Technology, Central 7, Higashi 1-1-1, Tsukuba, Ibaraki , Japan. (Received March 10, 2000; Accepted November 15, 2001)

Transcription

1 Geochemical Journal, Vol. 36, pp. 1 to 20, 2002 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes, southwest Japan: Evaluation of short-term variations and precision of the gas sampling and analytical techniques GENJI SAITO, 1 * HIROSHI SHINOHARA 1 and KOHEI KAZAHAYA 2 1 Institute of Geoscience, National Institute of Advanced Industrial Science and Technology, Central 7, Higashi 1-1-1, Tsukuba, Ibaraki , Japan 2 Research Center for Deep Geological Environments, National Institute of Advanced Industrial Science and Technology, Central 7, Higashi 1-1-1, Tsukuba, Ibaraki , Japan (Received March 10, 2000; Accepted November 15, 2001) Up to 15 fumarolic gas samples were collected successively within a few hours at Satsuma-Iwojima and Kuju volcanoes, southwest Japan, with the aim to evaluate the naturally occurring short-term variations in the chemical and isotopic compositions. Variations in the concentrations of the major gas components (CO 2, total S, SO 2 and H 2 S) and in the hydrogen, oxygen (H 2 O) and carbon (CO 2 ) isotopic compositions were relatively small and were within the analytical errors. Exceptions were HCl at both volcanoes, CO 2 at Satsuma-Iwojima, and D/H and 18 O/ 16 O ratios of water at Kuju volcano. Large variations were found for the minor gas components, i.e., N 2, Ar, CH 4 and CO. The observed variations in HCl and CO concentrations probably relate to sampling and analytical procedures. Variations in HCl, N 2, Ar and CH 4 concentrations, and D/H and 18 O/ 16 O ratios of water may result from processes such as gain or loss of HClrich liquid, sporadic additions of air and hydrothermally-derived CH 4 and change of mixing ratio between magmatic and meteoric waters. Precision associated with the gas sampling and analytical techniques used in this study were estimated by making the hypothesis that the chemical and isotopic compositions of the fumarolic gases did not change during the sampling period. The estimated precision was 1 3% for CO 2, H 2 S and SO 2, 6 11% for HCl, 1 6% for H 2, 4 6% for N 2, 5 9% for Ar, 2% for CH 4 and 7 27% for CO. The precision of the isotopic analyses was estimated to be for 13 C/ 12 C ratio of CO 2. Also the precision was 0.9 and for D/H and 18 O/ 16 O ratios of the condensates, respectively. The results in this study suggest that careful sampling and analysis of volcanic gases provide reliable chemical and isotopic compositions that can be used to monitor volcanic activities. INTRODUCTION Observations of chemical and isotopic compositions of volcanic gases were carried out at certain volcanoes with the aims to monitor their activity and to predict forthcoming eruptions (e.g., Gerlach and Casadevall, 1986; Allard et al., 1991; Tedesco et al., 1991). In order to interpret a real change with time in the gas composition from long-term observations, it is necessary to know 1) the compositional fluctuation during a shortterm period (a day) and 2) the precision of the gas collection technique. Systematic and periodic changes during a short period also may provide information on magma degassing processes. Only a few studies have been carried out about the compositional fluctuation of fumarolic gases that may occur within a day. Repeated sampling *Corresponding author ( saito-g@aist.go.jp) 1

2 2 G. Saito et al. of volcanic condensates within a few hours was conducted at Nasudake volcano (Wada, 1968), Showa-Shinzan volcano (Mizutani and Matsuo, 1959; Mizutani, 1978) and Satsuma-Iwojima volcano (Matsuo et al., 1974). Remarkable variations were observed in B, HF, SO 2, H 2 S, HCl and CO 2 concentrations at Nasudake volcano (Wada, 1968), and in HCl, HF and B concentrations and D/H and 18 O/ 16 O ratios of water at Showa-Shinzan volcano (Mizutani and Matsuo, 1959; Mizutani, 1978). These large variations were attributed to contamination of the gas samples by sublimates occurring in the fumaroles (Mizutani and Matsuo, 1959), or to addition of shallow groundwater during gas ascent towards the surface (Mizutani and Matsuo, 1959; Wada, 1968; Mizutani, 1978). In contrast, minor variations in CO 2, SO 2, HCl and HF concentrations were observed at Satsuma-Iwojima volcano (Matsuo et al., 1974). Field workshops on volcanic gases were organized and performed by the International Association of Volcanology and Chemistry of the Earth s Interior s (IAVCEI) commission on the chemistry of volcanic gases (CCVG) for the purpose of evaluation and improvement of techniques for geochemical monitoring of volcanoes (Giggenbach and Matsuo, 1991; Giggenbach et al., 2001). Geochemists from different laboratories collected volcanic gas samples from the same fumarole and the results of subsequent gas analyses were compared assuming that both chemical and isotopic compositions of the fumarolic gases did not change in the course of the sampling. Large differences, up to 50% for standard deviation/average values, in chemical and isotopic compositions of the gases were often reported (Giggenbach and Matsuo, 1991; Shinohara, 1999). Giggenbach and Matsuo (1991) suggested that the large scatter was due to sampling artifacts. The problem evoked include chemical reaction of acid gases with sampling equipments (e.g., Ti tube), or gas condensation and subsequent gas loss within the sampling train. Therefore, it is important to evaluate the error associated with each gas sampling technique. In this study, up to 15 fumarolic gas samples were collected successively within a few hours at Satsuma-Iwojima and Kuju volcanoes in order to evaluate (1) short-term, natural variations in the gas composition and (2) the artificial fluctuation caused by the sampling techniques. GEOLOGICAL BACKGROUND Satsuma-Iwojima volcano Satsuma-Iwojima is a volcanic island comprising a part of the rim of a largely submerged Kikai caldera (e.g., Ono et al., 1982), which is located about 50 km south of Kyushu, Japan. Satsuma- Iwojima volcano is currently discharging a large amount of high temperature volcanic gases from the summit crater of a rhyolitic dome, known as Iwodake. The maximum temperature (900 C) and the flux of the volcanic gases ( t/d SO 2 ) have shown little variations for the last 10 years (Shinohara and Kazahaya, 1997; Kazahaya et al., 1997). The chemical and isotopic compositions of the high temperature fumarolic gases suggest a magmatic origin for the gases (Shinohara et al., 1993). To explain the steady gas discharge, magma convection in a conduit was proposed as a mechanism of gas transport from a deep volatile-rich magma chamber to the near surface (Kazahaya and Shinohara, 1996). Geophysical observations further support this model (Ohminato and Ereditato, 1997). These observations showed that long-period seismic pulses synchronized with a regular amplitude modulation of volcanic tremor repeated with a period of about 50 minutes and a shallow source of the pulses was about 40 m below the bottom of the summit crater. Ohminato and Ereditato (1997) demonstrated that the source of the volcanic tremor was probably the rapid gas flow discharged from magma within a narrow conduit, while the repetitive long-period pulses resulted from replacement of the dense degassed magma by the ascent of gas-rich magma at the magma column head. Petrological studies at Satsuma-Iwojima volcano (Saito et al., 2001) suggest occurrence of a stratified magma chamber, consisting of a basaltic layer underlying a rhyolitic layer. The rhyolitic magma is supplied continu-

