Conduit system, degassing, and flow dynamics of a rhyolite lava: A case study of the Shiroyama lava on Himeshima Island, Japan

This study presents a description of a rhyolite lava-forming eruption, including the conduit system, degassing history during the lava flow dynamics. We examined the Pleistocene Shiroyama rhyolite lava on Himeshima Island, Japan. The lava is mainly characterized by locally developed obsidian. Based on the structural variation, the obsidian lithofacies correspond to the shallow conduit. The geological investigation and FTIR analyses showed that gas removal from the conduit magma proceeded via vesiculation, fracturing, and brecciation, allowing formation of the dense obsidian. Since the lava originally maintained some extent of water, the lava effervesced just after the effusion. This vesiculation resulted in pervasive bubble coalescence and the formation of abundant permeable pathways. The volcanic gasses escaped via those pathways, allowing collapse of the bubbles and deflation of the lava. AMS (anisotropy of magnetic susceptibility) results indicate that the lava spread concentrically.


Introduction
Effusion of lava has resulted in various volcanic disasters that frequently lead to severe damage to houses, farms, and traffic networks. To understand the potential hazards, the eruption, degassing, and flow dynamics of lavas have been continuously studied. Volcanic hazards associated with lavas are usually diversified owing to the effect of composition on magma properties. In basaltic, andesitic, and dacitic lavas, there have been many opportunities for the direct observations, and their eruption, flow dynamics, and potential hazards are relatively well understood [e.g. Gregg 2017;Nakada et al. 2019]. However, it seems to be difficult for volcanologists to predict the flow behavior and the envisaged hazards precisely during effusion of rhyolite lava. This is because rhyolite lava effusion has rarely been directly observed, and little is known about its lava emplacement. The direct observations of rhyolite lavas are restricted to the 2008 eruption of Chaitén volcano [Castro and Dingwell 2009;Lara 2009] and the 2011-2012 eruption of Cordón Caulle volcano [Schipper et al. 2013;Tuffen et al. 2013]. These events have provided limited but valuable opportunities for understanding rhyolite lavas [Farquharson et al. 2015;Magnall et al. 2018;Schipper et al. 2013;Tuffen et al. 2013]. Because these are infrequent events, investigation has instead generally relied on the exposure of dissected lavas and conduits, or a combination of numerical and Corresponding author: kfuru@vega.aichi-u.ac.jp experimental work.
Non-explosive silicic volcanism, which often manifests as obsidian lava [e.g. Sano et al. 2015;Shields et al. 2016;Tuffen et al. 2013], has been considered to require sufficient outgassing during magma ascent, and the outgassing mechanism has been controversial. Taylor et al. [1983] and Eichelberger et al. [1986] explain how the connected bubbles formed during magma ascent act as a permeable network through which volatiles escapes, and the foam collapses to produce a dense lava. Castro et al. [2014] and Cabrera et al. [2015] proposed that degassing and outgassing occur via transient fracturing networks formed in the magma during ascent, and the fractures heal to form a dense lava. Wadsworth et al. [2020] recently suggest that rhyolitic magma generally fragments during ascent through the upper crust and that effusive eruptions result from conduit blockage and sintering of the pyroclastic products of deeper cryptic fragmentation. The silicic conduit structure has been revealed via investigation of spines and dissected conduits and is likely composed of the fault gouge zone, cataclastic and sheared zone, and massive zone from outer to center [Pallister et al. 2012]. Erupted rhyolite lavas usually show complex internal structures, and the development processes have been studied [e.g. Furukawa et al. 2019;Manley and Fink 1987;Stevenson et al. 1994]. Fink and Manley [1987] and Manley and Fink and Manley [1987] classified the vertical structural variation of a rhyolite lava in terms of the vesiculation and cooling processes into finely vesicular pumice (FVP), dense obsidian (OBS), diapiric coarsely vesicu-!"#$%& lar pumice (CVP), and crystalline rhyolite (RHY) from upper to center. Formation processes of surface ogives [e.g. Andrews et al. 2021;Fink 1980] and spherulites in obsidian lithofacies [e.g. Furukawa et al. 2019;Watkins et al. 2008] of rhyolite lavas have also been discussed. Potential hazards such as surface explosions and flow front collapses have also been pointed out from ancient rhyolite lavas [Castro et al. 2002a;Fink and Manley 1987]. Because such hazards are potentially caused by the overpressure of inner volcanic gasses, the degassing and outgassing systems of rhyolite lavas have been examined [Furukawa et al. 2010;Ryan et al. 2019;Shields et al. 2016]. The flow dynamics of rhyolite lavas have been deduced by researchers based on textural analyses, magnetic experiments, and numerical studies [Bullock et al. 2018;Castro et al. 2002a;Furukawa and Uno 2015;Magnall et al. 2017].
We examined Shiroyama rhyolite lava [Itoh 1989] on Himeshima Island, SW Japan ( Figure 1). The lava is characterized by the presence of obsidian, whose exposure is restricted in the Kannonzaki area ( Figure 2A). An accurate feeder of Shiroyama lava was not previously identified, but our geological investigation revealed that the obsidian lithofacies correspond to the shallow conduit. This means that Shiroyama lava is observable from the conduit to its distal extent. Because of those good exposures, the rhyolite lava emplacement can be revealed via studies of Shiroyama lava. Here, we examined the geological characteristics, densities, water contents, and magnetic characteristics of Shiroyama lava and discussed the conduit system, eruption, degassing history during the lava flows, and flow dynamics.
Obsidian in Shiroyama lava will make a significant contribution to volcanology as well as archeology. Stone tools made from the obsidian in the prehistoric period have been widely excavated in southwestern Japan. Due to its high natural and cultural value, the Kannonzaki area (Figure 2A), which includes the obsidian site, has been designated as a national natural monument. It is used for education and tourism as a geopark site. Owing to this designation, the rocks, animals, and plants are strictly protected in the Kannonzaki area. Thus, we conducted a non-destructive survey there.

