Relation between microlite textures and discharge rate during the 1991–1995 eruptions at Unzen, Japan

Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Contents lists available at ScienceDirect

Journal of Volcanology and Geothermal Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j vo l g e o r e s

Relation between microlite textures and discharge rate during the 1991–1995 eruptions at Unzen, Japan Satoshi Noguchi a,b,c,⁎, Atsushi Toramaru a, Setsuya Nakada b a b c

Department of Earth and Planetary Sciences, Graduate School of Sciences, Kyushu University, 6-10-1 Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan Volcano Research Center, Earthquake Research Institute, University of Tokyo, 1-1-1 Yayoi, Bunkyo, Tokyo 113-0032, Japan Technology and Research Center, Japan Oil, Gas and Metals National Corporation, 1-2-2, Hamada, Mihama-ku, Chiba-city, Chiba, 261-0025, Japan

A R T I C L E

I N F O

Article history: Accepted 24 March 2008 Available online 12 April 2008 Keywords: Unzen 1991–1995 eruption microlite textures discharge rate crystal size distribution degassing conduit dimension conduit flow

A B S T R A C T We conducted microlite textural analyses for the 1991–1995 eruptions at Unzen, Japan, in order to clarify the relationship between the discharge rate (exit velocity) and the kinetics of the microlite crystallization processes. The temporal variations in the plagioclase microlite crystallinity and the average size are negatively correlated with the variation in the magma discharge rate. On the other hand, the variation in the microlite number density (MND) exhibits a positive correlation with the discharge rate without a significant time lag. Groundmass microlites contain calcic plagioclase microlite (An45–65) and pargasite, suggesting that the microlite crystallization occurred at a depth (70–160 MPa) in the region of stability of the pargasite and plagioclase. The MND is determined at a depth (nucleation depth) with a certain effective undercooling rate (dT/dt) that is proportional to the exsolution rate of H2O from magma (dCH2O/dt). According to the MND water-exsolution rate meter, dCH2O/dt is calculated as 3.5 to 34.6 × 10− 6 wt.%/s at the final stage of the nucleation depth (zn: 70–100 MPa). By assuming that H2O exsolution occurs in equilibrium, the temporal variation in the decompression rate can be associated with the ascent velocity, provided the conduit flow is steady. The calculated ascent velocity of magma at zn (70–100 MPa; 0.8–7.6 cm/s) is higher than the exit velocity of magma (0.1 MPa; 0.2–3.9 mm/s). Using reasonable bulk densities, the difference in the velocities at the nucleation depth and surface results in the cross-section of the conduit (ellipse elongated dykes 2:5) at the nucleation depth (from 3 to 21 m) to be smaller than that at the surface (20 × 50–100 m) according to the mass conservation law along the conduit. The estimated temporal variation in the ascent velocity at zn (70– 100 MPa) is positively correlated with the observed discharge rate of magma, except at the final stage of eruptive activity. Furthermore, the variation in the conduit dimension at zn is correlated with the variation in the exit velocity, thereby implying that the observed discharge rate of magma is controlled by the ascent velocity at zn (70–100 MPa) and the conduit dimension. With regard to the final stage of eruptive activity, a decrease in the conduit dimension from 19 to 3 m indicates that the conduit is finally closed due to the termination of the magma supply. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Groundmass microlites are commonly observed in various types of eruption products, and they are considered to be associated with the effective undercooling (ΔT) that results from an increase in the melting point due to the H2O exsolution from the magma. The experimental studies that were recently conducted provide some insights into the kinetics of nucleation and the growth processes in decompression-induced microlite crystallization (Geschwind and Rutherford, 1995; Blundy and Cashman, 2001; Hammer and Rutherford, 2002; Couch, 2003; Couch et al., 2003a; Martel and Schmidt, ⁎ Corresponding author. Present address: Technology and Research Center, Japan Oil, Gas and Metals National Corporation, 1-2-2, Hamada, Mihama-ku, Chiba-city, Chiba, 261-0025, Japan. Tel.: +81 43 276 4317; fax: +81 43 276 4063. E-mail address: [email protected] (S. Noguchi). 0377-0273/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.jvolgeores.2008.03.025

2003). Furthermore, the textural and compositional analyses of microlites in the volcanic products are considered to be beneficial for understanding the dynamics of volcanic eruption (Cashman and Marsh, 1988; Cashman, 1992; Klug and Cashman, 1994; Hammer et al., 1999, 2000; Noguchi et al., 2006). H2O exsolution and microlite crystallization are time-dependent processes and they can significantly alter the physical properties (i.e., bulk density and viscosity) and the physical condition (temperature-pressure path) of magma. In fact, it is suggested that during dome eruptions, the coupling between crystallization (Melnik and Sparks, 1999, 2005) and vesiculation (Wylie et al., 1999) controls the dynamics of volcanic eruptions. The dacite lava eruptions during 1991–1995 at Unzen, Japan, resulted in the development and subsequent collapse of a series of lava domes, thus generating pyroclastic flows (block-and-ash flow). Many geophysical and geological observations during the dome eruptions have been documented in detail (e.g., Nakada and Fujii,

142

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

1993; Kusakabe et al., 1999; Nakada et al., 1999; Nishi et al., 1999; Yamashina and Shimizu, 1999). Although the petrological aspects of Mt. Unzen have been extensively studied (e.g., Nakada et al., 1995; Nakamura, 1995; Nakada and Motomura, 1999; Sato et al., 1999), several aspects of the dynamics of the dome evolution within a conduit, such as vesiculation, degassing, and the microlite crystallization processes, are still poorly understood. In particular, the quantitative description of the vesicle and microlite textures preserved in the eruption products during the 1991–1995 eruptions and their relation to a geophysical observation (i.e., discharge rate) have not been carried out. Therefore, in the present study, we conducted a detailed textural and compositional analysis of the volcanic products in order to clarify the relationship between the temporal change in the discharge rate (exit velocity) and the variation in the microlite textures during the 1991–1995 eruptions. In this paper, we present the results of the quantitative textural and compositional analyses and provide some constraints of the conduit flow process by using a microlite number density (MND)exsolution rate meter developed by Toramaru et al. (2008-this issue). First, we present the methods and results of the quantitative textural and compositional analyses and correlate the resultant textural data with the fluctuations in the discharge rate. Second, we discuss the microlite nucleation depth (zn) from the viewpoint of the compositional data of plagioclase microlites. Third, we assess the ascent velocity of magma at zn using the MND water-exsolution rate meter developed by Toramaru et al. (2008-this issue) and consider the conduit dimension and timing of vesiculation, degassing, and microlite crystallization processes. Finally, we discuss the conduit flow process by proposing a simple conduit flow model based on these observations. 2. Background of the 1991–1995 eruptions at Unzen Since the appearance of the first dome on May 20, 1991, the blockand-ash flow of Unzen 1991–1995 eruptions repeatedly occurred more than 10,000 times owing to dome collapses. Prior to the dome growth, phreatic and phreatomagmatic eruptions occurred (early May 1991). In the early stages of the dome growth, small explosive events such as pelean (June 8, 1991) and vulcanian (June 11, 1991) eruptions occurred. However, from the mid-1991 eruption to the end of the eruptive activity (May 1995 eruption), the eruptions continuously produced a series of lava domes that intermittently collapsed to produce block-and-ash flow (Nakada et al., 1999). Most of the samples were poorly vesiculated (bulk vesicularities 10–30 vol.%) except for the eruptions on June 8 and 11, 1991 (Kueppers et al., 2005). The rock formed during the 1991–1995 eruptions at Unzen is dacite with similar bulk SiO2 content (from 64.5 to 66 wt.%) throughout the eruptive activity (Nakada and Motomura, 1999). The phenocrysts ranged from 20 to 30 vol.% with plagioclase, amphibole, biotite, quartz, titanomagnetite, ilmenite, and the rare orthopyroxene and clinopyroxene. Several disequilibrium features (e.g., reverse zoning of plagioclase, hornblende, and magnetite; partial dissolution of quartz; and pargasite reaction rims of biotite) suggest that the magma mixing between phenocryst-rich rhyodacite and aphyric mafic magma occurred just before the eruption (Nakada and Fujii, 1993; Nakamura, 1995; Nakada and Motomura, 1999; Sato et al., 1999; Venezky and Rutherford, 1999). Nakamura (1995) calculated the time scale of magma mixing by using the zoning profiles of magnetite phenocrysts. The typical diffusion time was estimated to be a few months – from May 1991 to May 1993 – irrespective of the effused sequence. Other mixing signatures such as the disequilibrium phenocryst and unstable biotite that coexist with the post-mixing magma during effusion, also suggest that the magma mixing (mafic magma was injected into felsic magma) occurred throughout the entire eruption period. Most of the phenocryst minerals belonged to low-temperature magma, and the equilibration temperature ranges from 780 to 800 °C in the core