3 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes 3 ously with CO 2 -rich gases from the underlying basaltic magma. Based on these geophysical and petrological observations, fluctuations in the chemical composition of the volcanic gases, in particular CO 2, are expected to occur. Kuju volcano Kuju volcano is located in the active center of the Hohi volcanic zone in central Kyushu, Japan. The volcano comprises 16 hornblende-andesite lava domes and associated lava flows, five pyroxene-andesite cones, and one olivine-basalt cone. The youngest magmatic activity erupted andesitic rocks about 1600 years ago (Kamata and Kobayashi, 1997). Occurrence of ash layers on top of scoria-rich deposits indicate that phreatic eruptions sometimes followed the magmatic activity (Itoh et al., 1996). Many geothermal manifestations (fumaroles, steaming grounds and hot springs) are present in the vicinity of the volcano (Ono, 1963; Ono et al., 1977; Ehara, 1994; Kamata, 1997). Intense fumarolic activity at Kuju- Iwoyama, on the northeastern flank of the Hosshozan andesitic lava dome, has been continuous for at least 500 years (Ehara et al., 1981; Imura and Kamata, 1996). Recent phreatic eruption on 11 October 1995 created new craters named a, b, c, d and e series (Sudo et al., 1998). Water flux from these craters decreased from 98,600 t/d in October 1995 (Hirabayashi et al., 1996) to 10,000 t/d in March 1998, whereas fumarolic temperature at b crater increased from 130 C in October 1995 to 260 C in March Long-term variations in chemical and isotopic compositions of the volcanic gases at Kuju- Iwoyama were observed from 1959 to 1984, and were attributed to the mixing of magmatic gas with meteoric water (Mizutani et al., 1986). Temporal changes in the isotopic compositions of volcanic condensates collected from a crater formed by the 1995 eruption also may have resulted from a similar mixing process (Hirabayashi et al., 1996). In addition, short-term fluctuations in the CO 2 /H 2 gas ratios were observed in a soil gas near the crater in 1996 (Kazahaya et al., 1996). These observations suggest that the volcanic gas discharge system at this volcano are unstable and therefore, fluctuations in the volcanic gas composition are expected to occur. GAS SAMPLING AND ANALYTICAL METHODS Gas sampling procedures Successive gas sampling was conducted twice at the Ohachi fumarolic field located in the southern part of the summit crater of Iwodake, Satsuma- Iwojima, on 2 November 1997 and 18 March 1998 (Table 1). The fumarolic gas was collected every 6 7 minutes during a period of 100 minutes on 18 March Although we did not record the sampling time on 2 November 1997, five gas samples were collected during one hour. Successive gas sampling was made at the Kuju-Iwoyama and Kuju b crater in October 1996, October 1997 and March 1998 (Tables 2 and 3). Gas sampling methods are essentially the same as those reported by Giggenbach and Goguel (1989). A Ti tubing was used to conduct the gases from fumaroles to sampling bottles at Kuju volcano, whereas a silica glass tubing was used for high-temperature gas sampling at Satsuma- Iwojima to prevent reaction of HCl with Ti at high temperature (>400 C) (Giggenbach and Matsuo, 1991). The gases were collected in Giggenbach s bottles which are 140 ml-evacuated glass bottles filled with 20 ml of 5N NaOH solution. The Satsuma-Iwojima high-temperature volcanic gas contains a high concentration of H 2 (Shinohara et al., 1993), which does not condense in the sampling bottle. As a result, less than 13 g of fumarolic gases could be collected with the 140 ml sampling bottle. Therefore, sampling bottles with a larger volume (330 ml) also were used at Satsuma- Iwojima in order to increase amount of sample mass. At Kuju volcano, volcanic condensates were collected with double water-cooled traps in series in order to determine the hydrogen and oxygen isotopic compositions of water. In 1997, sampling of gas and volcanic condensates were carried out alternately (Tables 2 and 3). Gas collection at Satsuma-Iwojima on 18 March 1998 was performed under maximum sam-

4 4 G. Saito et al. Table 1. Chemical and carbon isotopic compositions of volcanic gases from Satsuma-Iwojima volcano Chemical composition is given in units of (µmol/mol). S t : total S; n S : the average oxidation state of sulfur; n.a.: not analyzed; n.m.: not measured; n.d.: not detected. CV (%): Coefficient of variation (see text).

5 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes 5 Table 2. Chemical and carbon isotopic compositions of volcanic gases from Kuju volcano Chemical composition is given in units of (µmol/mol). S t : total S; n S : the average oxidation state of S; n.a.: not analyzed; CV (%): Coefficient of variation (see text). *Range (%) is the difference between the mean and the value devided by the mean.

