**6. Discussions**

#### *6.1. Relationship between Observed Geothermal Activities and Underlying Geothermal System*

Recently, MT inversion analysis showed that an extremely conductive layer of resistivity less than several Ω-m existed beneath 200 to 700 m depth (elevation 1100 m to 600 m) from the surface of the Ioyama [25]. This conductive layer was interpreted as composed of a hydrothermally altered clay-rich layer containing copious amounts of smectite [20]. Kagiyama et al. [60] also reported the presence of a conductive layer of 3- to 30- Ω-m resistivity with a thickness of 120 to 150 m beneath the area

around the Ioyama. Inverse modeling, based on repeatedly operated and precise leveling survey data, revealed an inflation source at a depth of 600 to 700 m, just under the conductive layer [61]. Thus, it is suggested that the conductive layer behaved as a caprock preventing the up-flow of pressurized hydrothermal fluids [25]. This caprock, likely corresponding to a low permeable layer, can explain the activity sequence from Stage-1 to Stage-4. Furthermore, the geological setting identified Ebinokogen Ioyama Volcano can be considered to represent a Type 1 system described by Stix and Moor [11].

Another deeper pressure source, likely related to a magma reservoir, is present between 8 and 10 km below Ebinokogen. It appeared that this source was inflated before the eruption at Shinmoedake in 2011 and was deflated during the early large eruptions (subplinian and lava-accumulation stages) in 2011 [47]. Further, it was re-inflated after the early large eruption stages, and almost stagnated around January 2012 at Shinmoedake [62,63]. Seismicity then increased at the northern side of Karakunidake from December 2013 [48], implying the possibility that magmatic fluids started to be supplied from the deeper parts under Karakunidake to the shallow hydrothermal system under Ebinokogen Ioyama Volcano. The volcanic tremor event that took place on 20 August, 2014, may represent the beginning of the activation of the shallow hydrothermal system under the conductive caprock in Stage-1. As no manifestation was observed at the surface in this stage, it is hypothesized that the fluid-driven heating and pressurization of the shallow hydrothermal system beneath the Ioyama-fumarole area developed gradually (Figure 10). Incidentally, the deep system is the magmatic reservoirs in Kirishima Volcano Group [62,64,65].

The temperatures of fumaroles during Stage-2 were almost constant at around 96 ◦C (Figure 4), the boiling point temperature at an altitude of 1300 m. The relationship between the δD and the δ<sup>18</sup> O ratio showed that the water from the spring K3 at this stage was derived from the meteoric water (Figure 9). Most likely the high-temperature hydrothermal fluids accumulated below a depth of 600 to 700 m in Stage-1 started to percolate through the conductive caprock layer during Stage-2. Thus, fumaroles began to appear at the ground surface (Figure 10).

The successive period was characterized by the appearance of mud pots and jet fumaroles between March and April 2017 in Stage-3. The stage was defined by a rapid expansion of the thermal anomaly area from early 2017, and a temperature rose in the spring K3 in early April (Figure 4). The mud pot A appeared just before/on 19 March, and the mud pot F appeared between 19 and 21 March (Figure 5). The jet fumarole H appeared about 30 m east of the mud pot F between 15 and 18 April, and the jet fumarole A appeared about 10 m northwest of the mud pot A on 26 April. Hydrothermal minerals of the blowout deposit from the jet fumarole A contained mostly quartz and alunite (soda) and almost no smectite (Table 1). The mineral assemblage of the blowout deposit well matches the surficial alteration near Ioyama, characterized by the alunite zone with quartz, cristobalite and tridymite [44]. In other cases, the lower limit of the alunite zone is at a 380 m depth in the Hatchobaru geothermal area in Kuju Volcano [66,67], as well alteration zones containing alunite develop in very shallow parts of Hakone and Ontake volcanoes [68,69]. The XRD analysis for a drilling core at Owakudani in Hakone showed that alunite was distributed up to a depth of 100 m from the surface [23]. A key information for estimating depth of hydrothermal fluids in the Ebino-fumarole area comes from the geothermal drilling carried in the 1950s, in which blowout occurred at two wells while drilling at depth between 18.3 and 18.4 m below the surface [41]. The hot groundwater (aquifer) at a depth of 20 m in the western part of Ioyama with the same topographical gradient is considered to have been at a 50 to 70 m depth below the summit of Ioyama in the 1950s. Currently, the height difference between the spring K3 and the summit is about 50 m. The temperature of the spring K3 at the time of the blowout event was relatively low, 40 ◦C, and it was probable that there was a hot confined aquifer located deeper than 50 m. Alunite is widely stable in a low pH environment, and its deposition or accumulation is known to form altered layers of low-permeability [70,71]. Recently, alunite filling among mineral grains was reported [72]. Therefore, the 2017 blowout event at the jet fumarole A with the products abundant in alunite can be considered as a phenomenon that originated from low pH hydrothermal liquids at a very shallow depth (mostly up to a depth of 100 m). Alunite, quartz, cristobalite, tridymite, kaolinite, smectite and

halloysite were found on the surface around the Ioyama-fumarole area in 1950s [44]. It was probable that the alunite accumulation had formed low permeability layers (or spots) before the year 2008, under which developed pockets of hydrothermal fluids or confined groundwater (Figures 10 and 11).