compositions of iron-titanium oxide phenocrysts (Venezky and Rutherford, 1999). On the other hand, the temperature calculated from an analysis of the rims of orthopyroxene and clinopyroxene ranges from 1030 to 1130 °C (1055 ± 75 °C; Holtz et al., 2005). The mafic enclaves in Unzen 1991–1995 eruption products exhibit a wide range of bulk composition (basaltic to andesitic; Nakada and Motomura, 1999); however, the average temperature was 1050 ± 26 °C, which is close to that of the high-temperature magma. In contrast, the groundmass of the orthopyroxene and clinopyroxene thermometer determined by Holtz et al. (2005) suggested that the temperature of the post-mixing magma is 930 °C; this is in good agreement with the estimation by oxide thermometry (900 ± 30 °C) by Venezky and Rutherford (1999). Holtz et al. (2005) also carried out an experimental study to determine the phase relations and pre-eruptive conditions of the dacite formed in 1991–1995 at Unzen. They suggested that 35 wt.% of the high-temperature magma is an almost aphyric pyroxene bearing andesitic magma (62–64 wt.% SiO2 at 1055 °C ± 75 °C; 4 ± 1 wt.% H2O in the melt), whereas 65 wt.% of lowtemperature magma contains a high fraction of phenocrysts (N35 wt.%) with a melt of 75 wt.% SiO2 below 775 °C. A glass inclusion analysis also suggests that the H2O contents of the low-temperature magma could be as high as 8 wt.% (saturation pressure: 300 MPa), while that of post-mixing magma is estimated to be 6 ± 1 wt.% (saturation pressure: 190 ± 30 MPa) by a simple mass balance calculation (mixing of 50% rhyolitic melt with 50% andesitic melt). This value is similar to the value obtained by the melt inclusion analysis carried out by Nishimura et al. (2005), and it ranges from 5.1 to 7.2 wt.%. Sato et al. (2005) have studied the chlorine-hydroxyl exchange partition coefficient between hornblende and melt and suggested that the variation in the Cl/OH ratio in the coexisting melt is caused by the 1.2– 1.8 wt.% degassing of H2O from the magma, thereby implying the occurrence of repeated influx and convective degassing of the fluid phase in the magma chamber. GPS measurements detected the ground deformation caused by magma intrusion and lava discharge during the 1991–1995 eruption. Ishihara (1993) and Nishi et al. (1999) estimated the pressure source beneath Unzen volcano, which is located at depths of 11 km (290 MPa) and 7–8 km (210 MPa), using the inflation and deflation of Mogi's source model (a point source model). This calculation is in good agreement with the above mentioned arguments (Holtz et al., 2005; Sato et al., 2005). They also suggested that the pressure source shifts from beneath Chijiwa bay (west) to beneath the summit of the lava dome (east) implying that the magma ascends through the inclined conduit. The location of the hypocenter of the seismic activity also confirmed that the volcano-tectonic (VT) earthquakes migrated eastward to the summit and became shallower, similar to the path of the pressure source (Nakada et al., 1999; Umakoshi et al., 2001). 3. Analytical method 3.1. Samples In this study, we used 59 samples from 21 events that occurred during the 1991–1995 eruptions at Unzen. Most of the samples are the same as those used by Nakada and Motomura (1999) except for the 1995 eruption (lava spine). These samples were directly taken from hot blocks in the block-and-ash flow deposit within few days of the deposition. Some samples are scarce (b20–30 g) and cannot be obtained from the present outcrops. The lava spine was sampled at the summit of Unzen lava dome. We fabricated thin sections with polished surfaces from all the samples. All the samples are listed in Table 1. 3.2. Textural analysis The quantitative measurement of the textural characteristics (groundmass crystallinity, average microlite size, MND, and crystal

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

143

Table 1 Result of the textural analysis Sample

Clasts No. of microlites Measured area PI length Na

1991.0527 1991.0608 1991.0611 1991.0915 1992.0108 1992.0311 1992.0426 1992.0601 1992.0813 1992.0928 1992.1229 1993.0214 1993.0309 1993.0401 1993.0429 1993.0501 1993.0624 1993.0703 1994.0204 1994.0821 1995.Spine

4 4 5 2 2 2 4 3 2 3 1 1 2 2 1 2 8 1 2 4 4

2353 1993 1355 849 1245 1318 1489 2256 1238 1862 313 690 1375 1215 892 1207 5324 948 1343 2874 3766

Nv log

(mm2)

(mm)

(/mm2) (/mm3)

0.53 0.38 0.24 0.21 0.20 0.32 0.40 0.53 0.19 0.43 0.10 0.12 0.29 0.21 0.11 0.23 0.85 0.16 0.20 0.44 0.46

0.014 0.007 0.013 0.009 0.006 0.030 0.037 0.043 0.042 0.026 0.064 0.013 0.015 0.018 0.027 0.012 0.005 0.009 0.027 0.011 0.009

4595 5287 5584 4100 6160 4114 3691 4218 6403 4344 3261 5815 4776 5744 8196 5206 6249 5890 6791 6499 8239

n0a

1/R0(−)b R2(EXP)c 2nd area of PI microlites

5.62 (0.14) 18.5 137 5.84 (0.17) 19.2 230 5.77 (0.27) 18.8 147 5.68 (0.20) 19.2 259 5.24 (0.45) 17.8 71 5.29 (0.06) 17.2 68 5.12 (0.09) 17.8 49 5.20 (0.08) 17.2 61 5.34 (0.19) 17.1 49 5.20 (0.05) 17.3 69 4.87 15.5 51 5.59 18.7 141 5.55 (0.11) 18.5 130 5.22 (0.01) 18.5 113 5.63 18.0 73 5.62 (0.18) 19.2 176 6.07 (0.15) 20.2 321 5.87 19.7 220 5.54 (0.03) 17.6 66 5.79 (0.09) 19.7 202 5.99 (0.14) 20.0 229

0.99 0.99 0.99 0.97 0.98 0.98 0.88 0.99 0.98 0.99 0.94 0.98 0.99 0.99 0.98 0.99 0.98 0.96 0.99 0.99 0.99

V⁎: 3D PI crystallinity

(vol.%)

(vol.%)

16.6 (0.8) 17.8 (4.5) 15.2 (1.0) 16.9 (0.8) 20.8 (0.6) 16.4 (1.0) 18.9 (1.5) 15.6 (0.9) 20.0 (0.1) 16.5 (0.6) 17.8 29.7 22.8 (0.2) 24.1 (0.5) 21.7 23.6 (1.8) 19.9 (4.1) 18.3 19.6 (1.3) 23.4 (0.6) 22.1 (0.6)

15.4 (3.8) 7.1 16.0 (2.3) 11.6 15.9 (5.6) 12.3 18.4 (0.7) 10.2 19.7 (2.7) 19.7 19.9 (0.9) 19.9 18.8 (1.6) 18.8 18.2 (2.3) 18.2 19.5 (2.9) 19.5 22.0 (1.7) 22.0 21.0 21.0 24.7 24.7 23.4 (2.7) 12.8 20.3 (2.5) 18.0 24.4 14.3 26.1 (1.1) 15.0 22.6 (2.3) 9.2 17.6 8.6 27.5 (2.0) 27.5 26.2 (6.7) 16.5 26.1 (2.9) 17.6

(vol.%)

Note: Numbers in parentheses show the standard deviation (1σ). V∞ and V⁎ are the 3D crystallinity, which integrated the size from 0 to 0.1 mm and from 0 to R⁎, respectively. a intercept of CSD curve. b slope of CSD curve. c correlation coefficient.

size distribution (CSD)) of the plagioclase microlites (length: b50 µm, area: 1000 µm2) was performed on the basis of the binary images obtained by a microscope (×500 magnification) with a reflected light source. We obtained digital microphotographs that were stored in a computer. Then, we manually traced the outline of each microlite within the groundmass (the bubble-free region) using an application software (Canvas). The number, shape, size, and volume of the microlites were measured by the NIH Image program (freeware from National Institute of Health http://rsb.info.nih.gov/nih-image/index. html). We then converted the number density from two dimensions to three dimensions. For the calculation of 3D CSD, we used the CSD correction method proposed by Higgins (2000, 2002). The MND N is obtained by the integration of the 3D CSD: nðRÞ ¼ n0 eR=R0 ;

ð1Þ

where n (R) is the population density, which is obtained by dividing the 3D number density by the size interval. Z N¼

R⁎ 0

nðRÞdR;

ð2Þ

where R⁎ is the upper limit of the integration. In our samples, the CSD plot is not linear (not exponential distribution) at upper ranges of R. We defined the maximum length (R⁎) as the intercept on the size axis for the exponential law with the highest value of the correlation coefficient (R2 exponential law) using a regression line, as shown in Table 1. The MND calculated according to the exponential law up to R⁎ is similar to the total MND up to the maximum size because the MND is mainly controlled by microlites of smaller size. Consequently, we regard the total MND as the integration of the straight CSD line from 0 to the upper limit of the exponential law. The average microlite length was also calculated using the reciprocal of the CSD slope. In addition, the volumetric proportion of plagioclase (Vi) can be calculated by integrating the volume of all the crystals as determined by Higgins (2002): Z Vi ¼ r

0

l

nðRÞR3 dR;

ð3Þ

where σ is the shape factor of plagioclase; the shape factor can be further expressed in a more practical form as follows: h  pi r¼ 1x 1 IS=L2 ; 6

