Time-averaged SO2 fluxes of subduction-zone volcanoes: Example of a 32-year exhaustive survey for Japanese volcanoes

Authors


Abstract

[1] All available SO2 flux data for 32 years (1975–2006) of Japanese volcanoes, accounting for about 10% of the world's arc volcanoes, were compiled to evaluate the temporal variation of the flux of each volcano and to estimate the time-averaged SO2 flux. The compiled data revealed that 6 volcanoes (Tokachi, Asama, Aso, Sakurajima, Satsuma-Iwojima, and Suwanosejima volcanoes) out of 17 significantly degassing volcanoes usually contributed more than 94% of the total flux. The time-averaged annual flux was 2.2 Tg a−1, which includes intense degassing of Miyakejima volcano after 2000, which raised the figure from 1.4 Tg a−1, indicating that a single huge emitter is capable of significantly skewing regional time-averaged degassing totals and indicating that the time-averaged flux assessments for infrequent huge emitters are important for accurate estimation. The regional SO2 flux distribution in cumulative frequency-flux plot does not obey a power law distribution. It shows a roll-off curve bending at about 500 t d−1, implying that it is misleading to assume the power law distribution for estimation of the global flux. Because the contribution of the major degassing volcanoes including the six volcanoes and additional sporadically degassing volcanoes during eruptive and posteruptive periods to the total flux is more than 95%, measurement of all large flux volcanoes can approximate the global flux.

1 Introduction

[2] Volcanic gas emission is the major process which advects volatiles from the Earth's interior to the atmosphere. On a geological timescale, this emission has contributed to the formation of the Earth's atmosphere and has strongly affected the Earth's environment. Consequently, determination of global volcanic gas flux is necessary to understand atmospheric evolution. In subduction zones, volatiles in the mantle are lost to the atmosphere through the volcanoes. At the same time, volatiles are injected into the mantle by subduction processes. Some of them are recycled to the atmosphere through magmatic processes. For investigation of such volatile recycling, determination of volatile output at individual arcs is regarded as an important measure [Hilton et al., 2002; Wallace, 2005].

[3] Many scientists have estimated global SO2 flux from volcanoes. The reported global SO2 flux ranges from a few Tg a−1 to 50 Tg a−1 [e.g., Stoiber et al., 1987; Andres and Kasgnoc, 1998]. These studies revealed that the SO2 flux of noneruptive continuous degassing is about an order of magnitude greater than that of explosive eruptions [Andres and Kasgnoc, 1998; Bluth et al., 1993; Shinohara, 2008]. The time-averaged SO2 emissions for explosive eruptions can be estimated from worldwide observations by Total Ozone Mapping Spectrometer [e.g., Krueger et al., 1995] and Ozone Monitoring Instrument (OMI) [e.g., Krotkov et al., 2006], together with the empirical relation between magnitude, frequency, and SO2 output of past explosions [Pyle et al., 1996]. Among the global continuous SO2-emitting subaerial volcanoes, emissions from subduction volcanoes account for nearly 95% of the total [Andres and Kasgnoc, 1998; Shinohara, 2008]. Based on the considerations presented above, more detailed and accurate estimation for continuously degassing volcanoes, especially for arc volcanoes, is important to ascertain the global SO2 flux of volcanoes.

[4] Previous studies have conducted SO2 flux measurement of continuously degassing volcanoes based on ground-based or airborne correlation spectrometer (COSPEC) measurements [Stoiber et al., 1983]. For estimating global SO2 fluxes of continuously degassing volcanoes, compilation of observed SO2 fluxes is necessary. An early estimate of global SO2 flux for continuously degassing volcanoes by Stoiber et al. [1987] was retrieved by questionnaires asking 35 volcanologists for data concerning about 100 volcanoes. They used empirical relations between the SO2 flux obtained by COSPEC and the visible plume size. Although the visible plume size might be a first-order estimate of the SO2 flux, it is strongly influenced by weather conditions such as temperature and humidity. Consequently, the compiled fluxes are only crude estimates. Andres and Kasgnoc [1998] compiled continuous degassing data for 49 volcanoes observed by the COSPEC and estimated the annual flux. The top five emitting volcanoes listed in their compilation yielded more than 50% of the global flux. Conversely, preclusion of a few volcanoes with high SO2 emission into the compilation might lead to a significant underestimation of the global flux [Shinohara, 2008]. In fact, at least three volcanoes continuously emitting more than a few thousand t d−1 of SO2 were reported since the compilation by Andres and Kasgnoc [1998]: Miyakejima, Japan [Kazahaya et al., 2004], Popocatépetl, Mexico [Witter et al., 2005], and Ambrym, Vanuatu [Bani et al., 2012]. Therefore, the global flux might have been influenced considerably and might have been underestimated by a few unmonitored large flux volcanoes in remote areas.

[5] To account for unmeasured volcanic fluxes, Brantley and Koepenick [1995] made an assumption that the distribution of volcanic SO2 and CO2 fluxes could be approximated by a power law function, which they used to calculate the global SO2 and CO2 fluxes. The power law assumption was recently used for estimation of SO2 flux for each arc [Hilton et al., 2002]. This approach using the power law distribution might have been useful considering the difficulties of measuring fluxes of all volcanoes. Nevertheless, it remains an open question whether the volcanic SO2 flux distribution obeys the power law distribution or not.

