https://doi.org/10.1130/G51763.

Manuscript received 29 September 2023

Revised manuscript received 12 December 2023
Manuscript accepted 3 January 2024

© 2024 Geological Society of America. For permission to copy, contact [email protected].

Published online 18 January 2024

Increase in magma supply to Sakurajima volcano’s (Japan)

shallow magma chamber over the past 500 years
Christian Huber1,* and Atsushi Toramaru2
1
 epartment of Earth, Environmental and Planetary Sciences, Brown University, Providence, Rhode Island 02912, USA
D
2
Department of Earth and Planetary Sciences, Kyushu University, Fukuoka 819-0395, Japan

ABSTRACT Seismic surveys (Iguchi et al., 2013; Per-

Constraining the magma supply to a subvolcanic reservoir is crucial to characterizing mana et al., 2022) and geobarometry (Araya
unrest and evaluating the potential for a forthcoming eruption. With the advent of GPS, tilt- et al., 2019) point to a shallow (4–6 km depth)
meters, and satellite interferometry, it has been possible to infer changes in the supply rate chamber fed by a deeper reservoir (∼10 km). The
of magma to shallow reservoirs over time scales of as much as decades. While these methods connection between the two reservoirs is poorly
have impacted our ability to characterize volcanic unrest, they only probe magmatic activ- constrained, although studies of crystal size dis-
ity over short time scales. Here, we constrain changes in magma supply rate at Sakurajima tribution reveal that the crystallization environ-
volcano (Kyushu, Japan) over the past five centuries. The combination of thermo-mechanical ment in the shallow system has remained steady
modeling, documented eruption history, and textural analyses of pumices allows us to tightly over the past five centuries, while the conditions
constrain the long-term rate of magma supply to the subvolcanic chamber through time. for the supply as inferred from crystal size distri-
Specifically, we find that magma supply rate has increased by roughly an order of magnitude bution have likely changed over time, correlating
over the past 500 years and that the bubble content in the magma chamber has increased with a geologically estimated increase in magma
over time, explaining the changes in eruption volume between the Bunmei (ca. A.D. 1470), supply rate (Yamashita and Toramaru, 2020).
An-ei (ca. A.D. 1780), and Taisho (A.D. 1914) eruptions.
METHODS
INTRODUCTION 1914, use textural analyses on pumice samples, We use the model developed by Degruyter
While the development of geodetic tech- and apply a thermo-mechanical model to invert and Huber (2014) and modified by Townsend
niques has enabled us to constrain changes in for the magma supply history that best matches et al. (2019), where a magma chamber is open
magma supply rate at shallow magma chambers, the eruptive history. to recharge and eruptions are triggered when
monitoring data sets can go back only a few a critical overpressure is reached. The model
decades. Erupted samples provide a complemen- GEOLOGICAL BACKGROUND builds on (1) the conservation of total mass and
tary set of information pertaining to the state Sakurajima is part of the Aira caldera com- enthalpy of the magma, and (2) the conservation
of the system near the time of the onset of past plex south of the island of Kyushu, Japan. Over of mass of volatile species H2O and CO2. The
eruptions, yet the connection of these observa- the past five centuries it has produced three magma chamber is coupled to the host rocks by
tions to the mechanical state of the chamber dur- massive eruptions, each of which generally heat loss by conduction and a visco-elastic stress
ing repose is ambiguous at best. comprised a Plinian phase and lava flows. The balance at the chamber-wall interface. However,
While magmatic plumbing can respond to Bunmei eruption (A.D. 1471–1476) produced given the shallow depth (5 km) and the short
short-term perturbations, the growth and ther- a dense rock equivalent (DRE) volume of ∼0.7 duration (500 years), viscous creep of the rela-
mal and mechanical evolution of a magma body km3, followed nearly 300 years later by the An-ei tively cold host is negligible in our simulations.
take place over significantly longer time scales. eruption (A.D. 1779–1782) DRE volume of 1.8 The melting curve and volatile solubility of the
Probing these systems over longer time scales km3, the Taisho eruption (A.D. 1914–1915) DRE magma are computed from rhyolite-MELTS
requires a multidisciplinary approach where volume of 1.5 km3, and finally the significantly (Gualda et al., 2012; details in Cordell et al.,
eruption history, petrology, and geophysical smaller and effusive Showa eruption (A.D. 1946) 2022).
constraints are integrated into a thermo-mechan- DRE volume of 0.1 km3 (Ishihara et al., 1981; The model is initialized at the critical over-
ical model of magma chambers through time. Kobayashi et al., 2013; Nishihara et al., 2022). pressure (20 MPa) and proceeds with the onset
We study Sakurajima volcano (Japan), which Bulk rock composition evolved from dacitic of the first eruption (Bunmei). The initial tem-
has a well-documented eruption history over the (Bunmei) to a mixture of dacitic and andesitic perature is set to match the phenocryst content
past 500 years. We focus our attention on the components over time. The phenocryst content of the Bunmei eruption products (Yanagi et al.,
three major eruptions between A.D. 1470 and roughly doubled from Bunmei (∼15 vol%) to 1991), assuming an initial volatile content of 3.6
Taisho and Showa (∼30 vol%) (Yanagi et al., wt% H2O and 200 ppm CO2, in general agree-
1991). The melt inclusion record suggests a ment with Araya et al.’s (2019) data from the
Christian Huber https://orcid.org/0000-0001​
-6518-710X maximum water content of 3.5 wt% and a CO2 most volatile-rich melt inclusions. The volume
*christian_huber@brown​.edu content of ∼30 ppm (Araya et al., 2019). of the first eruption is not prescribed but depends