ð4Þ

where ω, I, S, and L are the roundness factor determined by Higgins (2000), the real intermediate dimension (3D intermediate length), the real shortest dimension (3D width), and the real longest dimension (3D length), respectively. For the parallelepiped object of our sample, we determined the roundness factor as 0. Further, in this study, we also chemically confirmed the maximum size of the plagioclase microlites due to decompression (see Section 4.2 Compositional analyses). As a result, we found that the transition between the phenocryst and microlite size is between 0.1 and 0.3 mm in length. Consequently, in order to calculate the 3D groundmass crystallinity, we calculated the integration from the size of 0 to a maximum size of 0.1 mm (minimum transition length: V∞). We also estimated the 3D volume at the dominant MND, which can be integrated from 0 to R⁎ (VR⁎) of the microlites of small size (i.e., VR⁎ ∼ 3D volume of the microlites of smaller size) (see Eq. (2) and below). Thus, these values are relatively smaller than that of V∞ (see Section 4.1 Microlite textural analysis). These data are listed in Table 1. 3.3. Compositional analyses The chemical compositions are analyzed using a JEOL 8800 electron probe micro-analyzer (EPMA) at the Department of Earth Sciences, Kanazawa University, and a scanning electron microscope (SEM; JEOL JSM-5800LV) at the Department of Earth and Planetary Sciences, Kyushu University. For the EPMA analysis, we used a focused beam (3 µm) with an accelerating voltage of 15 KV and a beam current of 10 nA utilizing a ZAF X-ray intensity reduction routine. The SEM was performed using an accelerating voltage of 20 KV, a beam current of 0.5 nA and a count time of 100 s. The groundmass H2O analysis was performed at the Department of Earth and Planetary Sciences, Kyushu University, using a Karl Fischer Titration VA-21 type manufactured by the Mitsubishi Chemical Company. First, we powdered the samples to a size of 1 mm. Next, we hand-picked a groundmass (including microlites and groundmass

144

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

glasses) of approximately 1 g in order to eliminate the phenocrysts. The samples were heated to 150 °C for 1 day for the evaporation of the absorbed H2O. Finally, we heated the samples to a temperature of

1000 °C for 1.5 h in order to calculate the amount of dissolved H2O. For each sample, we carried out the measurements three times in order to ensure reproducibility.

Fig. 1. Images of the representative microlite textures in the eruption products from the 21 events during the 1991–1995 eruptions of Unzen volcano. These images are transformed from photomicrographs under reflected light. The microlite texture varies temporally with the eruptive activity. The plagioclase microlites are shown in black; the bubbles and other microlites, in gray; and the groundmass glasses, in white.

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

4. Results 4.1. Microlite textural analysis The groundmass of the 1991–1995 eruption products contains plagioclase, pargasite, orthopyroxene, clinopyroxene, Fe–Ti oxide, apatite, and glass; further, its textures temporarily vary with the eruptive activity (Fig. 1). The eruption products in early 1991 contain abundant microlites of small sizes; however, until late 1992, the total number of microlites decreases with an increase in the microlite size. In the eruption products from 1993 to early 1995, the number of microlites again increases with a decrease in the microlite size. The average plagioclase length increases from early 1991 to late 1992, and attains its maximum value in late 1992; further it decreases from 1993 to early 1995 (Fig. 2a). The microlite crystallinities determined by the thin section (2D) and the CSD (3D) slightly increase with a change in the eruptive activity except for the eruption products in early 1993 (Fig. 2(b)). On the contrary, the MND decreases from early 1991 to late

145

1992 and reaches its maximum value in mid-1993. The MNDs gradually increase from 1993 to early 1995. The relationship between the MND and microlite length of the eruption products exhibits a negative correlation (Fig. 3a), while the 2D microlite crystallinity does not correlate with the MND and average microlite length (Fig. 3b and c). On the other hand, the variation in VR⁎ (3D volume of the microlites of smaller size) is roughly correlated with the MND and microlite length, implying that VR⁎ appears to increase with the population of a relatively larger microlite size. Fig. 3d also indicates the correlation between the 2D crystallinity determined by the thin section and the 3D crystallinity calculated from the CSD. Both the crystallinities are roughly linearly correlated (broken line as shown in Fig. 3d); however, the 3D crystallinity tends to be larger than the 2D crystallinity. The differences between the 2D and 3D crystallinities may result from the fact that some assumptions with regard to the bin size for the particle length are made while determining the CSDs (e.g., 3D particle length can be obtained from the crystal shape provided that all the particles have the same aspect ratio (e.g., 1:2:5), which is determined from the average aspect ratio of the 2D particle size). The slope of the CSD curve (−1/R0) for the early 1991 eruption becomes steeper with an increase in the intercept (n0), whereas the CSD slope of the late 1992 eruption becomes flat and low with the low n0 that corresponds to a decrease in the MND (Fig. 4, Table 1). From 1993 to early 1995, the CSD curves become steeper as the intercept increases. 4.2. Compositional analyses

Fig. 2. Textural variations in the microlite number density, crystallinity, and average length as a function of time during the eruptive activity. The average length increased from early 1991 to late 1992; it attained its maximum value in late 1992. However, the lengths decreased from 1993 to mid-1995. In contrast, the microlite crystallinities determined by the thin section (2D) and the CSD (3D) rather increase with time. On the other hand, the MND decreases from early 1991 to late 1992, and it attains its maximum value in mid-1993. Then, the MND gradually increases from early 1993 to mid-1995. The discharge rate can be obtained from the volumetric change recorded using aerial photographs taken by the Geographical Survey Institute (GSI), Public Work Research Institute (PWRI), Geological Survey of Japan (GSJ), and the calculations for the shortterm volume increments in the lava added into the dome carried out by the Geologic Party, Joint University Research Group (JURG). A detailed explanation is given by Nakada et al. (1999).

The H2O contents of the groundmass of all the eruption products are nearly constant at 0.1–0.2 wt.% except for the eruptions on June 8 and 11, 1991 (pelean and vulcanian eruption, respectively) (Table 2). The H2O contents (0.1–0.2 wt.%) show that in the Unzen dome eruptions, the degassing proceeded nearly at equilibrium in the conduit and at the surface (0.1 MPa). The compositions of the plagioclase microlites and overgrowth rims reveal a high An content (An40–70) in comparison to the phenocryst core and interior rim (i.e., inside of the dusty zone) (Fig. 5). In addition, the FeO⁎ content (total Fe as FeO) of the microlites and the overgrowth rim are relatively high due to the mixing of mafic magma (e.g., Sato, 1996; Nakada and Motomura, 1999; Sato et al., 1999). Depending on the area (size) of the plagioclase crystals, the An content of the plagioclase microlites core indicates that the phenocrysts and microlites can be clearly distinguished based on their compositional trends at certain scales between 100 and 300 µm in length (Fig. 6). The boundary of these different compositional trends does not appear to depend on time during the eruptive activity. The An content of the core in the plagioclase microlites decreases with the microlite size from An70 to An45–40. A comparison of the eruption products among different stages shows that the plagioclase microlite cores of the 1991.0915 and 1993.0624 samples have a higher An content (average: An61, minimum: An50), while those of the 1992.1229 and 1995 spine samples exhibit a lower An content (average: An57, minimum: An40–45). Although the maximum An content does not seem to correlate with the discharge rate, the minimum An content roughly depends on the discharge rate. Nakada and Motomura (1999) also report that the average An content decreases with decreasing the magma discharge rate; this tendency is in good agreement with our data. 4.3. Correlation between the textural data and discharge rate The average length is roughly correlated with the discharge rate (negative correlation) except at the final stage of eruption (Fig. 7). The microlite crystallinity also has a negative correlation with the discharge rate. The MND has a positive correlation with the discharge rate except at the final stage of eruption. In addition, the temporal variation in the textural data has a correlation with that of the

146

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Fig. 3. Correlations among the microlite number density, crystallinity, and average length. (a) The MND increases with a decrease in the average microlite length, while the MND and average length do not correlate with the variation in the 2D crystallinity (b and c). VR⁎ is estimated by the integration of the straight CSD line from 0 to R⁎ in length (see Fig. 4) and its value is roughly correlated with the MND and the average length. (d) The relationship between the 2D area of the crystallinity determined by the thin section and V∞ (3D volume of the crystallinity) determined by the integration of the CSD curve from a length of 0 to 0.1 mm. The 2D area of crystallinity and V∞ reveal relatively similar values.

