First timeseries record of a large-scale silicic shallow-sea phreatomagmatic eruption

Phreatomagmatic eruptions are one of the most common styles of volcanic eruptions on Earth 1–2 . Recent studies have highlighted the importance of the eruption depth and magma discharge rate on the eruptive behaviour of underwater volcanoes 2–7 . Even though voluminous silicic eruptions in shallow-water environments are likely to be intense and hazardous, such eruptions mostly appear in geological records 7–12 and the nature of this type of eruption is therefore poorly understood. Here, we show the �rst timeseries record of a large-scale silicic phreatomagmatic eruption that occurred at the Fukutoku-Oka-no-Ba volcano, Ogasawara, Japan, on 13 August 2021. The eruption started on the sea�oor at a depth of < 70 m and breached the sea surface, resulting in a 16-km-high, steam-rich sustained eruption column. The total magma volume was ~ 0.1 km 3 , including the subaerial tuff cone and the 300-km 2 pumice raft, most of which can be explained by the effective accumulation of pyroclasts near the vent resulting from interactions between the eruption plume and the ambient water. This eruption provides a rare opportunity to investigate the process of a large-scale phreatomagmatic eruption in a shallow sea and contributes to our understanding of the nature, dynamics, and hazards of submarine volcanism.


Introduction
Phreatomagmatic eruptions, caused by the interaction of magma and external water, are one of the most hazardous types of volcanic eruptions on Earth 1,2 .Such eruptions can signi cantly impact areas around volcanoes by generating high-energy pyroclastic density currents (PDCs), strong pressure waves and tephra-laden jets [13][14][15][16] .The explosivity of phreatomagmatic eruptions increases when a certain amount of external water is incorporated and a higher energy-exchange e ciency from magmatic heat to mechanical energy is achieved [17][18][19][20][21][22] .Therefore, the mixing conditions and the ratio of external water to magma are important to understand the explosivity and eruption style of subaqueous eruptions.The explosivity and eruption style are also related to the water depth of the eruption and the magma discharge rate [2][3][4][5][6][7] .In general, shallow-water environments result in more explosive eruptions than deepwater or dry conditions.Even small-volume eruptions with low magma discharge rates can be explosive, as often observed in Surtseyan-type eruptions [13][14][15][16] .For large-volume eruptions, the explosivity and eruption style may change dramatically if the eruptions occur in a shallow-water environment where external water is e ciently involved [7][8][9][10][11][12] ; conversely, if the eruptions occur in a deep-water environment, the explosivity is suppressed by the high pressure [4][5][6] .However, direct observational data of such largescale phreatomagmatic eruptions are limited.Only the geological record can provide information concerning the possible surface phenomena of these eruptions [7][8][9][10][11][12] , although recent remote-sensing tools, including satellites and the global infrasound monitoring network, have captured some smaller cases [23][24][25][26][27] .The explosive submarine eruption that occurred in Tonga on 15 January 2022 may be one of the examples of large-scale phreatomagmatic eruptions in the shallow sea; however, the physical and chemical processes and related parameters of the eruption are under debate.Accordingly, how such eruptions proceed and their impacts in real space and time are poorly constrained.The 2021 Fukutoku-Oka-no-Ba (FOB) eruption may be the rst to reveal the processes involved with a large-scale eruption of this type.
The FOB volcano is one of the active volcanoes in the Izu-Ogasawara arc (Fig. 1a).On 13-15 August 2021, a large explosive eruption occurred at this volcano (Fig. 1b).Prior to the eruption, the depth of the summit of the volcano was 40-50 m b.s.l.(below sea level) and a slightly deeper ssure (~ 70 m b.s.l.), where the 2021 eruption began, existed on its northern side (Fig. 1c).We analysed the process of this eruption using satellite imagery, aerial photos, infrasound, plume modelling, and geochemistry.The eruption began with a sustained plume, which breached the overlying sea water and reached a height of 16 km a.s.l.(above sea level) in hours 28 .As a result, the eruption reclaimed the shallow sea and produced a tuff cone (Fig. 1c).The eruption also produced a 300-km 2 pumice raft (Fig. 1d), which was dispersed by ocean currents more than 1,000 km west of the volcano.The pumice raft reached the coastal areas of the Paci c Ocean along the Japanese archipelago and caused damage to coastal infrastructure.A chemical analysis of the pumice clasts indicates that the eruption consisted of trachytic-trachydacitic magma with glass compositions reaching 68 wt.% SiO 2 .Therefore, the eruption was a large-scale silicic explosive phreatomagmatic eruption, the rst ever recorded in modern history.