CITATION: Huber, C., and Toramaru, A., 2024, Increase in magma supply to Sakurajima volcano’s (Japan) shallow magma chamber over the past 500 years:
Geology, v. 52, p. 235–239, https://doi.org/10.1130/G51763.1

Geological Society of America | GEOLOGY | Volume 52 | Number 4 | www.gsapubs.org 235

Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/52/4/235/6322051/g51763.1.pdf

by Univ of Ryukyu Toshokan 1 Senbaru Nishihara cho user
 on the initial volume of the magma chamber, use Markov chain Monte Carlo (MCMC) meth- dratic power law with an initial recharge rate A
overpressure, and the effective compressibility ods to retrieve the recharge rate parameters A, (at the onset of the Bunmei eruption) between
of the magma–host rock system (Huppert and B, and c, defined as: 1 × 10−3 and 3 × 10−3 km3/yr.
Woods, 2002; Townsend et al., 2019). We adjust Figure 2 shows one converged Markov
the initial magma chamber volume (to ∼20 km3) Vin = A + Bt c , (1) chain solution; the most likely solutions for
to yield an eruption of volume comparable to each chain yield very similar results. The
the volume of the Bunmei eruption (1.2 km3). where Vin is the volume injected per unit time pressure increases during repose because of
The model is run for 500 years and generates an (km3/yr), A is the initial recharge rate, and B recharge and drops rapidly during an erup-
eruption history that is modulated by the magma and c control the temporal dependence of the tion because of the mass loss to the surface
recharge rate. The modeled eruption history is magma input rate into the chamber. The magma (Fig. 2A). The steeper pressure loading with
compared to the cumulative volume erupted at injected is assumed to have a constant tempera- time during repose after each eruption shows
Sakurajima over that time span (Bunmei, An-ei, ture (initial temperature of the magma body) that the recharge rate is increasing with time.
and Taisho). We do not consider Showa, which and constant phase volume fraction (the same Figure 2B shows the evolution of the average
is significantly smaller and did not involve an as initial conditions in the magma chamber for phenocryst fraction over time. While the ini-
explosive phase and therefore was likely not simplicity). We run nine Markov chains with tial conditions are slightly undersaturated with
comparable to the other three (more in the different initial parameter values (A0, B0, c0) and volatiles, the decompression associated with
Supplemental Material1). conduct a random walk through the parameter the Bunmei eruption drives volatile satura-
Our objective is to calculate magma input space where A, B, and c are each drawn from a tion for all simulations. The volume fraction
rates over time that minimize the error between uniform prior distribution. For more informa- of phenobubbles (magma chamber bubbles;
the data and modeled cumulative eruption vol- tion about the MCMC applied to inversions in Toramaru, 2022) is initially low (∼0.1 vol%)
ume over the past 500 years. This requires a volcanology, we refer readers to Anderson and and increases until the second (An-ei) eruption
nonlinear inversion, and we leverage the speed Segall (2013) and Florez et al. (2021). After a (Fig. 2C). While volatile exsolution during syn-
of the forward magma chamber model solver to burn-in period (∼100 iterations), the posterior eruptive decompression dominates the trend,
distribution of each parameter is sampled a large the increasing recharge rate with time leads
number of times (∼104 MCMC iterations per ultimately to a net decrease in phenobubble
1
Supplemental Material. The Supplemental chain) to provide the best fit values for A, B, volume fraction during subsequent repose peri-
Material contains two important additions to the and c and a measure of how well constrained ods. Figure 2D illustrates the fit of the cumula-
main manuscript. First, it provides a more detailed these values are (uncertainty and covariance). tive volume erupted with that from the model
discussion of the rationale for leaving the Showa using the best-fit parameters (A, B, and c) for
inversion out of our modeling effort. The second
addition concerns the pumice samples analyzed for RESULTS the recharge rate. Although the model (dashed
crystal volume fraction and importantly for their All nine chains converged to satisfying solu- line) underestimates the volume of the An-ei
bubble size distribution. The latter is instrumental tions (i.e., matching the cumulative erupted vol- and Taisho eruptions, the overall fit is excellent.
in determining whether phenobubbles are present ume data). Visualization of the posterior dis- We note that we do not prescribe the number
and how their content varied from one eruption to
tributions of each of the parameters and their of eruptions, their timing, nor their volume; all
the next. Please visit https://doi​.org​/10​.1130​/GEOL​
.S.24986067 to access the supplemental material; covariance is provided in Figure 1. We see that three are computed from the conservation laws
contact editing@geosociety​.org with any questions. the most likely solutions involve a nearly qua- in response to the magma supply rate.