Geological setting
Himeshima Island is located in the north of the Kyushu area in SW Japan (Figure 1). The island is mainly composed of seven Pleistocene volcanoes called the Himeshima volcanic group [Kasama and Huzita 1955], characterized by silicic magmatism of 65-75 wt.% SiO 2 [Itoh 1990]. The magmas were generated by the subduction of the Philippine Sea Plate. The Himeshima volcanic group is underlain by lower to middle Pleistocene sediments, comprising the Maruishibana formation, Kawashiri gravel bed, and Karato formation in chronological order [Itoh et al. 1997;Kasama and Huzita 1955]. The sediments characteristically contain rounded andesite clasts, which are considered to be derived from the adjacent Early Pleistocene Futago volcanic group via a fluvial system [Itoh 1989].

Shiroyama lava
Shiroyama lava is part of the Shiroyama volcano ( Figure 2), which is located in the western part of Himeshima Island. The lava is characterized by the presence of obsidian. Several ages for the lava have been reported using various methods. Kaneoka and Suzuki [1970] reported lava ages 0.320 .05 Ma and 0.34˘0.05 Ma based on K-Ar and fission track datings, respectively. Kamata [1988] reported an age of 0.2˘0.1 Ma based on K-Ar dating. Matsumoto et al. [2010] reported an age of 0.104˘0.007 Ma based on Ar-Ar dating. The ages correspond to the Chibanian age (middle Pleistocene) or the Upper Pleistocene. An accurate feeder of Shiroyama lava was not previ- ously identified. The length of Shiroyama lava is approximately 480 m from north to south. The bottom of the lava is not exposed, and the maximum thickness of the exposed part is 52 m [Itoh 1989]. The lava is broadly overlain by Shiroyama pyroclastic rocks (Figure 2A). The bulk chemical composition of the lava is 74-75 wt.% SiO 2 [Itoh 1990]. The lava contains garnet phenocrysts approximately 0.2 mm in size, and the matrix is composed of glass (>95 vol. %) and groundmass of plagioclase and biotite [Itoh et al. 1997]. Itoh [1989] recognized the Kannonzaki and Shiroyama craters within the distribution area of Shiroyama lava ( Figure 2A) and proposed that both craters were formed after the emplacement of Shiroyama lava. The activities of the Shiroyama crater commenced with the destruction of the pre-existing Shiroyama lava, and a pyroclastic cone composed of Shiroyama pyroclastic rocks were eventually formed on Shiroyama lava [Itoh 1989]. The Shiroyama crater was subsequently filled by lake deposits. Itoh [1989] also considered that the Kannonzaki crater was formed by the destruction of Shiroyama lava. In contrast to the sequence proposed by Itoh [1989], our results showed that the effusion of Shiroyama lava was preceded by the activity of the Kannonzaki crater. The crater was eventually filled by the Kannonzaki pyroclastic rocks, which were supplied from the crater. The distribution of the Kannonzaki pyroclastic rocks is restricted within the crater, unlike that of the broadly scattered Shiroyama pyroclastic rocks (Figure 2A). Although a pyroclastic cone does not currently exist, Itoh [1989] considered that it had been formed on the crater.

Geological description
Shiroyama lava is mainly divided into obsidian and vesicular lithofacies (Figure 2A). While most of the interior part of the lava is massive, the marginal part tends to be brecciated. We evaluated Shiroyama lava and the Kannonzaki pyroclastic rocks from three areas: the Kannonzaki, Quarry, and Teishi areas ( Figure 2A).

Page 109
Conduit system, degassing, and flow dynamics of a rhyolite lava

Kannonzaki Area
The Kannonzaki area is composed of Shiroyama lava and the Kannonzaki pyroclastic rocks. The obsidian lithofacies of Shiroyama lava is distributed in a narrow area of the tip of Cape Kannonzaki (Figure 2A). Distribution of the Kannonzaki pyroclastic rocks is restricted in the bay of Cape Kannonzaki. The shape of the Kannonzaki crater roughly follows the outline of the bay (Figure 2A and 2B).

Kannonzaki pyroclastic rocks
The Kannonzaki pyroclastic rocks are exclusively distributed in the bay, which is approximately 100 m in diameter [Itoh 1989] ( Figure 2A). Most of the rock surfaces are usually below sea level, and they are accessible only at low tide. The Kannonzaki pyroclastic rocks are composed of alternate beds of tuff, lapilli tuff, and tuff breccia. Clasts of the pyroclastic rocks are composed of obsidian lava, perlitic lava, pyroclastic rocks including perlitic clasts, and rounded andesite (Figure 3A and 3B). While the clasts of the obsidian and perlitic lavas and pyroclastic rocks are juvenile materials, rounded andesite clasts are reported to be derived Presses universitaires de �rasbourg from the underlying Pleistocene sediments [Itoh 1989]. Strikes of the bedding planes of the pyroclastic rocks are concentrically arranged along the margin of the bay, and they dip 10-30°toward the center of the bay [Itoh 1989]. Matrix ash often shows a reddish color. Based on the occurrences, Itoh interpreted that the bay is a crater filled by pyroclastic rocks. There is a rotated block of the Kannonzaki pyroclastic rocks in the north margin of the bay, showing a bedding plane strike of N80°W and 70-90°southward dip-ping ( Figure 3C and 3D). The block has an approximate width of 20 m, length of 10 m, and height of 5 m. The pyroclastic rocks near contact with the obsidian lithofacies are strongly deformed by faults and folds (Figure 3E). The bomb sag structures ( Figure 3F) observed in the block show that the north side of the block was originally the base. Since the block was unlikely rotated over 90°towards the north, it is reasonable to consider that the block was rotated 70-90°towards the south.