[6] Reassessment of arc-scale SO2 flux have been done recently for Central American volcanoes with ground-based measurements [Mather et al., 2006] and for Papua New Guinea volcanoes using OMI data [McCormick et al., 2012]. As described in this paper, we compiled all available SO2 flux data from volcanoes in Japan during 1975–2006 to quantify the total annual SO2 flux of Japanese volcanoes accurately. Based on other monitoring results such as plume height and sizes, we also estimated the variation of SO2 flux during the period when SO2 measurements were not conducted. Based on these, we estimated the time-averaged SO2 flux of Japanese volcanoes for 32 years during 1975–2006. We also discuss the problems underlying methods used for estimating the global SO2 flux.

2 Method

2.1 Volcanoes in Japan

[7] Of the 1406 subaerial volcanoes in the world listed as Holocene volcanism (< 10,000 years), 80% are related to subduction volcanoes (Smithsonian Institution, Global Volcanism Program: http://www.volcano.si.edu/index.cfm). In Japan, 93 volcanoes correspond to about 8% of the world's subduction-related Holocene volcanoes. Considering the eruptive activity occurring after 1900 (D1 and D2 time frames in the list of the Smithsonian Institution), 38 Japanese volcanoes are counted among the 320 world volcanoes, which correspond to about 12% of recent eruptive volcanoes. The total length of subducting margins or trenches is about 43,500 km [von Huene and Scholl, 1991]. The length of trenches related to Japanese volcanoes (Japan, Nankai, Izu-Bonin, and Ryukyu) adds up to 3950 km, corresponding about 9% of the total trench length. These values illustrate that Japanese volcanoes occupy about 10% of world subduction-related volcanism in both numerical terms and by trench length. The advantage of dealing with Japanese currently active volcanoes is that activities for these volcanoes have been well recorded at least for the last 30 years, except for underwater volcanoes. These records include SO2 flux, seismicity, crustal movement, plume height, and gas composition. This huge amount of information related to approximately 10% of subduction-related volcanoes in one area of the world is expected to be meaningful for estimating global subduction-related volcanism.

2.2 Sulfur Dioxide Flux Measurements in Japan

[8] Volcanic gas flux provides valuable information to evaluate the amount of magma engaged in volcanic activities. The volcanic gas flux is commonly quantified as SO2 flux because of its absorption cross section in UV wavelength range and its low concentration in the atmosphere, unlike other major volcanic gas components such as H2O and CO2. The history of quantitative SO2 flux measurements in the volcanological community started with the introduction of COSPEC in the early 1970s [Stoiber and Jepsen, 1973; Okita and Shimozuru, 1975]. From the 2000s, miniature UV spectrometry came to be used for SO2 flux measurements. It has now replaced COSPEC [Galle et al., 2002; Horton et al., 2006; Mori et al., 2007].

[9] In Japan, Moffat et al. [1972] and Okita and Shimozuru [1975] started SO2 flux measurements at Asama and Izu-Oshima volcanoes in the early 1970s. To the present day, SO2 fluxes are measured regularly at Sakurajima, Aso, and Asama volcanoes. During eruptive and posteruptive periods of many volcanoes including Usu (1977–1978), Izu-Oshima (1986), Unzen (1990–1995), and Miyakejima (2000) volcanoes, the SO2 fluxes were monitored at 11 volcanoes using COSPEC through 2003 (Table 1). Since 2003, the compact UV spectrometer system has been used to monitor the SO2 flux of volcanoes in Japan [Mori et al., 2007]. This system is a Japanese version of the mini-DOAS [Galle et al., 2002] or FLYSPEC [Horton et al., 2006]. Actually, the UV spectrometer-based systems are much lighter and more compact than COSPEC. Therefore, measurements have been carried out even at small fumaroles at the summit crater using a walking traverse method [McGonigle et al., 2003; Oppenheimer et al., 2004; Mori et al., 2006a]. We visited 18 volcanoes after 2003, of which data of eight volcanoes were newly measured using the new instrument.

Table 1. Sulfur Dioxide Flux Data of Japanese Volcanoes Compiled for 1975–2006
VolcanoNo. of Observation DaysMax Fluxa (t d−1)Min Fluxa (t d−1)Time-Averaged Fluxb (t d−1)Referencesd
  1. a

    Maximum or minimum daily flux values recorded during 1975–2006.

  2. b

    Time-averaged SO2 flux of six-monthly data for 1975–2006.

  3. c

    n.d.: not detected. Detection limit is 0.1–5 t d−1 depending on the distance from the plume.

  4. d

    a, This study; b, Mori et al. [2006a]; c, Hirabayashi et al. [2000]; d, Kyushu Univ. and Hokkaido Univ. [1979]; e, Ota et al. [1984b]; f, Ota and Matsuwo [1981]; g, Moffat et al. [1972]; h, Okita and Shimozuru [1975]; i, Kyushu Univ. and Univ. of Tokyo [1984]; j, Kyushu Univ. and Univ. of Tokyo [1986]; k, Ohwada et al. (Sulfur dioxide emissions related to volcanic activity at Asama volcano, Japan, submitted to Bulletin of Volcanology, 2013); l, Kyushu Univ. [1991]; m, Kazahaya et al. [2004]; n, http://www.seisvol.kishou.go.jp/tokyo/320_Miyakejima/320_SO2emission.htm; o, Kamada and Ota [1977]; p, Ehara et al. [1981]; q, Kyushu Univ. [1997]; r, Ota et al. [1978]; s, Ota et al. [1984a]; t, Hirabayashi et al. [1995]; u, Shinohara et al. [2008]; v, Kamada and Kubota [1975]; w, Kamada et al. [1980]; x, Kamada et al. [1982]; y, Kamada et al. [1986]; z, Kamada et al. [1988]; A, Ota and Shimizu [1989]; B, Hirabayashi et al. [1998]; C, Okita et al. [1977]; D, Kazahaya et al. [2002].