**Figure 10.** Conceptual sketch showing the different stages of geothermal activity in Ebinokogen Ioyama Volcano. Numerical values in Stage-1 denote Ω-m [25].

The high heat and gas emission released at the jet fumaroles H and A resulted in a decrease in the thermal anomaly area since August 2017. The temperature of the K3 spring rose again in February 2018, with the simultaneous appearance of the fumarole S. In this stage, a continuous uplift of the summit area was observed by the JMA tilt meter from 2018 mid-March. The Global Navigation Satellite System (GNSS) observations showed inflation prior to the 2018 eruption. The geophysical pressure source was estimated to be around 950 m altitude (about 350 m below the ground) [53]. The 2018 eruption deposits contained dominant quartz and alunite (soda) similar to those in the 2017 event deposits, and both probably came from similar hydrothermal environments. The smectite content of the 2018 eruption deposit was about 1 wt%, implying that the April 2018 eruption originated from a smectite depleted hydrothermal environment. The surface altered areas at the Ioyama contained minor amounts of the smectite mineral [44]. Less kaolinite (about 10 wt%) from the April 2018 eruption were derived from the hydrothermal region in an acidic environment. Part of the kaolinite or minor smectite may also have been taken from very shallow altered host rocks. The hydrothermal pond Y1 appeared on 7 April, and a large amount of hydrothermal fluids gushed out. The relationship between δD and δ18O ratio of the water in the Y1 pond just before the eruption indicated that the pond liquid was a mixture of meteoric and magmatic liquids, creating an environment in which both end member liquids probably mixed. Thus, geological and geochemical observations show that the main excavation of the 2018 eruption may have been at a very shallow depth around the top of conductive layer. It is probable that the pressure source of about 350 m, indicated by low-resistivity implies the stagnation of magmatic–hydrothermal water. Seki et al. [73] suggested that a low-resistivity layer in the shallow part was formed due to the percolation of acidic hydrothermal fluid into the clayey lacustrine sediment at Midagahara Volcano in Tateyama. The resistivity >10 Ω-m was observed down to a depth of about 150 m (1150 m altitude) beneath Ioyama [60]. The similar interpretation is also possible here for the aquifer of hydrothermal fluids at depths of between 150 m and 350 m. We defined those systems above the 350 m depth as the very-shallow hydrothermal system (Figure 10). The main zone of acid alteration is the kaolinite zone without dickite which registered temperatures of up to 120 ◦C [13]. The temperature of 130 ◦C at the jet fumarole H in Stage-3 may sugges<sup>t</sup> a leakage from this hydrothermal system (Figure 11).

The geology of Ioyama and the surrounding area consists of thin debris avalanche and thick welded pyroclastic deposits from Karakunidake [29,32]. The borehole survey at 1.1 km west of Ioyama [74] indicated that an upper boundary of the welded pyroclastic deposit was about 50 m depth. This deposit is exposed on the surface near the Ioyama-south craters. The elevation di fference between the two sites is about 170 m. Assuming horizontally flat deposition, the welded pyroclastic deposit is distributed down to about 170 m depth under the Ioyama-south craters, corresponding to the distribution of a moderate resistivity layer down to 200 m depth. This welded pyroclastic deposit is interfingered pumice fall layers [75]. It is considered that pumice layers behave as permeable layers for ground water, where meteoric and magmatic liquids mix.

The conductive caprock layer with a low resistivity of <1 Ω-m at depths from 200 to 700 m was considered to have smectite as its main altered mineral [25]. In the laboratory experiments, higher contents of smectite achieved a lower resistivity [76]. Smectite exceeding 10% of maximum intensity by the XRD analyses was contained in the drilling core from the conductive layer in Hakone [23]. However, the lithological resistivity measured in the laboratory indicated a conductivity of about one order of magnitude higher than those in borehole observation. This di fference was possibly a ffected by the pore fluid conductivity [24]. Additionally, the results of the e ffect of acidity on the conductivity were significant [77]. We will need further discussions about the morphology of the conductive layer.