Obsidian lithofacies
The obsidian lithofacies of Shiroyama lava are mainly distributed in the north and west sides of the Kannonzaki crater (Figure 2A). The obsidian lithofacies grade gradually into the vesicular lithofacies. The obsidian is characterized by well-developed flow banding defined by the difference in vesicularity ( Figure 4A). In the outcrops, the part with high vesicularity shows a relatively light color ( Figure 4B and 4C). The obsidian lithofacies are classified into three zones: brecciated obsidian, sheared brecciated obsidian, and massive to brecciated obsidian zones from south to north ( Figure 5A). They are distributed roughly in the east-west strike direction.
The brecciated obsidian zone is distributed along the adjacent rotated block of the Kannonzaki pyroclastic rocks ( Figure 5B and Figure 6A). Strike of the zone is slightly undulated, and the width is varied between mostly 20 and 80 cm. The boundary with the rotated block of the Kannonzaki pyroclastic rocks is relatively sharp ( Figure 6A). While most clasts are obsidian and chiefly 5-10 cm in diameter, the reddish clasts of the Kannonzaki pyroclastic rocks can be easily recognized ( Figure 6B). The long axis of the clasts tends to be aligned in a nearly vertical orientation ( Figure 6B). The matrix is mostly reddish in color.
The sheared brecciated obsidian zone, approximately 13 m in width, is distributed between the southern brecciated obsidian and northern massive to brecciated obsidian zones ( Figure 5A and 5B). The obsidian lithofacies distributed west of the Kannonzaki crater is entirely composed of sheared brecciated obsidian. The clasts are entirely composed of obsidian and are vertically elongated and flattened, exhibiting a platy shape (Figure 7). The matrix is composed of fine-sized obsidian, and the obsidian particles exhibit ductile deformation, indicating comminution above the glass transition temperature [Ishikawa and Kamada 2009]. The matrix of the sheared brecciated obsidian near the side of the brecciated obsidian zone tends to show a red color (Figure 7). Foliation developed in the sheared brecciated obsidian zone due to the alignment of the platy obsidian clasts ( Figure 5B). On the flattened surface of the platy obsidian clasts, parallel striations with a nearly vertical orientation are frequently recognized ( Figure 8A). These observations indicate that the sheared brecciated obsidian records both of folia-Presses universitaires de �rasbourg Page 111 Conduit system, degassing, and flow dynamics of a rhyolite lava tion and lineation. Since the motion of obsidian may be recorded in the structure, we measured the attitudes of foliation and lineation using a compass clinometer from the sheared brecciated obsidian distributed north and west of the Kannonzaki crater ( Figure 8B), obtaining 10 data points from both sides. The results show that both foliation and lineation indicate steep southward dipping. Furthermore, the attitudes are slightly different between both sides of obsidians. Lineations of the west and north obsidians dip to the east and west, respectively.
The massive to brecciated obsidian zone, approximately 22 m in width, occurred in the farthest area from the Kannonzaki pyroclastic rocks ( Figure 5A). A large part of the zone is composed of massive obsidian ( Figure 9A), while brecciated obsidian is partly recognized ( Figure 9B). In the brecciated part, the clast alignment does not show a preferred orientation, unlike other obsidian zones. This zone characteristically contains rounded andesite clasts ( Figure 9C). The oc-currences of the clasts resemble those of the andesite clasts in the Kannonzaki pyroclastic rocks. This indicates that the andesite clasts were derived from the underlying Pleistocene sediments. In the massive obsidian part, a patchy pumiceous structure is often developed ( Figure 9D). Within the pumiceous parts, the interiors are more vesicular than the exteriors. Because the outlines of the pumices are always unclear, the pumiceous part appears to gradually change into obsidian. Although most of this zone is not fragmented, the fracture network is frequently developed ( Figure 10A). The fractures are mostly <5 cm in width. The fractures, called tuffisites, are filled with sintered or welded glassy fragments Saubin et al. 2016;Tuffen and Dingwell 2004]. The observations indicate that fracturing occurred above the glass transition temperature [e.g. Tuffen and Dingwell 2004]. The black and light gray colors of the obsidian are recognized around the fracture network. The obsidian along the fracture is black, while the obsidian away from the fracture is light gray in color ( Figure 10A in the color attributes to the vesicularity; the bubblefree part is black. The fractures are often obscured and disappear, while the two-colored obsidian remains preserved ( Figure 10B). In some cases, the two-colored obsidians are gradually flattened by ductile deformation, and flow banding is developed ( Figure 10C). The pervasively developed flow banding of the obsidian is considered to have been formed by this process ( Figure 4A).

Vesicular lithofacies
The vesicular lithofacies of the lava are gradually developed from obsidian lithofacies and override on the rotated block of the Kannonzaki pyroclastic rocks . The marginal part of the vesicular lithofacies is mostly brecciated. In the brecciated clasts, polyhedral and radial joints are frequently recognized ( Figure 11B and 11C). Although the matrix is mainly composed of juvenile materials, reddish mud deposits, showing soft sediment deformation, occasionally occur in the matrix ( Figure 11D). The mud deposits are interpreted as peperites. The polyhedral and radial joints of the clasts and peperites are interpreted to be developed by interaction with water and wet sediments [Cas and Wright 1987;Kano 1991;McLean et al. 2016;Yamagishi and Goto 1992].
In vesicular lithofacies around the Kannonzaki area,

Presses universitaires de �rasbourg
Page 113 Conduit system, degassing, and flow dynamics of a rhyolite lava Furukawa et al., 2021 ig. 9 Furukawa et al. the perlite is widely developed, being generally 1-4 cm in diameter, resembling the macro perlite of Yamagishi and Goto [1992] ( Figure 11E). Perlitic texture has been considered to result from the ingress of external water into fractures of a glassy lava [Denton et al. 2012;von Aulock et al. 2013]. Clastic dikes, approximately 1 m in width, are also recognized in this area. The dikes are usually clast-supported and filled by mostly glassy lava clasts with perlitic texture ( Figure 11F). The clasts of- ten show ductile deformation and adhere to each other, indicating entrainment above the glass transition temperature ( Figure 11G). Itoh [1989] interpreted dikes as spiracles. Spiracles are inferred to have been created when flowing lava crossed wet sediments, and trapped water within the sediments is explosively converted to steam [Taniguchi 1982;Tolan et al. 2009]. The monolithologic lava clasts and adhered clasts of the clastic dike are consistent with the formation during lava flow.

