Content uploaded by Karen Fontijn
Author content
All content in this area was uploaded by Karen Fontijn on Sep 18, 2017
Content may be subject to copyright.
RESEARCH ARTICLE
A 5000-year record of multiple highly explosive mafic eruptions
from Gunung Agung (Bali, Indonesia): implications for eruption
frequency and volcanic hazards
Karen Fontijn
1,2,3
&Fidel Costa
1
&Igan Sutawidjaja
4
&
Christopher G. Newhall
1,5
&Jason S. Herrin
1,6
Received: 28 January 2015 /Accepted: 27 May 2015
#Springer-Verlag Berlin Heidelberg 2015
Abstract The 1963 AD eruption of Agung volcano was one
of the most significant twentieth century eruptions in
Indonesia, both in terms of its explosivity (volcanic
explosivity index (VEI) of 4+) and its short-term climatic
impact as a result of around 6.5 Mt SO
2
emitted during the
eruption. Because Agung has a significant potential to gener-
ate more sulphur-rich explosive eruptions in the future and in
the wake of reported geophysical unrest between 2007 and
2011, we investigated the Late Holocene tephrostratigraphic
record of this volcano using stratigraphic logging, and geo-
chemical and geochronological analyses. We show that
Agung has an average eruptive frequency of one VEI ≥2–3
eruptions per century. The Late Holocene eruptive record is
dominated by basaltic andesitic eruptions generating tephra
fall and pyroclastic density currents. About 25 % of eruptions
are of similar or larger magnitude than the 1963 AD event, and
this includes the previous eruption of 1843 AD (estimated
VEI 5, contrary to previous estimations of VEI 2). The latter
represents one of the chemically most evolved products
(andesite) erupted at Agung. In the Late Holocene, periods
of more intense explosive activity alternated with periods of
background eruptive rates similar to those at other subduction
zone volcanoes. All eruptive products at Agung show a tex-
turally complex mineral assemblage, dominated by plagio-
clase, clinopyroxene, orthopyroxene and olivine, suggesting
recurring open-system processes of magmatic differentiation.
We propose that erupted magmas are the result of repeated
intrusions of basaltic magmas into basaltic andesitic to andes-
itic reservoirs producing a hybrid of bulk basaltic andesitic
composition with limited compositional variations.
Keywords Agung .Tephrostratigraphy .Eruptive history .
Basaltic andesite .Magma mixing .Magma mingling
Introduction
A detailed view of the eruptive history of a volcano provides
crucial information for hazard assessment at different time-
scales. Historical records of volcanic activity typically only
span a few hundred years, depending on the geographic loca-
tion. In the case of frequently erupting volcanoes, these his-
torical records are sufficient for a relatively good understand-
ing of the typical style of activity at a specific volcano in its
current state and thus short-to-mid-term hazards, e.g. the case
for twentieth century dome-forming eruptions at Merapi
Editorial responsibility: J.E. Gardner
Electronic supplementary material The online version of this article
(doi:10.1007/s00445-015-0943-x) contains supplementary material,
which is available to authorized users.
*Karen Fontijn
Karen.Fontijn@earth.ox.ac.uk
1
Earth Observatory of Singapore, Nanyang Technological University,
50 Nanyang Avenue, N2-01, Singapore 639798, Singapore
2
Department of Earth Sciences, University of Oxford, South Parks
Road, Oxford OX1 3AN, UK
3
Department of Geology and Soil Science, Ghent University,
Krijgslaan 281–S8, 9000 Ghent, Belgium
4
Centre for Volcanology and Geological Hazard Mitigation,
Geological Agency, Jalan Diponegoro 57, Bandung 40122,
Indonesia
5
Present address: Mirisbiris Garden and Nature Center, Salvacion, Sto
Domingo, Albay 4508, Philippines
6
Facility for Analysis Characterisation Testing Simulation, School of
Materials, Science and Engineering, Nanyang Technological
University, Nanyang Drive, N4.1 B4-10, Singapore 639798,
Singapore
Bull Volcanol (2015) 77:59
DOI 10.1007/s00445-015-0943-x
(Voight et al. 2000). Even at frequently active volcanoes, how-
ever, historical records may be far from complete, as was
recently demonstrated for Villarrica volcano in Chile, based
on a high-resolution temporal record of tephra and lahar de-
posits entrained in lacustrine sediments (Van Daele et al.
2014). Mid-to-long-term hazard assessment especially re-
quires information from geological, typically
tephrostratigraphic, studies which may highlight (long-term)
temporal variations in magma composition, eruptive style
and/or frequency and associated hazards (e.g. for Merapi:
Andreastuti et al. 2000; Gertisser et al. 2012; Newhall et al.
2000). At dormant volcanoes with only few or no historically
documented eruptions, tephrostratigraphy is the primary
source of information for volcanic hazard assessment, which
feeds into risk management and mitigation plans. In this paper,
we present the first tephrostratigraphic record for Gunung
(Indonesian for hill or mountain) Agung volcano in eastern
Bali, to understand the past and potential future behaviour of
this highly explosive basaltic andesite volcano.
Gunung Agung is part of the Sunda volcanic arc and forms
the highest peak (3142 m a.s.l.) on Bali, surrounded by the
major calderas of Bratan (Ryu et al. 2013)andBatur(Reubi
and Nicholls 2004a; Sutawidjaja 2009) to the west (Fig. 1),
and the Rinjani–Samalas complex (Lavigne et al. 2013)on
Lombok to the east. Agung is of sacred importance to the
Balinese people, who hold ceremonies and make offerings
to the volcano on a regular basis. Historical texts describing
Balinese history mention eruptions in the sixteenth to eigh-
teenth centuries, with a notably calamitous event in 1711 AD
that caused several hundreds of deaths and destruction on all
sides of the volcano (Hägerdal 2006). Agung also had explo-
sive eruptions in 1808 AD, 1821 AD (uncertain, possibly
mistaken for an eruption from neighbouring Batur) and
1843 AD, for which little or no detail is known from historical
accounts. All two/three nineteenth century eruptions are con-
sidered VEI (volcanic explosivity index; Newhall and Self
1982) 2 events (Siebert et al. 2010), although we provide
evidence here that the 1843 AD eruption must have been
significantly larger (VEI 5, i.e. three orders of magnitude larg-
er than previously thought). The 1843 AD eruption was re-
ported to be preceded by felt earthquakes and to have dis-
persed Bsand, ash and stones^(Zollinger 1845, reported in
Piip et al. 1963; Zen and Hadikusumo 1964).
In 1963, Agung had one of the biggest twentieth century
eruptions worldwide (Self and Rampino 2012). After a few
days of felt earthquakes and small ash eruptions in February
1963, ~0.1 km
3
of andesite lava was extruded until 17th
March, when a large explosive eruption generated pyroclastic
density currents (PDCs) and lahars and devastated a wide area
predominantly north and south of the volcano. A second ex-
plosive phase of similar intensity occurred 2 months later, on
18th May, producing more PDCs and lahars. The total death
toll of the eruption is estimated between 1100 and 1900 (Self
and Rampino 2012; Tanguy et al. 1998; Zen and Hadikusumo
1964). Intermittent, smaller, explosions continued until 20th
June 1963 (Piip et al. 1963), and secondary lahars were gen-
erated for several months after the end of the eruption. A total
estimated magma volume of ca. 0.4 km
3
was erupted (Self and
Rampino 2012), corresponding to a magnitude 5.0 eruption
(Pyle 2000); the bulk volume of pyroclastics (ca. 0.8 km
3
; Self
and Rampino 2012) makes this a high-end VEI 4 event
(Newhall and Self 1982). Self and King (1996)inferredfrom
petrological and geochemical data that a basaltic intrusion
mixed into an andesitic reservoir. The 1963 AD eruption
was one of the first witnessed volcanic eruptions that had a
short-lived climatic impact due to vast amounts of sulphur
being injected into the higher atmosphere, where it formed
10–12 Mt of H
2
SO
4
aerosols (Self and Rampino 2012).
