) on the surface of the flow units C, D, E, F, L, and M of the 1783 pyroclastic flow deposits. Masses per unit area were determined by measuring numbers and size of blocks in outcrops of more than 280 2,500 m 2. Pie graphs showing the number ratio of types 1, 2, and 3 blocks are provided for the flow units depleted in large blocks (units C and L) blocks and the population of different types of block varies from one flow unit to another (Fig. 11). Units C and L are depleted in large blocks. Unit M is more enriched in type 3 blocks than units E and F. Units D and E predominantly contain types 1 and 2 blocks. The diverse population of block types may suggest that different units of pyroclastic flows were generated by different mechanisms. For example, partial collapse of the cone may have played a significant role in generation of unit M, whereas generation of unit D was due to a magmatic explosion. The total area covered by the pyroclastic flow deposits is 21 km 2. The deposit is thickest in the eastern part of the medial area (see Fig. 8). The maximum thickness is estimated to be up to 40 m on the basis of geophysical surveys using an air gun (Shimozuru 1981). The average Lava flows of the 1783 eruption are grouped together as the Onioshidashi lava flow (Aramaki 1956). The lava flows have the surface morphology of typical andesitic lava flows, such as lava levees, lateral cliffs, and terminal cliffs. Three flow units are recognized and they are defined as L1, L2, and L3 in the order of generation (Inoue 1998; Fig. 8). L1 and L2 are the main flows extending to the north and L3 is the northeastern branch. L1 and L2 can be traced to the northern, shallow depression on the outer slope of the cone just below the summit crater (Fig. 6). L3 can not be traced to the crater, but to a collapsed depression on the northeastern outer slope of the cone, suggesting that L3 is a rootless flow that originated during partial collapse of the outer slope of the cone. L1 is covered by the third member of the pyroclastic flow deposits (unit B) at the eastern part of mid-stream (Fig. 8). On the other hand, L3 overlies the third member of the pyroclastic flow deposits (units D, E, F, and O; see Fig. 8), and is not covered by any pyroclastic flow deposits, suggesting that L3 formed after all the pyroclastic flows had been deposited. The lava flows have unique surface and internal structures as follows. Meso-blocks of the bedded cone deposits commonly occur on the surface, particularly at the base of scarps in the upstream region (cf. Sumner 1998). The uppermost part (approximately 10 m from the surface) is composed of an alternation of welded and nonwelded pyroclastic layers (Inoue 1998). A borehole core section shows that the upper half consists of porous reddish brown, strongly oxidized materials, in which massive gray densely welded zones about 2 m thick are intercalated at ten horizons (Fig. 12b). Non-welded parts show stratification. The degree of welding tends to increase downward at a given locality. The lower half consists mostly of massive light gray-colored lava. The base (bottom 20 cm) is composed of angular, glassy breccia. Macroscopic and microscopic eutaxitic textures with reddish-brown, pale gray, and dark gray fiamme are common in the densely welded parts throughout the section including the lower half (Fig. 13a). These surface and internal structures suggest that the lava flows were originally formed by rheomorphism of pyroclastic materials rather than having been erupted as a coherent lava [i.e., clastogenic lava flow described by Cas and Wright (1987) and Sumner (1998)].

