Behavior of magmatic components in fumarolic gases related to the 2018 phreatic eruption at Ebinokogen Ioyama volcano, Kirishima Volcanic Group, Kyushu, Japan

Direct sampling and analysis of fumarolic gas was conducted at Ebinokogen Ioyama volcano, Japan, between December 2015 and July 2020. Notable changes in the chemical composition of gases related to volcanic activity included a sharp increase in SO2 and H2 concentrations in May 2017 and March 2018. The analyses in March 2018 immediately preceded the April 2018 eruption at Ioyama volcano. The isotopic ratios of H2O in fumarolic gas revealed the process of formation. Up to 49% high-enthalpy magmatic vapor mixed with 51% of cold local meteoric water to generate coexisting vapor and liquid phases at 100–160 °C. Portions of the vapor and liquid phases were discharged as fumarolic gases and hot spring water, respectively. The CO2/SO2 ratio of the fumarolic gas was higher than that estimated for magmatic vapor due to SO2 hydrolysis during the formation of the vapor phase. When the flux of the magmatic vapor was high, effects of hydrolysis were small resulting in low CO2/SO2 ratios in fumarolic gases. The high apparent equilibrium temperature defined for reactions involving SO2, H2S, H2 and H2O, together with low CO2/SO2 and H2S /SO2 ratios were regarded to be precursor signals to the phreatic eruption at Ioyama volcano. The apparent equilibrium temperature increased rapidly in May 2017 and March 2018 suggesting an increased flux of magmatic vapor. Between September 2017 and January 2018, the apparent equilibrium temperature was low suggesting the suppression of magmatic vapor flux. During this period, magmatic eruptions took place at Shinmoedake volcano 5 km away from Ioyama volcano. We conclude that magma sealing and transport to Shinmoedake volcano occurred simultaneously in the magma chamber beneath Ioyama volcano.


Introduction
Although phreatic eruptions are generally small in scale, they can nonetheless cause injuries and fatalities. The phreatic eruption at Ontake Volcano Japan in 2014 killed 58 tourists at the summit area (Oikawa et al. 2016). Prior to the eruption, some volcanic earthquakes had been observed by the Japan Meteorological Agency (JMA); however, the eruption occurred in advance of the announcement of an evacuation advisory to tourists. A skier was also killed by falling rocks ejected during the phreatic eruption of Moto-Shirane Volcano Japan in 2018 (Japan Meteorological Agency 2020a). No precursory volcanic earthquake was observed prior to that eruption.
The detection of volcanic earthquakes is in some cases insufficient to mitigate disasters resulting from phreatic eruptions (Barberi 1992;Stix and de Moor 2018), and it is, thus, necessary to evaluate the eruptive potential of such volcanoes on the medium and long term using methods in addition to seismic observation. Phreatic eruption involves the explosion of hydrothermal fluid reservoirs that develop in the relatively shallow crust. These reservoirs comprise groundwater from aquifers, which absorbs high-enthalpy magmatic vapor (MV). SO 2 and CO 2 are the typical constituent chemical species of MV components. By understanding the behavior of MV, the potential of medium-to long-term phreatic eruptions can be evaluated. Geochemical observations of fumarolic gas could improve our understanding of MV behavior, because such gases dominate the gas phase in hydrothermal reservoirs that leak to the surface (e.g., Ossaka et al., 1980;Taran et al. 2002;Tassi et al. 2003;Chiodini et al. 2010;Fischer et al. 2015;Ohba et al. 2019a, b). In addition to the MV component, fumarolic gases include components formed within the hydrothermal reservoir (HR). Here, HR components are composed of reducing chemical species such as H 2 S and CH 4 . Those H 2 S and CH 4 are thought to be reduced from the magmatic SO 2 and CO 2 , respectively, by the Fe 2+ contained in the crustal rocks contacting with the fluid in hydrothermal reservoir (Giggenbach 1997). The coexistence of MV and HR components in fumarolic gases was identified at White Island, New Zealand (Giggenbach 1987). Before the phreatic eruption at Hakone Volcano in 2015, volcanic earthquakes occurred, and the ratio of MV to HR components increased according to the frequency of earthquakes (Ohba et al. 2019a). Such changes in fumarolic gas can be attributed to the injection of MV into the HR. This injection induces increases in fluid pressure in the HR and may initiate volcanic earthquakes (Ohba et al. 2019a).