Estimates of the global average temperature decrease as a
result of these aerosols in the months to years after the erup-
tion, vary from 0.1 to 0.4 °C (Angell and Korshover 1985;
Canty et al. 2013; Rampino and Self 1982;Selfetal.1981;
Wigley et al. 2005).
In 1989, weak solfataric activity and a few volcanic earth-
quakes were reported (Global Volcanism Program, http://
volcano.si.edu). Apart from those, Agung has been largely
quiet since 1963. Based on ALOS Interferometric Synthetic
Aperture Radar data, Chaussard et al. (2013) reported inflation
centred on the volcano’s summit at a rate of 7.8 cm/year be-
tween mid-2007 and early 2009, followed by slow deflation
(and no eruption) at a rate of 1.9 cm/year until mid-2011 (the
last acquired data). The 2007–2009 inflation was modelled to
result from a shallow source ca. 4.4 km below Agung’ssum-
mit, or 1.9 km below the average base level of the volcano,
and could reflect the addition of new magma in the shallow
storage levels (Chaussard and Amelung 2012). The deflation
may be explained by cooling and thermal contraction of the
magma (Chaussard et al. 2013) or by passive degassing of
magma that was injected up until 2009 (Girona et al. 2014).
Based on the evidence of active deformation, in 2012 and
2013, a GPS and new seismic network was installed at
Agung by the Indonesian Centre for Volcanology and
Geological Hazard Mitigation and the US Geological Survey
Volcano Disaster Assistance Program. To date, little signifi-
cant deformation or seismicity has been detected (J Pallister,
personal communication 2015).
Previous studies of Agung have focused mostly on the
1963 AD event (Self and King 1996;SelfandRampino
2012) and its climatic impact (e.g. Rampino and Self 1982).
From the geological map by Nasution et al. (2004), it is clear
that Agung has a history of repeated eruptions generating lava
flows, PDCs and lahars. Unlike at its immediate neighbours,
there is no geological evidence to suggest that Agung had a
caldera-forming event in its past. This study evaluates the
Holocene eruptive frequency–magnitude relationships at
Agung using the stratigraphy of pyroclastic deposits, and aims
59 Page 2 of 15 Bull Volcanol (2015) 77:59
to put constraints on the magma plumbing system from petro-
logical and geochemical data on products from some major
eruptions.
Methodology
Stratigraphic logging and sampling of pyroclastic fall and
PDC deposits was carried out on all sides of Agung in 52
locations. Sequences of pyroclastic fall deposits were mostly
found on the western side of the volcano, in the saddle be-
tween Agung and Batur (Fig. 1). Stacks of PDC deposits were
mainly studied in quarries on the southern and northern flanks
of Agung. Samples of well-preserved deposits were taken for
petrological and geochemical analysis. Charcoal enclosed in
PDC deposits and palaeosols between fall deposits was sam-
pled for radiocarbon dating, performed at Beta Analytic Inc.
(FL, USA). Thirty samples were dated, either by the conven-
tional radiometric method or by accelerator mass spectrome-
try, depending on the amount of material available (Table 1).
Polished thin sections of individual scoria and pumice la-
pilli from fall deposits, scoriaceous bombs from PDC de-
posits, and a few pieces of lava, were prepared at High Mesa
Petrographics (NM, USA). Bulk geochemical analysis was
performed at Activation Laboratories Ltd. (Ontario, Canada).
When possible, individual scoria or pumice lapilli were select-
ed for these bulk analyses. In other cases, multiple lapilli were
grouped to obtain enough material for a representative bulk
analysis. Samples were first powdered in a mild steel mill,
after which loss on ignition was determined. Major elements
and selected trace elements were analysed with inductively
coupled plasma–optical emission spectroscopy (ICP-OES)
using lithium-metaborate flux melting to dissolve the pow-
ders. Other trace elements were analysed with ICP–mass spec-
troscopy (ICP-MS). Bulk geochemical data are presented in
Supplementary Table 1.
Major and minor element composition of phenocrysts was
obtained on thin sections and grain mounts of selected sam-
ples with a JEOL JXA 8530-F Field Emission Electron
Microprobe (EMP) equipped with five tuneable wavelength-
dispersive spectrometers at the Facility for Analysis,
Characterisation, Testing and Simulation at Nanyang
Technological University, Singapore. Point analyses on py-
roxene, olivine and magnetite were acquired using a focused
beam, 20 nA beam current and 15 kV accelerating voltage.
Beam current was reduced to 10 nA for plagioclase and glass
(melt inclusions), which were analysed with a beam diameter
of 3 and 5 μm, respectively. Cl and S were measured at a 50-
Fig. 1 SRTM DEM at
3 arcsecond (~90 m) resolution
(Jarvis et al. 2008)ofeastBali
showing the two active
volcanoes, Gunung Batur and
Gunung Agung. Gunung Seraja is
considered inactive and of Lower
Pleistocene age (Purbo-
Hadiwidjojo 1971). Black stars
indicate visited outcrops. Named
outcrops refer to stratigraphic
columns in Fig. 3
Bull Volcanol (2015) 77:59 Page 3 of 15 59
Tabl e 1 Radiocarbon dates of palaeosols between (scoria) fall deposits and charcoal entrained in PDC deposits
Sample Beta
lab code
Method Material Pre-treatment Date (a BP, 1 ) 13CCal. a BP (95.4 %) Cal. ka BP ( ± ; 95.4 %) Stratigraphic Unit
AG001bisD 310663 AMS Charred wood Acid/alkali/acid 790 ± 30 −27.1 672 –760 0.71 ± 0.02 Scoria fall covering PDC
(AG001bisE)
AG001bisA 310672 Radiometric Charcoal Acid/alkali/acid 670 ± 30 −24.6 559 –600 (42.3 %)