Figure 1. Outputs of
Markov chain Monte
Carlo (MCMC) inversions
for nine Markov chains
(90,000 forward models).
Plots on the diagonal
show marginal posterior
distributions for param-
eters A, B, and c (see text
for parameter descrip-
tions); off-diagonal plots
show covariance (warm
colors represent more
likely solutions, here like-
lihood normalized).

236 www.gsapubs.org | Volume 52 | Number 4 | GEOLOGY | Geological Society of America

Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/52/4/235/6322051/g51763.1.pdf

by Univ of Ryukyu Toshokan 1 Senbaru Nishihara cho user
 A B Figure 2. Time series for
the lowest misfit solu-
tion from one chain. (A)
Magma pressure time
series. Sudden pressure
drops represent erup-
tions (An-ei and Taisho
are marked with red ver-
tical dotted lines). (B)
Average volume frac-
tion of phenocrysts
over time. (C) Volume
fraction of exsolved
volatiles in the magma
chamber over time. Note
that exsolution takes
C D place shortly after the
onset of Bunmei, during
the eruption. (D) Mod-
eled (blue dashed line)
and observed (solid red
line) cumulative erupted
volume time series. Note
that number, timing, and
volume of eruptions are
not prescribed but result
from the evolution of state
variables in the magma
chamber.