From the viewpoint of general volcanic disaster prevention, it is important to investigate MV components in fumarolic gases prior to phreatic eruptions at volcanoes such as Hakone volcano. A phreatic eruption took place in April 2018 at Ebinokogen Ioyama volcano (hereafter "Ioyama volcano") (Tajima et al. 2020). Between December 2015 and July 2020, we repeatedly collected and analyzed fumarolic gases at Ioyama volcano to investigate the behavior of MV components in fumarolic gases before and after the eruption, and to extract geochemical parameters that could be used to denote precursor events in advance of the main eruption.

Geological settings
The Kirishima volcanic group is located near the volcanic front on Kyushu Island Japan (Fig. 1a). Ioyama volcano (Io) is among the volcanoes in this group, and is surrounded by many volcanic craters and cones formed during the Quaternary period (Fig. 1b). The formation ages Fig. 1 a Location of Ebinokogen Ioyama volcano (Io) on Kyushu Island (KI), Japan. The Philippine Sea Plate (PP) is subducting beneath the Eurasian Plate (EP) on which KI is situated. PP and EP are separated by the Nankai trough (NT). Circles and broken line indicate active volcanoes and the volcanic front, respectively. b Io surrounded by Ebinodake volcano (Ed), Karakunidake volcano (Kr), Shiratori-yama crater (Sy), Rokkannonmiike crater (Ro), and Fudoike crater (Fd). Kd is a debris avalanche deposit. P and W denote ponds and collapse walls, respectively. Eq denotes the region in which the volcanic earthquake hypocenters are distributed. Sm, Ss, Oy, Ks and On denote Shinmoedake volcano, Shishidodake volcano, Ohatayama volcano, Koshikidake volcano and Onamiike volcano, respectively. Bold red curves indicate the crater rim. c Location of fumaroles a, b, c and h. Phreatic eruptions took place at e1 and e2 on April 16 and 26, 2018, respectively. Hot spring water was discharged at K3 and Y1. R denotes roads of Ebinodake volcano (Ed), Karakunidake volcano (Kr), Shiratori-yama crater (Sy), Rokkannonmiike crater (Ro) and Fudoike crater (Fd) were estimated to be 70-300 ka, 15 ka, 18-25 ka, 25 ka, and 6.3 ka, respectively (Imura and Kobayashi 2001). A part of northwest flank of Kr has collapsed and spread a debris avalanche deposit (Kd). The andesitic Ioyama volcanic body was developed on Kd during a magmatic eruption in 1768 (Imura and Kobayashi 2001). During January 2011, subplinian eruptions (VEI = 3) occurred at Shinmoedake volcano (Nakada et al. 2013), which is located 5 km southeast of Ioyama volcano. Following the 2011 events, magmatic eruptions took place at Shinmoedake volcano in October 2017, March 2018 and April 2018. Prior to the 2017 eruption, crustal deformation was observed in the area of the Kirishima volcano group (Japan Meteorological Agency 2020b). This deformation was interpreted to result from the combination of shrinking beneath Ioyama volcano and inflation beneath Shinmoedake volcano, suggesting that the magma transport had occurred from a magma chamber beneath Ioyama volcano to Shinmoedake volcano.
To clarify the magma plumbing system of the Kirishima volcanoes, a magnetotelluric (MT) survey was conducted (Utada et al. 1994). Based on the one-dimensional (1D) analysis of these results, the emplacement of a magma chamber was suggested at − 10 km from the surface (Utada et al. 1994). A low-resistivity region was also detected approximately 5 km beneath Ioyama volcano, which was interpreted to be a small magma chamber or magmatic gas reservoir. Three-dimensional (3D) analyses of these MT survey data confirmed the depth of magma chamber at − 10 km depth, located 10 km west of Ioyama volcano as shown schematically in Fig. 2 (Aizawa et al. 2014). These analyses also detected a vertical electric conductor beneath Ioyama volcano, which was interpreted to be a magmatic vapor reservoir zone (MVR in Fig. 2) with temperatures greater than 400 °C. This vertical conductor extends between − 5 and − 3 km, and was estimated to be a magma pathway that was partially Fig. 2 Schematic illustration of the magma plumbing system revealed by a previous 3D-MT survey (Aizawa et al. 2014). MV and M refers to magmatic vapor and magma, respectively. Io and Sm denote the Ioyama and Shinmoedake volcanoes, respectively. MVR, SMC and DMC are the magmatic vapor reservoir, shallow magma chamber and deep magma chamber, respectively formed before magmatic eruptions at Shinmoedake volcano (Aizawa et al. 2014).