Meakan4122.66a, b
Atosanupuri1n.d.cn.d.0a
Tokachi2210140175a, b
Tarumae5244.615a, b
Kuttara1n.d.n.d.0a
Usu21169n.d8b, c, d, e
Iwate10.10.10a
Azuma1n.dn.d6f
Nasu1n.dn.d0f
Asama1414,60025360a, g, h, i, j, k
Hakone1n.d.n.d.0a
Izu-Oshima65203030h, l
Miyakejima280199,9001,1002,119m, n
Kuju16260327a, o, p, q
Aso1233,75019410a, r, s
Unzen192300.114t, u
Kirishima320n.d.8a, o, y
Sakurajima1075,4801101,641a, o, v, w, x, y, z, A, B
Satsuma-Iwojima161,760260574a, y, C, D
Kuchinoerabujima348431a, y
Nakanoshima140.4040a
Suwanosejima141,130300579a, y
Total   6,013 

2.3 Estimation of Total Flux

[10] We compiled all available SO2 flux data obtained during 1975–2006 to evaluate the variation of SO2 fluxes of respective volcanoes in Figure 1. Most SO2 flux data were obtained during persistent degassing activity. The SO2 flux was measured during effusive dome eruption of Unzen volcano during 1990–1995, but no data are available for periods of explosive eruptive activity. The frequency of the SO2 flux measurements has been quite varied. To compare SO2 fluxes of different volcanoes and different periods, an average SO2 flux for every 6 months is calculated and shown for 17 volcanoes significantly degassing SO2 (Figure 2 and Table A1). Only recently, frequent SO2 flux measurements have been performed at only a few volcanoes. Because measured data are not available for most periods of most volcanoes, we estimated the SO2 flux variation by simple interpolation or extrapolation of the measured data when marked changes of degassing activity have not been reported. When important changes of degassing activity are reported, the SO2 flux variation is estimated based on other data, such as plume height and size or based on reports about the degassing activity for each volcano [Japan Meteorological Agency (JMA), 2005]. Criteria of these flux estimates vary depending on available data and information. Detailed procedures and criteria are given in Appendix A. These procedures might introduce a considerable error in each estimate, but they are less likely to cause a significant error on the total flux estimate for each volcano because frequent measurements are commonly performed during large SO2 flux emission periods (Figure 2).

Figure 1.

Map of Japan showing locations of the 22 volcanoes presented in Table 1.

Figure 2.

Sulfur dioxide flux variation of all significantly degassing Japanese volcanoes, where diamonds show the 6 monthly averages of the measured data; the curve without the diamonds is estimated using interpolated and extrapolated variation with consideration of other information related to volcanic activity changes (see the text): (a) Hokkaido volcanoes, (b) Honshu and Izu Islands volcanoes, (c) Kyushu volcanoes, and (d) Ryukyu Islands volcanoes.

Table A1. Estimated 6 Monthly SO2 Flux (t d−1) of 17 Degassing Volcanoes in Japan for 1975-2006a
 MeakanTokachiTarumaeUsuAzumaAsamaIzu-OshimaMiyake jimaKujuAsoUnzenKirishimaSakurajimaSatsuma-IwojimaKuchinoerabujimaNakanoshimaSuwanosejima6 Monthly Total
  1. a

    Bold-italic numbers correspond to the 6 months which had flux observations during the period. The numbers correspond to 6 monthly average flux values of the period except for those of Meakan, Tokachi, Tarumae, and Kirishima volcanoes. For these volcanoes, see the text in Appendix A. Other numbers were estimated based on the observed data and other information (see text).