631 –677 (53.1 %)
0.62 ± 0.04 PDC (AG001bisB)
AG024G 310674 Radiometric Charcoal Acid/alkali/acid 1200 ± 30 −23.4 1010 –1022 (1.4 %)
1055 –1185 (87.8 %)
1206 – 1236 (6.2 %)
1.13 ± 0.05 PDC (AG024D-E)
AG024A 310673 AMS Charcoal Acid washes 5190 ± 40 −23.1 5773 –5780 (0.3 %)
5795 –5803 (0.4 %)
5891 –6018 (91.3 %)
6081 –6108 (2.3 %)
6157 – 6172 (1.1 %)
5.95 ± 0.05 PDC (AG024B)
AG006bisB 310665 AMS Palaeosol Acid washes 1720 ± 30 −23.8 1560 –1703 1.63 ± 0.04 Between scoria falls
(above AG006bisC)
AG006bisG 309172 AMS Palaeosol Acid washes 1920 ± 30 −24.1 1746 – 1751 (0.3 %)
1812 –1948 (95.1 %)
1.87 ± 0.04 Between ash falls
(AG006bisF-H)
AG006bisJ 309173 AMS Palaeosol Acid washes 2910 ± 30 −26.3 2960 –3158 3.05 ± 0.05 Between scoria falls
(above AG006bisK)
AG006bisP 309174 AMS Palaeosol Acid washes 2860 ± 40 −24.5 2865 – 3078 (93.2 %)
3095 – 3106 (1.2 %)
3129 –3138 (1.0 %)
2.98 ± 0.06 Between scoria falls
(below AG006bisO)
AG006bisU 309175 AMS Palaeosol Acid washes 2900 ± 30 −24.5 2953 – 3083 (78.8 %)
3090 –3156 (16.6 %)
3.04 ± 0.05 Between scoria falls
(below AG006bisT)
AG006bisAC 310664 AMS Palaeosol Acid washes 3040 ± 30 −26.4 3165 –3346 3.25 ± 0.05 Between scoria falls
(below AG006bisAB)
AG006C 310666 AMS Palaeosol Acid washes 3830 ± 30 −26.0 4101 – 4113 (1.5 %)
4147 – 4300 (81.4 %)
4305 – 4317 (1.1 %)
4324 – 4356 (6.2 %)
4367 –4405 (5.2 %)
4.24 ± 0.07 Underlying scoria fall
(AG006B) and Batur
pumice fall (AG006A)
μσσδ
AG006P 309176 AMS Palaeosol Acid washes 2980 ± 30 24.2 3061 –3246 (94.3 %)
3308 – 3321 (1.1 %)
3.15 ± 0.05 Between ash falls
(below AG006O)
AG006T 309177 AMS Palaeosol Ac id washes 4480 ± 30 −23.4 4979 –5008 (4.6 %)
5036 –5290 (90.8 %)
5.16 ± 0.09 Between scoria falls
AG008C 309178 AMS Palaeosol with
charcoal
Acid/alkali/acid 1990 ± 30 −25.0 1878 –1998 1.94 ± 0.04 Between scoria falls
AG008J 310667 AMS Palaeosol Acid washes 1380 ± 30 −25.7 1270 –1344 1.30 ± 0.02 Between scoria falls
(below AG008I)
AG021J 310671 AMS Palaeosol Acid washes 2650 ± 30 −23.5 2740 –2799 (89.5 %)
2818 – 2844 (5.9 %)
2.77 ± 0.02 Between scoria falls
(AG021K-I)
AG021F 310670 AMS Palaeosol Aci d washes 2570 ± 30 −26.5 2508 –2529 (2.5 %)
2537 –2590 (10.2 %)
2616 –2634 (5.2 %)
2697 – 2758 (77.5 %)
2.70 ± 0.07 Between scoria falls
(above AG021E)
AG021A 309180 AMS Palaeosol Acid washes 3010 ± 30 −24.4 3077 –3096 (4.4 %)
3106 –3259 (78.9 %)
3289 – 3335 (12.2 %)
3.20 ± 0.06 Underlying bottom ash fall
(AG021B)
AG042A 310676 AMS Charcoal Acid/alkali/acid 150 ± 30 −26.3 … –41 (17.7 %)
60 –154 (30.0 %)
167 –233 (31.4 %)
241 –284 (16.3 %)
0.15 ± 0.08 PDC (AG042B; 1843 AD)
AG042C 310678 Radiometric Charcoal Acid/alkali/acid 650 ± 30 −26.0 556 –607 (51.9 %)
625 –670 (43.5 %)
0.61 ± 0.04 PDC (AG042D; block-and-
ash flow)
AG042H 310679 AMS Charcoal Acid/alkali/acid 1070 ± 30 −24.5 929 –1010 (74.5 %)
1022 –1055 (20.9 %)
0.98 ± 0.04 PDC (AG042I)
AG042bisC 310677 Radiometric Charcoal Acid/alkali/acid 1130 ± 30 −25.1 962 – 1090 (86.6 %)
1108 – 1145 (5.6 %)
1159 –1173 (3.2 %)
1.04 ± 0.05 PDC (AG042bisD)
AG012D 309169 Radiometric Charred wood Acid/alkali/acid 110 ± 30 −25.9 12 –148 (65.5 %)
187 –205 (2.8 %)
211 –270 (27.1 %)
0.13 ± 0.08 PDC (AG012E; 1843AD)
−
AG017G 309170 Radiometric Charcoal Acid/alkali/acid 1140 ± 30 −25.7 969 – 1096 (78.6 %)
1103 – 1149 (11.3 %)
1158 – 1174 (5.5 %)
1.05 ± 0.05 PDC
AG018E 309171 Radiometric Charred wood Acid/alkali/acid 123.0 ± 0.5 pMC −13.9 Modern Modern PDC (possibly 1963 AD)
AG018D 309179 AMS palaeosol Acid washes 103.5 ± 0.3 pMC −13.0 Modern Modern Between PDCs
AG020bisG 310669 AMS Palaeosol Acid washes 2450 ± 30 −26.2 2360 –2544 (53.2 %)
2558 – 2619 (15.5 %)
2630 – 2703 (26.7 %)
2.54 ± 0.10 Between scoria fall (above,
AG020bisF) and stream
deposits (below)
AG20bisA 310668 AMS Palaeosol Acid washes 3850 ± 40 −25.2 4153 – 4410 4.27 ± 0.08 Underlying scoria fall
(AG020bisB)
AG028A 310675 Radiometric Charcoal Acid/alkali/acid 150 ± 30 −25.5 … –41 (17.7 %)
60 –154 (30.0 %)
167 –233 (31.4 %)
241 –284 (16.3 %)
0.15 ± 0.08 PDC (AG028B; 1843AD)
AG045E 310680 Radiometric Charcoal Acid/alkali/acid 200 ± 30 −26.0 … –25 (19.3 %)
140 –222 (51.2 %)
261 –304 (24.9 %)
0.17 ± 0.09 PDC (AG045D, 200 a BP)
Samples are grouped by section, separated by grey horizontal bars. Analysis was performed at Beta Analytic Inc. (FL, USA), either by conventional
radiometric or AMS (accelerator mass spectrometry) methods. Dates were calibrated using the IntCal13 calibration curve (Reimer et al. 2013) in OxCal
4.2 (Bronk Ramsey 2009). Calibratedages reported at the 95.4 % probability range. Samples are listed in stratigraphic order for each (composite) section,
with the stratigraphically youngest sample on top. BStratigraphic unit^column indicates relation to immediately underlying, immediately overlying or
enclosing deposits which were analysed for bulk geochemical composition (Supplementary Table 1,Fig.3)
59 Page 4 of 15 Bull Volcanol (2015) 77:59
nA current in glass at the end of each analysis. Peak and off-
peak counting times varied between 20 and 40 s, depending
on expected concentration. Results were quantified using
well-characterised natural and synthetic external calibration
standards and a modified ZAF matrix correction procedure.
A time-dependent intensity correction was applied to analyses
of plagioclase and glass to correct for beam modification of
the sample during analysis. Amphibole was not present in
sufficient quantities for reliable analysis. The groundmass
of many samples was too crystal-rich to allow for ma-
trix glass analysis. EMP data are presented in
Supplementary Table 2a–e.