An area of active fumarolic activity has been observed around the summit of Ioyama volcano since the early of 1900s (Funasaki et al. 2017). During the 1970s, the fumarolic activity was particularly intense, and the temperature of fumarolic gases reached 247 °C (Kagiyama et al. 1979). In 1994, the temperature of fumarolic gas decreased to 101 °C (Ohba et al. 1997), and after the 1990s, the fumarolic activity gradually decreased. After 2007 the discharge of fumarolic gas around the summit area of Ioyama volcano ceased completely (Funasaki et al. 2017), and the seismic activity has been continually low (Japan Meteorological Agency 2020b).
Since mid-2015, however, volcanic earthquakes have been observed sporadically. In December 2015, fumarolic gas discharge was confirmed to occur for the first time since 2007 (position a, Fig. 1c). Since December 2015, fumarolic gas emissions at Ioyama volcano have become increasingly intense. Tajima et al (2020) reported superficial phenomena occurring at Ioyama volcano in details from 2015 to 2018. According to their report, the geothermal surface area with temperatures higher than 50 °C was approximately 2000 m 2 in May 2016. This geothermal area gradually increased to around 10,000 m 2 in June 2017. In May 2017, a strong fumarole characterized by a roaring sound developed (position h, Fig. 1c). Following the appearance of this fumarole, the geothermal surface area temporarily decreased to about 4000 m 2 in October 2017. This reduced geothermal area was maintained until around February 2018, at which time the number of earthquakes increased sharply. On April 19 and 26, 2018, phreatic eruptions occurred (positions e1 and e2 in Fig. 1c, respectively). The eruption on April 19 2018 began at 15:39 and lasted for about 15 h, and the height of the eruption column was about 500 m at maximum (Japan Meteorological Agency 2020b). The total volume of the ejecta was estimated to be 1500 m 3 , and no juvenile materials were detected in the ejecta (Tajima et al 2020).
Volcanic earthquakes on Ioyama volcano were observed to occur mostly below the summit crater and in the area extending to the south (Fig. 1b). The depth of their epicenters was distributed in the range of − 3 to − 1 km below sea level (Japan Meteorological Agency 2020b). According to Tajima et al (2020), the shallow crust beneath Ioyama volcano is estimated to have the following layers.
Volcanic earthquakes at Ioyama volcano occurred mostly in Layer-5. The point in Layer-5 just below the summit crater was a pressure source of crustal deformation since July 2017 (Japan Meteorological Agency 2020b).

Sampling and analysis of fumarolic gases
Fumarolic gases were repeatedly sampled at positions a, b and c (see Fig. 1c). Samplings were obtained at 17 occasions between December 2015 and July 2020. In addition to the above three fumaroles, one fumarolic gases was sampled at position h in July 2020 (Fig. 1c). The outlet temperature of fumarolic gas was measured using a thermocouple with a K-type sensor (Yokogawa Electric Corp., TX-10). Fumarolic gas was introduced using a titanium pipe inserted directly into the fumaroles from which samples were obtained. The end of the titanium pipe was connected to a rubber tube, which was in turn connected to a 120 ml pre-evacuated Pyrex glass bottle with an airtight stop cock (Giggenbach 1975), where 20 ml of 5 M KOH solution was introduced. Water vapor and acidic gases (CO 2 , SO 2 , H 2 S, etc.) were absorbed by the KOH solution whereas the residual gases (hereafter R-gas), e.g., N 2 , O 2 , Ar, He, H 2 , and CH 4 were maintained in the headspace of the glass bottle. The chemical species in the KOH solution were analyzed according to the methods of Ozawa (1968). The H 2 S/SO 2 ratio in the fumarolic gas was determined using a KI-KIO 3 solution (Lee et al. 2016) and the total molar amount of R-gas was determined by the head space volume of the bottle and the inner pressure of the head space at room temperature. Based on the molar amount of H 2 O, CO 2 , H 2 S, SO 2 and R-gas, the relative concentration (ppm) of each gas was calculated ( Table 1). The analysis of R-gas by gas chromatography followed the methods of Ohba et al. (2019a, b).