discharge rate (Fig. 2). It is noteworthy that there is no time lag between the correlated temporal change in the textural data and the discharge rate. In other words, this means that the textural characteristics found in the conduit are simultaneously reflected in the surface phenomenon or the discharge rate. 5. Discussions 5.1. Microlite nucleation depth The products of the 1991–1995 Unzen eruptions exhibit a high An content in the plagioclase, ranging from An40 to NAn70, similar to the Soufrière Hills volcano at Montserrat (e.g., Murphy et al., 2000; Couch et al., 2003a). Couch et al. (2001) suggested based on their experiment that calcic plagioclase overgrowth rims and microlites in Montserrat are produced at substantially high temperatures and high PH2O values (Fig. 8). Based on the melt temperature (880 °C) and the stability of groundmass pargasite for Unzen dacite, Sato et al. (1999) also suggested that the groundmass microlite nucleation occurred at N70 MPa, corresponding to a depth of approximately 3 km in the conduit. Fig. 8 shows the phase diagram of Montserrat by Couch et al. (2001) combined with the stabilities of pargasite, plagioclase, and clinopyroxene in Unzen dacite determined by Sato et al. (1999). In general, the An contents of plagioclase depend on the H2O contents (water saturation pressure) and temperature (Housh and Luhr, 1991; Couch et al., 2003a; Takagi et al., 2005) as well as the Ca/Na and Si/Al ratios in the coexisting melt. Although the starting compositions of Unzen dacite and the Montserrat are somewhat different, the liquidus lines of plagioclase in both the cases show similar trends with a

decrease in water-saturated pressure. Furthermore, the An contours in P–T diagrams of Unzen and Montserrat exhibit similar lines according to the calculation by the MELTS program (Ghiorso et al., 1983; Ghiorso and Sack, 1995) and the melt-plagioclase thermometer developed by Putirka (2005). The reason for the similar liquidus lines in Unzen and Montserrat may be due to the counterbalance of the groundmass composition: the low CaO content of the groundmass at Unzen reduces the An content in comparison to that at Montserrat, while the higher Al2O3 content with low Na2O content of the groundmass at Unzen may increase the An content. According to the nucleation pressure obtained by the An content of plagioclase from the phase diagram by Couch et al. (2001, 2003a), it is inferred that the plagioclase microlite crystallization in Unzen dacite occurred between 160 and 70 MPa (for the 1995 spine: from 160 to b50 MPa). This is consistent with the stability field of the groundmass pargasite and quartz determined by Sato et al. (1999) and Couch et al. (2001, 2003a). Although the range of the microlite nucleation pressure determined by the plagioclase composition is relatively wide, we suggest that the microlite nucleation effectively occurred at a shallow level of the conduit (70–100 MPa, for the 1995 spine: b50 MPa). This is because the MND is mostly determined by the large amount of small microlites under 20 µm, which record a shallower pressure in their composition. Based on the initial H2O content (saturation pressure: 190 ± 30 MPa), it is considered that the microlite nucleation began at 160 MPa under a condition of low undercooling (ΔT). Then, the microlite nucleation proceeded until a shallow depth with an increase in ΔT and attained its climactic stage (70–100 MPa). It should be noted that the microlite crystallization proceeds more or less at disequilibrium. However, the previous experiment suggests that the An

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

147

Fig. 4. Graph depicting the variation in the CSD of the microlites from 21 samples. The broken line indicates the limit of the exponential distribution; this limit is determined by a survey of the correlation coefficient (R2: exponential) for regression lines. The CSD slope (−1/R0) becomes flat with a lower intercept (n0) for the eruption from early 1991 to late 1992. However, during early 1993, the CSD curve becomes steeper with an increase in the intercept.

content of plagioclase is almost the same irrespective of the equilibrium or decompression experiments (Couch et al., 2003b; Fig. 5). Thus, we assume that the depth obtained from the An content of plagioclase is reasonable. The minimum An content of the plagioclase microlites (smaller microlites and microlite rim) at the final stage of eruption (1995 spine) exhibits relatively lower values (∼ An40) in comparison to other eruption products (An45 An50). The characteristics of the eruption product in the 1995 spine only reveal that quartz can be sparsely observed in the groundmass. According to the experimental work (Sato et al., 1999; Venezky and Rutherford, 1999; Holtz et al., 2005), the quartz is the last phase (close to the solidus) to crystallize during

the crystallization sequence. Several experiments of the dacitic melt composition reveal that the quartz saturation occurs at a pressure below ∼50 MPa with a temperature of ca. 850 ± 30 °C (Johannes and Holtz, 1996; Blundy and Cashman, 2001; Couch et al., 2003a). The presence of the groundmass quartz and the low An content of plagioclase in the 1995 spine indicate that the groundmass crystallization continued below the quartz stability region (∼ 50 MPa), unlike other eruption products. On the other hand, a few large plagioclase microlites (100–300 µm) in the 1991.0915 samples exhibit a higher An content (NAn70) in comparison to most of the microlites. These microlites cannot be produced by the isothermal decompression of the post-mixed magma

148

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Table 2 H2O contents in the groundmasses of the analyzed samples Sample

1991.0527 1991.0608 1991.0611 1991.0915 1992.0108 1992.0601 1992.0901 1992.0928 1993.0309 1993.0401 1993.0501 1993.0624 1994.0204 1995.spine

H2O

Average H2O

wt.%

wt.%

0.19 0.36 0.40 0.11 0.04 0.16 0.21 0.21 0.12 0.14 0.04 0.21 0.10 0.17

0.15 0.29 0.35 0.09 0.13 0.16 0.19 0.22 0.17

0.18 0.30 0.35 0.13 0.44 0.16 0.20 0.28 0.26

0.13

0.14 0.21 0.05 0.17

0.13 0.18 0.29 0.15

0.18 0.16

0.11 0.15 0.16

0.14

0.10

0.11

0.16 (0.03) 0.32 (0.04) 0.37 (0.03) 0.11 (0.02) 0.19 (0.17) 0.16 (0.00) 0.20 (0.01) 0.24 (0.04) 0.17 (0.06) 0.14 0.12 (0.06) 0.19 (0.02) 0.15 (0.13) 0.15 (0.03)

Note: The groundmass H2O analysis is performed at the Department of Earth and Planetary Sciences, Kyushu University using a Karl Fischer Titration. The numbers in parentheses show the standard deviation (1σ).

because post-mixing magma at an inferred temperature must have crystallized at a level deeper than that at 190 MPa (magma chamber). Couch et al. (2001) interpreted this calcic plagioclase as a crystal formed in the high-temperature magma by the effective cooling during convective mixing. In the case of Unzen dacite, a high An content (NAn70) of such plagioclase microlites (microphenocryst size) also may be produced before the pre-mixed mafic magma. 5.2. Relationship between MND and discharge rate—application of MND water-exsolution rate meter to textural data Microlite nucleation occurred at some depth (nucleation depth: zn) at a certain decompression rate (i.e., H2O-exsolution rate). If we consider a single nucleation event, the MND is determined at zn because the nucleation rate, which is expressed as a function of ΔT, primarily controls the MND (Hammer and Rutherford, 2002; Couch et al., 2003b). As ΔT increases with the depletion of H2O (CH2O) from the magma (Hammer and Rutherford, 2002; Couch et al., 2003b), the nucleation of microlites eventually occurs at a certain ΔTn. It is important that ΔTn is controlled by the rate of H2O exsolution. Hence, the different MNDs correspond to different H2O-exsolution rates. The relationship between the effective cooling rate dT/dt and the number density of crystals N is expressed by a simple formula proposed by Toramaru (2001):  N ¼ ad

dT dt

the nucleation pressure determined by the composition of the An content of the plagioclase, the MND is mostly determined at the final stage of nucleation at zn, corresponding to a nucleation pressure of 70– 100 MPa. If the H2O exsolution proceeds at equilibrium, its rate can be controlled by the decompression rate (dPH2O/dt) at zn as shown below:

j

j j

j

dPH2O dCH2O ¼c ; dt dt

ð7Þ

where c is 11.2 × 106 ×CH2O, which is determined by the H2O solubility relation. The decompression rate can be attributed to the ascent velocity. In a simple case of steady flow, the decompression rate can be written as follows:

j

dPH2O dt

j j j j j j j j j Zn

Y

dP dt

Zn

¼

AP Az

Zn

d

Az At

Zn

Y

dP dt

Zn

¼ qd gd UZn ;

ð8Þ

where z, ρ, g, and U are the depth from the surface along the conduit, the density at the nucleation depth (zn), the acceleration due to gravity, and the ascent velocity at zn, respectively. From Eqs. (5), (6), and (7), from which the value of dCH2O/dt for the MND data can be determined, the calculated ascent velocity of the magma at zn is estimated to be in the range of 8–76 mm/s for ρ = 1825 kg/m3 of vesiculated magma in a closed system (melt density: 2500 kg/m3). On the other hand, the exit velocity (z = 0) of magma is satisfactorily determined by the GPS observations and daily monitoring by a helicopter (Nakada et al., 1995; Nakada et al., 1999; Yamashina and Shimizu, 1999). The conduit dimension of the elongated dyke is observed to be approximately 1000–2000 m2 (about 20 × 50–100 m) during the eruptive activity. According to the variation in the discharge rate and the cross-sectional area of the dyke (1000–2000 m2), the calculated exit velocity (z = 0.1 MPa) ranges from 0.2 to 3.9 mm/s. Fig. 10 exhibits the temporal variation in the ascent velocities at zn (70–100 MPa) and z = 0 (0.1 MPa). Although the variation in the ascent velocity at zn exhibits a trend similar to that of the exit velocity (Fig. 10), the calculated ascent velocity at zn is relatively higher than the exit velocity by one order of magnitude. It should be noted that we used 3.5 wt.% of the CH2O (corresponding pressure: 70 MPa) for all the eruption products in terms of the An content of plagioclase (An45) as shown in the phase diagram by Couch et al. (2001, 2003a). However, the An content of plagioclase at