6 6 G. Saito et al. Table 3. Hydrogen and oxygen isotopic compositions and chlorine contents of volcanic condensates from Kuju volcano No. Sampling start time (h:m) Collection time (min) Sampling* rate (g/min) δd SMOW of H 2 O ( ) δ 18 O SMOW of H 2 O ( ) Cl content (mg/l) Kuju b crater on 10 October 1996 (176 C) KCB1-1 11: KCB1-2 12: KCB1-3 13: KCB1-4 14: Mean Standard deviation Kuju b crater on 30 October 1997 (243 C) KCB2-1 10: KCB2-2 10: KCB2-3 11: KCB2-4 11: KCB2-5 11: KCB2-6 12: KCB2-7 12: KCB2-8 12: KCB2-9 13: KCB : Mean Standard deviation Kuju-Iwoyama on 30 October 1997 (351 C) KCI-1 12: KCI-2 12: KCI-3 12: KCI-4 13: KCI-5 13: KCI-6 14: KCI-7 15: Mean Standard deviation *Sampling rates were approximates due to rough estimates of the sample mass. pling rate conditions, i.e., the valve of the sampling bottle was fully opened each time, resulting in similar rates of sampling (8 11 g/min, Table 1). A similar technique was used on 2 November Maximum sampling rate was also intended at Kuju volcano in 1997 and 1998, leading to similar sampling rates for each series, except for KI-4 on 30 October 1997 (Table 2). In 1997, fast sampling of volcanic condensates was achieved using a hand-held vacuum pump. A similar sampling rate was obtained except for KCI-4 (Table 3). Variable sampling rates were applied in 1996, with the aim to detect the effects of sampling rate on the gas composition (Table 3). Gas analysis The methods used to analyze the volcanic gas sample were essentially the same as those reported by Giggenbach and Goguel (1989). Carbon dioxide was extracted from the oxidized alkali solution by sulfuric acid, and its content was measured by volumetry (Shinohara and Matsuo, 1986). The 13 C/ 12 C ratio of CO 2 was measured by mass spectrometry. The total S content in the alkali so-

7 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes 7 Table 4. Analytical errors estimated by repeated analyses of volcanic gas samples *Two analyses were carried out for each sample. **Range (%) is given by the difference between the two analyses for each sample devided by the mean. S t : total S. lution was determined as sulfate by ion chromatography following oxidization of the dissolved sulfur species by H 2 O 2. The average oxidation state of S (n S ) was obtained by back-titration with thiosulfate after oxidation of the alkaline solution with I 2. Assuming that SO 2 (n = +4) and H 2 S (n = 2) are the predominant S species occurring in fumarolic gases, n S corresponds to ( ) () n = 4X 2X / X 1 S SO2 H2S St where X SO2, X H S 2, X St are the molal concentrations of SO 2, H 2 S and total S in fumarolic gases, respectively (Giggenbach and Matsuo, 1991). In addition, the following expressions are derived X X X = + ( 2) St SO H S 2 2 N = 2X + 8X () 3 m SO H S 2 2 where N m is the sulfur normality of samples, which was calculated from the amount of I 2 necessary for oxidation of sulfur species in the sample to S(+6). The relation between N m and n S is n S = 6 N m /X St. (4) The concentrations of SO 2 and H 2 S are calculated based on n S and X St, ( ) ( ) XSO = 2+ ns X 2 St / 6 5 X = 4 n X / 6. 6 H S S St 2 ( ) ( ) The concentration of HCl was measured on the oxidized alkaline solution by colorimetry and that of HF by ion chromatography. Helium, H 2, N 2, CH 4 and CO were analyzed by gas chromatography with Ar as the carrier gas, a TCD detector, and a 5A Molecular Sieve separation column. A similar setting was used to detect Ar and N 2, but in this case H 2 was the carrier gas and a 5A Molecular Sieve combustion column heated at 300 C was used for the conversion of O 2 to H 2 O. Carbon monoxide concentration in gas phase in the sampling bottles decreases with time as it reacts with the strong alkali solution (Giggenbach and Matsuo, 1991). In order to obtain the initial CO concentration at the time of sampling, the gas phase in the bottles was analyzed for CO twice over a certain time interval, and the initial CO concentration was extrapolated back to the sampling date. The total amount of fumarolic H 2 O in the sampling bottle was calculated from the difference in weight measured between before and after sampling, and taking into account the contribution of the other components. Repeated analyses of CO 2, total S, HCl, He, H 2, N 2, Ar, CO and CH 4 were run on 2 samples of Satsuma-Iwojima and 9 samples of Kuju gases in order to evaluate the analytical errors (Table 4). The analyses were carried out twice on each sample having various chemical compositions, and the difference between the two analyses (the range), was calculated. In order to compare the ranges obtained from these analyses (Table 4), the range (R) associated with each sample was divided by the mean value (M) to give the R value (%) defined by:

8 8 G. Saito et al. R (%) = 100R/M. (7) In the following discussion, the error associated with the gas analysis is taken as the maximum value of R (%) obtained for each gas component (Table 4). The errors associated with the analyses of SO 2 and H 2 S depend on the precision of the total S concentration and average sulfur oxidation state determinations. The total S concentration and sulfur normality in a sample are defined to be X St and N m, respectively. The ratio of the measured value containing the analytical error to the true value is given as, α = X St /X St (8) β = N m /N m (9) where α and β are factors representing the analytical errors. Equations (4), (5), (6), (8) and (9) give the SO 2 and H 2 S concentrations of the sample according to: ( ) ( ) X = 8( α β ) / ( 2+ n )+ 1 X 10 SO2 S SO2 ( ) ( ) X = 2( β α )/( 4 n )+ β X. 11 H S S H S 2 2 These equations indicate that errors of SO 2 and H 2 S concentrations depend on the n S value at constant α and β. When H 2 S is dominant in the sample, our methods cause large error in the SO 2 determination because (2 + n S ) approaches 0. On the other hand, when SO 2 is dominant, error in the H 2 S concentration becomes large because (4 n S ) approaches 0. Repeated analyses of N m for Satsuma-Iwojima and Kuju fumarolic gas samples revealed that precision in the determination of N m is ±1.2%. Errors of SO 2 and H 2 S concentrations can be estimated from the analytical errors of N m and X St (Table 4) using the n S values of the samples. For example, sample S-1 (9373 µmol/mol, n S = 3.7) has indicates errors of 4.5% and 29% for SO 2 and H 2 S concentrations, respectively. In the case of Kuju-Iwoyama volcanic gas, the errors for sample KB-2 (1235 µmol/mol, n S = 1.32) are 49% for SO 2 and 3% for H 2 S and those of KI- 1 (2458 µmol/mol, n S = 1.53) are 71% for SO 2 and 3% for H 2 S. The analytical errors of SO 2 and H 2 S concentrations of each sample estimated above are shown in Figs. 1, 2 and 3. The analytical uncertainties associated with concentrations of gas components depend on not only chemical analysis of the gas and liquid phase in the bottle but also determination of weight of sample in the bottle. Precision in the weighing operation was ±0.01 g. Hence the expected error of weight of sample becomes, Error of weight of sample (%) = 0.01/W 100(12) where W is weight of sample in the bottle. In the case of the Satsuma-Iwojima gas samples, the error of weight of sample was expected to be 0.034% for samples with a mass of ~30 g collected using the large sampling bottles and 0.10% for samples with ~10 g collected using the small sampling bottles. The error can be expected to be 0.015% for samples with a mass of ~70 g from Kuju b crater in 1997 and Kuju-Iwoyama and 0.020% for samples with a mass of 50 g from Kuju b crater in These estimated errors ( %) were very small compared to the analytical errors estimated from repeated analyses of gas samples (1.1 27%; Table 4). Thus, we disregard the error of weight of sample in the following discussion. The precisions of the hydrogen, oxygen, and carbon isotopic determinations were ±1 (1σ), ±0.1 (1σ), and ±0.02 (1σ), respectively. Reproducibility in δ 13 C value by repeated analyses of one gas sample from Kuju-Iwoyama was within 0.1. RESULTS Chemical and isotopic compositions of Satsuma- Iwojima volcanic gases The Satsuma-Iwojima volcanic gases in 1997 and 1998 are characterized by high SO 2, HCl and H 2 and low H 2 S concentrations (Table 1), similar to previous analyses obtained in 1990 (Shinohara

9 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes H2O H2 Large-volume sampling bottles Small-volume sampling bottles 4200 CO Total S H2S Periodic? N2 Ar SO CH HCl CO HF δ 13 C of CO Fig. 1. Variations in the chemical and CO 2 isotopic compositions of Satsuma-Iwojima volcanic gases. Chemical composition is given in units of µmol/mol. The isotopic composition is given in. Open and closed symbols indicate samples collected on 2 November 1997 and 18 March 1998, respectively. Error bars represent the analytical uncertainties estimated from repeated analyses (see text). et al., 1993). The chemical and isotopic compositions of the gas collected on 2 November 1997 were slightly different from those of the gas sampled on 18 March In 1997, the gas showed slightly higher H 2 O and lower SO 2 and N 2 concentrations, and a higher carbon isotopic ratio of CO 2 than in 1998 (Table 1). The chemical and isotopic compositions corresponding to the smallvolume (140 ml) sampling bottles did not differ from those obtained with the large-volume (330 ml) sampling bottles (Table 1 and Fig. 1). This suggests that a difference in the sample mass ranging from 10 to 40 g did not affect the gas composition collected with the Giggenbach s bottles. Coefficients of variation (CV) were calculated in order to evaluate the variations in gas concentrations in each successive gas sampling (Table 1). The CV value is given by CV (%) = 100σ/χ, (13) where χ and σ are the arithmetic mean for samples in each series and the standard deviation, respectively (Miller and Miller, 1988). The CV val-

10 10 G. Saito et al H2O 250 H CO N Total S Ar H2S CH SO CO 200 HCl δ 13 C of CO Fig. 2. Variations in the chemical and CO 2 isotopic compositions of Kuju b crater volcanic gases. Chemical composition is given in units of µmol/mol. The isotopic composition is given in. Open and closed circles are samples collected on 29 October 1997 and 8 March 1998, respectively. Error bars represent analytical uncertainties estimated from repeated analyses of volcanic gases (see text). ues of total S, SO 2, H 2 S, H 2, CO and He concentrations in each successive gas sampling were within, or comparable to the analytical error (Table 1), while the CV values of CO 2, HCl, N 2, Ar and CH 4 concentrations exceeded the analytical errors. Variations in the CO 2 concentration in 1997 and 1998 and in the carbon isotopic composition in 1997 series were slightly larger than the analytical error (Fig. 1). Sample S-3 of the 1997 series had a higher CO 2 concentration and 13 C/ 12 C ratio of CO 2 than the other samples. In addition, fluctuations in the CO 2 concentration in 1998 seem to be periodic: samples S-6 and S-7 and samples S-13 and S-14 had slightly higher concentrations (Fig. 1). In contrast, the variations in Ar, N 2, and CH 4 concentrations were not periodic; Ar and N 2 increased suddenly in S-5 and CH 4 increased suddenly in S-3 (Fig. 1). The variations in the HCl concentration seems to be inversely correlated with that observed for H 2 O (Fig. 1): S-4 had the lowest HCl concentration and showed the highest H 2 O concentration. Similarly, S-16 had the highest HCl concentration and showed the lowest H 2 O concentration. Matsuo et al. (1974) collected volcanic gas samples (487 C) in a water-cooled alkaline solu-