For isotopic analysis of water vapor in fumarolic gas, this gas was cooled using a double-tube condenser made of Pyrex glass. The isotopic ratio of the condensed water was determined using an IR-laser cavity ring down analyzer (Picarro Inc., L2120-i). The analytical precision of the analyzer was ± 0.12‰ and ± 0.05‰ for δD and δ 18 O, respectively. The isotopic ratio of H 2 in the R-gas was determined using a continuous flow system combined with a mass spectrometer (Thermo Fischer Scientific Delta V) (Tsunogai et al. 2011). The analytical precision of the analyzer was ± 0.8‰ for δD of H 2 .   Table 1 lists the data obtained on fumarolic gas, including outlet temperature, chemical composition, isotopic ratios (δD and δ 18 O) of H 2 O and the isotopic ratios (δD) of H 2 . The outlet temperatures were predominantly close to the boiling point (95.8 °C) of water under standard air pressures at the altitude of the fumaroles (1300 m) (Fig. 3a). The outlet temperature of fumarole "a" was 106 °C in September 2017, 10 °C higher than the boiling temperature. The outlet temperatures of fumaroles a and c have dropped significantly below their boiling point since November 2019. The histograms in Fig. 3a, b indicate the number of earthquakes observed monthly by the JMA. Vertical lines indicate the phreatic eruption of Ioyama volcano (April 19, 2018) and the magmatic eruptions at Shinmoedake volcano (October 11, 2017, March 1, 2018, and April 5, 2018. The fumarolic gases were mostly dominated by H 2 O vapor, except the fumaroles a and c after November 2019 (Fig. 3b). Figure 4 shows temporal changes in the logarithmic concentrations (ppm) of CO 2 , H 2 S, SO 2 , He, H 2 , O 2 , N 2 , CH 4 , and Ar. After November 2019, the concentrations of N 2 , O 2 , and Ar in fumaroles a and c increased significantly, indicating air contamination. The SO 2 and H 2 concentration of fumaroles a, b, and c increased significantly during May 2017 and March 2018 (Fig. 4c, e).
In the following, we divide the observation period based on the isotopic ratio of H 2 O (Fig. 5a, b).
Period-I: December 2015 to December 2016; Period-II: January 2017 to May 2018; Period-III: June 2018 to July 2020. During Period-1, δD increased from approximately − 90‰ to the around of − 50 ~ − 40‰, while the δ 18 O increased from about − 14‰ to the range of − 4 ~ − 1‰ (Fig. 5a, b). The number of earthquakes during this period was relatively small, and the activity of Ioyama volcano is regarded to be in its preparatory stage.
During Period-II, the δD and δ 18 O maintained relatively high values. The highest values of δD and δ 18 O were recorded at fumarole b in March 2018 (Fig. 5a,  b). Throughout this period, the volcanic activity was enhanced. Shinmoedake volcano exhibited three magmatic eruptions, and phreatic eruptions occurred at Ioyama volcano. More earthquakes occurred in Period-II than in Period-I.
In Period-III, the δD decreased to − 100‰ and the δ 18 O decreased to − 18‰. The number of earthquakes was high at the beginning of this period but gradually decreased. No eruptions occurred in this period at Ioyama volcano.
For the δD of H 2 , the highest values were recorded at fumarole "a" in March 2018 (Fig. 5c). In terms of the δD of H 2 , the time variation at the three fumaroles did  not harmonize. For example, considering the changes between August 2016 and January 2017, the δD at fumarole "a" increased significantly, while that at fumarole b decreased. The δD at fumarole c was low between October 2018 and July 2020, whereas the δD of fumarole b was 200‰ higher than that of fumarole c for the same period. The three fumaroles are located close to one another (Fig. 1c); thus, their distinct δD implies that the phenomena controlling the δD of H 2 operated at a shallow depth. If δD of H 2 was fixed at a large depth, the δD is expected to be similar for fumaroles located close each other.