3=2 ;

ð5Þ

where a is a constant (Toramaru et al., 2008-this issue). Toramaru et al. (2008-this issue) studied the relationship between the H2Oexsolution rate (dCH2O/dt) and the MND. The H2O-exsolution rate (dCH2O/dt) has a linear relationship with the effective cooling rate (dT/ dt) in the given liquidus curve of the microlite mineral. As a result, dCH2O/dt can be calculated as a function of MND as follows:

j

j

dCH2O ¼ kd N2=3 ; dt

ð6Þ

where k is a constant that generally depends on the chemical composition and solubility curve of H2O. ΔCSi =CSi − 50, and CSi denotes the silica content in wt.%. According to the solubility curve of H2O at the final nucleation pressure (70–100 MPa), for the plagioclase microlites of the 1991–1995 eruptions at Unzen, CH2O is approximately 3.5 wt.% (see later). Using ΔCSi = 18, we have k = 1.2× 10− 12 − 0.23 ΔCsi + 0.43 CH2O. The H2O-exsolution rates calculated from the MND values reported in the present study range from 3.5 to 34.6× 10− 6 wt.%/s (Fig. 9). According to

Fig. 5. Graph showing the An and FeO⁎ contents of the plagioclase phenocrysts, microlites, and overgrowth rims analyzed by EPMA. The compositions of the An and FeO⁎ contents of the plagioclase microlite and overgrowth rim are higher than those of the phenocryst core and interior rim (just inside the dusty zone). Although some cores of the plagioclase phenocrysts in the eruption products of Unzen dacite often show a higher An content (An60–80), the FeO⁎ contents are significantly lower than the compositions of the plagioclase microlite (Sato, 1996; Nakada and Motomura, 1999).

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

149

Fig. 6. The An content of plagioclase is plotted against the length of plagioclase. The compositions of plagioclase are obtained using SEM-EDS, while the length of the microlites is determined by the area of the microlites using textural analysis. At certain scales between a length of 100 and 300 µm, we can clearly distinguish the crystals into phenocrysts and microlites according to their compositional trends. The An content of the plagioclase microlite core decreases with its size. The compositions of the plagioclase microlites with high magma discharge rates (1991.0915 and 1993.0624 samples) exhibit a higher An content, while those with low magma discharge rates (1992.1229 and 1995 spine samples) exhibit a lower An content.

the final stage of eruption (1995 spine) exhibits lower values (An40). This fact suggests that the nucleation pressure (water saturation pressure) might be relatively low (∼ 50 MPa), which corresponds to approximately 3.0 wt.% of CH2O. If we applied this value to the final stage of eruption (1995 spine), the ascent velocity at zn is estimated to be 31 ± 7 mm/s. Although this value is somewhat lower than the calculated ascent velocity of the 1995 spine using CH2O as 3.5 wt.% (Uzn = 60 ± 13 mm/s in Fig. 10); consequently, we suggest that the values of the ascent velocities at zn are always higher than the exit velocity. If we assume that the H2O exsolution behaves as a closed system with the saturation pressure at the initial H2O content of magma (6 ± 1 wt.%: Nishimura et al., 2005; Holtz et al., 2005), the vesicularity at the microlite nucleation pressure (70–100 MPa) can be estimated as 30–40 vol.%, which was determined by Villemant and Boudon (1998). On the other hand, most of the samples at the surface are poorly vesiculated at around 10–30 vol.% (Kueppers et al., 2005), which expresses the melt vesicularity of ca. 12–39 vol.%. If we consider a simple spatially uniform flow with a constant conduit dimension (ρU = const.), the ascent velocities at the nucleation depth and the surface differ by one order of magnitude. This implies that the bulk density of the surface changes significantly to a value approximately ten times higher than that at the nucleation depth. This contradiction implies that the constant conduit radius cannot account for the estimated velocities at the nucleation depth and surface and the bulk densities of the eruption products. Hence, it is plausible that the

conduit dimension changes with the depth. The conservation of mass is expressed as follows: qZn UZn AZn ¼ qZ¼0 UZ¼0 AZ¼0 ¼ QZ¼0 ;

ð9Þ

where A is the cross-sectional area of the conduit. When the values of ρZn, ρz = 0, Uzn, Uz = 0, and Az = 0 are set as 1825 kg/m3, 2000 kg/m3, 8– 76 mm/s, 0.2–3.9 mm/s, and 1000–2000 m2, respectively, Azn can be estimated to range from 4.3 to 233 m2. The differences in the conduit dimensions, ascent velocities, and melt densities between the nucleation depth and the surface constrain the dynamics of conduit flow and the permeable flow of gas through the conduit region (see the next section). 5.3. Conduit flow model for the 1991–1995 eruptions at Unzen 5.3.1. Conduit dimension The ascent velocities at two different depths, which are estimated by two different methods, reveal a correlation without time lag except for the eruption at the last stage (late 1994–1995 eruption) (Fig. 10). The calculation of the travel time of the magma from 100 MPa (approximately 4 km in depth) to the surface with a given ascent velocity at zn (8–76 mm/s) or a given exit velocity (0.2–3.9 mm/s) revealed that it ranged from a few days to one week. These values of the travel times indicate that the time intervals of the sampling are too long to resolve such a short time variation in the microlite crystallization process.

150

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Fig. 7. Correlation between the textural characteristics and the magma discharge rate. The MND has a positive correlation with the discharge rate except at the final stage of the eruption. The temporal change in the microlite crystallinity and the average size are negatively correlated with the discharge rate.

When we account for the several conduit shapes in the form of a column with a circular cross-section and elliptically elongated dykes (2:5) estimated on the basis of the cross-sectional area of the conduit (Fig. 11), the temporal variation in the conduit dimension at zn can be roughly classified into two regimes. From the early 1991 eruption to the mid-1993 eruption, the variation in the conduit dimension corresponds to a change in both the ascent velocities at zn and z = 0.

Fig. 8. Graph depicting the phase diagram of Montserrat by Couch et al. (2001). We schematically illustrate the stabilities of pargasite, plagioclase, and clinopyroxene in Unzen dacite determined by Sato et al. (1999). The stability of the pargasite in Unzen dacite indicates pressures that are lower than those at Montserrat, while the stability of plagioclase in Unzen dacite indicates values similar to that at Montserrat. According to the An content in the microlites of Unzen dacite, the microlite nucleation due to the saturation pressure occurs at a depth corresponding to 160–70 MPa.

A typical conduit dimension (elliptically elongated dykes, 2:5) ranges from 16 to 21 m in length; this dimension increases with the ascent velocities. This regime signifies that the discharge rate of the magma is determined by the ascent velocity at zn and the conduit dimension. On the other hand, from mid-1993 to the end of the eruptive activity, the conduit dimension decreases gradually from 19 to 3 m, corresponding to a change in the different ascent velocities between zn and z = 0 (two orders of magnitude). A decrease in the conduit dimension during the final stage of eruption also implies that the conduit is finally closed; this corresponds to the diminution in the supply rate of magma. It is noteworthy that the different characteristics in the textural results (MND) and the discharge rate of magma at the final stage of eruption reflect the change in the conduit dimension, which also constrains the

Fig. 9. Graph of the H2O-exsolution rate estimated by a MND water-exsolution rate meter (Toramaru et al., 2008-this issue) as a function of time during the eruptive activity at Unzen. The calculated H2O-exsolution rate ranges from 3.5 to 34.6 × 10−6 wt.%/ s and its variation is rather consistent with the fluctuation in the magma discharge rate except at the final stage of eruption.

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Fig. 10. Temporal variation in the calculated ascent velocities at zn (70–100 MPa) and z = 0 (0.1 MPa). The variation in the ascent velocity at zn exhibits a trend similar to that of the exit velocity (z = 0), except at the final stage of eruptive activity. However, the calculated ascent velocity at zn is higher than the exit velocity by one order of magnitude.

geometry of the plumbing system at the final stage of the 1991–1995 Unzen eruptions. Hence, we conclude that the observed variation in the exit velocity at z = 0 corresponds to the changes in the conduit dimension and ascent velocity at the nucleation pressure (70– 100 MPa), except for the final stage of the eruption. 5.3.2. Degassing process The maximum melt vesicularity at the nucleation pressure (70– 100 MPa) can be estimated as 30–40 vol.%, provided that the H2O exsolution proceeds as a closed system. On the other hand, most of the samples at the surface have a similar vesicularity of approximately 10– 30 vol.% (melt vesicularity = 12–39 vol.%; Kueppers et al., 2005). These results suggest that in order to ensure effective degassing, the open system must operate from the depth corresponding to the nucleation pressure (70–100 MPa) to the surface (0.1 MPa). Using natural or experimental products, recent laboratory experiments have revealed

Fig. 11. Graph of the temporal variation in the calculated conduit dimensions at zn (70– 100 MPa). In the present study, we account for several conduit shapes such as columns and elongated dykes (2:5). The temporal variation in the conduit dimensions at zn can be roughly classified into two regimes. Based on the eruptions from early 1991 to mid1993, it is observed that the variations in the conduit dimension correlate with a change in the ascent velocities at both zn and z = 0. Typically, the length of the conduit (elongated dikes, 2:5) ranges from 16 to 21 m and the length increases with the ascent velocities. On the other hand, from mid-1993 to the end of the eruptive activity, the conduit dimension gradually decreases from 19 to 3 m, corresponding to the change in the velocities between zn and z = 0 (two orders of magnitude). A decrease in the conduit dimension during the final stage of eruption also implies that the conduit was finally closed, thereby corresponding to the diminution in the supply rate of magma.