DISCUSSION vastly increase the compressibility of the magma short time span, recharge provides only a rela-
At Sakurajima, the shortening of the repose (Townsend et al., 2019). As a result, subsequent tively small contribution to the overall mass and
time from the Bunmei to Taisho eruptions sug- eruptions (An-ei and Taisho), which start satu- heat balance and therefore does not dramatically
gests a net increase in magma recharge rate rated with volatiles according to our model, are alter the CSDs. Interestingly, CSDs of the more
with time, but the short time span studied (500 significantly more voluminous than Bunmei mafic and high-temperature plagioclase pheno-
years) is insufficient to attribute the larger because bubble expansion and exsolution dur- crysts originating from the recharging magmas
eruption volumes to a significant growth of ing the eruption buffer the pressure drop caused systematically increase the slope and intercept
the magma chamber. The volume of eruptions by magma withdrawal. of log-linear CSDs (i.e., log population density
depends mostly on the volume of magma avail- Pumice textures from the Bunmei and vs. linear crystal size) with time, consistent with
able (chamber volume), the compressibility of Taisho eruptions seem to support an increase the geologically inferred recharge rate increase
the magma, and the overpressure at the onset of in magma compressibility between the Bunmei, between the Bunmei and Showa eruptions
eruptions (Huppert and Woods, 2002; Townsend An-ei, and Taisho eruptions. Bunmei pumices (Yamashita and Toramaru, 2020).
et al., 2019): show few phenobubbles (<10 vol%), and most The model results, supported by textural
of the porosity is composed of matrix bubbles analyses, suggest that volatile exsolution dur-
Ver = V β ∆ Pc , (2) nucleated during magma ascent to the surface ing or shortly after the Bunmei eruption is key
(Toramaru, 2022). Taisho pumices are signif- to explaining the increase in eruption volume
where Ver and V are respectively the erupted and icantly different and contain a larger amount between the Bunmei (0.7 km3) and An-ei and
chamber volumes, β is the magma compress- (30–50 vol%) of phenobubbles. Using an ideal Taisho eruptions (∼1.5–1.8 km3). However, the
ibility, and ΔPc is the critical overpressure trig- gas equation of state, we predict that the pheno- presence of phenobubbles decreases the rate of
gering eruptions. As mentioned above, the short bubble volume fraction at chamber conditions pressure loading caused by recharge because
time span and the shallow depth of the magma during the Taisho eruption was about an order of of the increased compressibility of the magma
chamber (relatively cold crust) prevent any sig- magnitude greater than for the Bunmei eruption (Townsend et al., 2019). Yet the repose time
nificant non-elastic deformation of the host rock (Supplemental Material). between An-ei and Taisho is about half of that
to accommodate a fast-growing magma cham- The crystal size distributions (CSDs) for each between Bunmei and An-ei, which requires that
ber. If one considers the triggering overpressure eruption were studied recently by Yamashita and the increasing magma recharge rate more than
to remain roughly similar between consecutive Toramaru (2020) and suggest that the condi- compensates for the compressibility change.
eruptions, the main factor that controls eruption tions for crystallization have remained mostly Figure 3A shows the favored recharge rate
volume is magma compressibility. unchanged over the past 500 years. This may (lowest cumulative-volume-erupted misfit) and
We note that at the time of the onset of the appear to contradict our findings of increased its standard deviation for one of the MCMC
Bunmei eruption, the conditions are volatile magma recharge rate with time, but we note chains. Figure 3D shows all the chains (no stan-
undersaturated but close to saturation given our that the volume of magma erupted during the dard deviation shown for clarity). Even though
solubility model parameterized from rhyolite- Bunmei, An-ei, and Taisho events each range the power-law exponent can vary in the pos-
MELTS (Ghiorso and Gualda, 2015). Therefore, from 3% to 6% of the volume of the estimated terior distributions (see Fig. 1), the modeled
as the Bunmei eruption proceeds, phenobubble magma chamber. The time-integrated volume recharge rates are very similar over the 500-year
exsolution is predicted in all simulations. Even of magma recharge over the 500 years amounts span. Importantly, the calculated recharge rate
if only <1 vol% phenobubbles are present, they to 15% of the volume of the chamber. Over this increases by nearly an order of magnitude over

Geological Society of America | GEOLOGY | Volume 52 | Number 4 | www.gsapubs.org 237

Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/52/4/235/6322051/g51763.1.pdf

by Univ of Ryukyu Toshokan 1 Senbaru Nishihara cho user
 A B

Figure 3. (A) Comparison

of cumulative volume
erupted and magma
recharge over time for
one of the converged
solutions, with its stan-
dard deviation from the
posterior distribution
(shaded area). (B) Best
solution for each of the
nine Markov chain Monte
Carlo (MCMC) chains
(dotted lines).