Stratigraphy and bulk geochemistry of pyroclastic
deposits
Individual tephra fall and PDC units were correlated wherever
possible using a combination of stratigraphic, petrological,
geochemical and geochronological constraints. We assumed
at first, and later proved geochemically, that scoria fall de-
posits in sections west of Agung result from Agung due to
dominantly easterly winds dispersing tephra to the west.
Almost no tephra fall deposits are preserved on the eastern
flank of the volcano. PDC deposits found in quarries on all
sides of the volcano are also interpreted to originate from
Agung, based on their spatial distribution and their geochem-
ical composition which is distinct from Batur deposits.
Schematic logs of representative sections exposing tephra fall
and PDC deposits (Fig. 2), with suggested correlations, are
shown in Fig. 3.
The oldest radiocarbon date we obtained on a palaeosol in
the tephra fall sequences is 5.16± 0.09 cal. ka BP (Sample
AG024B; Table 1, Fig. 3).Thecombinedtephrafallsections
provide a record of Late Holocene explosive eruptions that
were large enough to be preserved. The oldest radiocarbon
date on charcoal from a PDC deposit is 5.95± 0.05 cal. ka
BP (Table 1, Fig. 3). The latter is however separated by a
Fig. 2 Representative photos of outcrops. Samples taken for chemistry in yellow, for radiocarbon dating (red stars)inblue. See Fig. 3, Table 1and
Supplementary Table 1
Bull Volcanol (2015) 77:59 Page 5 of 15 59
stratigraphic discordance (stream deposits) from the deposits
higher up in the sequence. Most PDC deposits studied here
were emplaced within the last 1200 years.
Agung’s Late Holocene stratigraphy is dominated by scoria
fall deposits of medium-K basaltic to basaltic andesitic com-
position (Fig. 4; classification following Peccerillo and Taylor
(1976)). Occasional andesitic pumice fall deposits or dis-
persed fine pumice lapilli in palaeosols also occur. These pum-
ice fall deposits are usually poorly preserved, and we could
only obtain geochemical data for the three most recent ones,
including the one we interpret to be associated with the
1843 AD eruption (see further). There is no evidence for large
Plinian-style fall or ignimbrite deposits at Agung, in contrast
to Batur (Reubi and Nicholls 2004a; Sutawidjaja 2009).
PDC deposits are generally more evolved than scoria fall
deposits, spanning a compositional range from basaltic andesite
to andesite, similar to that displayed by the 1963 AD eruptive
products, from the lava flow to the most mafic scoria fall de-
posits (Fig. 4; Self and King 1996). This compositional spread
between types of deposits, the lack of a glassy groundmass and
the compositional range displayed by one eruption (1963 AD)
makes it difficult to firmly correlate units. In addition, the tephra
fall sequences tend to sample a different timeframe than the
PDC sequences. The only Agung deposits we confidently cor-
relate between multiple locations, apart from the 1963 AD de-
posits which occur in the topsoil, are the underlying pumice fall
and PDC units of the 1843 AD eruption. This correlation is
based on the characteristic andesitic chemical composition
(Supplementary Table 1) as well as radiocarbon ages that are
consistent with the historical age of the eruption (Table 1). The
120±40
14
C year BP Aap5 PDC unit defined by Nasution et al.
(2004;
14
C date after Doust 2003), occurring to the northeast of
Agung, most likely also corresponds to the same event.
Nasution et al. (2004) also defined a pyroclastic fall unit with
scoria, pumice and lithic lapilli (Ajp1), which stratigraphically
underlies the 1963 AD PDC deposits (Aap6). Based on its
stratigraphic position and description, this Ajp1 unit likely cor-
responds to the 1843 AD fall deposit, although its thickness
distribution is poorly constrained. The chemical composition
of the second most recent lava flow, Al13 (Nasution et al.
2004), also matches well with that of our 1843 AD deposits
(Doust 2003). Our new localities where the 1843 AD deposits
are found significantly expand their previously known spatial
distribution (Fig. 5). PDC deposits occur on all sides of the
volcano, and suggest a similarly widespread distribution than
that of the 1963 AD PDC deposits. The 1843 AD fall deposit
Fig. 3 Schematic logs of the most complete stratigraphic sections. The
sections west of Agung represent Agung’s history of eruptions generating
tephra fall deposits in approximately the last 3–5 ky. The sections south
and southeast of Agung show a history of repeated PDCs in
approximately the last 1 ky. Correlations are suggested and units named
where possible, but are hampered by the repetitive, dominantly basaltic
andesitic bulk composition of the deposits. Diamonds indicate bulk SiO
2
content; those delineated in black represent samples for which we
acquired mineral compositional data. A useful marker horizon is the
1257 AD pumice fall deposit from Samalas (Lavigne et al. 2013).
Radiocarbon ages are given as the mean and standard deviation of the
calibrated range at the 95.4 % probability range (Table 1). Sample names
refer to those samples forwhich analyses are available, either radiocarbon
dating (black,Table1) or bulk geochemistry (grey, Supplementary
Tab le 1). For brevity, BAG0^was omitted from the sample name
59 Page 6 of 15 Bull Volcanol (2015) 77:59
thicknessof40to70cmalsosuggestsaneruptionsizeatleast
comparable to, and probably larger than, the 1963 AD one (Self
and Rampino 2012).
The chemical composition of the 1843 AD products is
similar to that of the 1963 AD lava flow (Figs. 4and 6),
which was the first to erupt before scoria falls and PDCs
(Self and King 1996). This could suggest that the
1963 AD andesitic lava flow was a leftover of the
1843 AD magma, and would be consistent with the trig-
gering mechanism for the 1963 AD eruption proposed by
Fig. 5 Spatial distribution of 1963 AD and 1843 AD deposits,
suggesting similar magnitudes for both events. Base map data from
Jarvis et al (2008), 500-m contour intervals. Volcano names other than
Agung given for reference. Isopach contours of 1963 AD fall deposit
(Ajp2, units 1–3, corresponding to first explosions and paroxysmal phase
of 17th March) after Self and Rampino (2012); 1963 lava flow (Al14),
PDC (Aap6) and lahar (Alh3) deposits after Nasution et al. (2004). We
interpret Al13 (lava flow), Ajp1 (tephra fall) and Aap5 (PDC) units
defined by Nasution et al. (2004) to correspond to the 1843 AD eruption,
of which we identified deposits on all flanks of the volcano
Fig. 4 K
2
O–SiO
2
classification diagram (Peccerillo and Taylor 1976).
Individual data points represent bulk analyses of Agung rocks, Batur and
Samalas 1257 AD pumice. Some Agung 1963 AD data after Self and
King (1996). Melt inclusion (MI) data represent EMP point analyses of
MI in pyroxene or olivine in Agung scoria fall samples. Compositional
fields for Batur (Reubi and Nicholls 2004b,2005), Rinjani (Foden 1983),
Bratan (Ryu et al. 2013) and Merapi (Costa et al. 2013; Gertisser and
Keller 2003a; Preece et al. 2013) given for reference. Two separate
Merapi fields represent medium- and high-K series, typically the older
(>1.9 ka) and younger (<1.9 ka) Merapi deposits respectively, as defined
by Gertisser and Keller (2003a)
Bull Volcanol (2015) 77:59 Page 7 of 15 59
Self and King (1996), i.e. mixing of basaltic magma into
an existing andesitic reservoir.