Presses universitaires de �rasbourg
In the bay of the Cape Kannonzaki, both the vesicular lithofacies of Shiroyama lava and the Kannonzaki pyroclastic rocks are observed ( Figure 12A). Their boundaries are highly deformed. The Kannonzaki pyroclastic rocks are deformed along the lava margin ( Figure 12B) and occasionally infiltrate into the interspace of the lava. The Kannonzaki pyroclastic rocks along the lava margin often show a reddish color ( Figure 12C), and the infiltrated pyroclastic rocks tend to show darker red ( Figure 12D).

Quarry area
A quarry is established in the inner area of the lava (Figure 2B) and exposes a north-south section of the lava. Massive vesicular lithofacies are entirely distributed in the quarry. The maximum thickness of the quarry is approximately 45 m. Obsidian blocks, mostly <10 cm in diameter, are often recognized from there. They would have been derived from the obsidian lithofacies of the Kannonzaki area as entrained clasts and/or ballistic ejecta. Although the structural development of massive obsidian layers below a surficial pumiceous layer has been reported from rhyolite lavas [Furukawa et al. 2019;Furukawa et al. 2010;Manley and Fink 1987], obsidian layers have not been recognized in Shiroyama lava. Perlitic texture, spiracles, and polyhedral and radial fractures, which imply interaction with water, were not identified from the quarry area, unlike the Kannonzaki area. Flow banding was extensively developed ( Figure 13A). The black color band is defined by a low vesicularity part similar to that of obsidian. The lava margin, covered by the Shiroyama pyroclastic rocks, is exposed in the southern part of the quarry. The lava margin is characterized by a layering structure formed by the platy joints ( Figure 13B). The joints show buckle folding with a wavelength of 1-2 m ( Figure 13B). Buckle folds are common in rhyolite lavas [Castro and Cashman 1999] and formed via layerparallel compression. This lava margin is partly brecciated, and en echelon cracks, indicating shear stress, are frequently recognized within the platy joints (Figure 13C). These occurrences show that the compressive and shear stresses continued after emplacement and likely resulted from the pressure from the inner moving lava. Vesicularity varies in the quarry and tends to be higher on the north side. We impregnated thin sections with yellow resin to observe the detailed bubble morphology. In the northernmost part of the quarry, the vesicularity, which is estimated from the area ratios of the photomicrographs, of the lava is 70-80 %. The shape of the bubbles is relatively spherical (Figure 13D). Structures indicating bubble coalescence, such as bending and dimpling [Castro et al. 2012b;Martel and Iacono-Marziano 2015], were observed. In the central part of the quarry, the vesicularity of the lava is 45-60 %. The bubbles are highly flattened, indicating significant flow-induced shear ( Figure 13E). The streak texture resulting from bubble collapse is frequently observed ( Figure 13F). The texture of the southern margin of the lava resembles that of the Teishi area, as described in the following section.

Teishi area
In the Teishi area, an entirely brecciated lava margin of the vesicular lithofacies is exposed. The lava is directly covered by the Shiroyama pyroclastic rocks. Perlitic texture, spiracles, and polyhedral and radial fractures were not recognized from the Teishi area. Dense flow bands are usually developed in the lava ( Figure 14A). The black band in the hand sample is clear under a microscope ( Figure 14B), is defined by the low vesicularity part similar to those of the obsidian and vesicular litho-facies of the quarry ( Figure 14A and Figure 13A). The vesicularity of the lava is 15-25%, which is measurably lower than those of the vesicular lithofacies in other areas.

Magnetic susceptibility measurement
In the brecciated obsidian zone distributed in the southern margin of the obsidian lithofacies, clasts from Presses universitaires de �rasbourg the adjacent Kannonzaki pyroclastic rocks are incorporated. We measured the magnetic susceptibility to confirm the mixing of the Kannonzaki pyroclastic rocks into obsidian lithofacies.
Magnetic susceptibility was determined in the field using a Tanaka geological corporation WSL-C portable magnetic susceptibility meter and investigated at three sites: matrices of the sheared brecciated obsidian, brecciated obsidian zones, and adjacent Kannonzaki pyroclastic rocks. Measurements were taken 15-20 times at each site. The error in the measurements is less than 2 %.

Anisotropy of magnetic susceptibility (AMS) measurement
The AMS of a lava sample reflects the alignment of magnetic minerals and can record the flow direction and internal strain of a lava and pyroclastic flow deposit [e.g. Cañón-Tapia and Mendoza-Borunda 2014; Furukawa and Uno 2015;Shields et al. 2016]. Therefore, AMS has contributed to research in the field of volcanology. In this study, AMS measurements were conducted to reveal the detailed deformation history of Shiroyama lava. The AMS samples were collected from 5 sites (sites 2, 5, 6, 8, and 9; Figure 15A). All samples were hand samples. The samples were oriented in the field using a tripod-mounted magnetic compass. The specimens were drilled in the laboratory, yielding cores 25 mm in diameter and 22 mm in length. In total, 11-36 oriented specimens were analyzed from each sampling site.
We measured the AMS of the samples using a susceptibility bridge (KLY-3, AGICO, Inc.) at Okayama University to evaluate the bulk magnetic fabric. AMS can be represented by an ellipsoid with three principal axes denoted as k max > k int > k min . The results were projected onto an equal-area projection and analyzed using Jelínek statistics. The shape parameter T and the degree of anisotropy P were calculated using Jelínek's method.

Density measurement
The apparent densities of the obsidian and vesicular samples were measured using the glass-beads method [Sasaki and Katsui 1981]. We can analyze the apparent densities easily, quickly, and non-destructive by the method. Measurements were taken from 9 sites, with 3-16 samples each ( Figure 15A). Site 1 is the obsidian, and sites 2-9 are the vesicular lithofacies. Obsidian samples were collected from the quarry. Because obsidian is considered to be derived from obsidian lithofacies in the Kannonzaki area, we selected this area as a sampling site (site 1). All samples were washed with water and dried at 60°C for three days. Each sample was placed in a graduated cylinder filled with glass beads. The using glass beads are 0.8 mm in diameter, which is slightly larger than that of vesicles. In a size lower than that, glass beads infill the vesicles, and the results will be inaccurate. The increased volume corresponded to the bulk volume of the sample. The apparent density of each sample was estimated from its volume and weight.