151

Fig. 12. Schematic graph showing the permeability evolution during magma ascent in the permeability–vesicularity relationship. As the vesicularity reaches its maximum values (ϕmax), the permeability drastically increases by two orders of magnitude (Takeuchi et al., 2005). However, the permeability of the natural samples (e.g., Klug and Cashman, 1996) decreases logarithmically with a decrease in the vesicularity. The relatively constant permeability of the natural samples regardless of the decrease in the vesicularity is interpreted to result from the change in the vesicle textures with a decrease in the aspect ratios. These data can be interpreted as an evolutionary trend during the magma ascent.

some constraints with regard to the permeability–vesicularity relationship. Takeuchi et al. (2005) used the experimentally quenched products to examine the permeability development during the decompression of magma. They showed that the permeability of the experimental products and the previous measurements of natural samples exhibit considerably different trends (i.e., Klug and Cashman, 1996; Rust and Cashman, 2004). The permeability of the natural samples increases at a low vesicularity ranging from 2 to 40 vol.% (Fig. 12-P2), while the permeability of vesiculating magma drastically increases during decompression with an increase in the vesicularity ranging from 45 to 80 vol.% (Fig. 12-P1). We suggest that these different trends are closely related to the two regimes corresponding to the change in the vesicle texture (Saar and Manga, 1999) during the magma ascent. For the P1 process in the vesiculating magma during decompression, the connectivity of vesicles requires sufficient volume and time to produce a permeable gas network. Thus, the coalescence of vesicles may occur effectively in the magma with moderatevesicularity. For the P2 process, when ϕ decreases from the maximum level (ϕmax, as shown in Fig. 12-P2), the connectivity of the vesicles is maintained even if there is a decrease in the aspect ratio of the vesicle shape. This implies that effective degassing (the escape of gas or bubble collapse), which continues to a shallow depth to produce magma with lower vesicularity, is responsible for the typical domeforming eruptions. The actual conduit drilled by the Unzen scientific drilling project (USDP) indicates that all the dykes with various eruption ages collected from a depth of approximately 1.4 km (saturation pressure: 40 MPa) contain poorly vesiculated rocks with a porosity of b20% (Nakada et al., 2005; Noguchi et al., 2008-this issue; Sakuma et al., 2008-this issue). If the magma vesiculates in a closed system, the vesicularity at 40 MPa is expected to be high—approximately 70 vol.%. According to the survey of the historical eruptions of Fugendake volcano by Hoshizumi et al. (1999), all the lava domes developed due to the eruption and their subsequent collapses resulted in the generation of the block-and-ash flow. These facts may suggest that all the lava domes from the magma plumbing system at Unzen – including those of the 1991–1995 eruption – always ascend with lowvesicularity magmas due to effective degassing. Consequently, the low vesicularity from the depth corresponding to the nucleation pressure (100–70 MPa) and to the surface (0.1 MPa) implies that effective degassing continuously proceeds up to the surface (Fig. 13). Ohba et al. (2008-this issue) monitored the SO2 and CO2 gases at the summit of Unzen dome during the 1991–1995 eruptions. For the

152

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

Fig. 13. Schematic model showing the magma ascent process in the conduit and the temporal variation in the conduit dimensions during Unzen 1991–1995 eruptions. The vesiculation of the magma occurred at 190 ± 30 MPa (6 ± 1 wt.% H2O by Kusakabe et al., 1999; Venezky and Rutherford, 1999; Holtz et al., 2005; Nishimura et al., 2005). The microlite nucleation begins at a depth corresponding to 160 MPa and ceases at a depth corresponding to 70–100 MPa (for the final stage of eruption: zn ∼50 MPa). The ascent velocity at zn is estimated as 8–76 mm/s, while the exit velocity is 0.2–3.9 mm/s. The decrease in the ascent velocity from zn to the surface is reflected in the increase in the conduit dimension at a shallow depth. Furthermore, an effective degassing at this deep level also decreases the ascent velocity of magma. Our model also suggests that the low-vesicularity magma can easily connect the bubble network due to the abundance of phenocrysts and microlites in the lower proportion of the melt.

concentration of the CO2/H2O ratio, effective degassing is required at a depth of around 5 km (saturation pressure of 125 MPa), suggesting that a gas pocket and pipe were formed along the conduit at around this depth. Further, this evidence of the high concentration of CO2 gas is consistent with our proposed model. We conclude that this depth corresponding to the effective degassing of low-vesicularity magma, controls the ascent velocity; this resulted in mostly non-explosive eruption during 1991–1995 at Unzen dacite. 5.3.3. Implication for theoretical models of discharge oscillation With regard to the origin of the oscillatory behavior of the magma discharge rate during the dome-forming eruption, there are two models proposed by Ida (1996) and Maeda (2000) and Melnik and Sparks (1999; Barmin et al., 2002). Melnik and Sparks (1999) suggested that the variation in the viscosity of magma due to the variation in crystallinity destabilizes the steady flow; this is due to the nonlinear effects of crystallinity on the ascent velocity and the overpressure. A low flow rate results in a significant amount of microlite crystallization; this in turn results in an increase in the viscosity, and consequently, the magma extrusion rate becomes low. On the other hand, Ida (1996) proposed the cyclic patterns of the magma outflow rate using the Poiseuille flow with a constant Newtonian viscosity and elastic conduit, which is nonlinear with respect to the cross-sectional area of the conduit. This model produces a variation in the pressure of the magma chamber; this variation corresponds to the oscillatory behavior with regard to the discharge rate. Maeda (2000) examined a model with a more realistic oscillatory solution with regard to the outflow rate; this model consists of viscous magma and a visco-elastic conduit with the given magma supply rate as a function of time. In this model, a specific supply rate of magma results in a discharge rate with multiple pulses, as observed during

Unzen 1991–1995 eruptions. In our samples, the temporal variation in the discharge rate directly corresponds to the change in the ascent velocity at zn (70–100 MPa) and the conduit dimensions (Fig. 13). The characteristic of this conduit behavior seems to be consistent with the Ida and Maeda model. Hence, in the following discussion, we deal with the conduit deformation process of the 1991–1995 eruptions at Unzen in terms of the Ida and Maeda model, and examine the applicability of the model by comparing the results of the model to those of the present study. Fig. 14a represents the trajectory of the ascent velocity and the cross-sectional area of the conduit determined by the calculation of the Ida and Maeda model (Maeda, 2000). In this model, since the ascent velocity and pressure are determined by the mass balance of the magma chamber, the ascent velocity and the cross-sectional area of the conduit rotate the orbit around one equilibrium point. On the other hand, the temporal variation in the ascent velocity and cross-sectional area of the conduit at Unzen 1991– 1995 eruption determined by a MND water-exsolution rate meter (Fig. 14b) indicates that equilibrium point can be found at more than two points. Further, the variation in the ascent velocity and conduit cross-sectional area rotates the orbit around several equilibrium points except at the final stage of eruption. With regard to the sequence of the ascent velocity and conduit dimension, although some flow behaviors of the magma in the conduit seem to be controlled by the pressure fluctuations in the magma chamber, our model is not completely identical to the Ida and Maeda model. These results may indicate that the origin of the oscillatory behavior of the magma discharge rate is not determined only by the viscous coupling between the magma flow and the conduit deformation that occurs with the pressure fluctuation at the magma chamber. Thus, we need to examine the conduit flow model further by considering into account new factors for the conduit flow.

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

153

(z,t) and z0(t). The discharge rate Q(t) is given by −dVcon(t)/dt. From Vcon(t) = Vcon(z0(t),t), the following equation is derived: Z Q ðt Þ ¼ Aðz0 ðt Þ; t Þdvðt Þ 

Fig. 14. (a) Graph of the trajectory of the ascent velocity and the cross-sectional area of the conduit determined by the calculation of the Ida and Maeda model (Maeda, 2000). The variations in the ascent velocity and conduit cross-sectional area rotate the orbit around one equilibrium point. (b) The temporal variations in the ascent velocity and cross-sectional area of the conduit at the zn (70–100 MPa) at Unzen 1991–1995 eruptions determined by MND water-exsolution rate meter. The variations in the ascent velocity and cross-sectional area of the conduit rotate the orbit around several equilibrium points, except during the final stage of eruption.