Andesitic to (trachy)dacitic pumice fall and PDC deposits
interbedded within the scoria fall sequence (Fig. 3)are
interpreted to originate from Batur Plinian-style eruptions, as
confirmed by their chemical composition (Figs. 4and 6,
Supplementary Table 1; Reubi and Nicholls 2005).
Although the major element compositions of some Batur sam-
ples are similar to those of some Agung samples, especially in
the basaltic andesite range (Fig. 4), certain trace element ra-
tios, e.g. Zr/Nb, are useful to distinguish deposits from Agung
and Batur (Fig. 6). We attribute these subtle differences in
trace element signature between Agung and Batur to
differences in the source rock components and/or the
degree of partial melting generating the parent melts
beneath both volcanoes (e.g. Reubi and Nicholls
2004b). The trace element signatures confirm that all
scoria fall deposits and the few andesitic units we iden-
tified in the studied sequences, all downwind of Agung
and upwind of Batur, originate from Agung. The latter
include pumice at the surface and a PDC unit south of
Agung (samples AG001bisH, AG045D; Supplementary
Tab le 1,Fig.3), which chemically corresponds well
with the Aap3 unit defined on the geological map
(Doust 2003;Nasutionetal.2004) This andesitic PDC
deposit represents the most evolved composition found
for Agung and is chemically clearly distinct from the
1843 AD products. It was dated at 0.17 ±0.10 ka cal
BP (sample AG045E; Table 1) and could correspond
to the 1808 AD eruption.
A twin fine-grained cream-coloured pumice fall unit which
clearly stands out from Agung deposits has proven a useful
marker horizon in several key sections. We interpret this unit
as the 1257 AD pumice fall deposit from the neighbouring
Rinjani–Samalas complex on Lombok. This is consistent with
both its geochemical composition and radiocarbon ages
bracketing the deposit age (Fig. 3, Table 1, Supplementary
Tab le 1; Lavigne et al. 2013).
Petrography and mineralogy
All studied samples from Agung are petrologically complex,
with signs of open-system processes (Figs. 7and 8). Scoria
fall samples generally show a porphyritic texture, with only
few microphenocrysts in a moderately vesicular brown to
black groundmass (Fig. 7a). Scoriaceous bombs and blocks
from PDC deposits have a seriate texture with a crystallised,
poorly to moderately vesicular groundmass and little glass
(Fig. 7b, d). The most evolved samples (>55 % SiO
2
)contain,
in varying proportions, porphyritic inclusions with a lighter-
coloured and more vesicular groundmass than the bulk of the
material, but with similar mineralogical assemblage (Fig. 7c).
The mineralogical assemblage is dominated by plagioclase,
followed by clinopyroxene, orthopyroxene and
titanomagnetite. Most samples also contain olivine and acces-
sory apatite (sometimes included in pyroxene or
titanomagnetite). Rare amphibole isfound in the most evolved
samples and tends to have breakdown rims a few tens of
micrometres wide.
To quantify mineral compositions using EMP analyses, we
selected samples from the three most recent PDC deposits
from section AG042-042bis (Fig. 3), spanning the composi-
tional range for PDCs found at Agung, ca. 53–56 % SiO
2
,
including the 1843 AD deposits. For the scoria fall samples,
we focus on the major deposits from section AG008-021 as it
is the most complete section for the Late Holocene and we can
correlate it well to section AG042-042bis (Fig. 3). Again, this
Fig. 6 Selected trace element variation diagrams for Agung and Batur. a
Zr/Nb vs. Nb; bCe vs. Nb. Individual points represent bulk analyses of
Agung, Batur and 1257 AD Samalas samples. Some Agung 1963 AD
data after Self and King (1996).RN04 Reubi and Nicholls (2004b), RN05
Reubi and Nicholls (2005). Grey arrows indicate fractional crystallisation
paths modelled with Igpet (Carr 2012) using appropriate partition
coefficients from the GERM database (http://earthref.org/KDD/),
suggesting clear source differences between both volcanoes
59 Page 8 of 15 Bull Volcanol (2015) 77:59
selection of samples spans almost the entire compositional
range of fall deposits found at Agung, ca. 51.5–57.5 % SiO
2
.
Plagioclase
Plagioclase shows a variety of zoning patterns, including nor-
mal, reverse and oscillatory zoning (Fig. 8a, b). The larger
crystals, up to 1–1.5 mm in size, often have a sieve-textured
core and a clear oscillatory zoned rim, especially in the basal-
tic to basaltic andesitic samples. Sieve-textured plagioclase is
less abundant in andesitic samples and in the light-coloured
vesicular inclusions. Some plagioclase crystals show dissolu-
tion zones between a clear core and rim (Fig. 8b). Clear,
unzoned, sub- to euhedral plagioclase needles occur in the
groundmass of all samples. Compositions range from An
41
to An
93
(Fig. 9). Core compositions span this entire range in
the most evolved samples (≥55 wt% SiO
2
), but tend to display
a more narrow range in more mafic samples, from around
An
50–60
to An
90
. In these latter samples, plagioclase cores tend
to be more calcic than rims (Fig. 9a, b). Plagioclase rims in
nearly all samples display a more narrow range than cores, of
~An
50–80
(Fig. 9), a wider range than that reported for the
1963 AD products by Self and King (1996).
Pyroxene
Clinopyroxene and orthopyroxeneoccur in roughly equalpro-
portions but are less abundant than plagioclase. Some samples
contain glomerocrysts with pyroxenes, olivine and
titanomagnetite. Most clinopyroxenes have large cores, often
with a patchy appearance and narrow oscillatory zoned rims
(Fig. 8c, d). Apatite, titanomagnetite and melt inclusions are
common. Most clinopyroxenes are sub- to euhedral, but in the
most mafic samples, they have some rounded edges or show
signs of rim dissolution. Compositions range from En
39-
49
Wo
34–45
and thus classify as augite (Morimoto et al. 1988).
Mg# [=100×Mg/(Mg+Fe*); Fe* is total iron] varies between
68 and 86 (Fig. 10a, b), without a clear correlation with bulk
SiO
2
content. There is very little variation between core and
rim compositions (Fig. 10a, b). Al
2
O
3
contents generally vary
between ca. 1.0 and 3.5 wt%, with a few exceptions up to
5.9 wt% in the more evolved samples.
In the most evolved samples, orthopyroxene mostly oc-
curs with large cores, often with a patchy appearance,
surrounded by relatively broad oscillatory zoned rims
(Fig. 8c). Apatite, titanomagnetite and melt inclusions some-
times occur in the patchy cores. The contact between core
and rim is usually sharp but can be irregular. In more mafic
samples, crystal rims tend to be narrower or even absent and
sometimes show signs of dissolution or rounding.
Compositions generally vary between En
65-74
Wo
3–4
, with
Mg#= 67–79 (Fig. 10c, d), except for the cores in the more
evolved samples (especially the PDC ones), which tend to
be less Mg-rich (as low as Mg#=61, Fig. 10d) and show a
range in composition (En
59-71
Wo
3–4
). Orthopyroxene in the
1963 AD products was also commonly found to be reversely
Fig. 7 Selection of optical
microscope images taken under
cross-polarised light. aSample
AG012I, representing the
1963 AD scoria fall deposit. b
Sample AG001bisB, representing
abasaltic–basaltic andesitic PDC
deposit which occurs under the
1257 AD Samalas deposit
(Fig. 3), and with a variety of
plagioclase textures. cSample
AG001bisE, representing a
basaltic andesitic PDC deposit
above the 1257 AD Samalas
deposit (Fig. 3) and with a lighter-
coloured inclusion in a darker
groundmass. dSample AG012E,
representing the 1843 AD PDC
deposits of andesitic composition,
with dominantly a relatively
light-coloured groundmass and
few sieve-textured plagioclase
crystals
Bull Volcanol (2015) 77:59 Page 9 of 15 59
zoned by Self and King (1996). In the other samples, core
and rim compositions show little or no variation.