Timeseries data
Himawari-8 satellite and infrasound remote observations suggest that the FOB eruption began at 05:55 JST on 13 August 2021 (Fig. 2a, b).The eruption occurred in four phases.Phase 1 consisted of a continuous plume phase that began at 05:55 JST on 13 August and lasted for ~ 14 h with uctuations until ~ 20:00 JST on 13 August.The period 12:00-19:00 JST was more intense with sustained plumes.
Phase 2 was a 14-hour pulsating phase, characterised by frequent strong infrasound signals.The period 05:30-06:30 JST on 14 August was more intense with a sustained plume.Phase 3 consisted of 24 h of intermittent weak explosions and sparse strong explosions until 09:00 on 15 August.In Phase 4, the activity decayed.Phase 1 was the most powerful and produced the major eruptive products.It generated a continuous, white-coloured vigorous plume directed to the west.At approximately 08:00 JST on 13 August the pumice raft was con rmed to be spreading circularly from the source (Fig. 2c).The development of the pumice raft was observed for the rst 4 h until the source area was covered by the eruption plume in the satellite view.After the cessation of Phases 1-3, the pumice raft drifted westward and was carried by ocean currents.

Plume characteristics of Phase 1
In the periods of 12:00-19:00 JST on 13 August and from 05:30-06:30 JST on 14 August, sustained eruption columns developed (blue bars in Fig. 2b).The most vigorously erupted columns formed in the 1 h following 14:00 JST and in the 20 min from 12:45 JST on 13 August.These eruption columns formed thin laterally spreading clouds with a 15-20-km radius at 16 km a.s.l.(Fig. 2a), which corresponds to the tropopause height (SI).The eruption column was entirely white-coloured, indicating a steam-rich eruption.
The thin laterally spreading clouds (Fig. 1b) were similar to an 'anvil cloud' or 'incus', which is often observed when the upper portion of a strong cumulonimbus spreads out in the shape of an anvil along the tropopause 29 .These clouds differ from those observed in (Sub-)Plinian eruptions in which more vigorous, thicker pyroclast-laden 'umbrella clouds' with a grey/brown colour develop 30 .The FOB eruption column may have contained ne-grained ash; however, its shape, colour, and spreading behaviour along the tropopause did not provide evidence for a large amount of pyroclasts being suspended in the column.
The area covered by the entire eruption cloud reached ~ 10 5 km 2 at approximately 15:00 JST on 13 August.However, there is no report or evidence of ashfall on any ships or boats in the downwind area or on the neighbouring Minami Io To island, 6 km south-southwest of the source.

Thermal anomaly
No thermal anomaly was detected in the near-source region, even when the root of the eruption column was clearly observable from Himawari-8.An aerial observation by the Japan Coast Guard (JCG) with an infrared camera at 15:00-15:30 JST on 13 August captured hot pyroclastic material being ballistically ejected; however, the thermal anomaly was likely too small to be detected by satellites.A reasonable interpretation is that most of the heat issued from the rising magma was consumed in the rapid vaporisation of seawater before it could be detected by thermal monitoring.

Tuff cone and pumice raft formation
The eruption formed a new tuff cone with an ~ 1-km-diameter crater around the vent; however, this cone was rapidly eroded by waves and separated into western and eastern islands.The cone can be seen in a satellite image taken on 14 August.The height of the cone was ~ 15 m at maximum, and the cone components were massive, poorly sorted loose pyroclastic units, suggesting multiple depositional processes.These islands disappeared by early 2022.From 15:00-15:30 JST, 13 August, during the most intense phase, the JCG airplane repeatedly observed laterally spreading PDCs at the source.Therefore, a major component of the new islands was likely formed via near-vent depositional processes such as partial collapses of the eruption column.The volume of the tuff cone is estimated to have been 0.04-0.07km 3 .
A brown-coloured pumice raft began to form at 08:00 JST on 13 August and spread against the direction of the wind and ocean currents.The edge of the pumice raft reached ~ 4 km southeast from the source by 12:00 JST on 13 August (Fig. 2c), suggesting a spreading speed of ~ 1 km/h upstream before being drifted by ocean currents.The area of the pumice raft reached ~ 300 km 2 at 01:00 JST on 15 August.Most of the pumice raft appeared to originate from the vent location.The growth of the pumice raft cannot be explained by the direct deposition of fallout from the eruption plume because the upwind area was never covered by the plume as a result of the strong easterly wind during the observation period.In the downwind direction, the formation of the pumice raft was not observed, indicating that the eruption plume did not contain a large amount of pumice clasts dispersed in the distal direction.Therefore, most of the pumice raft is thought to have been directly generated from the vent during Phase 1.The volume of the pumice raft was estimated to be 0.1-0.3km 3 .Therefore, the sum of the volumes of the tuff cone and pumice raft is 0.1-0.4km 3 (0.04-0.1 km 3 dense rock equivalent, DRE).