Water concentration measurement
Water concentrations were estimated from obsidian (site 1; Figure 15A) and vesicular lithofacies (sites 2, 6, and 9; Figure 15A) samples to help elucidate the degassing process. In the obsidian sample, a black streak cut across the sample was recognized ( Figure 16A). The black streak is composed of a bubble-free part probably developed along the healed fracture as well as the occurrence of Figure 10B. We measured the water concentration from both the massive and the black healedfracture parts of the obsidian sample.
The water content was measured by Fourier transform infrared (FTIR) spectroscopy. The analyses were  conducted at Kwansei Gakuin University using a Shimadzu IRPrestige-21 spectrometer with an AIM-8800 infrared microscope. Infrared spectra were collected using a Ge-coated KBr beamsplitter. The results were obtained over a wavelength range of 4000-400 cm -1 , with a liquid-nitrogen cooled MCT detector and an aperture set at 15 µm square. Samples were prepared as doubly polished wafers with a thickness of 43-150 µm.
The concentration of total dissolved water (H 2 O t ) and water dissolved as molecular water (H 2 O m ) was determined from the absorbance bands at 3550 and 1630 cm -1 , respectively [Wysoczanski and Tani 2006] using the Beer-Lambert law: where M is the molecular weight (18.02 g mol´1 for water), A is the height of the spectral band of interest, ρ is the density (g L´1), d is the sample thickness (cm), and ε is the molar absorptivity coefficient for the ab-sorbance band of interest (L mol´1 cm´1). Absorbance peak heights were determined using a straight baseline.  [Mitchell et al. 2018]. Giachetti et al. [2015] suggested that samples with high connected porosity and ages of approximately 1000 years may contain a significant amount of Presses universitaires de �rasbourg Page 119 Conduit system, degassing, and flow dynamics of a rhyolite lava water. Because Shiroyama lava meets these conditions, it may have been subject to rehydration. Therefore, a comparison of H 2 O t or H 2 O m to discuss the degassing process may not be suitable. The H 2 O m /OH ratio varies with magma ascent velocity (cooling rate) [Castro et al. 2012a]. Because all samples were collected from the same lava, the H 2 O m /OH ratios were expected to be initially equal. Therefore, we compared OH concentration to discuss the qualitative degassing process because it is not controlled by the rehydration process, unlike H 2 O m concentration.

Magnetic susceptibility measurement
The results of magnetic susceptibility varied between 1.9ˆ10´5 and 99.5ˆ10´5 SI (Figure 17). The values of the sheared brecciated obsidian, brecciated obsidian zones, and adjacent Kannonzaki pyroclastic rocks were 3.1˘0.7ˆ10´5 SI, 17.2˘7.4ˆ10´5 SI, and 71.51 4.8ˆ10´5 SI, respectively. The values of each lithology were statistically different from each other within error limit, and it was possible to distinguish them from each other by the magnetic susceptibility value. The magnetic susceptibility of the brecciated obsidian zone shows intermediate values. The results of magnetic susceptibility together with geological observation indicate that the Kannonzaki pyroclastic rocks are somewhat entrained into the obsidian lithofacies at the matrix and clast scales.

Anisotropy of magnetic susceptibility (AMS) measurement
The results of the AMS measurement are shown in Table A1. The T and P of Jelínek varied between´0.955 and 1.150 and between 1.012 and 1.299, respectively. The two groups were recognized based on the T -P diagram ( Figure 18A). For the group composed of sites 2, 8, and 9, the T value indicates that the shapes of AMS ellipsoids are broadly distributed between oblate and prolate. For the group composed of sites 5 and 6, the T value suggests that the shapes of AMS ellipsoids are conspicuously oblate, and the P 1 value is distinctly higher than that of the other groups. The higher P 1 values in the samples with highly elongated bubbles are

ble A1
Presses universitaires de �rasbourg Page 120 Volcanica 4(2): 107 -134. doi: 1 .3 9 9/vol. 4. 2.1 7134 consistent with the results obtained by Shields et al. [2016]. The orientation of the principal susceptibility axes of all specimens is shown in Figure 18B-F. The group composed of sites 2, 8, and 9 indicate that the principal susceptibility axes have a broad distribution (Figure 18B, E, and F). At these sites, the confidence limits of the axes intersect. These characteristics of the principal susceptibility axes are likely relevant to the low degree of anisotropy (low P 1 value). Conversely, the principal susceptibility axes of a group of sites 5 and 6 tend to be clustered ( Figure 18C and D). The plane defined by the k max -k int axes dips eastward, and the k max axis indicates the EW direction. The flow-foliation planes of sites 5 and 6 can be recognized, although it is difficult to measure the attitude of the flow-foliation planes of sites 2, 8, and 9 owing to their obscureness and complicated folding. The foliation planes of sites 5 and 6 are also shown in Figure 18C and D. The results show that the plane defined by the k max -k int axes (the AMS foliation plane) coincides with their flow-foliation planes.

Density measurement
The apparent density results varied between 0.80 and 2.59 g cm´3 ( Figure 15B). The obsidian samples (site 1) showed a higher apparent density relative to those of the vesicular lithofacies samples (sites 2-9). The apparent density of site 2, which is in the northernmost part of the quarry, was lowest, as can be expected from samples with the highest vesicularity. The results of sites 2-9 indicate that apparent density gradually increases toward the south.