11 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes H2O 90 H CO Total S H2S SO HCl N Ar CH CO δ 13 C of CO Fig. 3. Variations in the chemical and CO 2 isotopic compositions of Kuju-Iwoyama volcanic gases. Chemical composition is given in units of µmol/mol. The isotopic composition is given in. Error bars represent analytical uncertainties estimated from repeated analyses (see text). tion with a time interval of 10 minutes during 2 hours in 1967 at Satsuma-Iwojima. Their results did not indicate any periodic variations in the concentrations of acid gas components. The CV values associated with the CO 2 and H 2 S concentrations measured in this study were smaller than those obtained by Matsuo et al. (1974; 10 and 28% for CO 2 and H 2 S). On the other hand, the CV values calculated for HCl in this study were larger than those reported in the 1967 analyses (2%), while the CV values of SO 2 and HF concentrations were similar to those in 1967 (2 and 4%). Chemical and isotopic compositions of Kuju volcanic gases Compared to Satsuma-Iwojima gas, Kuju b crater and Kuju-Iwoyama gases were characterized by higher H 2 O and lower SO 2 and HCl contents (Table 2). The 1997 gas samples from Kuju b crater showed that the CV values for the gas concentrations and the standard deviation value associated to the δ 13 C composition of CO 2 were within the analytical errors. Notable exceptions occurred for HCl and CO concentrations. The gases collected in 1998 showed variations in CO 2, HCl, SO 2 and CO concentrations exceeding the analytical errors (Fig. 2). The gas samples form Kuju-Iwoyama behaved similarly to the 1997 Kuju b crater (Fig. 3). In this series, the sampling rate ranged from 11 (KI-2 and KI-3) to 20 g/min (KI- 4) (Table 2). The small variations observed in the chemical and carbon isotopic compositions of the Kuju-Iwoyama volcanic gases suggest that the sampling rate did not seriously affect the gas com-

12 12 G. Saito et al. Fig. 4. Hydrogen (D/H) and oxygen ( 18 O/ 16 O) isotopic ratios (a) and chlorine contents (b) of volcanic condensates from Kuju volcano. Closed symbols indicate samples collected at Kuju b crater on 10 October 1996, and the other symbols are as in Figs. 2 and 3. The error bar represents the standard deviation (1σ) for isotopic analyses and analytical uncertainties for Cl estimated from repeated analyses of the samples. A closed square in a small figure in (a) indicates D/H and 18 O/ 16 O ratios of magmatic water estimated from analyses of high temperature fumarolic gases from Kuju-Iwoyama (Mizutani et al., 1986), which are identical to those of high-temperature fumarolic gases in Japan (Kusakabe and Matsubaya, 1986) and andesitic water named by Giggenbach (1992). An open rectangle indicates D/H and 18 O/ 16 O ratios of local meteoric water (Hirabayashi et al., 1996; Saito et al., 1996). The mixing line between the magmatic and the meteoric water proposed by Hirabayashi et al. (1996) is also shown in (a). The broken lines with arrows in (a) and (b) indicate the shift in the isotopic and chemical compositions which are obtained by 11% addition of the meteoric end member (δd = 74 ( ) and δ 18 O = 11.6 ( ); Hirabayashi et al., 1996), assuming that the meteoric water has no chlorine. An arrow in (a) indicates the shift in the isotopic composition obtained by 18% loss of condensate formed at 100 C from the KCB1-2, which is obtained from mass balance calculations using the hydrogen and oxygen isotopic fractionation factors between vapor and liquid from Friedman and O Neil (1977). Arrows in (b) indicate changes in the oxygen isotopic composition and chlorine concentration which are obtained by 0.61%, 0.25% and 0.12% losses of condensate formed at 100 C from the KCB1-2, KCB2-6 and KCI-3, respectively.

13 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes 13 position. The D/H and 18 O/ 16 O ratios of the volcanic condensates from Kuju b crater and Kuju- Iwoyama were between those of the magmatic water of Kuju volcano (Mizutani et al., 1986) and those of the local meteoric water (Mizutani et al., 1986; Hirabayashi et al., 1996; Saito et al., 1996; Fig. 4). They lie along a mixing line between the magmatic water and the local meteoric water. The Kuju-Iwoyama gases had higher D/H and 18 O/ 16 O ratios than the Kuju b crater gases, possibly indicating a higher contribution of magmatic component. The volcanic condensates of Kuju b crater collected on 10 October 1996 exhibited variations in D/H and 18 O/ 16 O ratios larger than the 1σ analytical error (Fig. 4). They also displayed a large variation in the chlorine concentration (Fig. 4). In particular, KCB1-4, which was collected with the lowest sampling rate (1 ml/min; Table 3) had the lowest D/H and 18 O/ 16 O ratios and chlorine concentration. The 1997 volcanic condensates from Kuju b crater and Kuju-Iwoyama showed small variations in D/H and 18 O/ 16 O ratios within the analytical error (1σ), while the variation in the chlorine concentration exceeded it (Fig. 4). SOURCES OF VARIATIONS IN THE CHEMICAL AND ISOTOPIC COMPOSITIONS OF VOLCANIC GASES CO 2 concentrations in the Satsuma-Iwojima gases Variations in CO 2 concentration exceeding analytical errors were observed in both the 1997 and 1998 gas sample series from Satsuma-Iwojima (Fig. 1). In 1998, these variations appeared to be periodic: the CO 2 concentration was higher in S- 6 and S-7 and in S-13 and S-14, while it was lower in S-8 to S12 and S-15 to S-20 except for S-16. The sampling time interval between S-6 and S-7 and S-13 and S-14 is 43 minutes (Table 1). This corresponds to the time period measured between repetitive long-period seismic pulses of Satsuma- Iwojima (Ohminato and Ereditato, 1997). The source of the long-period seismic cycle may relate to replacement of the dense degassed magma by the ascending CO 2 -rich magma in the volcanic conduit (Kazahaya et al., 1997; Ohminato and Ereditato, 1997). Hence, the observed fluctuations in the CO 2 concentration may reflect magma convection processes in the conduit. HCl concentrations in the Satsuma-Iwojima and Kuju gases Large variations in HCl concentration were observed in all successive gas sampling series from Satsuma-Iwojima and Kuju volcanoes (Figs. 1, 2 and 3). However, the data indicated constant C/S ratios (Fig. 5). At Satsuma-Iwojima, the apparent inverse correlation between HCl and H 2 O (Fig. 6) suggests (1) gain or loss of HCl-rich condensate during the sampling procedures, and/or (2) short-term changes in the gas composition. Condensation within the sampling train leads to the formation of liquid highly enriched in soluble constituents. Since HCl is readily incorporated into the volcanic condensate, gain or loss of this liquid phase during sampling will result in an increase or decrease in the HCl gas concentration, but with minor changes in other gases (Giggenbach and Matsuo, 1991). Changes in the chemical composition of Satsuma-Iwojima and Kuju gases caused by gas condensation were calculated using the partial pressures of HCl and H 2 O in equilibrium with an aqueous solution of HCl (Perry, 1950), and assuming that condensation occurs at 100 C and 1 bar in the sampling tube. If the condensation in S-5 gas make S-4 gas and a liquid, the liquid coexisting with S-4 gas has HCl/ H 2 O ratio of according to the partial pressure data. The mass balance calculation results indicate that 3.0% loss of the liquid can decrease HCl concentration from 7156 µmol/mol to 5544 µmol/mol. This calculation also indicates that 3.0% addition of the liquid, which is in equilibrium with S-4, can change a HCl/H 2 O ratio of S-4 to that of S-5 gas. Thus, the difference between samples S-4 and S-5 can be explained by a 3.0%- gain or -loss of a HCl-rich condensate (Fig. 6). Similarly, a 3.6%-gain or -loss of a HCl-rich condensate accounts for the different HCl concentrations in S-16 and S-7 (Fig. 6).