5.3.4. Factors controlling the discharge rate and conduit closure In this section, we address the origin of the counter-intuitive behavior of the conduit dimension at the final stage of the eruptions. The discharge rate Q(t) at time t can be related to the ascent velocity U (z, t) at depth z by the following mass conservation equation: Q ðt Þ ¼ Aðzn ; t  sðt ÞÞdqðzn ; t  sðt ÞÞdU ðzn ; t  sðt ÞÞ;

ð10Þ

where ρ and A are the density of magma and the cross-sectional area of the conduit as functions of time and depth. τ is the time taken by the depth zn to reach the surface. In the case of non-explosive eruptions, the density of magma is assumed to be nearly constant owing to effective degassing. As long as we consider the behavior of the discharge rate only on the basis of this equation, U may increase with time when the decreasing rate of A is larger than that of Q. However, the relation between A and U in the Poiseuille-type flow postulates that A and U have the same sense of change, i.e., U must decrease with Q. This argument suggests the further consideration of the control of the discharge rate and conduit dynamics at the stage of conduit closure. In general, the volume Vcon of magma filling the conduit can be expressed by Z Vcon ðt Þ ¼

z0 ðt Þ 0

Aðz; t Þdz;

ð11Þ

where z0(t) is the depth of conduit termination or the length of the conduit at t; Vcon(t) is a function of time, which is also a function of A

z0 ð t Þ 0

AAðz; t Þ dz; At

ð12Þ

where v(t) is the upward propagation velocity of the conduit closure. The first term in the RHS is the contribution by a change in the conduit length, and the second term is the contribution by a change in the cross-sectional area of the conduit. Setting Q(t)∝t− m, v(t)∝t− n, A (t)∝t− k, and U(t)∝t− l, and taking into account the mass conservation, the second term shows the time dependence of t− k − 1 ≈ t− m − 1 + l. It is likely that when the cross-sectional area at the closure is constant, i.e., A(z0(t),t) = const., the second term shows a time dependence of t− m. If the second term controls the discharge rate, the time dependence of the discharge rate equals the time dependence of the second term, i.e., m must be equal to 1 + l or l = 1. This indicates that the ascent velocity decreases with time. This is not the case during the final stage (1994–1995) of Unzen eruptions. However, if the discharge rate is controlled by the first term, i.e., the upward propagation of the conduit closure, the time dependence of the upward propagation velocity of the conduit closure should equal the time dependence of the discharge rate (m = n). Then, the second term can freely take any value for the exponent controlling the temporal change in the ascent velocity. In this case, the discharge rate and the ascent velocity at the depth can be decoupled. Thus, l can be a negative value (−l N 0) so that the ascent velocity increases with time. Although the specific process is unclear at this point, the conduit deformation would have to be accelerated with time when the conduit closes. In such a case, the magma flow might be inversely proportional to or independent of the cross-sectional area of the conduit, as suggested in Fig. 14b. 6. Conclusions We quantitatively analyzed the decompression-related texture and chemical composition of the plagioclase microlites in the 1991–1995 eruptions at Unzen, Japan, in order to examine their relationship to the magma discharge rate. The temporal variation in the plagioclase microlite crystallinity and average microlite size exhibits a negative correlation with the variation in the magma discharge rate. The presence of pargasite and calcic plagioclase microlites (An45–65) in the groundmass indicates that microlite nucleation occurred at a substantially high partial pressure of water (160–70 MPa). On the other hand, the variation in the MND has a positive correlation with the discharge rate. According to the MND water-exsolution rate meter, dCH2O/dt is calculated to be 3.5 to 34.6 × 10−6 wt.%/s. Assuming that the exsolution of H2O takes place at equilibrium, the variation in the decompression rate would be related to the ascent velocity, provided the conduit flow is steady. The ascent velocity of magma calculated at the final stage of the microlite nucleation (depth corresponding to 70–100 MPa; 8–76 mm/s) is greater than the exit velocity of magma (0.2–3.9 mm/s), although the temporal variation in the ascent velocities at zn (70–100 MPa) is reflected in the surface phenomenon or the discharge rate without time lag. Considering the maximum vesicularity as a criterion for a closed system, the vesicularity at zn (70–100 MPa) is estimated to be 30–40 vol.%. This value is relatively similar (or higher) than the vesicularity at the surface (10–30 vol.%). Due to this difference in the velocities between the nucleation depth and surface with relatively similar vesicularities, it is required that the conduit dimension at the nucleation depth (cross-section of conduit 4.3–233 m2) is smaller than that at the surface (1000–2000 m2). Further, the temporal variation in the conduit dimension at zn corresponds to the variation in the exit velocity, implying that the observed discharge rate of magma is controlled by the ascent velocity at zn (70–100 MPa) and

154

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155

the conduit dimension. Furthermore, for the final stage of eruptive activity, the decrease in the conduit diameter from 19 to 3 m indicates that the conduit is finally closed due to the diminution of the supply of magma. The low-vesicularity magma from zn (70–100 MPa) to z = 0 (0.1 MPa) suggests that effective degassing continued up to the surface with a change in the vesicle texture. Consequently, the effective degassing of the low-vesicularity magma at this depth affected the ascent velocity, thereby resulting in the non-explosive eruption during 1991–1995 at Unzen dacite. Acknowledgments We would like to thank Drs. H. Shimizu, T. Matsushima, T. Kobayashi, T. Fujii, J. Hirabayashi, S. Takarada, and M. Ohno for providing samples with regard to the 1991–1995 eruptions at Unzen. The EPMA analysis was performed under the guidance of Drs. K. Tazaki and T. Morishita. We also thank Drs. H. Sato, T. Ikeda, and T. Miyamoto for their helpful discussions and encouragement during this work. The critical comments and suggestions from Drs. M. Nakamura, I. Miyagi, T. Koyaguchi, S. Takeuchi, and N. Geshi greatly enhanced this study. Constructive review and comments of Dr. H. Sato and anonymous reviewer significantly improved the paper. This work was financially supported by the Sasagawa Scientific Research Grant from The Japan Science Society and a Grant-in-Aid for Scientific Research from MEXT (No. 14080202, No. 17340131, and No. 18749003). References Barmin, A., Melnik, O., Sparks, R.S.J., 2002. Periodic behavior in lava dome eruptions. Earth. Planet. Sci. Lett. 199, 173–184. Blundy, J., Cashman, K.V., 2001. Ascent-driven crystallization of dacite magmas at Mount St Helens, 1980–1986. Contrib. Mineral. Petrol. 140, 631–650. Cashman, K.V., 1992. Groundmass crystallization of Mount St. Helens dacite, 1980– 1986: a tool for interpreting shallow magma processes. Contrib. Mineral. Petrol. 109, 431–449. Cashman, K.V., Marsh, B.D., 1988. Crystal size distribution (CSD) in rocks and the kinetics and dynamics of crystallization, II, Makaopuhi lava lake. Contrib. Mineral. Petrol. 99, 292–305. Couch, S., 2003. Experimental investigation of crystallization kinetics in a haplogranite system. Am. Mineral. 88, 1471–1485. Couch, S., Sparks, R.S.J., Carroll, M.R., 2001. Mineral disequilibrium in lavas explained by convective self-mixing in open magma chamber. Nature 411, 1037–1039. Couch, S., Harford, C.L., Sparks, R.S.J., Carroll, M.R., 2003a. Experimental constraints on the conditions of formation of highly calcic plagioclase microlites at the Soufriere Hills volcano, Montserrat. J. Petrol. 44, 1455–1475. Couch, S., Sparks, R.S.J., Carroll, M.R., 2003b. The kinetics of degassing-induced crystallization at Soufriere Hills volcano, Montserrat. J. Petrol. 44, 1477–1502. Geschwind, C.H., Rutherford, M.J., 1995. Crystallization of microlites during magma ascent: the fluid mechanics of 1980–1986 eruptions at Mount St Helens. Bull. Volcanol. 57, 356–370. Ghiorso, M.S., Sack, R.O., 1995. Chemical mass transfer in magmatic processes IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contrib. Mineral. Petrol. 119, 197–212. Ghiorso, M.S., Carmichael, I.S.E., Rivers, M.L., Sack, R.O., 1983. The Gibbs free energy of mixing of natural silicate liquids; an expanded regular solution approximation for the calculation of magmatic intensive variables. Contrib. Mineral. Petrol. 84, 107–145. Hammer, J.E., Rutherford, M.J., 2002. An experimental study of the kinetics of decompression-induced crystallization in silicic melt. J. Geophys. Res. 107. doi:10.1029/2001JB000281. Hammer, J.E., Cashman, K.V., Hoblitt, R.P., 1999. Degassing and microlite crystallization during pre-climactic events of the 1991 eruption of Mt. Pinatubo, Philippines. Bull. Volcanol. 60, 355–380. Hammer, J.E., Cashman, K.V., Voight, B., 2000. Magmatic processes revealed by textural and compositional trends in Merapi dome lavas. J. Volcanol. Geotherm. Res. 100, 165–192. Higgins, M.D., 2000. Measurement of crystal size distributions. Am. Mineral. 85, 1105–1116. Higgins, M.D., 2002. Closure in crystal size distributions (CSD), verification of CSD calculations, and the significance of CSD fans. Am. Mineral. 87, 171–175. Holtz, F., Sato, H., Lewis, J., Behrens, H., Nakada, S., 2005. Experimental petrology of the 1991–1995 Unzen dacite, Japan. Part I: phase relations, phase composition and preeruptive conditions. J. Petrol. 46, 319–337. doi:10.1093/petrology/egh077. Hoshizumi, H., Uto, K., Watanabe, K., 1999. Geology and eruptive history of Unzen volcano, Shimabara Peninsula, Kyushu, SW Japan. J. Volcanol. Geotherm. Res. 89, 81–94.