Water concentration measurement
The results are shown in Table A2 and Figure 15C. The total H 2 O concentration of obsidian was approximately 0.8 wt.%, which is slightly higher than that of typical of obsidian lava [0.1-0.4 wt.% Shields et al. 2016]. The OH concentrations of obsidian (site 1) and vesicular lithofacies (sites 2, 6, and 9) samples varied between 0.38 and 0.54 wt.% and between 0.07 and 0.12 wt.%, respectively. In the vesicular lithofacies of the three sites, OH concentrations were nearly constant. The discrepancy of the OH concentrations between the obsidian and vesicular lithofacies of 0.3-0.4 wt.% indicates the removal of volatiles from the lava. This result shows that obsidian lithofacies develop into vesicular lithofacies with degassing. In the obsidian sample, the OH concentration of the healed-fracture part tended to be broad and <0.16 wt.% lower than that of the massive part ( Figure 16B).

Source of Shiroyama lava
Shiroyama lava is roughly enclosed by brecciated lithofacies, while the inner part of the lava is massive. Because brecciated lithofacies are usually developed in the flow margin, the original distribution of Shiroyama lava is considered to have not been modified by significant erosion.
The distribution of obsidian lithofacies is restricted to a narrow area of the tip of Cape Kannonzaki (Figure 2A), while the vesicular lithofacies comprises a large part of the lava. Rhyolite lavas are usually effused as dense obsidian and vesiculate as they flow away from the vent [Castro et al. 2005;Fink et al. 1992;Furukawa et al. 2019;Ramsey and Fink 1999]. We showed that the sheared brecciated obsidian records both of foliation and lineation indicating steep southward dipping ( Figure 8B). Since the attitudes reflect the motion of the lithofacies, the obsidian is considered to have been transported nearly vertically from below the Kannonzaki crater. Thus, the obsidian lithofacies of the Kannonzaki area were determined to be the source of Shiroyama lava. The lava spreads southward with vesiculation. The subsequent vesiculation is consistent with the results of density measurements and FTIR analyses. The spreading resulted in a fan-shaped flow field. The fan-shaped spreading is a usual method for rhyolite lavas, such as the Big Glass Mountains in California [Fink 1980], Takanoobane rhyolite lava in Japan [Furukawa and Uno 2015], and the obsidian flow of the Cordón Caulle volcano in Chile ].

Emplacement environment of Shiroyama lava
Shiroyama lava involves structures indicating interaction with water and wet sediments such as polyhedral and radial joints, peperite, perlitic texture, and spiracles. Specifically, spiracles demonstrate that only the base of the lava was soaked in water and in contact with wet sediment [Taniguchi 1982]. The Kannonzaki pyroclastic rocks exhibit a reddish matrix, indicating hightemperature oxidation in a subaerial environment. This reveals that the lava was effused at sea with shallow water, likely at a water depth of under several meters. However, the lack of structures indicating interaction with water, except for the Kannonzaki area, means that the wet environment was restricted in the source area. From this, it can be concluded that the lava was effused around the coastline and flowed southward on the land surface.

Emplacement process of the obsidian lithofacies
We demonstrated that obsidian lithofacies developed in the source area, which is the tip of Cape Kannonzaki. The obsidian lithofacies were classified into brecciated, sheared brecciated, and massive to brecciated zones from south to north ( Figure 5). The degree of breccia-Presses universitaires de �rasbourg Page 121 Conduit system, degassing, and flow dynamics of a rhyolite lava tion tends to increase toward the south, and the south margin is in contact with the rotated block of the outer Kannonzaki pyroclastic rocks, at which the outer Kannonzaki pyroclastic rocks are intermingled in the brecciated zone. In the brecciated and sheared brecciated obsidian zones, the vertical foliation and lineation are conspicuous.
We consider that the obsidian lithofacies in this study correspond to the shallow conduit. The silicic conduit structures have been presented via investigation of spines and dissected conduits. The horizontal sections are well investigated in Mount St. Helens [Gaunt et al. 2014;Pallister et al. 2012] and Chaos Crags [Ryan et al. 2020], USA. The studies show that the silicic conduit structures are composed of the fault gouge zone, cataclastic and sheared zone, and massive zone from outer edge to the center. The marginal fault zone has also been reported from spines of the 1991 to 1995 eruption of Unzen volcano Lamb et al. 2015;Nakada et al. 1999] and the 1996 eruption of Soufrière Hills [Sparks et al. 2000]. Since the conduit margin is damaged due to strain localization during magma ascent [Gaunt et al. 2014], the degree of brecciation tends to increase toward the conduit margin. The horizontal structural variation is very similar to that of the obsidian lithofacies of Shiroyama lava, and the brecciated lithofacies in southern part should be the conduit margin. Parallel striations with nearly vertical orientation are interpreted as slickensides formed by faulting during magma ascent such as Pallister et al. [2012], Simões et al. [2018], and Walker et al. [2017]. This is supported by intermingling of the wall rocks in the brecciated zone. Pallister et al. [2012] also suggest entrainment of the wall rocks in the conduit margin based on chemical compositions. These observations strongly suggest that the obsidian lithofacies of Shiroyama lava correspond to a dissected-shallow conduit or a spine inherited the structure from the conduit. In either case, the obsidian lithofacies of Shiroyama lava must preserve the structure of the shallow conduit.
The foliation and lineation of both sides of the obsidians appear to have been transported nearly vertically from below the Kannonzaki crater and away from each other ( Figure 8B), indicating that they underwent petallike expansion during extrusion. In the inner massive obsidian zone, the concentration of the larger andesite xenoliths is more observable than in the marginal brecciated obsidian zone ( Figure 9C). This suggests that the xenoliths were segregated into the interior of the conduit due to the frictional-strain gradient during the magma ascent.