Modelling a steam-rich eruption plume
The magma discharge rate required to form the 16-km-high eruption column in Phase 1 was estimated using a one-dimensional eruption plume model 31,32 that includes the effect of the phase change of the external water.The effect of an amount of pumice that should have provided thermal energy to the plume but did not rise in the plume was also considered in this study (SI).The results indicate that a magma discharge rate of 3-6 × 10 5 kg/s is su cient to explain the observed plume height if only a fraction (0.3-3 × 10 5 kg/s) goes into the plume (Extended data Fig. 6).Assuming a nine-hour sustained plume (blue bars in Fig. 2b), the erupted mass is estimated to be 1-2 × 10 10 kg, corresponding to 0.004-0.008km 3 in DRE.Because some fraction of the eruptive material is deposited in the proximal area, the contribution of the magma to the distant fallout tephra is signi cantly reduced compared with these values.

SO 2 emissions
The SO 2 emissions were observed by the TROPOMI instrument installed on the Sentinel-5 Precursor satellite.The mass of SO 2 emitted during the 15 h of activity was 2.1 × 10 7 kg.We also analysed the SO 2 concentrations of the silicic matrix glass of the pumice and the melt inclusions (MIs) of plagioclase, which is a major phenocryst in the products.The degassed SO 2 was estimated to be 73.3 ppm from the difference between the SO 2 concentrations of the matrix glass and the MIs.Using this SO 2 concentration and the observed amount of SO 2 , the mass of the erupted magma is estimated to be 2.9 × 10 11 kg, corresponding to 0.11 km 3 DRE.Therefore, the total erupted volume estimated from the SO 2 balance can be mostly explained by the sum of the geology-based and the model-based tephra volume estimates without assuming any other source.

Discussion And Summary
The eruptive volume of the FOB eruption has large uncertainties; however, a comparison between the volumes estimated from the geology and from the SO 2 emissions indicates that it likely reached ~ 0.1 km 3 DRE.This eruptive scale, the high eruption plume, and the voluminous pumice raft caused by continuous magma discharge differ from the features of well-observed phreatomagmatic explosions, such as Surtseyan eruptions, which are characterised by a series of discrete events with a relatively low magma discharge rate 16,19 .Plumes from Surtseyan eruptions rarely rise to high altitudes (generally less than 10 km) because their thermal ux is generally low 2,19 .Instead, this type corresponds to Phases 2 and 3 of the FOB eruption.The magma discharge rate of Phase 1 is similar to those of (Sub-)Plinian eruptions, which are characterised by sustained explosive eruptions, discharging hot pyroclast and gas mixtures with a tall eruption column and resulting in the widespread dispersion of large amounts of pyroclasts [33][34][35] .However, the style of the FOB eruption also differs from these types because most of the pyroclasts accumulated near the vent and were consumed to form the tuff cone and pumice raft.Infrasound also does not indicate features of sustained explosive eruptions.
The features of Phase 1 of the FOB eruption may be explained by the effective decoupling of coarse pyroclasts from the eruption plume caused by the interaction between the eruption plume and the seawater.In this process, the gas-pyroclast mixture above the submarine vent (50-100 m b.s.l.) penetrates the seawater and atmosphere.First, the mixture ingests the ambient seawater and develops an eruption plume, possibly in a manner similar to deep submarine eruptions [4][5][6] .Then, the plume reaches the sea surface that acts as the boundary between two ambientes with different densities, where it may split into two parts (Fig. 3): (1) a breaching part into the atmosphere caused by the upward migration of the mixture and heating, where the seawater ashes to steam and is rapidly accelerated, and (2) a remaining part in the water that is decelerated by the increasing density resulting from mixing with liquid water.Underwater, lighter parts may form density currents composed of a slurry of hot water and pyroclasts along the sea surface, and denser parts may form submarine density currents along the sea oor 36,37 .After ejection of the material into the atmosphere, there may be a further decoupling of the large poorly fragmented clasts and the wet, cold gas-pyroclast mixture, which may result in partial column collapse and the generation of PDCs.The remaining buoyant parts of the plume rise and form a steam-rich convective plume that carries highly fragmented ne material.A large amount of oating pumice might result from sedimentation from subaerial PDCs and/or directly from the slurry gushing from the submarine vent (Fig. 3).The circular spread of the pumice raft from the source indicates that the slurry discharge was su cient to drive an upstream current.The volume of the pumice raft is estimated to have reached 50-90% of the total erupted volume.This is similar to a silicic deep submarine eruption in which ~ 70% of the eruptive products were dispersed as a pumice raft 4 .
Phase 1 was followed by the Phase 2 and 3 Surtseyan eruptions, re ecting both the upward migration of the vent position and the decrease in the magma discharge rate (Fig. 3).The transition of the eruption style was likely signi cantly affected by the eruption depth and magma discharge rate.Phase 1 demonstrated an eruption style that appears when certain conditions of eruptions (depth of < 100 m, magma discharge rate of < 10 7 kg/s) are met.While a buoyant steam-rich plume was generated, tephra dispersal processes might have been signi cantly affected via the interaction of magma and seawater in the shallow-water environment.It is di cult to categorise this eruption into previously de ned eruption styles such as 'Surtseyan' or 'Plinian' [33][34] .'Phreatoplinian' may potentially describe this type of eruption; however, this eruption was de ned based on its deposit characteristics, such as its extensive dispersion of voluminous ne-grained ash, because of its intense fragmentation and perhaps higher magma discharge rate 8-11 .Conversely, we have no evidence of a large amount of ne-grained ash being generated in the FOB eruption and the proposed near-vent processes may be different.The large-scale silicic phreatomagmatic eruption with a steam-rich sustained eruption column observed in the FOB can be called the 'Ultra-Surtseyan' eruption.
The FOB eruption provides an important opportunity to explore the processes of large-scale silicic submarine eruptions in shallow-water environments.The surface phenomena, eruptive products, and emplacement processes of such eruptions may signi cantly differ from those of deep submarine eruptions and dry explosive eruptions.To examine these problems, we need to comprehensively survey submarine deposits.Such information is essential to constrain the eruption and enhance our knowledge of submarine volcanism.