1975 1st5.51751500460002280018.51,1605700404703,016
1975 2nd5.51751500400002280018.51,3005100404703,032
1976 1st5.51751500330002280018.51,4504500404703,049
1976 2nd5.51751500270002280018.51,5903800404703,065
1977 1st5.51751500200002250018.51,6503200404702,974
1977 2nd5.51751512001400022110018.58602600405702,331
1978 1st5.517515120200110002380018.51,0702700405602,684
1978 2nd5.5175154520080002480018.52,0602700405503,565
1979 1st5.51751540050002580018.51,2902800405402,559
1979 2nd5.5175153702500261,000018.51,3502800403003,273
1980 1st5.517515330150002670018.51,3202900404702,615
1980 2nd5.517515300280002620018.57803000404702,163
1981 1st5.517515260400002620018.52503100404801,764
1981 2nd5.517515220530002620018.51,0603200404802,713
1982 1st5.517515180570002640018.51,8703300404903,592
1982 2nd5.517515140630002640018.51,4703300404903,262
1983 1st5.517515100450002640018.51,7003400405003,328
1983 2nd5.51751560270002640018.51,9403500405003,393
1984 1st5.51751520250002640018.52,1803600405103,613
1984 2nd5.517515002200026910018.52,3603700405104,649
1985 1st5.5175150027000261,180018.52,5403800405205,165
1985 2nd5.517515003100026200018.52,7303800405204,426
1986 1st5.517515001700026200018.52,6202900405304,193
1986 2nd5.517515003030026530018.52,5004000405304,313
1987 1st5.517515006030026410018.52,4104100405404,132
1987 2nd5.517515008030026300018.52,3004200405403,950
1988 1st5.51751500120350026190018.52,1704300405504,079
1988 2nd5.51751500160520026470018.52,0404400405504,459
1989 1st5.51751500210480026750018.51,8104400405604,520
1989 2nd5.517515002504300261,320001,5804500405604,865
1990 1st5.51751500300600261,140001,3504600405704,144
1990 2nd5.5175150035000261,470001,1204700405704,248
1991 1st5.51751500320002643014009004900405803,113
1991 2nd5.51751500290002643018006705100405802,918
1992 1st5.51751500270002643014004405300405902,654
1992 2nd5.5175150024000261,9509005305600405904,223
1993 1st5.5175150021000261,49013006205800406003,881
1993 2nd5.51751500180002641012007106000406002,886
1994 1st5.5175150016000264109001,0106200406103,155
1994 2nd5.517515001300026410001,3106400406103,356
1995 1st5.517515001000026410001,6001,2000406104,189
1995 2nd5.5175150012000186580001,9001,7600406205,402
1996 1st5.517515001300078450002,7101,1100406205,334
1996 2nd5.517515001500072310003,5104600406305,369
1997 1st5.517515001700044140003,3901,3700406305,984
1997 2nd5.5175150018000411,010003,2701,0200406406,404
1998 1st5.517515002000039140003,1506700406405,084
1998 2nd5.517515002200036140003,0303200406504,634
1999 1st5.517515002300033140002,9203900406504,604
1999 2nd5.517515002500030140002,8004600406604,574
2000 1st5.517515402700028140002,5704800406604,384
2000 2nd5.51751520280036,07025140002,34049004067040,256
2001 1st5.51751500300026,72022140002,11051004067030,713
2001 2nd5.51751500300016,10019140001,89052004068019,882
2002 1st5.51751500300012,42017140001,32058004069015,704
2002 2nd5.517515001,17007,250141400076064004069010,917
2003 1st5.517515001,37005,930111400087098004070010,240
2003 2nd5.5175150033007,00081,400009801,32004083012,093
2004 1st5.5175150025006,3606540005801,1700404909,624
2004 2nd5.517515002,44004,9803820005701,01004090010,967
2005 1st5.517515002,43003,7903610005809800403909,015
2005 2nd5.517515001,00004,29334700052095015409408,428
2006 1st5.5175150076002,35032800081095030406706,083
2006 2nd5.5175150018002,38036100079095046406705,858
Time-averaged61751586360302,119274101481,6415741405796,013

[11] The SO2 measurements are biased by intensely degassing volcanoes. Few measurements have been made of small fluxes. Fumarolic activities are reported for about 50 volcanoes in Japan [Geological Survey of Japan, 1998]. Most fumarolic areas discharge low-temperature gases with low SO2 content. The small fluxes of numerous volcanoes, however, can also contribute to the total flux of the region. To evaluate the contributions of small flux emissions, we conducted a survey to quantify the contributions of the small flux emissions. Among the 50 fumarole-bearing volcanoes, we selected seven volcanoes with intense fumarolic activity: Atosanupuri, Kuttara, Usu, Iwate, Hakone, Kirishima, and Nakanoshima. These volcanoes were selected based on the reported plume height information [JMA, 2005], previously reported chemical composition of fumarolic gases [Geological Survey of Japan, 1998], and our experience at the volcanoes.

3 Results

3.1 SO2 Flux Between 1975 and 2006

[12] Table 1 shows a list of volcanoes compiled for SO2 flux. The data were obtained using the traverse and scanning methods. They are listed with no discrimination. Using the compiled data, we estimated SO2 flux variations of each volcano for every 6 months during 1975–2006 based on the strategy explained in section 2.3. Detailed methods used for the estimation and some additional volcanic activity information are explained in Appendix A. Among the 22 volcanoes presented in Table 1, six (Atosanupuri, Kuttara, Iwate, Nasu, and Hakone) were not compiled in Figure 2 because the SO2 flux of these volcanoes was regarded as negligible during 1975–2006.

[13] Each of the reported SO2 flux data compiled in our data set has certain amount of uncertainty. Stoiber et al. [1983] discussed the errors in the COSPEC measurements by considering uncertainties in calibration cell concentration, record reading, distance determination, and wind speed determination, and derived a total uncertainty of 13–23% for the general case and 42% for the worst case. Detailed consideration for errors of the flux values obtained using the UV spectrometer-based systems have been made by various researchers (e.g., 30–40% [Mather et al., 2006]; 26%–54% (good condition and fair condition, respectively) [Galle et al., 2010]). From the time of COSPEC, determination of the plume velocity has been the major source of the uncertainty in the flux measurements. The respective flux data compiled are expected to have uncertainty of the same level as above. In addition, there is a considerable source of uncertainties related to atmospheric scattering [Millan, 1980; Mori et al., 2006b; Kern et al., 2009] and to dense aerosol in the plume [Kern et al., 2012], which might make uncertainties huge. Since the degree of the uncertainty depends on the observation method and conditions, reevaluation of the uncertainties for the entire data set is impossible. Moreover, SO2 fluxes reported as daily values are average of one to several tens of measurements which are observed during several hours of the day. Because the SO2 flux often varies considerably during a day, there is always concern whether the averaged value really represents the flux of the day. Considering above, it is not certain whether the uncertainties estimated by the error propagation are meaningful in our compiled data. Thus, we considered the reported averaged daily values as the representative value of the observation day and did not give uncertainties for the estimated values in this study.