13 255 Fig. 12a f Summary of features of a borehole core sample from the Onioshidashi lava flow. a Map showing horizontal variation in bulk SiO 2 composition within the lava flow (Aramaki and Takahashi 1992). Borehole site (star) is also shown on the map. b Columnar section of the borehole core. c Vertical variation in density. d Bulk SiO 2 content. e Phenocryst content. f Proportion of broken plagioclase phenocrysts (volume of broken plagioclase phenocryst/total volume of plagioclase phenocryst) Another important feature is that most of phenocrysts were mechanically broken after their growth (Fig. 13b). Those crystals are not euhedral or subhedral, but have flat and/or curved surfaces that cut compositional zones. Rarely, the complete outline of an original phenocryst can be reconstructed from neighboring broken crystals like a jigsaw puzzle. Sometimes such broken crystals are aligned around a single vesicle, suggesting that violent vesiculation caused the breakage of the crystal. The size of individual broken crystals varies from several tens of microns to several millimeters. Fine crystals in the matrix of the lava flows are mostly composed of fragments of broken crystals rather than euhedral groundmass crystals. The proportion of the broken plagioclase phenocrysts in the lava flows (0.7 throughout the flow in Fig. 12f) is much higher than typical andesitic lava flows, but are as high as those of aggulutinate layers from other Japanese volcanoes (Yasui, in progress). The presence of such abundant broken crystals supports the idea that the lava flows are clastogenic. Phenocryst content in the lava flows varies from 14 to 29 vol% for 27 thin sections (Fig. 12e) and is substantially lower than those in juvenile fragments in the pyroclastic fall and flow deposits (>30 vol%). Phenocryst-poor samples in the lava flows tend to be rich in fine broken crystals. It is suggested that the phenocryst content in the clastogenic lava flows does not represent that in magma, but is largely affected by fragmentation of crystals; the phenocryst-poor samples are interpreted as welded ash. The total area covered by the lava flows is 7 km 2. Thickness of the lava flows varies from less than 10 m to more than 65 m (e.g., the borehole site in Fig. 12a). Average thickness of the deposits in the proximal area including the slope of Maekake-yama (one third of the total area) is estimated to be 35 m. Average thickness of the deposits in the medial to distal area, on the other hand, is estimated to be 50 m. From these values, total volume is calculated to be 0.31 km 3. Densities of the lava vary from 1,400 to 2,680 kg/m 3. On the basis of the data from the borehole section in Fig. 12c, we assume that average density of the lava flow is 2,400 kg/m 3, from which DRE volume of the lava flows is estimated to be 0.30 km 3 (see Table 2 for volume estimates for individual flow units). Petrological features The petrology of the juvenile materials of the 1783 eruption are diverse. The diverse features are useful to reconstruct the stratigraphic relationship between deposits of different modes of emplacement. In this section, we briefly describe the petrological features of the juvenile materials for this purpose rather than discuss detailed magma genesis. The bulk SiO 2 content of juvenile materials varies from 60 to 64 wt%. Macroscopic streaky and/or patchy textures, caused by differences in color of glass and groundmass crystallinity, are common. Under the microscope, the heterogeneous groundmass is composed of three types of domain; brown glassy part, clear glassy part, and small amounts of dark brown cryptocrystalline part. These differences correspond to a contrasts in chemical composition; the brown is around SiO 2 67 wt%, the clear is around SiO 2 73 wt%, and the dark brown is around SiO 2 64 wt% (cf. compositions of the glass shards in the fine ash layers). The compositions of brown and clear glass are considered to represent those of melts judging from their low microlite content (<10 vol%), whereas the composition of the dark brown part may be modified due to groundmass crystallization. Textures and color change gradually in some places, but sometimes the boundary between areas of different texture or color is sharp.