Olivine
Olivine is found in most samples, except in some of the most
evolved ones. It occurs as sub- to anhedral crystals, sometimes
embayed, and often with reaction rims of orthopyroxene
(En
68-71
Wo
4–5
, Mg#=74–78; Fig. 8e, f). Reaction rims are
thin in the most mafic sample (AG042D). Sample AG042I,
compositionally between AG042D and AG042B, contains
embayed olivine with thick reaction rims of orthopyroxene,
as well as subhedral olivine without reaction rims.
Compositions generally vary from Fo
71
to Fo
80
. Scattered
olivines in scoria fall samples show compositions mostly of
Fo
71–74
, with some exceptions (Fo
57
,Fo
62
) in one of the most
mafic scoria fall samples. Cores are generally slightly more
Mg-rich than rims (excluding reaction rims).
Melt inclusions
A few melt inclusions were found in pyroxene and olivine of
the tephra fall samples. The inclusions do not contain vapour
bubbles or secondary crystals. The glass composition is gen-
erally basaltic andesitic to andesitic, and more evolved than
the bulk composition of the host magma (Fig. 4). Sulphur
contents vary from 120 to 740 ppm, with a tendency for the
higher values to occur in the more mafic melt inclusions.
Chlorine contents range from 1900 to 3300 ppm, with higher
values in the more evolved melt inclusions. These values are
similar to those reported by Self and King (1996)inthe
1963 AD products.
Intensive variables
Pre-eruptive temperatures of touching pairs or intergrowths of
clino- and orthopyroxene were estimated using QUILF
Fig. 8 Selection of backscatter
electron images of mineral phases
analysed by EMP
(Supplementary Table 2). White
X: approximate locations of
analysed points. aSample
AG042B (1843 AD), plagioclase
24, with calcic sieve-textured
core: core An
87–89
,rimAn
67
.b
Sample AG042I, plagioclase 13
showing complex zoning with An
content ranging from 42 to 70 %
(higher An values are lighter
grey). cSample AG042B,
pyroxene pair 20 (clinopyroxene
left, orthopyroxene right);. d
Sample AG042I, clinopyroxene
11 with large core (Mg#= 74–75),
broad intermediate zone (Mg# =
86) and narrow outer rim (Mg#=
79). eSample AG042I, olivine 1
(Fo
80–75
) with orthopyroxene rim.
fSample AG042I, subhedral
olivine 6 (left,Fo
80–79
)and7
(right,Fo
80
)
59 Page 10 of 15 Bull Volcanol (2015) 77:59
(Andersen et al. 1993). In absence of other constraints on
pressure, we set the input pressure at 123 MPa, corresponding
to 4.4 km depth below the summit (using a crustal density of
2800 kg/m
3
), i.e. the depth of the modelled source of inflation
Fig. 9 Histograms of plagioclase
core and rim An composition,
grouped by deposit type and host
rock composition. Less evolved
samples (a,b<55 % SiO
2
)showa
more restricted range of
plagioclase core compositions
than more evolved samples (c,d
≥55 % SiO
2
),andtendtohave
high-calcic cores
Fig. 10 Histograms of pyroxene
core and rim Mg#, grouped by
host rock composition. Both
clino- and orthopyroxene
generally show very little
variation between core and rim
composition. Only in the more
evolved samples, especially of
PDC deposits, orthopyroxene
rims tend to be more Mg-rich than
cores. a,b<55 % SiO
2
.c,d
≥55 % SiO
2
Bull Volcanol (2015) 77:59 Page 11 of 15 59
from 2007 to 2009 by Chaussard and Amelung (2012).
Pyroxene pairs were tested for equilibrium following Putirka
(2008). Estimated temperatures for touching pairs in equilib-
rium vary between 976 and 1085 °C, with a general tendency
for temperatures to be lower in the more evolved samples.
Even the most mafic sample shows relatively low tempera-
tures as well, but we ascribe this to the limited number of
touching pairs that were found in each sample (between 3
and 10 pairs per sample).
Discussion
Magmatic differentiation
Based on our dataset representing compositions erupted at
Agung during the Late Holocene, we construct a generalised
conceptual model of recurring magmatic processes at Agung
leading to highly frequent basaltic andesitic explosive activity.
All samples show petrographical evidence of systematic mag-
ma mingling/mixing, generating the range of erupted
magmas. This is particularly evidenced by the complex zon-
ing patterns and wide range of compositions of plagioclase,
suggestive of crystallisation, and occasional dissolution, under
variable conditions of host melt composition, temperature,
water content and/or pressure. Plagioclase crystals tend to
look more pristine in andesitic samples than in more mafic
samples, and rim compositions in all samples display a more
narrow range than cores. Pyroxene compositions are less var-
iable, but their textures are suggestive of a mafic component
showing disequilibrium features and a more evolved compo-
nent with large cores and rims of a different composition.
Olivine is unstable and recrystallizing to orthopyroxene in
basaltic andesitic samples, but only shows a limited amount
of dissolution in more mafic samples. All these textural obser-
vations together suggest that mingling of a basaltic (andesitic)
magma and an andesitic magma to form a hybrid basaltic
andesite is a ubiquitous process at Agung. During mingling,
both end-member magmas partially re-equilibrate with
the new liquid composition and varying proportions of
mingling components may lead to variations in the bulk
composition of the hybrid magma over a limited com-
positional range.
The few analysed melt inclusions in olivine and pyroxene
from Late Holocene fall deposits, suggest similar pre-eruptive
levels of S and Cl in the melt to those reported for the
1963 AD eruption, which produced an estimated 6.5 Mt of
SO
2
(Self and King 1996; Self and Rampino 2012). If 25 % of
eruptions at Agung produce an amount of H
2
SO
4
aerosols of
the same order of magnitude (Section BEruption frequency
and rate^), Agung is a significant source of regular, strong
atmospheric perturbations.
Eruption style—hazards
The valleys south and southeast of Agung are subjected to PDCs
every ca. 250–300 years (Figs. 1and 3). These are minimum
estimates influenced by erosion, slope morphology and eruption
particularities. The geological map suggests that the N-NE and
SW-SE flanks are most prone to PDCs. Lava flows occur on all
flanks of the volcano. The most silicic ones are generally
restricted to the upper flanks, whereas the most mafic ones have
reached the northeast coastline (Nasution et al. 2004).
Most tephra fall deposits are more mafic than the PDC
deposits, and it is not clear how the two types of deposits are
linked in Agung’s history. The fact that different sequences
sample different time intervals complicates interpretations. In
1963 AD, the PDCs were more mafic than the lava flow and
early fall deposits (Self and King 1996), and in the case of the
1843 AD eruption, all deposits have a similar geochemical
composition. Those two eruptions were both associated with
a lava flow, tephra fall and PDCs, but for older deposits, we
cannot constrain whether all PDC deposits have associated fall
deposits, and vice versa, or whether eruptions typically start
with the effusion of a lava flow, as in 1963 AD.