3.2 Characteristics of SO2 Emissions From Japanese Volcanoes

[14] Annual SO2 fluxes for 1975–2006 from Japanese volcanoes were calculated using the data presented in Table A1 (Figure 3a). The average annual flux for 1975–2006 is 2.2 Tg a−1 with a maximum of 9.2 Tg a−1 in 2001 and a minimum of 0.8 Tg a−1 in 1981. The significant increase of the annual flux after 2000 is caused by the flux of Miyakejima volcano (Figure 2b). The average annual flux without the flux of Miyakejima volcano for 1975–2006 is 1.4 Tg a−1, with a maximum of 2.3 Tg a−1 in 1997 and a minimum of 0.8 Tg a−1 in 1981 (Figure 3b). The variation in Figure 3b is strongly dominated by variations of large emitters such as Sakurajima, Asama, and Aso volcanoes.

Figure 3.

(a) Annual total SO2 flux variation of Japanese volcanoes during 1975–2006. (b) Annual SO2 flux variation of Japanese volcanoes excluding the fluxes of Miyakejima volcano. (c) Percentage of SO2 fluxes of southwestern Japan (squares), Kanto and Izu Islands volcanoes (diamonds), and the six major continuously emitting volcanoes (circles).

[15] Sakurajima, Suwanosejima, Satsuma-Iwojima, Asama, Aso, and Tokachi volcanoes are the only volcanoes regarded as emitting more than 100 t d−1 for most of the time during the 32 years. In addition to the six volcanoes described above, Kuju, Izu-Oshima, Unzen, Usu, Azuma, and Miyakejima volcanoes emitted more than 100 t d−1 at least for a half-year period (Figure 2). These emissions are related to eruptive and posteruptive periods of the respective volcanoes. These volcanoes, excluding Kuju volcano, do not emit substantial amounts of SO2 during quiescent periods (Figure 2). Meakan, Tarumae, Kuju, and Nakanoshima volcanoes continuously emit SO2 of less than several tens of t d−1 (Figure 2). Their influence to the total flux is negligible.

[16] Along the volcanic arc of Japan, the SO2 emissions are not equally distributed. Four out of the six volcanoes with the large flux of more than 100 t d−1 of SO2 are in southwestern Japan, including Kyushu Island and Ryukyu Islands (Figure 1). Except for 1975, 1981, 1982, and after 2000, more than 80% of the flux was emitted from volcanoes in southwestern Japan (Figure 3c). The decreased percentages of the southwestern Japanese volcanoes reflect decreased SO2 emissions from Sakurajima volcano in the early 1980s and Miyakejima's increased emissions after 2000 (squares in Figure 3c). More than 94% of the flux is produced by the top six continuously emitting volcanoes (Tokachi, Asama, Aso, Sakurajima, Satsuma-Iwojima, and Suwanosejima volcanoes) for 18 years out of 32 years during 1975–2006. The years which fell below 94% were related to the eruptive and posteruptive periods of Usu, Azuma, Izu-Oshima, Unzen, and Miyakejima volcanoes. With the time-averaged consideration, the flux of the six volcanoes contributes 62% and 96% of the total flux including and excluding that of Miyakejima volcano, respectively (Table 1). Thus, we can consider that the SO2 flux of Japanese volcanoes can be quantified by measuring the fluxes of the six volcanoes with consideration of sporadic large fluxes during eruptive or posteruptive periods of other volcanoes.

4 Discussion

4.1 Comparison With Previous Estimates

[17] Andres and Kasgnoc [1998] compiled flux data of 49 continuously degassing volcanoes and 25 sporadically degassing volcanoes around the world to retrieve a time-averaged volcanic sulfur flux. They obtained a global SO2 flux of 9.66 Tg a−1, of which continuously emitting volcanoes account for 9.56 Tg a−1, and reported that the flux ratio of continuous to sporadic emissions is 100:1. Eight Japanese volcanoes, Asama, Sakurajima, Satsuma-Iwojima, Aso, Izu-Oshima, Kuju, Unzen, and Usu volcanoes, are included in their compilation with a sum of approximately 3500 t d−1 (1.28 Tg a−1). The two large emitters—Tokachi and Suwanosejima volcanoes—are not included. In contrast, Andres and Kasgnoc [1998] included the high SO2 flux data of posteruptive periods of Izu-Oshima, Kuju, Unzen, and Usu volcanoes; thereby, the fluxes are overestimated considering the time-averaged flux for these volcanoes. Although the total SO2 flux of Japanese volcanoes by Andres and Kasgnoc [1998] is similar to our time-averaged estimate of approximately 1.4 Tg a−1, this similarity is attributable to a coincidence in the balance between overestimation and underestimation.

[18] Using the compiled SO2 flux data of Andres and Kasgnoc [1998], Hilton et al. [2002] estimated SO2 fluxes of individual arcs by extrapolation of observed fluxes assuming a power law distribution as proposed by Brantley and Koepenick [1995]. The extrapolated flux estimations for the arc segment of Japan were 1.89 Tg a−1 and 1.49 Tg a−1. The two different values are attributable to the use of two different parameters for the slope of the power law distribution: one is a global value and the other is a calculated value for the Japanese arc [Hilton et al., 2002].