For magma with a temperature of 870 ◦C such as the 2011 Shinmoedake silicic andesite magma [64] located at a depth of 600 to 700 m, the conductive, smectite-bearing caprock would not be observed around the inflation source because of the instability of smectite at temperatures above 200 ◦C [13]. It is known the high electrical conductivity is enhanced by the higher temperature e.g., [24,76]. Therefore, it was reasonable to assume that the main component of the inflation source was magmatic–hydrothermal

fluids at a temperature slightly higher than 200 ◦C. Thus, we tentatively propose a hydrothermal reservoir system at the inflation source at around 600 to 700 m in depth (Figure 11). A high-temperature fumarole of 247 ◦C had previously developed at the summit of Ioyama in the 1970s [34], where gases probably originated from this depth.

#### *6.2. A Kick Sign before the April 2018 Eruption*

In order to forecast a steam-driven eruption, it is essential to understand how an aquifer of hydrothermal system is depressurized or heated and what triggers an eruption.

During the events of March to April 2017, the jet fumarole A blew out on 26 April, after the appearance of the mud pot A just before/on 19 March. The jet fumarole A blew out about a ten meter away from the mud pot A. Furthermore, welling-up of a large amount of fluids in the hydrothermal pond Y1 continued for 12 days immediately before the steam-driven eruption on 19 April, 2018 at Ioyama south craters (4–19 vents). The locations of the upwelling of hydrothermal fluids and of the eruption were about 60 m apart.

The characteristic feature in both the blowout and eruption is that the upwelling of hydrothermal fluids began just before the events. The technical report on drilling describes a phenomenon referred to as a "kick", in which fluid leaks into a borehole, which typically occurs before blowouts of pressurized fluid [3]. It was known that two boreholes blew out after reaching 18.3 to 18.4 m below the surface in the Ebino-fumarole area during a geothermal survey [41]. In this event, hot water welled up and steam blowout occurred after sudden increase of the borehole temperature. In the geothermal excavation, the pressurized hot water leaked, resulting in blowing through the wells. In addition, it is known that the cyclic behaviors of the lake water level and temperature in a hydrothermal system can be explained by the "two-phase boiling systems" phenomenon observed at the Inferno Crater Lake in New Zealand [78,79].

**Figure 11.** The schematic image of the shallow (ca. 600 m) and very shallow (ca. 350 m) systems beneath Ebinokogen Ioyama Volcano (See more details in Sections 6.1 and 6.2).

*Geosciences* **2020**, *10*, 183

One question remains open about why hydrothermal fluids upwelling site and the explosion site were different. There appeared depressions of 3 to 10 m in the crater Y2a, 2 to 8 m in the crater Y2b, and 3 to 12 m in the crater Y3, a little time after the April 2018 eruption. The total depression volume was about 8400 m3. And the depression volume of the Y2a and Y3 craters was about 6500 m<sup>3</sup> (Table 1). Those values are clearly larger than the ejecta volume of about 1500 m3, assuming the density of the deposit as 1500 t [54] is 1000 kg/m3. Therefore, the mechanism such as the crushing of the host rock around the conduit as in a maar formation [80] is not plausible. A hydrothermal dissolution [81] may be possible by continuous hydrothermal water outflow after the eruption. However, it is unclear if dissolution of a large volume in a short time. Therefore, we conclude that a certain amount of space had existed underground (Figure 11). The coupling between the geyser conduit and a laterally offset reservoir system [82] in the geyser eruption may be constructed under Ebinokogen Ioyama Volcano. However, the geyser system does not eject directly from a laterally offset reservoir, so it is difficult to apply this model for the April 2018 eruption. Likely, when the relatively low-temperature hydrothermal fluids in the upper layer leak and reduce the groundwater level, the resultant decreasing of hydrostatic pressure causes boiling to start. An instance of a similar trigger was seen when the dropping water levels at Gengissig lake in Kverkfjöll Volcano (Iceland) were observed to decompress the pressured aquifers, resulting in explosions due to boiling [15]. Additional explanation for the separation between the upwelling and eruption sites may originate from the fluid percolation pattern dictated by the deposit heterogeneity in the very-shallow hydrothermal system. Therefore, several mechanisms are possible for this eruption.

In summary, the pressurization of hydrothermal system is considered to be caused first as a bottom–up. The increase in fumarolic activity and the spring K3 temperatures from February 2018 indicated that the magmatic liquids were supplied to the very-shallow hydrothermal system. Second, hydrothermal fluids were observed to comprise mixed magmatic and meteoric waters up-welled at the Y1 just before the eruption in April 2018. We conclude that an initial bottom–up fluid pressurization (kick) destabilized the shallow part of the hydrothermal system, producing a boiling-front that penetrated downwards into the hydrothermal reservoir, followed by the explosion according to a top–down mechanism [12,83]. The expanding steam fragmented and dispersed the broken host rocks until the boiling decreased and stopped supplying sufficient energy to break and ejected rocks from the crater. Such mechanisms may explain the 2018 Ebinokogen Ioyama eruptions. Based on our observations, we also conclude that sudden injection and up-welling of hot magmatic fluids may cause an increase in temperature of very-shallow hydrothermal systems, and in turn produce changes in geothermal features at the surface. Thus, rapid increases in the geothermal manifestation over a short period of time, as in the case of Ebinokogen Ioyama Volcano, should be taken as a sign for caution in geothermal areas.