the past 500 years, which is significant. We note of the system over time and promote hysteresis of magma supply rate over time scales that
that the values calculated here are consistent in the eruption cycle. far exceed what we can infer from geodetic
with geodetic estimates (Hickey et al., 2016). observations.
The time-integrated volume of magma sup- CONCLUSIONS
plied to the shallow magma chamber of Sakura- We combine field and textural data from ACKNOWLEDGMENTS
jima over the past 500 years yields similar values three large historical eruptions at Sakurajima We acknowledge the funding from the Japan Society
for the Promotion of Science (JSPS) and from the U.S.
(∼3 km3) irrespective of the choice of admis- with a numerical model to invert for the evolu- National Science Foundation (2121655) to Huber. The
sible solutions within our converged data set tion of the magma supply rate through time. The authors acknowledge two anonymous reviewers for
(see Fig. 4). We note that while the estimated results point to a well-constrained temporal evo- constructive comments that improved the manuscript.
recharge volume is well constrained, the dif- lution of the supply rate increasing by nearly an The authors thank Makoto Sekiguchi for the prepara-
tion and density measurements of pumice samples and
ferences in recharge rate functions caused by order of magnitude over the past 500 years. The
Okuyama Sangyo for sampling permission.
the trade-offs between A, B, and c can seriously modeling results are in excellent agreement with
undermine forecasting efforts if extrapolated to the eruption history at Sakurajima and consistent REFERENCES CITED
future scenarios. The model results show that with petrological analyses of pumice textures Anderson, K., and Segall, P., 2013, Bayesian inversion
the estimated volume of magma recharge over sampled from these eruptions. of data from effusive volcanic eruptions using
the past 500 years is close to but does not match This study shows the power of combining physics-based models: Application to Mount St.
Helens 2004–2008: Journal of Geophysical Re-
the cumulative volume erupted. Phase changes eruptive history, textural analyses, and thermo-
search: Solid Earth, v. 118, p. 2017–2037, https://
(exsolution) change the mechanical properties mechanical modeling to invert for evolution doi​.org​/10​.1002​/jgrb​.50169.
Araya, N., Nakamura, M., Yasuda, A., Okumura, S.,
Sato, T., Iguchi, M., Miki, D., and Geshi, N.,
2019, Shallow magma pre-charge during repeated
Plinian eruptions at Sakurajima volcano: Scien-
tific Reports, v. 9, 1979, https://doi​.org​/10​.1038​
/s41598-019-38494-x.
Cordell, D., Naif, S., Troch, J., and Huber, C., 2022,
Constraining magma reservoir conditions by in-
tegrating thermodynamic petrological models and
bulk resistivity from magnetotellurics: Geochem-
istry, Geophysics, Geosystems, v. 23, https://doi​
.org​/10​.1029​/2022GC010455.
Degruyter, W., and Huber, C., 2014, A model for erup-
tion frequency of upper crustal silicic magma
chambers: Earth and Planetary Science Letters,
v. 403, p. 117–130, https://doi​.org​/10​.1016​/j​.epsl​
.2014​.06​.047.
Florez, D., Huber, C., Milliken, R.E., and Berkson,
J., 2021, Modeling lunar pyroclasts to probe the
volatile content of the lunar interior: Journal of
Geophysical Research: Planets, v. 126, https://
doi.org/10.1029/2020JE006645.
Ghiorso, M.S., and Gualda, G.A., 2015, An H2O–CO2
mixed fluid saturation model compatible with
Figure 4. Mass budget in the open shallow magma chamber at Sakurajima inferred from the rhyolite-MELTS: Contributions to Mineralogy
model. Error bars show standard deviation from the sampled posterior distribution. Red sym- and Petrology, v. 169, 53, https://doi​.org​/10​.1007​
bols show calculated cumulative volume erupted; blue symbols show the volume of magma /s00410-015-1141-8.
injected over the 500 years simulations. Note that although the values are similar, more magma Gualda, G.A., Ghiorso, M.S., Lemons, R.V., and
is erupted than recharged over the past 500 years. MCMC—Markov chain Monte Carlo. Carley, T.L., 2012, rhyolite-MELTS: A modified

238 www.gsapubs.org | Volume 52 | Number 4 | GEOLOGY | Geological Society of America

Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/52/4/235/6322051/g51763.1.pdf

by Univ of Ryukyu Toshokan 1 Senbaru Nishihara cho user
 calibration of MELTS optimized for silica-rich, University, Kyoto, Japan), v. 24, no. B-1, p. 1–10 Townsend, M., Huber, C., Degruyter, W., and Bach-
fluid-bearing magmatic systems: Journal of Pe- [in Japanese with English abstract]. mann, O., 2019, Magma chamber growth during
trology, v. 53, p. 875–890, https://doi​.org​/10​.1093​ Kobayashi, T., Okuno, M., and Tameguri, K., 2013, intercaldera periods: Insights from thermo-me-
/petrology​/egr080. Kirishima, Sakurajima volcanoes and their source chanical modeling with applications to La-
Hickey, J., Gottsmann, J., Nakamichi, H., and Iguchi, calderas in southern Kyushu: Guidebook of Pre/ guna del Maule, Campi Flegrei, Santorini, and
M., 2016, Thermomechanical controls on magma Post-Conference Field Trips IAVCEI 2013, Aso: Geochemistry, Geophysics, Geosystems,
supply and volcanic deformation: Application ­Kagoshima, Japan, p. B5.1–B5.16. v. 20, p. 1574–1591, https://doi​.org​/10​.1029​
to Aira caldera, Japan: Scientific Reports, v. 6, Nishihara, A., Geshi, N., and Naruo, H., 2022, Long- /2018GC008103.
https://doi​.org​/10​.1038​/srep32691. term change of the eruption activities of Sakura- Yamashita, S., and Toramaru, A., 2020, Control of
Huppert, H.E., and Woods, A.W., 2002, The role of jima volcano, Japan, inferred from the fallout de- magma plumbing systems on long-term erup-
volatiles in magma chamber dynamics: Nature, posits: Frontiers in Earth Science, v. 10, https:// tive behavior of Sakurajima volcano, Japan:
v. 420, p. 493–495, https://doi​.org​/10​.1038​/ doi.org/10.3389/feart.2022.988373. Insights from crystal-size-distribution analy-
nature01211. Permana, T., Nishimura, T., Nakahara, H., and sis, in Vetere, F., ed., Dynamic Magma Evolu-
Iguchi, M., Tameguri, T., Ohta, Y., Ueki, S., and Na- Shapiro, N., 2022, Classification of volcanic tion: American Geophysical Union Geophysi-
kao, S., 2013, Characteristics of volcanic activity tremors and earthquakes based on seismic cor- cal Monograph 254, p. 89–98, https://doi​.org​/10​
at Sakurajima volcano’s Showa crater during the relation: Application at Sakurajima volcano, .1002​/9781119521143​.ch4.
period 2006 to 2011: Bulletin of the Volcanologi- Japan: Geophysical Journal International, Yanagi, T., Ichimaru, Y., and Hirahara, S., 1991, Petro-
cal Society of Japan, v. 58, p. 115–135, https:// v. 229, p. 1077–1097, https://doi​.org​/10​.1093​ chemical evidence for coupled magma chambers
doi.org/10.18940/kazan.58.1_115. /gji​/ggab517. beneath the Sakurajima volcano, Kyushu, Japan:
Ishihara, K., Takayama, T., Tanaka, Y., and Hirabaya­ Toramaru, A., 2022, Vesiculation and Crystallization Geochemical Journal, v. 25, p. 17–30, https://doi​
shi, J., 1981, Lava flows at Sakurajima volcano of Magma: Fundamentals of the Volcanic Erup- .org​/10​.2343​/geochemj​.25​.17.
(I)—Volume of the historical lava flows: Disas- tion Process: Singapore, Springer, 427 p., https://
ter Prevention Research Institute Annuals (Kyoto doi​.org​/10​.1007​/978-981-16-4209-8. Printed in the USA

Geological Society of America | GEOLOGY | Volume 52 | Number 4 | www.gsapubs.org 239

Downloaded from http://pubs.geoscienceworld.org/gsa/geology/article-pdf/52/4/235/6322051/g51763.1.pdf

by Univ of Ryukyu Toshokan 1 Senbaru Nishihara cho user