[19] Halmer et al. [2002] estimated the global SO2 flux by compiling a data set of explosive and quiescent degassing for 50 monitored volcanoes and then extrapolated to approximately 310 unmonitored volcanoes. The global SO2 flux was estimated as 6.0–8.4 Tg a−1 by quiescent degassing and 9.0–12.6 Tg a−1 by explosive degassing volcanoes and the flux from Japanese volcanoes as 1.9–2.7 Tg a−1 including both quiescent and explosive degassing volcanoes. The continuous SO2 flux of Japanese degassing volcanoes corresponds to 0.8–1.1 Tg a−1 if the flux ratio of the global explosive and quiescent degassing is the same as that of the Japanese volcanoes. They reported that their extrapolation method is based on the stage of activity, tectonic setting, and magma composition. However, they showed no original data or details of their extrapolation method. Therefore, we cannot discuss their estimates. Fujita et al. [1992] estimated the averaged SO2 flux of Japanese volcanoes as 1.1 Tg a−1 based on compiled COSPEC data obtained during 1976–1989 and estimates based on visual observations for three volcanoes in Hokkaido (Meakan, Tokachi, and Tarumae).

[20] The time-averaged SO2 flux of Japanese volcanoes for 32 years is estimated as 2.2 Tg a−1 in this study. This value includes the intensive degassing of Miyakejima, which raised the time-averaged flux by about 60% only by 7 years relative to the time-averaged flux without Miyakejima (1.4 Tg a−1). Hilton et al. [2007] pointed out that indiscriminate use of short-lived extraordinary high flux observed at actively degassing volcanoes might cause misleading of the time-averaged arc-scale flux estimation based on flux data of Anatahan volcano. To obtain a precise time-averaged flux, we must compile the data for a much longer period or to evaluate the frequency of intense degassing. Neither evaluation, however, is possible with the present data set, which remains as the future problem. For example, if the Miyakejima-scale degassing occurs once every thousand years and contributes 0.026 Tg a−1, the time-averaged flux is estimated as 1.4 Tg a−1.

4.2 Implications for Estimating the Global SO2 Budget From Volcanoes

[21] An accurate global SO2 flux of volcanoes should be based on a time-averaged compilation of SO2 flux data from all degassing volcanoes in the world. The flux data are biased to volcanoes in populated regions. Although the SO2 flux data from unmonitored volcanoes are increasing because of the ready portability of recently developed UV spectrometer-based systems [Oppenheimer, 2010], measuring all degassing volcanoes will be extremely difficult. Brantley and Koepenick [1995] recognized this limit and proposed that subaerial volcanic flux distribution for both CO2 and SO2 is a power law distribution, as in many geophysical phenomena [Turcotte, 1992]. The power law distribution of volcanic SO2 flux can be expressed as

display math(1)

where N represents the number of volcanoes with SO2 flux f or more, and a and c are constants, respectively, and global SO2 flux ftot is calculable if c < 1 [Brantley and Koepenick, 1995]. They used the data compiled by Stoiber et al. [1987] and evaluated the global SO2 flux as 2–3 × 10 11 mol/a (13–19 Tg a −1) with c value of approximately 0.8. Andres and Kasgnoc [1998] followed this assumption and retrieved 1.85 Tg a−1 in addition to the compiled data of 9.56 Tg a−1. Hilton et al. [2002] extended the power law approach to individual arcs using the data of Andres and Kasgnoc [1998] and estimated the global flux as 16.8–17.9 Tg a−1.

[22] We tested the assumption of the power law distribution based on the data set compiled in this study for Japanese volcanoes, which occupy about 10% of subduction-related subaerial volcanoes in the world. Because the present compilation covers almost all the Japanese volcanoes emitting SO2 at over 10 t d−1 for 32 years, the data set provides an accurate test of the power law distribution assumption. As represented by plots in Figure 4, all 32 cumulative frequency-flux plots of respective annual data for 32 years have common characteristics of being roll-off at around 100–800 t d −1. Moreover, they do not follow a single linear trend expected from the power law distribution. For the 32 years, the number of volcanoes simultaneously degassing SO2 has been 11–14. Therefore, it is difficult to constrain the flux at which roll-off occurs in the plots. To clarify the roll-off characteristics, a cumulative frequency-flux plot was shown using all the data in the present data set and using the data excluding the outlier flux of Miyakejima volcano (Figure 5). Neither shows a linear trend and rather shows a roll-off curve which bends at around 500 t d−1 (Figure 5).

Figure 4.

Cumulative frequency-flux plot showing the number (N) of volcanoes greater than or equal to the SO2 flux (f) shown against the flux on the log-log graph of the annual SO2 fluxes are shown for several selected years.

Figure 5.

Cumulative frequency-flux plot of all estimated 6 monthly data. Solid squares and circles, respectively, correspond to the data of all volcanoes and the data excluding Miyakejima volcano.

[23] The distributions of the SO2 flux data compiled by Stoiber et al. [1987] and by Andres and Kasgnoc [1998] also form a roll-off curve that bends around several hundred t d−1. Brantley and Koepenick [1995] pointed out that the distribution is linear only at the large SO2 flux range. They interpreted that the bending is caused by a lack of measurement of smaller emitters and concluded that the volcanic SO2 fluxes obey the power law distribution in nature. In contrast, the compiled data for Japanese volcanoes do not obey the power law distribution, and that fact is not attributable to the neglect of small emitters. Since the power law distribution is not valid in the regional scale, the deviation is likely to occur also in the global scale. Therefore, we conclude that the power law assumption should not be applied to the global flux estimation without supporting evidences of the power law distribution.