14 14 G. Saito et al. CO2 Kuju b crater on 8 March 1998 Kuju b crater on 29 October CO2/S ratio CO2/HCl ratio 5 Kuju-Iwoyama on 30 October 1997 Total S Satsuma-Iwojima in 1997 and HCl/S ratio HCl Fig. 5. Relative CO 2, total S and HCl molal contents of Satsuma-Iwojima and Kuju volcanic gases collected in 1997 and Symbols are as Figs. 1, 2 and 3. Each broken line corresponds to a constant CO 2 /total S ratio. Similarly, the effect of gas condensation on the composition of Kuju gases was estimated using liquid-vapor partitioning coefficients for HCl at 100 C and for HCl concentration of less than 400 µmol/mol in vapor phase (Simonson and Palmer, 1993). The variations observed for the Kuju b crater and Kuju-Iwoyama series also can be explained by a 0.21%, 0.12% and 0.13% gain or loss of a HCl-rich condensate, respectively (Fig. 6). These results strongly suggest that the observed variations in HCl concentration at Satsuma-Iwojima and Kuju volcanoes arose from gas sampling artifacts. Alternatively, the variations in HCl concentration of the Kuju gases may reveal actual shortterm fluctuations. Indeed, the fact that HCl concentration of the Kuju b crater gases decreased between 1997 and 1998 (Fig. 5) suggests that short-term changes also may occur. Mizutani et al. (1986) postulated that HCl may be lost or gained during reaction of the ascending gas with a hydrothermal system. According to our calculations using partitioning constant of HCl between aqueous solution and vapor at 250 C (Simonson and Palmer, 1993; Fig. 6) and assuming that the condensation occurred at the fumarolic temperature (243 C), a 2.9% loss of HCl-rich condensate can account for the variation in Kuju b crater on 1997 (Fig. 6). On the other hand, the compositional variation of the 1997 Kuju-Iwoyama gas samples cannot be explained by this process, because at the measured gas temperature (351 C), the vapor is enriched in HCl than the liquid (Simonson and Palmer, 1993). Recently, Ohba et al. (2000) have suggested that interaction of the liquid with mineral in rock, which decreases H + concentration of the liquid and promote Cl distribution from the vapor to the liquid phase, can cause low HCl/H 2 O in the gas. Such reaction may explain the variation in the HCl concentrations of Kuju-Iwoyama gases. Another possibility is that HCl-rich liquid, formed by condensation of gases at lower temperature near the surface, mixes with the rising volcanic vapors. The above results indicate that addition of a small fraction of HCl-rich condensate would account for the observed variations. N 2 and Ar concentrations in the Satsuma-Iwojima and Kuju gases Sample S-5 had larger N 2 and Ar concentrations than the other samples in the corresponding series (Table 1 and Fig. 1). Its N 2 /Ar ratio of 78 is similar to that of air (83.6), suggesting air contamination. N 2 addition of 37 µmol/mol with the

15 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes 15 HCl content (µmol/mol) a) b) S % loss of condensate at 100 C S-7 S % loss of condensate at 100 C S-4 HCl content (µmol/mol) KI-6 KI % loss of condensate at 250 C 0.13 % loss of condensate at 100 C 0.12 % loss of condensate at 100 C KB-9 KB-12 KB % loss of condensate at 100 C KB H2O content (µmol/mol) H2O content (µmol/mol) Fig. 6. Variations in the HCl and H 2 O molal contents for Satsuma-Iwojima (a) and Kuju (b) gases. Symbols are as in Figs. 1, 2 and 3. Error bars represent the analytical errors estimated from repeated analyses (see text). Arrows indicate the compositional paths corresponding to loss of condensate formed at 100 C from each gas (see text). N 2 /Ar ratio of air or air-saturated water to S-4 yields the N 2 /Ar ratio and N 2 and Ar concentrations of S-5 (Fig. 7). The fumarolic field in the Iwodake crater may be highly porous due to occurrence of highly hydrothermally-altered rocks. This promotes air circulation and air mixing with the ascending volcanic gases. The relative N 2, Ar and He concentrations of the 1997 Kuju b crater gases and Kuju-Iwoyama gases lie along a mixing line between KB-12 and air and along that between KI-5 and air-saturated water, respectively (Fig. 7). Similar to Satsuma- Iwojima gases, these results suggest that air contamination occurred at these series. CH 4 concentrations in the Satsuma-Iwojima gases The Satsuma-Iwojima volcanic gases showed sporadic changes in CH 4 concentration during each gas sampling series (Fig. 1). Sato (1999) also reported a 40-fold increase in the CH 4 concentration of a high-temperature fumarolic gas (882 C) within a time period of 30 minutes. Based on the carbon isotopic composition of CH 4 and on the CH 4 /C 2 H 6 ratio, this author concluded that the CH 4 at Satsuma-Iwojima has a mixed thermogenic, biogenic and magmatic origin with the net contribution of the magmatic component being less than 10%. Methane-rich gases can be formed under the low temperature hydrothermal conditions surrounding the high-temperature magmatic system, and thus may be a source of CH 4 in fumarolic gas (Giggenbach, 1987). Because Satsuma-Iwojima gases are relatively poor in magmatic CH 4, which is in chemical equilibrium with the other gas components at high temperature, contamination by secondary CH 4 probably cause the large variations observed. CO concentrations in the Kuju gases Large CV values (16 26%) associated with the CO concentration were observed in the Kuju gas sample series (Table 2). It is suspected that the procedure used to restore the initial CO concentration in the gas phase introduced an additional error, that increased the uncertainty on CO determination. In particular, gas samples with a low CO concentration, such as those from Kuju, will be more affected. Isotopic composition of the volcanic condensates from Kuju volcano Large variations in the δd and δ 18 O values were observed in the KCB1 series (Fig. 4). Although these isotopic compositions lie along the