Outgassing and structural development of the obsidian lithofacies
In the massive obsidian zone, the patchy pumiceous structure is characteristically developed ( Figure 9D). Its vague outlines demonstrate that the pumiceous parts do not originate from the entrained clasts of the wall rocks but in situ vesiculation of the obsidian, which would have been induced by decompression associated with magma ascent. Fracture networks developed in the obsidian lithofacies are mostly healed, and the healed fractures are subsequently deformed in a ductile manner into flow banding (Figure 10). In the sheared brecciated obsidian zone, the obsidian is brecciated, and the clasts are subsequently elongated. These occurrences are records of the repetition of brittle and ductile deformation of the obsidian and illustrate that their brittle deformation occurred above the glass transition temperature. The brittle and ductile behavior of a high-viscosity melt such as rhyolitic composition are governed by temperature as well as strain rate [Dingwell 1996;Tuffen and Dingwell 2004]. The obsidian is fractured during the glass transition by a high strain rate pulse. After the strain rate decreased, the fractures become welded. The repetition of melt fracturing and ductile deformation has been described from a shallow conduit and lava [e.g. Furukawa et al. 2019;Tuffen and Dingwell 2004]. Similarly, the obsidian of Shiroyama lava would be fractured by transient increase of the flow-induced shear stress within the conduit, resulting in the formation of the fracture network. The diffusive loss of water toward the fractures would have occurred, as suggested by Yoshimura and Nakamura [2010] and Cabrera et al. [2015], and the volcanic gases were removed, accompanied by pyroclastic materials, through the fractures [e.g. Okumura et al. 2010;Tuffen et al. 2003]. Transient pyroclastic channels called tuffisites have been reported [Castro et al. 2012a;Furukawa et al. 2019;Kendrick et al. 2016;Stasiuk et al. 1996;Tuffen et al. 2003]. When the fractures cut the volatile-filled pumiceous parts in the obsidian, outgassing from the bubbles also progresses. The gases are transported by permeable flow and are removed through wall rocks [Jaupart and Allègre 1991;Stasiuk et al. 1996]. Because the removal of hydrogen gas would promote oxidation accompanied by hematite formation [Furukawa et al. 2010], the reddish color, indicating hematite formation, of the matrix of the marginal brecciated and sheared brecciated obsidian zones would be caused by preferential outgassing.
We obtained the results of the lower water contents along the healed fracture of the obsidian (Figure 16). They must be the consequence of selective degassing ( Figure 10A and Figure 16A). The formation of thin obsidian layers along the fractures via degassing has been experimentally recognized [Yoshimura and Nakamura 2010]. Fracture healing has been promoted by decreasing the strain rate and increasing temperature by frictional heating [Tuffen and Dingwell 2004]. In the obsidian distant from the fractures, vesiculation was induced due to the higher water content resulting in the obsidian with a light gray color ( Figure 10A).

Presses universitaires de �rasbourg
Subsequently, the conduit obsidian would be deformed in a ductile manner under low strain rate conditions. The deformation caused elongation and folding of the bubble-free black-and vesicular light gray obsidians. Thus, the repetition of brittle and ductile deformations of the obsidian lead to development of the flow banding [Gonnermann and Manga 2003]. In Shiroyama lava, the flow banding defined by the difference in vesiculation is recognized not only from the obsidian lithofacies of the conduit but also the vesicular lithofacies of the lava. Hence, pervasive flow banding is considered to be inherited from the structure developed within the conduit. The formation process of the flow bands is different from that described by Furukawa and Uno [2015], who reported that they were originated from the ductile-brittle tearing of lava.
Although we cannot discuss from our data whether sintering of pyroclastic products of deeper cryptic fragmentation formed a dense magma at the conduit walls [Wadsworth et al. 2020], gas removal via vesiculation, fracturing, and brecciation of the magma in the conduit may have suppressed the explosivity, allowing formation of the dense obsidian.

Outgassing and structural development of the vesicular lithofacies
The density variation shows an abrupt decrease from the conduit obsidian (site 1 of Figure 15A) to the vesicular lava (site 2) and subsequent increase toward the distal part (sites 8 and 9). Because there are likely no chemical differences among the samples, the variation is consistent with vesicularity variation. Thus, vesiculation was promoted just after the effusion of the lava due to the pressure decreasing. The vesicularity of the lava in site 2 shows a high vesicularity of 70-80%. Such high vesicularity promotes bubble coalescence ( Figure 13D) and allows the lava to have high permeability [e.g. Eichelberger et al. 1986;Martel and Iacono-Marziano 2015;Westrich and Eichelberger 1994]. The abrupt decreasing of OH concentration (Figure 15C) strongly indicate that the volatiles of the lava would have been removed effectively via networks of interconnected bubbles as permeable flow [e.g. Eichelberger et al. 1986;Okumura et al. 2009;Rust et al. 2004]. Outgassing allows stress from overburden to collapse the foam [Eichelberger et al. 1986]. A gradual increase in the apparent density with the flow distance indicates the incremental foam collapse ( Figure 15B).

Flow dynamics of Shiroyama lava
The AMS fabric of site 2, which is near the vent area, shows that the principal susceptibility axes have a broad distribution ( Figure 18B). Because the AMS fabric of silicic lava is considered to mirror the crystal preferred orientations [Cañón-Tapia and Castro 2004], the AMS fabric of site 2 means that the lava did not move in one direction at the emplacement. We interpret that the AMS fabric resulted from the lava inflation occurring just after extrusion. Because lava inflation likely caused the complicated local strain, the crystals in the lava would not be aligned well. The inflation of the lava would have also caused the broadly scattered AMS ellipsoid between oblate and prolate in shape ( Figure 18A). At sites 5 and 6, the AMS foliation planes coincide with their flow-foliation planes ( Figure 18C and 18D). The results indicate that an imbricated microlite alignment against the flow-foliation plane did not develop in the sites. Flow types such as simple or pure shear can be identified from crystal fabrics [e.g. Castro et al. 2002b;Iezzi and Ventura 2002;Manga 1998]. In simple shear, the crystal preferred orientation tends to be oriented at an angle to the flow foliation [Castro et al. 2002b;Manga 1998]. Thus, the coincidence between the AMS foliation plane and the flow-foliation plane shows that the flow types at sites 5 and 6 were dominated by pure shear rather than simple shear. Since it has been experimentally confirmed that the AMS ellipsoids develop prolate shapes within simple shear flows [Arbaret et al. 2013], the oblate shape of the AMS ellipsoid at sites 5 and 6 ( Figure 18A) also supports their dominance of pure shear. The dominance of pure shear in a silicic lava has been recognized by Castro et al. [2002b] and Befus et al. [2015], and they interpreted that the pure shear strain was induced by the gravity collapse of the foams. This process is consistent with the results of microscopic observations and density measurement of Shiroyama lava. The k max axes of the sites indicate nearly the EW direction ( Figure 18C and 18D), which is normal to the expected main flow direction. Because the microlite long axes are likely to be aligned in the extension direction within the pure shear flow [Manga 1998], the lava was stretched perpendicular to the main flow direction. These characteristics are interpreted to be a result of the concentric spreading of the lava [Merle 1998], and the motion would have formed the fan-shaped morphology of the lava. Sites 8 and 9 correspond to the distal part of the lava. The characteristics of the AMS fabrics and ellipsoids are different from those of sites 5 and 6 ( Figure 18A, E, and F). The AMS fabrics of the sites show that the principal susceptibility axes have a broad distribution. The AMS ellipsoids are broadly scattered between oblate and prolate shapes. The alteration of the AMS is likely caused by pressure from the inner moving lava to the immobile distal lava. The systematic alignment of microlites generated within the flowing lava (sites 5 and 6) become disturbed in the distal pressurized lava. The AMS ellipsoids at sites 8 and 9 were partly developed to prolate in shape. The prolate shape may be attributed to the increasing of the simple shear component in the distal part of a lava as suggested by Smith and Houston [1994] and Castro et al. [2002b].