#### *6.3. Issues of Steam-Driven Eruptions for the Volcanic Disaster Prevention*

Information pertaining to the level of damage corresponding to steam driven eruptions originating in hydrothermal aquifer systems should be provided at tourist spots near and in geothermal areas. The magmatic–hydrothermal fluids storage system may have developed at two different levels in Ebinokogen Ioyama Volcano, contained by the conductive caprock layer (Figure 11).

Steam driven explosions from very-shallow levels are known to occur in some geothermal areas [84–87]. In a hydrothermal blowout event at the jet fumarole vent A, altered volcanic fragments of block size were scattered only a few meters from the vent. The magmatic hydrothermal eruption at Ioyama in April 2018 may have involved hydrothermal liquids shallower than 350 m depth. Based on past activity [32], larger steam-driven eruptions ejecting larger volumes of debris and producing tens-to-hundreds-meter-diameter craters at the Fudoike and Karakunidake north craters, occurred in Ebinokogen Ioyama Volcano. These explosions may involve greater depth, similarly to phreatic or phreato–magmatic eruptions, ye<sup>t</sup> likely have a steam-driven dominant component e.g., [88–90]. The question remains, what may be the signs of bigger eruptions. The geophysically detected source

of the 2014 phreatic eruption at Ontake Volcano, which scattered ballistics ejecta over 1 kilometer away [91], was thought to be from 600 to 1000 m deep [92,93]. Kato et al. [94] proposed that hot fluids infiltrated to cracks from the deeper magma chamber in middle September 2014 before the September 27 phreatic eruption in Ontake. The break of the seal of the conduit for hydrothermal fluids at a shallow level began just prior to the eruption [94], resulting in the decompression that led flashing of hydrothermal fluids from the top to downward. Recent research [95] supports this model by the finding of a small amount of juvenile material in the 2014 eruption products in Ontake. A steam-driven eruption at Sinmoedake Volcano in 2008 scattered ballistics as far as 800 m [31]. The geophysically detected source of the inflation and deflation in the eruption was estimated to be at a depth of 500 to 600 m below the volcano [96]. These are similar to the VLP events, 700 ± 200 m below the lake, in Kawah Ijen (Indonesia) and that of 850 ± 150 m below the crater at White Island (New Zealand) [97].

In case of Ebinokogen Ioyama in 2018, it took over two months for magmatic fluids to be supplied into shallow hydrothermal system before the eruption from the deep system, contrasting to about two weeks (11 to 16 days) in Ontake. In Ebinokogen Ioyama, the supplied amount and rate of magmatic fluids of the 2018 eruption may be much smaller than the 2014 eruption in Ontake. In summary, smaller events could be better understood as result of our study. And more energetic steam-driven eruptions may occur in this area if larger amount of magmatic fluid is injected rapidly and a wider area of the shallow hydrothermal system is destabilized. However, more studies are needed to improve our knowledge of triggers, mechanisms and potential precursors of steam-driven eruptions at Ebinokogen Ioyama. Additionally, we need to consider repetitive highly hazardous hydrothermal eruptions like that of Lake Okaro in New Zealand [12].

The technology to forecast very-shallow hydrothermal eruptions has developed by combining observations of hydrothermal phenomena such as increasing activity of geothermal areas, a welling-up kick, results of In-SAR and precise leveling observations. In the small eruption of Hakone 2015, In-SAR observed an uplift deformation area of approximately 200 m in diameter immediately before the eruption, along with a ground uplift of up to 30 cm. The vents appeared at the southern edge of deformation [98]. Locations of the fumarole S and the Y1 hydrothermal pond at Ebinokogen Ioyama may be possible sites for future eruption. A hydrothermal eruption occurred from the Ioyama-south craters on 19 April 2018. The next day, vents (Ioyama-west crater) formed 500 m away from the Ioyama-south craters. The impact area of the ejecta from the latter vents was as small as ~50 m. The outside of the alert level 2 area possibly impacted by ballistics, within a one-kilometer radius from the Ioyama crater, shown in the volcanic hazard map of Ebinokogen Ioyama [99,100], was not a ffected. Given that the tourist facility is relatively close to the potential venting area, in addition to addressing the 19 April vents, there is also the need to highlight the occurrence of the 20 April.