Housh, T.B., Luhr, J.F., 1991. Plagioclase-melt equilibria in hydrous systems. Am. Mineral. 76, 477–492. Ida, Y., 1996. Cyclic fluid effusion accompanied by pressure change: implication for volcanic eruptions and tremor. Geophys. Res. Lett. 23, 1457–1460. Ishihara, K., 1993. Continuous magma supply inferred from discharge rate of magma and ground-deformation rate at Mt. Unzen, Japan. Annu. Disas. Prev. Res. Inst., Kyoto Univ. 36, 219–230 in Japanese with English abstract. Johannes, W., Holtz, F., 1996. Petrogenesis and Experimental Petrology of Granitic Rocks. Springer, Berlin. 335 PP. Klug, C., Cashman, K.V., 1994. Vesiculation of May 18, 1980, Mount St. Helens magma. Geology 22, 468–472. Klug, C., Cashman, K.V., 1996. Permeability development in vesiculating magmas: implications for fragmentation. Bull Volcanol. 58, 87–100. Kueppers, U., Scheu, B., Spieler, O., Dingwell, D.B., 2005. Field-based density measurements as tool to identify preeruption dome structure: set-up and first results from Unzen volcano, Japan. J. Volcanol. Geotherm. Res. 141, 65–75. Kusakabe, M., Sato, H., Nakada, S., Kitamura, T., 1999. Water contents and hydrogen isotopic ratios of rocks and minerals from the 1991 eruption of Unzen volcano, Japan. J. Volcanol. Geotherm. Res. 89, 231–242. Maeda, I., 2000. Nonlinear visco-elastic volcanic model and its application to the recent eruption of Mt. Unzen. J. Volcanol. Geotherm. Res. 95, 35–47. Martel, C., Schmidt, B.C., 2003. Decompression experiments as an insight into ascent rates of silicic magmas. Contrib. Mineral. Petrol. 144, 397–415. Melnik, O., Sparks, R.S.J., 1999. Nonlinear dynamics of lava dome extrusion. Nature 402, 37–41. Melnik, O., Sparks, R.S.J., 2005. Controls on conduit magma flow dynamics during lava dome building eruptions. J. Geophys. Res. 110. doi:10.1029/2004JB003183. Murphy, M.D., Sparks, R.S.J., Barclay, J., Carroll, M.R., Brewer, T.S., 2000. Remobilization of andesite magma by intrusion of mafic magma at the Soufrière Hills Volcano, Montserrat, West Indies. J. Petrol. 41, 21–42. Nakada, S., Fujii, T., 1993. Preliminary report on the activity at Unzen volcano (Japan), November 1990–November 1991: dacite lava domes and pyroclastic flows. J. Volcanol. Geotherm. Res. 54, 319–333. Nakada, S., Motomura, Y., 1999. Petrology of the 1991–1995 eruption at Unzen: effusion pulsation and groundmass crystallization. J. Volcanol. Geotherm. Res. 89, 173–196. Nakada, S., Motomura, Y., Shimizu, H., 1995. Manner of magma ascent at Unzen Volcano (Japan). Geophys. Res. Lett. 22, 567–570. Nakada, S., Shimizu, H., Ohta, K., 1999. Overview of the 1990–1995 eruption at Unzen Volcano. J. Volcanol. Geotherm. Res. 89, 1–22. Nakada, S., Uto, K., Sakuma, S., Eichelberger, J.C., Shimizu, H., 2005. Scientific results of conduit drilling in the Unzen Scientific Drilling Project (USDP). Sci. Drill. 1, 18–22. Nakamura, M., 1995. Continuous mixing of crystal mush and replenished magma in the ongoing Unzen eruption. Geology 23, 807–810. Nishi, K., Ono, H., Mori, H., 1999. Global positioning system measurements of ground deformation caused by magma intrusion and lava discharge: the 1990–1995 eruption at Unzendake volcano, Kyushu, Japan. J. Volcanol. Geotherm. Res. 89, 23–34. Nishimura, K., Kawamoto, T., Kobayashi, T., Sugimoto, T., Yamashita, S., 2005. Melt inclusion analysis of the Unzen 1991–1995 dacite: implications for crystallization processes of dacite magma. Bull. Volcanol. 67, 648–662. Noguchi, S., Toramaru, A., Shimano, T., 2006. Crystallization of microlite and degassing during magma ascent: constraints on the fluid mechanical behavior of magma at Tenjo eruption at Kozu Island, Japan. Bull. Volcanol. 68, 432–449. Noguchi, S., Toramaru, A., Nakada, S., 2008. Groundmass crystallization in dacite dykes taken in Unzen Scientific Drilling Project (USDP-4). J. Volcanol. Geotherm. Res. 175, 71–81 (this issue). Ohba, T., Hirabayashi, J., Nogami, K., Kusakabe, M., Yoshida, M., 2008. Magma degassing process during the eruption of Mt. Unzen, Japan in 1991 to 1995: modeling with the chemical composition of volcanic gas. J. Volcanol. Geotherm. Res. 175, 120–132 (this issue). Putirka, K.D., 2005. Igneous thermometers and barometers based on plagioclase + liquid equilibria: tests of some existing models and new calibrations. Am. Mineral. 90, 336–346. Rust, A., Cashman, K.V., 2004. Permeability of vesicular silicic magma: inertial and hysteresis effects. Earth. Planet. Sci. Lett. 225, 93–107. Saar, M., Manga, M., 1999. Permeability–porosity relationship in vesicular basalts. Geophys. Res. Lett. 26, 111–114. Sakuma, S., Kajiwara, T., Nakada, S., Shimizu, H., Uto, K., 2008. Drilling and logging results and knowledge of conduit drilling well USDP-4. J. Volcanol. Geotherm. Res. 175, 1–12 (this issue). Sato, H., 1996. A model of the 1991 eruption of Unzen volcano inferred from petrographic textures of the ejects (in Japanese with English abstract). Mem. Geol. Soc. Japan, 46, 115–125. Sato, H., Nakada, S., Fujii, T., Nakamura, M., Suzuki-Kamata, K., 1999. Groundmass pargasite in the 1991–1995 dacite of Unzen volcano: phase stability experiments and volcanological implications. J. Volcanol. Geotherm. Res. 89, 197–212. Sato, H., Holtz, F., Behrens, H., Botcharnikov, R., Nakada, S., 2005. Experimental petrology of the 1991–1995 Unzen dacite, Japan. Part II: Cl/OH partitioning between hornblende and melt and its implications for the origin of oscillatory zoning of hornblende phenocrysts. J. Petrol. 46, 339–354. doi:10.1093/petrogy/egh078. Takagi, D., Sato, H., Nakagawa, M., 2005. Experimental study of a low-alkali tholeiite at 1–5 kbar: optimal condition for the crystallization of high-An plagioclase in hydrous arc tholeiite. Contrib. Mineral. Petrol. 149, 527–540. Takeuchi, S., Nakashima, S., Tomiya, A., Shinohara, H., 2005. Experimental constraints on the low gas permeability of vesicular magma during decompression. Geophys. Res. Lett. 32. doi:10.1029/2005GL022491.

S. Noguchi et al. / Journal of Volcanology and Geothermal Research 175 (2008) 141–155 Toramaru, A., 2001. A numerical experiment of crystallization for a binary eutectic system with application to igneous textures. J. Geophys. Res. 106, 4037–4060. Toramaru, A., Noguchi, S., Oyoshihara, S., Tsune, A., 2008. MND (microlite number density) water exsolution rate meter. J. Volcanol. Geotherm. Res. 175, 156–167 (this issue). Umakoshi, K., Shimizu, H., Matsuwo, N., 2001. Volcano-tectonic seismicity at Unzen Volcano, Japan, 1985–1999. J. Volcanol. Geotherm. Res. 112, 117–131. Venezky, D.Y., Rutherford, M.J., 1999. Petrology and Fe–Ti oxide reequilibration on the 1991 Mount Unzen mixed magma. J. Volcanol. Geotherm. Res. 89, 213–230.

155

Villemant, B., Boudon, G., 1998. Transition from dome-forming to plinian eruptive styles controlled by H2O and Cl degassing. Nature 392, 65–69. Wylie, J.J., Voight, B., Whitehead, J.A., 1999. Instability of magma flow form volatiledependent viscosity. Science 285, 1883–1885. Yamashina, K., Shimizu, H., 1999. Crustal deformation in the mid-May 1991 crisis preceding the extrusion of a dacite lava dome at Unzen volcano, Japan. J. Volcanol. Geotherm. Res. 89, 45–55.