16 16 G. Saito et al. a) b) N Kuju b crater on 29 October 1997 Kuju b crater on 3 March S-20 Sastuma-Iwojima in 1997 and SW 250 N2/He ratio 2000 NE Air 50 ASW 25 Kuju-Iwoyama on 30 October 1997 N2/Ar ratio N2/Ar ratio Air S-4 ASW S He Ar/He ratio Ar N2 content (µmol/mol) Fig. 7. Relative N 2, He and Ar molal contents of Kuju gases (a) and relationship between N 2 /Ar ratios and N 2 contents of Satsuma-Iwojima gases (b). Symbols are as in Figs. 1, 2 and 3. Symbols labeled Air, ASW, NE and SW in (a) are the compositions of the air, the air-saturated water and the volcanic gas end members estimated for the volcanoes in northeast and southwest Japan (Kiyosu, 1986; Kita et al., 1993). Broken lines in (a) are mixing lines of KB-2 with air and KI-5 with air-saturated water (ASW). Error bars in (b) represent the analytical errors estimated from duplicate analyses. The arrows simulate the effects of air addition on the N 2 /Ar ratios of S-4 and S-20. Broken arrows simulate the effects of additions of N 2 and Ar from air-saturated water on the N 2 /Ar ratios of S-4 and S-20. mixing line between magmatic and the meteoric waters, the variations cannot be attributed to a change in the mixing ratio during the sampling period, because this could not account for the observed changes in the chlorine concentration. Alternatively, these gas samples may have been affected by the low sampling rate (Table 3), which may promote gas condensation in the sampling tubing. The fact that KCB1-4, which was collected at the lowest sampling rate, showed minimum δd and δ 18 O values, supports the condensation hypothesis. Mass balance calculations suggest that a 18% loss of the condensate from KCB1-2 result in isotopic ratios similar to those of KCB1-4, assuming that condensation occurs at 100 C and using the data on hydrogen and oxygen isotopic fractionation factors from Friedman and O Neil (1977). However, this process leads to a much lower chlorine concentration in the vapor phase than that measured in KCB1-4. Indeed, a large part of HCl will dissolve into the liquid phase (Simonson and Palmer, 1993), i.e., only 0.61% loss of condensate from KCB1-2 (162 mg/l) at 100 C is enough to make the low chlorine concentration of KCB1-4 (67 mg/l; Fig. 4). Thus, combination of the mixing and condensation processes discussed above can explain the observed variations in both δd and δ 18 O values and chlorine concentrations. Mass balance calculation indicates that 11% addition of the meteoric water to KCB1-2 and 0.5% loss of the condensate at 100 C from the mixed gas result in a gas with chemical and isotopic compositions of δd = 52.2, δ 18 O = 3.5 and Cl = 67 mg/l. These values are similar to KCB1-4. Similarly, 10% addition of meteoric water to KCB1-2 followed by 8.5% loss of the condensate at 250 C from the mixed gas produces chemical and isotopic compositions of δd = 51.7, δ 18 O = 3.55 and Cl = 67 mg/l close to those of KCB1-4. These results suggest that the variation observed in the KBC1 series reflect a change in the mixing ratio together with loss of condensate at depth, or during sampling.

17 Successive sampling of fumarolic gases at Satsuma-Iwojima and Kuju volcanoes CV (%) CO2 St H2S H2 N2 Ar CH4 CO SO2 HCl HF He Gas components Fig. 8. Summary of coefficients of variation (CV; see text) of analytical results of Satsuma-Iwojima, Kuju and Kilauea volcanoes. Cross symbols are minimum and maximum CV values reported for Kilauea volcanic gases at the sixth field workshop on volcanic gases (Shinohara, 1999; see text). The other symbols are as in Figs. 1, 2 and 3. Variations in the chlorine concentrations of the KCB2 and KCI series exceeded the analytical error, though their δd and δ 18 O ratios were within or similar to the analytical precision (Fig. 4). Mass balance calculations indicate that a 0.25% loss of condensate at 100 C from KCB2-6 and a 0.12% loss of condensate at 100 C from KCI-3 explain the observed variations. These results are consistent with the gas analyses of KB series and KI series (Fig. 6). PRECISION OF THE GAS SAMPLING AND ANALYTICAL TECHNIQUES Small concentration CV values comparable to the analytical errors were obtained for major gas components (CO 2, total S, SO 2, H 2 S, H 2 ), except for HCl (Tables 1 and 2). In addition, the δd and δ 18 O values of the volcanic condensates showed standard deviations of 0.9 and (Table 3), respectively, except for the 1996 Kuju b crater gases. These standard deviations compared to the precision (1σ) of the mass spectrometry analyses. Carbon isotopic analyses of CO 2 (Tables 1 and 2) had a standard deviation in the range of , which is similar to the variation obtained for repeated analyses of one volcanic gas sample (0.1 ). In contrast, large CV values were obtained for HCl and minor gas components (N 2, Ar, CH 4 and CO). Among them, the variation in the CO concentration is likely due to analytical problems, while the large HCl CV values may result from the sampling artifact, although the possibility of a change in the volcanic gas composition cannot be discarded. On the other hand, the variations in N 2, Ar and CH 4 concentrations of the Satsuma-Iwojima gases are explained by air contamination and mixing with hydrothermally-derived CH 4 during gas ascent. Assuming that the chemical compositions of volcanic gases are constant during each gas sampling series, the estimated CV values (Fig. 8) can be considered to be the precision of the gas sampling and analytical techniques, except for those of N 2, Ar and CH 4 concentrations of Satsuma-Iwojima gases. The CV values for concentrations of gas components obtained by a field workshop on volcanic gases held at Kilauea volcano (Shinohara, 1999; Fig. 8) are about ten times larger than those obtained in this study, excluding the Ar and CH 4 concentrations in the Satsuma-Iwojima gases and SO 2 concentration of SO 2 -poor Kuju gases. Although the CV values of the Kilauea gases are probably sufficient for a broad classification of volcanic gases, and meaningful thermodynamic interpretation (Giggenbach and Matsuo, 1991; Giggenbach et al., 2001), they may cause a significant problem when different groups work in monitoring chemical compositions of volcanic gases in cooperation. Therefore, further evaluation and improvement of gas sampling techniques should be urged by the CCVG. There are three potential causes for the large disagreement observed in the Kilauea gas; 1) analytical errors, 2) different sampling conditions, and 3) large fluctuation in volcanic gas composition. In order to