The δ 18 O and δD values of H 2 O in fumarolic gas exhibit a linear distribution (Fig. 5d). K3 and Y1 are hot spring waters collected at Ioyama volcano (Tajima et al 2020); collection locations are shown in Fig. 1c. K3 represents local meteoric water (Tajima et al. 2020). Y1 is a hot spring water discharging near the eruptive point e1 (Fig. 1c), which appeared just before the phreatic eruption of Ioyama volcano, with a temperature on April 16, 2018 of 93 °C (Tajima et al. 2020). AMW is general isotopic composition of H 2 O degassed from andesitic magma (Taran et al. 1989;Giggenbach 1992). Some fumarolic gas samples are distributed near the lines connecting AMW and K3, but most are distributed in the area below these lines, suggesting that a simple mixing model between the local meteoric water and AMW is insufficient to explain the isotopic ratio of H 2 O in fumarolic gas.

Temporal change of apparent equilibrium temperatures (AETs)
The apparent equilibrium temperature based on sulfur species in fumarolic gas (hereafter AETS) is defined in terms of the following reaction (e.g., Ohba et al 1994): Assuming the equilibrium of reaction as follows: where K and f H2O are, respectively, the equilibrium constant of reaction (1) and the fugacity of H 2 O, both of which are functions of temperature. The fractions in parentheses on the right-hand side of Eq. (2) refer to the molar ratio of each gas species relative to H 2 O. Based on the temperature dependence of the standard equilibrium constants of formation of elements and the fugacity (1) of H 2 O in hydrothermal systems (Giggenbach 1987), the left-hand side (LHS) of Eq.
(2) is calculated to be as follows: where T is temperature in Kelvin. AETS is defined as the temperature that satisfies Eq. (3). It is useful to consider AETS because it is the qualitative parameter depending on conditions favoring the retrograde reaction in terms of reaction (1). During the transport of MV from the source to surface, the temperature decreases along the temperature profile of the passage. When the magmatic vapor rises rapidly, the chemical composition of magmatic vapor will be quenched at high temperature without a large change, then, a high AETS is calculated for the fumarolic gas composition. When the magmatic vapor rises slowly, the chemical composition of magmatic vapor will be quenched with a large change because an enough time is allowed to occur reactions at lower temperature among gas species or reactions between gas species and crustal rock, a low AETS is calculated for the fumarolic gas composition (Giggenbach 1987). Therefore, an increase in AETS is often observed during periods of enhanced volcanic activity (e.g., Ohba et al 2011). Another apparent equilibrium temperature (AETD) is defined in terms of the following reaction: Assuming equilibrium in reaction (4), we obtain the equation of Richet et al (1977): where AETD is in degree C and x is given by AETS and AETD were high in Period-II relative to Periods-I and -II (Fig. 6a, b). For example, the AETS in May 2017 were 752, 745, and 681 °C at fumaroles a, b, and c, respectively and the AETD in May 2017 was 926, 340, and 490 °C at fumaroles a, b, and c, respectively. Although the AETS at fumaroles a, b and c often show close values, the AETD of fumaroles a, b and c occasionally show large differences due to the correspondingly large differences in the δD of H 2 . Therefore, to estimate the qualitative velocity and flux of MV, changes in AETS are more reliable than AETD.

Behavior of MV
The behavior of MV during Period-II can be estimated by correlating changes in AETS with the eruptive events occurring at Shinmoedake and Ioyama volcanoes (Fig. 7). AETS increased significantly during May 2017, when a strong fumarole appeared at point h. No eruptions occurred at Shinmoedake volcano in May 2017. Consequently, it is estimated that a stagnant magma in the shallow magma chamber (SMC) released a large flux of MV (Fig. 7a). This increased flux shortened the residence time of MV, increasing the value of AETS. Such an increase in MV flux is supported by the enlarged geothermal area observed during this time (Tajima et al. 2020), which reached its maximum in June 2017. The AETS was low between September 2017 and January 2018, suggesting a decrease in MV flux (Fig. 7b). The simultaneous observed reduction in geothermal area (Tajima et al. 2020) is consistent with a decreased MV flux. During the period of low AETS, a magmatic eruption occurred in October 2017 at Shinmoedake volcano, suggesting that magma was transported from the SMC to Shinmoedake volcano. The decreased flux of MV may be explained by magma sealing (Fournier 1999). The number of earthquakes increased from January to March 2018 (Fig. 7c). Furthermore, AETS rose in March 2018. Multiple eruptions occurred at Shinmoedake volcano during this period, and it is, thus, highly probable that SMC discharged magma toward Shinmoedake volcano, and the flux of MV to Ioyama volcano increased simultaneously. Magma movement and the increased MV flux likely occurred due to   (Fig. 7c).