Declarations
The plume rise height is primarily determined by the heat ux or mass discharge rate from the source because the development of a plume is driven by the conversion of thermal energy to potential energy.Under high humidity conditions in the atmosphere, latent heat is released during the phase change from steam to liquid water and then to ice as the eruption plume rises, causing an increase in plume height.Another essential feature of this eruption was the existence of a large amount of pumice that should have provided thermal energy to the plume but did not rise in the plume.In plume modelling, we applied the user-friendly plume model 'Plumeria' 31,32 , which includes the effect of the phase change of water, to calculate the relation between the eruption parameters and plume height (see Supplementary Information for details).To incorporate the thermal energy from the pumice that did not rise in the plume, we adjusted the input parameters of the software to represent the water-rich high-enthalpy mixture at the vent (Extended Data Fig. 3-6).To achieve a plume height of 16 km, the necessary mass discharge rate was estimated to be 3-6 × 10 5 kg/s.The temperature dependence is small compared with that of the mass discharge rate.

Chemical analysis
Pumice samples from the 2021 FOB eruption were taken by the Japan Meteorological Agency at 25° 30.3 N, 138° 53.3 E on 22 August 2021 during a survey cruise and at Minami Daito on 4 October 2021; samples were also acquired by Minami Daito Village at Minami Daito on 8 October 2021 and by our research group on Okinawa Island on 20 November 2021.We performed microscope observations and conducted whole-rock major element analyses using X-ray uorescence spectrometry (ZSX Primus II, Rigaku Co., Ltd., Tokyo, Japan) and groundmass and mineralogical analyses using an electron probe microanalyser (EPMA, JXA-8800R, JEOL Ltd., Tokyo, Japan) with an acceleration voltage of 15 kV, a beam current of 12 nA, and a beam diameter of 10 μm at the Earthquake Research Institute, University of Tokyo.On the basis of the microscope observations, all the products from the 2021 FOB eruption include phenocrysts of plagioclase, clinopyroxene, Fe-Ti oxides, olivine, and apatite.Plagioclase is the most abundant phenocryst.The whole-rock chemical compositions for the 13 samples of the FOB products are 61.7-64.0wt.% SiO 2 , 1.3-2.5 wt.% MgO, and 9.6-11.1 wt.% Na 2 O + K 2 O and are classi ed as trachyte or trachydacite.The ranges of the chemical compositions of the 2021 products are the same as those of past eruptions.The chemical compositions of the groundmass (GM) glass vary over a linear trend ranging from 56-68 wt.% SiO 2 , re ecting a mixture of ma c magma, even though silicic GM is the main component.The silicic melt inclusions (MIs) in plagioclase have 65-67 wt.% SiO 2 .We also measured SO 3 for the silicic GM (n = 95) and silicic MIs in a plagioclase phenocryst (n = 32) using EPMA.The SO 3 concentrations of the silicic GM and silicic MIs were 334 ppm (1σ = 336 ppm) and 425 ppm (1σ = 178 ppm), respectively.The SO 3 concentrations were recalculated as SO 2 concentrations, and nally the degassed SO 2 was estimated to be 73.3 ppm from the difference between the SO 2 concentrations of the silicic GM and silicic MIs.