[24] We concluded that the power law distribution is inapplicable to estimate the global SO2 flux and that must be obtained by simple summation of the measured SO2 flux for all volcanoes. Total SO2 flux of Japanese volcanoes, however, is controlled by only several volcanoes. The contribution of major degassing volcanoes including the top six volcanoes and additional sporadically degassing volcanoes during eruptive and posteruptive period is always more than 95% (Figure 3 and Table A1). Therefore, a practical estimate of the global SO2 flux can probably be achieved if we can cover all the SO2 flux larger than a few hundred t d−1 in the world. Recent advances of satellite remote sensing techniques can enable quantification of the SO2 in the tropospheric plume by OMI [Krotkov et al., 2006; McCormick et al., 2012]. Detailed evaluation of OMI data can therefore be expected to help to cover all large SO2 flux volcanoes such as the assessment performed at the Vanuatu arc by Bani et al. [2012].

[25] The cumulative frequency-flux plot changes the slope at 200–800 t d −1 not only for the Japanese data set but also for other data sets obtained by Brantley and Koepenick [1995] and by Andres and Kasgnoc [1998]. The slope change of the distribution plot is likely to be controlled by the difference in the mechanism of the large flux and the small flux degassing, but discussion of the variation of degassing mechanisms remains beyond the scope of the present study.

5 Conclusions

[26] We compiled all available SO2 flux data for 1975–2006 for Japanese volcanoes, evaluated the temporal variation of the flux of each volcano, and estimated the time-averaged SO2 flux of Japanese volcanoes. The time-averaged annual flux was 2.2 Tg a−1 with a maximum of 9.2 Tg a−1 in 2001 and a minimum of 0.8 Tg a−1 in 1981. The intense degassing of Miyakejima volcano after 2000 drastically raised the time-averaged flux, which was 1.4 Tg a−1 before 2000, indicating that a single huge emitter is capable of significantly skewing regional time-averaged degassing totals. That fact is important for accurate estimation. In actuality, more than 94% of the total flux was contributed by the six volcanoes for most of the period: Tokachi, Asama, Aso, Sakurajima, Satsuma-Iwojima, and Suwanosejima volcanoes.

[27] The flux data of Japanese volcanoes, which account for about 10% of arc volcanoes in the world, demonstrated that the SO2 flux distribution in the cumulative frequency-flux plot does not obey a power law distribution, implying that global flux cannot be estimated with the power law assumption. The cumulative frequency-flux plot is represented not as a straight line but by a roll-off curve which bends most significantly at about 500 t d−1. Because the contribution of the major degassing volcanoes is more than 95% in Japan, we can estimate the global flux if all the large flux volcanoes are assessed properly.

Appendix A: Procedures for Estimating 6-Monthly Flux Data Using Compiled SO2 Flux Data

A1. Hokkaido Volcanoes

[28] SO2 fluxes were quantified for Tokachi, Meakan, Usu, and Tarumae volcanoes in Hokkaido (Figure 1). Because only two to four observed data sets exist for Meakan, Tokachi, and Tarumae volcanoes, we considered that the average of flux values for respective volcanoes represents the degassing activities of the volcanoes and that they were extrapolated for the past flux variations (Figure A1 and Table A1). During 1975–2006, small-scale eruptions occurred at Meakan (in 1988 and 1998), Tarumae (1978–1982), and Tokachi (1988–1989) [JMA, 2005]. During these activities, the plume heights for the volcanoes were 2–4 times higher than the recent heights for respective volcanoes. Therefore, the SO2 emissions from these volcanoes were likely higher than the recent values. Our estimation for these volcanoes might be slightly underestimated for the eruptive stages of these volcanoes.

Figure A1.

Compiled raw SO2 flux data (square) of the 17 Japanese volcanoes shown for 1975–2006. Error bars show maximum and minimum fluxes of the observation day. Solid curves show the estimated SO2 flux for every 6 months in Figure 2.

[29] For Usu volcano, several flux data were obtained during the eruptive periods of 1977–1978 and 2000, and a few data with negligible amounts during the quiescent periods. Based on the results of measurements taken during the quiescent periods, we estimated that the flux before the eruptive period 1977–1978 was negligible. Usu volcano started to emit 100 t d−1 with the onset of the 1977–1978 eruption, and it decreased exponentially in harmony with seismic activity [Ohta et al., 1988]. Tokachi volcano continuously emitted 200 t d−1 of SO2, which always accounts for most of the flux of this region.