Emplacement process of Shiroyama lava
The volcanic activity of rhyolite magma in the Shiroyama Volcano commenced around the coastline. We consider that a pyroclastic cone, composed of the unconsolidated Kannonzaki pyroclastic rocks, had been formed on the Kannonzaki crater, preceded by the emergence of Shiroyama lava. Deformation of the Kannonzaki pyroclastic rocks at the boundary with Shiroyama lava ( Figure 12B) would have been a consequence of pressure associated with the advancement of Shiroyama lava, and subsequent thermal transfer from the lava must have caused preferential hightemperature oxidation of the marginal and infiltrated parts of the Kannonzaki pyroclastic rocks. Shiroyama lava, overlies the Kannonzaki pyroclastic rocks, and the lack of the Kannonzaki pyroclastic rocks above Shiroyama lava also support the sequence. The proposed volcanic succession is different from that reported by Itoh [1989], who suggested that the activity of the Kannonzaki crater was initiated after the emplacement of Shiroyama lava.
The Kannonzaki pyroclastic rocks include clasts of obsidian lava, perlitic lava, and pyroclastic rocks containing perlitic lava clasts, indicating repeated emergence and destruction of the lava. The activity may resemble the hybrid explosive-effusive eruption, which was directly observed by Schipper et al. [2013] in the vent of the 2011-2012 eruption at Cordón Caulle volcano, Chile. They observed that Vulcanian blasts and effusion of the obsidian lava had occurred synchronously from the common vent system. The eruption system provided abundant ash and lava bombs simultaneously. The Kannonzaki pyroclastic rocks may have been formed by such eruption style. The volcanic activity of the Kannonzaki crater eventually formed a pyroclastic cone, mostly above sea level. The conduit magma subsequently changed the ascent direction to the north of the crater. The shift of the magma ascent direction probably caused concentric depression of the pyroclastic cone. The northern part of the pyroclastic cone was deformed with portions rotated to vertical orientations by the intrusion of the following magma. Removal of the volcanic gases via vesiculation, fracturing, and brecciation from the conduit magma may have played a role in suppressing the explosivity, and the magma could be effused as a dense obsidian lava. The lava flowed southward on wet sediments, and the pyroclastic cone was deformed by pressure from the advancing lava. The contact zone of the pyroclastic rocks with lava was oxidized.
Since the lava originally maintained some extent of water, the lava effervesced just after the effusion. Bubbles at site 2 show a spherical shape, and there are two possible explanations about the formation of the spherical bubbles. First, they were formed in the condition of no significant motion of the lava. Secondary, the elongated bubbles were initially formed by shear deformation of the lava. The lava was, subsequently, stagnated for sufficient time for the bubbles to relax to a spherical shape. In either way, the vesiculation led to pervasive bubble coalescence, resulting in the formation of abundant permeable pathways, and effective gas removal was promoted. As a result of the extensive outgassing, the inflation of the lava shifted to deflation due to the gravity collapse or the interfacial tension-driven relaxation [Otsuki et al. 2015] of the foams. The flow resulted in the concentric spreading of the lava. The bubbles were elongated by flow-induced pure shear. Because the elongated bubbles are comparatively easy to connect [Okumura et al. 2008], additional outgassing occurs via the interconnected bubbles. With the progression of the removal of gases during foam collapse, the density and viscosity of the lava might have increased, and the distal lava eventually lost mobility. The immobile distal lava was deformed during the glass transition in both brittle and ductile manners owing to pressure from the advancing inner lava.

Conclusions
We present a geological description, density variation, FTIR results, and magnetic characteristics of the Pleistocene Shiroyama rhyolite lava on Himeshima Island, Japan. We illustrated the conduit system, degassing and flow dynamics of Shiroyama lava.
The lava was mainly divided into locally developed obsidian and vesicular lithofacies. We interpreted that the obsidian lithofacies corresponded to the shallow conduit based on the structural variation. Removal of the volcanic gases via vesiculation, fracturing, and brecciation from the conduit magma may have played a role in suppressing the explosivity, and the magma could be effused as a dense obsidian lava. Since the lava originally maintained some extent of water, the lava effervesced just after the effusion. The extensive effervescence caused pervasive bubble coalescence, resulting in the formation of abundant permeable pathways. Therefore, the volcanic gasses were removed from the inflated lava. As a result, the lava shifted to deflation due to foam collapse. The flow resulted in the concentric spreading of the lava.    (Km, in SI) and Directions (D = Declination, I = Inclination, in degrees°) of the principal susceptibilities (K max , K int , and K min ) of samples from Shiroyama lava; T : shape parameter (T > oblate, T < prolate); P 1 : degree of anisotropy.