The majority of PDC deposits at Agung resulted from sco-
ria flows, containing scoriaceous breadcrust bombs (e.g.
AG042B). Some PDC deposits were identified as block-and-
ash flow deposits (e.g. AG042D), which would typically re-
sult from gravitational collapse of a growing lava dome, as
frequently happens at Merapi in central Java (e.g. Voight et al.
2000). The main juvenile component of these PDC deposits
consists of dense decimetre-scale blocks of basaltic andesitic
lava, set in a loose gritty ash matrix. Although the crystallinity
of blocks originating from a lava dome could be expected to
be higher than that of scoriaceous bombs due to slower
cooling rates in the former (e.g. Sparks et al. 2000), a qualita-
tive comparison of both types does not reveal significant dif-
ferences in texture. All blocks and bombs from Agung PDC
deposits however display similar crystallinities to blocks from
Merapi PDC deposits, which are clearly derived from lava
dome collapse (Preece et al. 2013). Despite similarities in
overall crystallinity, we did not spot any Ti-magnetite grains
with exsolution lamellae that would result from stagnation in a
shallow conduit (Turner et al. 2008), as seen for Merapi’s
2010 samples (Preece et al. 2013), in any of the thin sections
that were studied with backscatter secondary electron imag-
ing. We cannot rule out that a lava dome could occasionally
grow at Agung’s summit, but based on the current dataset, it
does not seem likely that lava dome formation is a common
process in Agung’sgeologicallyrecenthistory.
Eruption frequency and rate
Fifty-two tephra-fall-producing eruptions are recognised in
the last ca. 5.2 ky in combined sections AG008/021 and
59 Page 12 of 15 Bull Volcanol (2015) 77:59
AG006/006bis (Fig. 3). This represents an average eruptive
frequency of one eruption every 100 years, seemingly consis-
tent with the general frequency of eruptions described in histor-
ical texts covering the last ca. 500 years (Hägerdal 2006). We
assigned an approximate age to each event, assuming the aver-
age eruptive frequency of one event per century has not
changed significantly over time. This assumption seems valid
from a qualitative inspection of the tephra fall sections (Fig. 3).
With both reference sections AG008/021 and AG006/
006bis 8–10 km downwind of Agung and only 3.5 km apart,
we use deposit thickness and maximum grain size as a proxy
of eruption size relative to that of the 1963 AD eruption (thick-
ness ~45 cm, maximum grain size ~40 mm), and assign an
approximate VEI and associated volume to each event. This
simplified approach assumes that the sections west of Agung
capture the most complete history of tephra-fall-producing
eruptions for Agung due to prevailing easterly winds.
Sections east of Agung clearly contain much less, and much
more poorly preserved, tephra fall deposits, and sections north
and south mostly comprise PDC deposits unlikely to provide a
complete picture of eruptive histories due to potential erosion
in underlying units during emplacement. Deposits which are
clearly thicker and/or coarser than 1963 AD at the type local-
ity, e.g. 1843 AD, are assigned VEI 5 and 1 km
3
(the lower
VEI 5 limit; Newhall and Self 1982). Deposits of similar scale
to 1963 AD are assigned VEI 4 with 0.5 km
3
, except for
1963 AD itself, which is estimated to comprise 0.8 km
3
of
tephra (Self and Rampino 2012). Events smaller than
1963 AD are further subdivided into VEI 3 (0.05 km
3
)and
VEI 2 (0.005 km
3
) using an arbitrary deposit thickness thresh-
old of 10 cm.
In the Late Holocene, almost 10 % of eruptions are signif-
icantly larger than the 1963 AD one and more than 15 % are of
similar intensity. From our conservative volume and age esti-
mates for each event, we obtain a normalised cumulative vol-
ume vs. time evolution for fall-producing eruptions (Fig. 11),
suggesting that Agung experienced a ~10-fold increase in
magma eruptive rates between ~3.2 and 2.0 ka. Rates before
and after this time interval averaged ~0.2–0.3 km
3
/ky (dense
rock equivalent, DRE). During the ~3.2–2.0 ka interval, they
averaged ~2 km
3
/ky. The most recent period of activity at
Agung seems to be characterised by similarly high eruptive
rates. The composition of the magmas leading to these volu-
metrically more important eruptions is not more evolved than
in the calmer periods of activity, the only clear exception being
the 1843 AD eruption.
Reported eruptive rates for explosive activity over the last
0.1–10 ky at basaltic–andesitic subduction zone volcanoes
range from 0.1–0.7 km
3
/ky (e.g. Colima: Luhr and
Carmichael 1982; Kuju: Kamata and Kobayashi 1997;
Lamongan: Carn 2000;Aso:Miyabuchi2009). Apparently,
Agung has experienced periods of relatively high magma
eruptive rates in its recent past, alternating with rates
comparable to those seen at other similar volcanoes. The pro-
cesses controlling the variable eruptive rates remain unclear. A
regional tectonic influence resulting in partial blockage of the
conduit or deeper plumbing systems seems counterintuitive:
we would expect to see (i) more evolved compositions in the
higher-activity periods as a result of prolonged magmatic dif-
ferentiation during calmer periods (e.g. seen at Merapi,
Gertisser and Keller 2003b) and (ii) a similar response of
reduced/increased activity to regional tectonic stress changes
in neighbouring volcanoes. In contrast, it seems more likely
that the Late Holocene history of Agung is characterised by
periodic increased magma supply rates from depth, followed
by limited crustal residence times, and resulting in frequent
intense explosive activity of relatively mafic magmas.
Conclusions
The tephrostratigraphic record of Agung volcano shows that
its eruptive activity in the Late Holocene was dominated by
explosive eruptions generating moderately widespread tephra
fall deposits, mainly impacting the east of Bali. Juvenile scoria
of basaltic andesitic composition, as well as PDC deposits, are
typically confined to the northern and southern slopes. These
PDCs were commonly reworked to produce lahars nearer the
coastline. The compositional range of eruptive products at
Agung is limited to basaltic andesite, and occasionally andes-
ite, e.g. the 1963 AD lava flow and the 1843 AD eruption. The
latter is of at least similar magnitude to the 1963 AD events, as
are about 25 % of the documented deposits. The complex
petrographic relationships record evidence for systematic
open-system magmatic differentiation. The most abundant
mineral phases show textural and compositional evidence of
a repeated history of basaltic intrusions into a slightly more
Fig. 11 Cumulative deposit volume versus time plot of Late Holocene
Agung pyroclastic eruptions. Eruption ages are approximate, assuming a
constant eruptive frequency of 1 event per century (52 events in ~5.2 ky;
Fig. 3). Eruption volumes are estimated by assigning a VEI to each
recognised event, relative to the well-characterised 1963 AD eruption
(VEI 4). Main eruptions are highlighted (sample names/compositions:
Fig. 3, Supplementary Table 1). See text for further details
Bull Volcanol (2015) 77:59 Page 13 of 15 59
evolved reservoir, possibly andesitic, producing hybrid basal-
tic andesitic magmas. S-rich melt inclusions suggest that
Agung is a potential source of regular and significant atmo-
spheric perturbation by frequent emission of large amounts of
SO
2
. The tephra fall record suggests an average frequency of
one explosive eruption per century, with some of the deep
valleys along the southern flank of Agung subjected to
PDCs every few centuries. At the millennium scale, Agung
is characterised by periods of background eruptive rates sim-
ilar to other subduction zone volcanoes, alternated with pe-
riods of increased eruptive rates, ascribed to increased magma
supply rates from depth.