A2. Honshu and Izu Islands Volcanoes

[30] Figure A1 shows the SO2 flux for Honshu and Izu Islands volcanoes. In Honshu, Asama is the only volcano that was continuously emitting SO2 throughout the data compilation period. The volcano emits about 100 t d−1 during the quiescent period but emits more than 500 to several thousand t d-1 during the active period. Izu-Oshima volcano emitted considerable amounts of SO2 during its eruption in 1986 and posteruptive period [Kazahaya et al., 1994]. The volcano emitted 345 t d−1 of SO2 in 1971 [Okita and Shimozuru, 1975]. Plume activity decreased rapidly after the eruption in 1974. It disappeared completely from 1976 [Tanaka, 1978]. Therefore, we regarded the flux as negligible from the beginning of 1975 until the eruption in 1986. The flux after 1990 is also regarded as negligible for the volcano (Table A1). The SO2 flux of Miyakejima volcano before the eruption in 2000 was also regarded as negligible. Miyakejima erupted in October 1983, but the duration of eruption was only 15 h. Posteruptive gas emissions were limited to fumarolic activities with little SO2 [Hirabayashi et al., 1984]. Consequently, the SO2 emissions during the 1983 eruption are also regarded as negligible. After the eruption in 2000, Miyakejima has continued extensive degassing of SO2 up to the present (Figure A1). Monthly average SO2 flux was recorded as more than 50,000 t d−1 in December 2000 [Kazahaya et al., 2004]. The sum of SO2 emissions from Miyakejima reached 23 Mt by the end of 2006, which is greater than the 17 Mt SO2 emission by the explosive eruption of Pinatubo.

[31] Azuma volcano in the northern part of Honshu had only one flux measurement with a negligible amount in 1977. The volcano activity increased suddenly in 2008 and started to emit over several hundred t d−1 of SO2 when the plume was continuous over 300 m above the crater for several months. Because the active period was after 2006, the flux related to the period is not included in the compilation. However, at Azuma volcano, the plume height exceeded 300 m for the first and second half of 1978 continuously [JMA, 2005]. We inferred that Azuma volcano had SO2 flux of 200 t d−1 for the two consequent half years in 1978 (Table A1). Nasu volcano also had one flux measurement with a negligible amount in 1977. The plume height of the volcano has been in decreasing trend since late 1960s [JMA, 2005]. Reported chemical composition had high SO2 until 1960s but significantly decreased by late 1970s [Geological Survey of Japan, 1998]. Considering above, we concluded that Nasu volcano had a negligible amount of SO2 emission during 1975–2006. Although small-scale phreatic eruptions occurred during 1975–2006 at several volcanoes in Honshu, they are not presented in Figure 2 and Table A1 because no marked increase of the plume emission was observed in association with these activities [JMA, 2005]. We did not include these effects in our compilation.

A3. Kyushu Volcanoes

[32] Kyushu Island has five significantly degassing volcanoes during the compilation period. Among them, Sakurajima volcano is the largest emitter (Figure A1). The volcano has been emitting an average of 1600 t d−1 between 1975 and 2006. The emission rate sometimes decreased below 500 t d−1 when the activity was low. The recent low emissions recorded for 2003–2006 correspond to a decreasing number of explosions [JMA, 2005]. The emitting center of the volcano was Minamidake crater until June 2006. After reactivation of Showa crater in 2006, the gas emissions correspond to the sum of the plumes from Minamidake and Showa craters.

[33] Aso volcano emits 100–200 t d −1 of SO2 during the quiescent period. Flux of >1000 t d−1 was observed when the volcano was more active and the water level of the crater lake was very low. Kuju volcano is a small emitter during the quiescent period with a few tens to a few t d−1 of SO2. The flux increased to approximately 100 t d−1 immediately after the 1995 eruption [Kyushu Univ., 1997]. Unzen volcano emitted little SO2 until the last eruption started in 1990. During the eruption, the SO2 flux varied in proportion to the effusion rate of lava [Hirabayashi et al., 1995]. The SO2 emission ceased quickly immediately after lava dome activity stopped in 1995.

[34] Kirishima volcanoes consist of several volcanic centers. Iwoyama volcano is the only one with reported SO2 flux values until the end of 2006. The volcano emitted approximately 20 t d−1 in the 1970s and 1980s. The thermal activity at the summit quickly ceased at the beginning of the 1990s. We inferred that the SO2 flux decreased to a negligible level with the decrease of the thermal activity. Because only two observed data sets exist for Iwoyama volcano before 1990s, we considered that the average of the two flux values represents the degassing activities of the volcano during high thermal activity period. Shinmoedake volcano erupted in August 2008 and emitted several tens of t d−1 of SO2 [Geshi et al., 2010] for some time after the eruption. After the 2011 eruption, the volcano emitted significantly large amounts of SO2 over 10,000 t d−1 [Mori and Kato, 2013]. Because these data are that the SO2 flux of Shinmoedake volcano is negligible for our compilation.

A4. Ryukyu Islands Volcanoes

[35] There are three continuously SO2-emitting volcanoes: Satsuma-Iwojima volcano, which has continuously emitted high-temperature volcanic gas for more than 1000 years [Shinohara et al., 1993], and Suwanosejima volcano, which is an extremely eruptive volcano in Japan. Both volcanoes have emitted 500–1000 t d −1 of SO2 for the last 30 years. Nakanoshima volcano has little information on volcanic gas, but had had continuous plume degassing, and a relatively large plume was reported in 1949 and 1973 [JMA, 2005]. Recent measurement of SO2 flux by the traverse method revealed a flux of 40 t d−1. We considered that this volcano has been continuously emitting SO2 with a flux of 40 t d−1 during 1975–2006. The SO2 flux of Kuchinoerabujima volcano increased after 2005 in response to increasing seismicity and inflation of the summit area (Figure A1).

Acknowledgments

[36] We are grateful to a number of colleagues for their support and assistance during the field works. We deeply appreciate D. Hilton, A. McGonigle, and an anonymous reviewer for their constructive reviews and the journal Editor, J. de Gouw, and an Associate Editor.

Ancillary