14 256 The 1783 magma carries clinopyroxene (cpx), orthopyroxene (opx), plagioclase, Fe Ti oxides, and small amounts of olivine as phenocrysts. Total phenocryst content is more than 30 vol% in juvenile materials in the pyroclastic flow deposits, whereas it is up to 57 vol% in juvenile materials in the pumice fall deposits. Glomeroporphyritic texture is common. Chemical compositions of the cores of these phenocrysts show broad variation: Mg# of cpx varies from 65 to 85, Mg# of opx from 63 to 72, An content of plagioclase from 58 to 91, and Fo content of olivine from 65 to 82. On the basis of mineral assemblage of crystal clots, these phenocrysts are divided into three groups. The first group is composed of cpx with cores of Mg#=65 70, opx with cores of Mg#=63 66, plagioclase with cores An=58 68, and Fe Ti oxide. The constituent minerals of this group are commonly euhedral, and they sometimes contain small light brown glass inclusions. The second group is composed of cpx with cores of Mg#=70 72, opx with cores of Mg#=66 72, and plagioclase with cores of An= These crystals are characterized by a honeycomb texture containing irregularly shaped, large brown glass inclusions. In addition to these phenocrysts, small amounts of olivine (Fo=65 82) and magnesian cpx (Mg#=80 85) with sector zoning are sometimes found; such crystals form the third group. The phenocrysts of the first and second groups tend to be surrounded by clear silicic glass and brownish mafic glass, respectively. Chemical compositions of rims of the phenocrysts of these two groups systematically vary with the composition of surrounding groundmass glass. These observations, as well as linear relationships in all the element-element variation diagrams, suggest that the petrological variations in the juvenile materials are basically accounted for by binary mixing between a relatively mafic magma carrying phenocrysts of the second group and a silicic magma carrying phenocrysts of the first group (Table 3). The amount of the third component (i.e., the phenocrysts of the third group and the dark brown matrix) is so small that it does not affect the general trend of the variation. All the samples have disequilibrium phenocryst assemblages and/or heterogeneous textures (i.e., streaky and patchy textures) in hand specimens, and no pure endmember magmas are observed. The petrological features of the two end-member magmas can be crudely estimated from the modal, bulk chemical, and glass compositions. Figure 14 shows the relationships between the bulk chemical composition (SiO 2 content) and the ratio of the honeycombed high-an plagioclase in the second group and the clear low-an plagioclase in the first group ( MF# in Fig. 14) for 25 samples. The result shows that MF# correlates with the bulk chemical composition; it is consistent with the hypothesis of binary mixing. The Fig. 13a b Clastogenic features of the lava flows. a Photo and its sketch showing macroscopic eutaxitic texture in the borehole core sample (depth 38 m). An arrow in the sketch shows upward in the borehole; b photomicrograph of a broken plagioclase phenocryst in the borehole core sample (depth 32.7 m)

15 257 Table 3 Summary of petrological features of end-member magmas of the 1783 eruption (phenocryst assemblages, modal proportions of the phenocrystic minerals and SiO 2 content of glass). cpx Clinopyroxene; opx orthopyroxene; pl plagioclase; Mg# 100 Mg/ (Mg+Fe); An anorthite content; Fo forsterite content; tr trace End member Phenocryst Glass SiO 2 content (wt%) 1 First group 73 cpx (Mg#=65 70) opx (Mg#=63 66) pl (An=58 68) cpx:opx:pl=7:13:80 2 Second group 67 cpx (Mg#=70 72) opx (Mg#=66 72) pl (An=76 91) cpx:opx:pl=16:21:63 3 a Third group 64 b Olivine (Fo=65 82) cpx (Mg#=80 85) Olivine (tr) and cpx (tr) a Because the amount of the end-member 3 is negligibly small, the petrological features of the 1783 eruptive products are approximately explained by the binary mixing of the end members 1 and 2 b Glass composition of the end-member 3 is considered to have been modified to some extent due to groundmass crystallization Fig. 14 Relationship between MF# and SiO 2 content. MF#= 100 HPL/(HPL+LPL) where HPL is the modal abundance of high An plagioclase with honey-combed structure in the second group and LPL is that of low An plagioclase with clear, oscillatory-zoned texture in the first group. Open squares show the average compositions of brown and clear glass determined by EPMA. Error bars indicate the range of measured SiO 2 content of glass compositions. Solid