Model for the formation of vapor and liquid phases
The correlation between δD and δ 18 O of H 2 O in the fumarolic gases at the Hakone and Kusatsu-Shirane volcanoes has been explained by the mixing of high-temperature MV and cold local meteoric water (LW), followed by a singlestep phase separation (Ohba et al 2019a, b). The concept of this model is illustrated in Fig. 8. Herein, we aim to apply the same model. Assuming enthalpy conservation, the conservation of isotopic ratios, the equilibrium distribution of isotopes between the vapor phase (Vp) and liquid phase (Lp), and the Ar distribution between Lp and Vp, the following equations can be written:  Fig. 3 where H, δ, and C represent enthalpy, isotope ratio in δ-notation, and the and the amount of Ar, respectively. Equations 10 and 11 describe the equilibrium distribution between Vp and Lp in terms of stable isotopes and Ar, respectively. Using Eqs. 7-11, the isotopic ratios and Ar/H 2 O ratios of Vp and Lp were calculated. The numerical values necessary for this calculation are listed in Table 2. As shown in Fig. 9a, the calculated vapor phase lines at 100 °C and 160 °C (Vp-100 and Vp-160) are located near the point of the observed fumarole. Furthermore, Y1 is located between the calculated liquid phase lines at 100 °C and 160 °C (Lp-100 and Lp-160). These agreements suggest that MV and LW are mixed and that the phase-separated vapor phase (Vp) corresponds to fumarolic gas. The temperature of phase separation is estimated to be about 100-160 °C. To calculate the Vp and Lp lines in Fig. 9a, "f", which denotes the mixed fraction of MV, varied within the range of 0.10-0.6 at 100 °C. At 160 °C, "f" varied within the range of 0.15-0.6. If "f" has a value higher than these ranges, Lp is not formed, i.e., only Vp forms. If "f" has a value lower than these ranges, Vp is not formed; i.e., only Lp is formed. When "f" is fixed at its maximum, the isotope ratios of Vp and Lp are located at the right end of the Vp and Lp lines, respectively. When "f" is fixed at a minimum value, the isotope ratios of Vp and Lp are located at the left end of the Vp and Lp lines, respectively. The highest δ 18 O values were observed at fumarole "b" in March 2018. The "f " values producing the same δ 18 O on the Vp line were 0.49 and 0.48, at 100 °C and 160 °C, respectively. The fumarolic δD and δ 18 O increased rapidly during Period-I, and high values were maintained throughout Period-II (Fig. 5a, b). According to the above model for the formation of Vp and Lp, the increase in Vp isotope ratios is due to an increase in the mixing ratio of MV. Therefore, the increase in the isotope ratios of H 2 O denotes an increased flux of MV (under the assumption of a constant flux of LW).
Some fumarolic gases have isotopic ratios that are much lower than the Vp line (Fig. 9a). When a part of H 2 O in Vp is condensed and lost, the isotopic ratio of the remaining water vapor decreases. This direction of change by condensation has the same direction as the vector connecting L and V-100 in Fig. 9a, i.e., the isotope ratios of liquid and vapor equilibrated at 100 °C. At the moment of condensation, the isotope ratio of the vapor phase leaves the Vp line (Vp-100 or 160) and moves toward the lower left. Some fumarolic gases with isotope ratios lower than the Vp line can be explained by the partial condensation of water vapor (Ohba et al. 1997). Isotope ratios lower than the Vp line can be also explained by the addition of water vapor generated from LW (Ohwada et al. 2003). In natural fumaroles, the above two effects may occur simultaneously. In Period-III, the low δD and δ 18 O values can, therefore, be explained by the condensation of water vapor and or the influence of LW.