18 18 G. Saito et al. evaluate the importance of the first cause, an aliquot from the same gas sample, or from a standard solution should be analyzed by the different working teams. Giggenbach and Matsuo (1991) suggested that the second point account for most of the scatter observed at the workshops. These authors recommended rapid sampling of a large amount of gas, which reduces the risk of condensation within the sampling pipes. The small CV values determined in our study may be related to the use of a high and similar gas sampling rate (Tables 1 and 2). It appears that sampling conditions should be recorded at the workshop. Successive samplings of fumarolic gases carried out within a few hours at Satsuma-Iwojima and Kuju volcanoes indicate that: Variations in the concentrations of the major gas components (CO 2, total S, SO 2 and H 2 S) and in the hydrogen, oxygen (H 2 O) and carbon (CO 2 ) isotopic compositions were within, or comparable to analytical uncertainties. Only the CO 2 concentration at Satsuma-Iwojima and the D/H and 18 O/ 16 O ratios of H 2 O at Kuju volcano in 1996 departed from these behavior. In contrast, large variations in concentrations exceeding the analytical errors were observed for HCl and the minor gas components (N 2, Ar, and CH 4 at Satsuma- Iwojima and CO at Kuju volcano). The gas compositional and isotopic variations are attributed to (a) H 2 O condensation and subsequent absorption of HCl in sampling tubing, (b) reaction of CO with the alkali solution in the sampling bottle, (c) sporadic addition of atmospheric N 2 and Ar and of hydrothermally-derived CH 4, (d) a change in the mixing ratio between magmatic and meteoric waters, and (e) gain or loss of a HClrich liquid. According to our results, the precision associated with gas sampling and analysis are 1 3% for CO 2, H 2 S and SO 2, 6 11% for HCl, 1 6% for H 2, 4 6% for N 2, 5 9% for Ar, 2% for CH 4 and 7 27% for CO. As for the isotopic ratios, the precision are for δ 13 C of CO 2, and 0.9 and for δd and δ 18 O of the condensates, respectively. These results suggest that careful sampling and analysis of volcanic gases provide reliable chemical and isotopic compositions that can be used to monitor volcanic activities. Acknowledgments We thank the Nantou Opal company for permission to work in the crater area of Iwodake at Satsuma-Iwojima. We thank Dr. Nobuo Matsushima (National Institute of Advanced Industrial Science and Technology, Japan) for his assistance in the field. We thank Drs T. Ohba (Tokyo Institute of Technology, Japan), P. Delmelle (Catholic University of Louvain, Belgium) and K. Nagao (University of Tokyo, Japan) for helpful comments on the manuscript. CONCLUSIONS REFERENCES Allard, P., Maiorani, A., Tedesco, D., Cortecci, G. and Turi, B. (1991) Isotopic study of the origin of sulfur and carbon in Solfatara fumaroles, Campi Flegrei caldera. Jour. Volcanol. Geotherm. Res. 48, Ehara, S. (1994) Magmatic hydrothermal system developing just above a cooling magma A case study of Kuju volcano, central Kyushu, Japan. Mem. Geol. Soc. Japan 43, (in Japanese with English abstract). Ehara, S., Yuhara, K. and Noda, T. (1981) Hydrothermal system and the origin of the volcanic gas of Kuju- Iwoyama volcano, Japan, deduced from heat discharge, water discharge and volcanic gas emission data. Bull. Volcanol. Soc. Japan 26, (in Japanese with English abstract). Friedman, I. and O Neil, J. R. (1977) U.S. Geological Survey Professional Paper, 440-K. U.S. Government Printing Office, Washington, D.C. Gerlach, T. M. and Casadevall, T. J. (1986) Fumarole emissions at Mount St. Helens volcano, June 1980 to October 1981: Degassing of a magma-hydrothermal system. Jour. Volcanol. Geotherm. Res. 28, Giggenbach, W. F. (1987) Redox processes governing the chemistry of fumarolic gas discharges from White Island, New Zealand. Appl. Geochem. 2, Giggenbach, W. F. (1992) Isotopic shifts in waters from geothermal and volcanic systems along convergent plate boundaries and their origin. Earth Planet. Sci. Lett. 113, Giggenbach, W. F. and Goguel, R. L. (1989) Collection and analysis of geothermal and volcanic water and gas discharges. DSIR Chemistry, Rept. No Giggenbach, W. F. and Matsuo, S. (1991) Evaluation of results from second and third IAVCEI filed workshop on volcanic gases, Mt. Usu, Japan, and White