squares show the average of bulk SiO 2 content for phenocrysts of the first and second groups calculated from SiO 2 content of clinopyroxenes, orthopyroxenes, and plagioclase and their proportions (Table 3). Error bars show variation due to wide range of SiO 2 content of plagioclase phenocrysts in each group. Open and solid stars indicate estimated bulk SiO 2 contents of silicic and mafic end members petrological features at MF#=1 and 0 in Fig. 14 represent those of the two end-member magmas, respectively. According to modal and mineralogical analyses, the average SiO 2 contents of the phenocrysts of the first and second groups are 54 and 49 wt%, respectively. From Fig. 15 Bulk SiO 2 content of the 1783 eruptive products. Frequency diagrams of SiO 2 content for the deposits of pumice fall, pyroclastic flow, and lava flow. Most of the data used here were obtained by Aramaki and Takahashi (1992) these values, as well as the chemical compositions of the brown glass (SiO 2 67 wt%) and the clear glass (SiO 2 73 wt%), it is inferred that the phenocryst content of the mafic end member is approximately 40 wt%, whereas that of the silicic end member is 50 wt%; these estimates of phenocryst content are consistent with the results of modal analyses. The relatively small variation in the bulk SiO 2 content (4 wt%) compared with the variation in glass composition (7 wt%) can be explained by the high phenocryst content of the silicic end-member magma. Mixing ratios of the two end-member magmas systematically varies depending on the mode of emplacement. The bulk chemical compositions (e.g., SiO 2 content) of juvenile fragments of pumice fall deposits and pyroclastic flow deposits suggest bimodal distributions (Fig. 15). A high SiO 2 mode (62.5 wt%) is dominant in pumice fall deposits, whereas that of low SiO 2 content (61 wt%) is dominant in pyroclastic flow deposits. Systematic analyses of samples collected at 60 localities on the surface of the lava flows (Aramaki and Takahashi 1992) show that SiO 2 content varies slightly depending on flow units (Figs. 8 and 12a); L3 tends to have a lower SiO 2 content (<62 wt%) than that of L1 (>62 wt%). On the other hand, the bulk chemical compositions of samples from L2 show intermediate compositions between the other two flow units (Fig. 15). Boundaries separating groundmass of two distinct compositions are less sharp in the lava flows, suggesting that the lava flows resulted from more complete mixing compared with juvenile materials in the pumice falls or pyroclastic flow deposits. The variation of the mixing ratio between pyroclastic fall deposits and pyroclastic flow deposits as well as that within the lava flows, will be taken into consideration in reconstructing the sequence and the

16 258 Fig. 16 Schematic illustration of eruptive sequence of the 1783 eruption. See text for details dynamics of the climactic phase of the 1783 eruption in the next section. Eruptive style and sequence of the climactic phase There are at least seven key observations relevant to eruptive evolution: 1. The upper half of the ESE pyroclastic fall deposits (layer 21p) formed during the climactic phase. 2. Almost all the pyroclastic flows overlie the pyroclastic fall deposits (layer 21p). 3. The lower and upper parts of the ESE pyroclastic fall deposits can be stratigraphically correlated with subunits Ba and Bb of the crater wall, respectively. 4. Unit B of the crater wall can be topographically traced to the section of lava flows at the northern part of the cone. 5. The pyroclastic fall deposits have a most silicic composition (SiO 2 =62.5 wt%), whereas the pyroclastic flow deposits and the uppermost unit of the crater wall (subunit Bc) have a relatively mafic composition (SiO 2 =61 wt%). 6. The average chemical compositions of L1 and L3 flow units of the lava flows are similar to those of pyroclastic fall deposits and pyroclastic flow deposits, respectively, whereas L2 flow unit has an intermediate chemical composition. 7. Most of the lava flows show clastogenic textures. On the basis of these observations as well as the descriptions in old documents, we can reconstruct the sequence of episode 5 as Figs. 16 and 17. Repetitive Plinian eruptions with some minor pyroclastic flows became intensive beginning on 2 August, depositing pumice over a broad area, and starting construction of a stratified cone at the vent (subunit Ba). The climactic Fig. 17 Schematic illustration of eruptive style of the climactic phase of the 1783 eruption. See text for details