In Fig. 9b, the Ar/H 2 O molar ratio of fumarolic gas is calibrated on the horizontal axis via conversion to the Interactions between MV and LW produce the vapor phase (Vp) and the liquid phase (Lp). A portion of both the Vp and Lp are discharged at the surface as fumarolic gas (FG) and hot spring water (HS), respectively. A portion of the SO 2 in MV is converted to SO 4 2ion and H 2 S gas, which are preferentially distributed to Lp and Vp, respectively magnified Ar fraction in the Ar-H 2 O binary system, which is defined as follows: As shown in Fig. 9b, even for the relationship between mF(Ar) and δ 18 O, the calculated Vp curves at 100 °C or 160 °C (i.e., are closely linked to fumarolic gases. The increased atmospheric contamination drives the points to the right in the diagram. Fumarolic gases with δ 18 O higher than that of the Vp-160 curve are interpreted to have suffered the contamination by an atmospheric component. Fumarolic gases located beneath the Vp-100 curve are considered to have been affected by the partial condensation of water vapor and the addition of water vapor originating in (12) mF (Ar) = 10 6 Ar H 2 O 10 6 Ar H 2 O + 1 meteoric water (vLW in Fig. 9b). The Vp curve overlaps with the isotopic ratios of fumarolic gases, especially in the low mF(Ar) and high δ 18 O region, indicating the validity of the Vp and Lp formation model. In the above model, mixing between LW and MV occurs during a single stage, but, as shown in Fig. 8, mixing can occur at both the HR located under the caprock and in the aquifer. It is not possible to evaluate the relative contributions of these two mixing events based solely on the isotope ratio of fumarolic gases.

Correlation between concentrations of gaseous species
Examining the correlation among the concentrations of species in fumarolic gases is useful for estimating the origin of each species, as well as for estimating the state of the hydrothermal system. In the He-N 2 -Ar ternary system (Fig. 10a), it has been reported that a mixing relationship holds between two end members, i.e., magmatic and atmospheric components (Kita et al. 1993;Giggenbach 1997;Taran 2011;Ohba et al. 2019a, b). In this study, we also found that the composition of fumarolic gas was distributed between these two endmembers (Fig. 10a). The magmatic endmember (broken circle) has a fixed N 2 /He ratio and is poor in Ar. The atmospheric endmember is composed of air and the component dissolved in water saturated with air (ASW).
In the He-N 2 -CO 2 ternary system (Fig. 10b), the fumarolic gases are concentrated almost into a single group (broken circle in Fig. 10b). Since helium is a magmatic component, the grouping in the He-N 2 -CO 2 ternary system suggests that CO 2 is also a magmatic component. Some points are distributed in the direction of the angle of N 2 away from the group. Their distribution indicates contamination by atmospheric components.
In the He-N 2 -CH 4 ternary system, fumarolic gases can be divided into two groups, although the distance between these groups is small (Fig. 10c). Some points distributed near the N 2 corner indicate the contamination by atmospheric components. According to Fig. 11a, the CH 4 /He ratio is low during the latter half of Period-II, when eruptions occurred at the Shinmoedake and Ioyama volcanoes. At Periods-I and -III, the CH 4 /He ratio is generally high. The distribution of fumarolic gases in Fig. 10c suggests that CH 4 is essentially a magmatic component (broken circle). In addition to magmatic CH 4 , non-magmatic CH 4 component was observed; this is deemed to be contaminative. For non-magmatic CH 4 , a thermogenic origin is possible, i.e., generation via the decomposition of organic matter in the crust.
Based on the SO 2 -CO 2 -H 2 S ternary system, the type of the hydrothermal system can be estimated. According to Stix and de Moor (2018), a hydrothermal system with deep-seated degassing magma can be distinguished from a hydrothermal system with a shallowly intruded degassing magma by the SO 2 -CO 2 -H 2 S ternary composition of fumarolic gas. In addition to this distinguishment, the SO 2 -CO 2 -H 2 S ternary system enables the evaluation of interaction between MV and shallow aquifer. The observed CO 2 /St ratio (where St refers to the molar summation of SO 2 and H 2 S) of fumarolic gases are distributed in a large range (between 1.4 and 22) (Fig. 10d). Most fumarolic gases have H 2 S/SO 2 molar ratios greater than 4; such gases are located in the region considered to be hydrothermally dominated (HD; Fig. 10d). As shown in Fig. 11b, all fumarolic gases measured during Period-I are classified as HD-type. Some fumarolic gases with low H 2 S/SO 2 ratios are classified as hydrothermal magmatic types. Furthermore, the hydrothermal magmatic types can be divided into deep hydrothermal magmatic (DHM) and shallow Fig. 10 a Fumarolic gas compositions in the He-N 2 -Ar ternary system; broken circle indicates the expected magmatic end member. ASW is the composition of air dissolved in water. Broken lines indicate the mixing between the magmatic and atmospheric end members. b Fumarolic gas compositions in the He-N 2 -CO 2 ternary system; broken circle indicates the expected magmatic end member. Broken lines indicate the mixing between the magmatic and atmospheric end members. c Fumarolic gas composition in the He-N 2 -CH 4 ternary system; broken circle indicates the expected magmatic end member. d Fumarolic gas composition in the SO 2 -CO 2 -H 2 S ternary system classified as a hydrothermal system (Stix and de Moor 2018). SLS, HD, DHM, SHM, DM, SM and SR indicate the types of volcanic gas with the following geneses: "S loss scrubbing", "hydrothermally dominated", "deep hydrothermal-magmatic", "shallow hydrothermal-magmatic", "deep magmatic", "shallow magmatic" and "sulfur remobilization", respectively. PF indicates the estimated parental magmatic fluid in the Kirishima volcanic area (Ohba et al. 1997). The CO 2 /SO 2 and H 2 S/SO 2 molar ratios are calibrated on the left and bottom sides, respectively. The CO 2 /St isomolar ratio is indicated by broken lines where St is the summation of SO 2 and H 2 S hydrothermal magmatic (SHM) based on their CO 2 /SO 2 ratios. Fumarolic gases of the DHM type were observed during Periods-II and -III (yellow circles in Fig. 11b), whereas SHM gases were observed only observed during Period-II (red circles in Fig. 11b). The fumarolic gas "a" observed in May 2017 and March 2018 is classified into the SHM field, with a CO 2 /SO 2 ratio less than 4. The AETS of these two fumarolic gases were calculated as 752 °C and 611 °C (Fig. 6a), suggesting an increased MV flux in both May 2017 and March 2018. Therefore, the discharge of SHM-type fumarolic gases is regarded as an index for enhanced volcanic activity at Ioyama volcano.
PF in Fig. 10d shows the composition of the parental fluid released from the magma of Kirishima volcano. The CO 2 /St of PF was estimated at 0.25-0.44 (Ohba et al 1997). In Fig. 10d, the sulfur component in PF is considered to be SO 2 . The CO 2 /St ratio of fumarolic gases (1.4-22) is higher than the CO 2 /St ratio of PF. If the essential CO 2 /St ratio of MV is similar to the CO 2 /St ratio of PF, it means that MV loses sulfur-bearing gases during travel to the surface. As suggested in the previous section, fumarolic gases was comprised of the vapor phase formed during the interaction of MV and LW. The following hydrolysis reaction is expected to occur during this interaction (Kusakabe et al. 2000): Through the above reaction, some SO 2 in MV is absorbed by LW in the form of the SO 4 2ion. Indeed, near the fumaroles of Ioyama volcano, hot spring waters rich in SO 4 2-, such as K3 and Y1, were discharged (Tajima et al 2020). In addition, MV can be depleted in sulfur species through other reactions (e.g., Takano et al. 2004), such as The absorption of SO 2 by LW and the formation of native sulfur can explain the depletion of St in fumarolic gases relative to PF (Fig. 10d). Unlike SO 2 , CO 2 in MV is not readily dissolved in Lp; thus, most of the constituent CO 2 is distributed to Vp. Consequently, the CO 2 / SO 2 ratio of fumarolic gases reflects the degree of SO 2 removal.

Conclusions
This study has reported direct sampling and analysis of fumarolic gas at Ioyama volcano between December 2015 and July 2020. Notable changes in the chemical composition related to volcanic activity included increased concentrations of SO 2 and H 2 in May 2017 (13) 4SO 2 + 4H 2 O = 3HSO − 4 + H 2 S + 3H + (14) SO 2 + 3H 2 S = 3S + 2H 2 O Fig. 11 a Changes in the CH 4 /He molar ratio of fumarolic gas. b Changes in the SO 2 /H 2 S molar ratio of fumarolic gas. The y-axis values of 0.5 and 4 for H 2 S/SO 2 ratio were indicated by the two horizontal broken lines. c Changes in the SO 2 /CO 2 molar ratio of fumarolic gas. Fumarolic gases surrounded by red and yellow circles are classified as being of shallow hydrothermal magmatic (SHM) and deep hydrothermal magmatic (DHM) origins, respectively. Red histogram and vertical lines are the same as in Fig. 3