Acknowledgments We thank CVGHM for logistic support during
fieldwork and RISTEK for research permits. We are grateful to Anwar
Sidik, I Nengah Wardhana and Dewa Mertheyash from the Rendang
Volcano Observatory for their hospitality and help in the field. Ryuta
Furukawa is thanked for introductions to key outcrops. Tanya Furman
is kindly acknowledged for sharing the work by Doust (2003). Reviews
by John Pallister and Mary-Ann del Marmol, and editorial handling by
James Gardner were greatly appreciated. Fieldwork and laboratory anal-
yses were funded by the Earth Observatory of Singapore. Data interpre-
tation and writing was performed at Oxford (NERC grant NE/I013210/1)
and Ghent universities.
References
Andersen DJ, Lindsley DH, Davidson PM (1993) QUILF: a Pascal pro-
gram to assess equilibria among Fe–Mg–Mn–Ti oxides, pyroxenes,
olivine, and quartz. Comput Geosci 19:1333–1350
Andreastuti SD, Alloway BV, Smith IEM (2000) A detailed
tephrostratigraphic framework at Merapi Volcano, Central Java,
Indonesia: implications for eruption predictions and hazard assess-
ment. J Volcanol Geotherm Res 100:51–67
Angell JK, Korshover J (1985) Surface temperature changes following
the six major volcanic episodes between 1780 and 1980. J Clim
Appl Meteorol 24:937–951
Bronk Ramsey C (2009) Bayesian analysis of radiocarbon dates.
Radiocarbon 51:337–360
Canty T, Mascioli NR, Smarte MD, Salawitch RJ (2013) An empirical
model of global climate—part 1: a critical evaluation of volcanic
cooling. Atmos Chem Phys 13:3997–4031
Carn SA (2000) The Lamongan volcanic field, East Java, Indonesia:
physical volcanology, historic activity and hazards. J Volcanol
Geotherm Res 95:81–108
Carr (2012) Igpet for Windows. http://home.comcast.net/~carrvolcano/
site/?/page/Igpet_for_Windows_and_M
Chaussard E, Amelung F (2012) Precursory inflation of shallow magma
reservoirs at west Sunda volcanoes detected by InSAR. Geophys
Res Lett 39, L21311
Chaussard E, Amelung F, Aoki Y (2013) Characterization of open and
closed volcanic systems in Indonesia and Mexico using InSAR time
series. J Geophys Res Solid Earth 118:3957–3969
Costa F, Andreastuti S, Bouvet de Maisonneuve C, Pallister JS (2013)
Petrological insights into the storage conditions, and magmatic pro-
cesses that yielded the centennial 2010 Merapi explosive eruption. J
Volcanol Geotherm Res 261:209–235
Doust R (2003) Volcanic hazard assessment Gunung Agung, Bali,
Indonesia. Unpublished MSc Thesis, Pennsylvania State University
Foden JD (1983) The petrology of the Calcalkaline Lavas of Rindjani
Volcano, East Sunda Arc: a model for island arc petrogenesis. J
Petrol 24:98–130
Gertisser R, Keller J (2003a) Trace element and Sr, Nd, Pb and O isotope
variations in medium-K and high-K volcanic rocks from Merapi
Volcano, Central Java, Indonesia: evidence for the involvement of
subducted sediments in Sunda arc magma genesis. J Petrol 44:457–
489
Gertisser R, Keller J (2003b) Temporal variations in magma composition
at Merapi Volcano (Central Java, Indonesia): magmatic cycles dur-
ing the past 2000 years of explosive activity. J Volcanol Geotherm
Res 123:1–23
Gertisser R, Charbonnier S, Keller J, Quidelleur X (2012) The geological
evolution of Merapi volcano, Central Java, Indonesia. Bull Volcanol
74:1213–1233
Girona T, Costa F, Newhall C, Taisne B (2014) On depressurization of
volcanic magma reservoirs by passive degassing. J Geophys Res
Solid Earth 119:8667–8687
Hägerdal H (2006) Candrasangkala: the Balinese art of dating events.
University of Växjö, Sweden, 212 pp
Jarvis A, Reuter HI, Nelson A, Guevara E (2008) Hole-filled seamless
SRTM data V4, International Centre for Tropical Agriculture
(CIAT), available from http://srtm.csi.cgiar.org
Kamata H, Kobayashi T (1997) The eruptive rate and history of Kuju
volcano in Japan during the past 15,000 years. J Volcanol Geotherm
Res 76:163–171
Lavigne F, Degeai J-P, Komorowski J-C, Guillet S, Robert V, Lahitte P,
Oppenheimer C, Stoffel M, Vidal CM, Surono, Pratomo I, Wassmer
P, Hajdas I, Hadmoko DS, de Belizal E (2013) Source of the great
A.D. 1257 mystery eruption unveiled, Samalas volcano, Rinjani
Volcanic Complex, Indonesia. Proc Natl Acad Sci 110:16742–
16747
Luhr JF, Carmichael ISE (1982) The Colima volcanic complex, Mexico:
III. Contrib Mineral Petrol 80:262–275
Miyabuchi Y (2009) A 90,000-year tephrostratigraphic framework of
Aso Volcano, Japan. Sediment Geol 220:169–189
Morimoto N, Fabries J, Ferguson AK, Ginzburg IV, Ross M, Seifert FA,
Zussman J, Aoki K, Gottardi G (1988) Nomenclature of pyroxenes.
Am Mineral 73:1123–1133
Nasution A, Haerani N, Mulyadi D, Hendrasto M (2004) Geological map
of Agung volcano, Bali. Directorate of Volcanology and Geological
Hazard Mitigation, Indonesia
Newhall CG, Self S (1982) The volcanic explosivity index (VEI): an
estimate of explosive magnitude for historical volcanism. J
Geophys Res 87-C2:1231–1238
Newhall CG, Bronto S, Alloway B, Banks NG, Bahar I, del Marmol MA,
Hadisantono RD, Holcomb RT, McGeehin J, Miksic JN, Rubin M,
Sayudi SD, Sukhyar R, Andreastuti S, Tilling RI, Torley R, Trimble
D, Wirakusumah AD (2000) 10,000 Years of explosive eruptions of
Merapi Volcano, Central Java: archaeological and modern implica-
tions. J Volcanol Geotherm Res 100:9–50
Peccerillo A, Taylor SR (1976) Geochemistry of Eocene calc-alkaline
volcanic rocks from the Kastamonu area, Northern Turkey.
Contrib Mineral Petrol 58:63–81
Piip BI, Suyehiro S, Tonani F (1963) Report of the UNESCO volcano-
logical mission to study the Agung volcano. UNESCO, 65 pp
Preece K, Barclay J, Gertisser R, Herd RA (2013) Textural and micro-
petrological variations in the eruptive products of the 2006 dome-
forming eruption of Merapi volcano, Indonesia: implications for
sub-surface processes. J Volcanol Geotherm Res 261:98–120
Purbo-Hadiwidjojo HM (1971) Geological map Bali, scale 1:250.000,
Geological Survey of Indonesia
Putirka KD (2008) Thermometers and barometers for volcanic systems.
Rev Mineral Geochem 69:61–120
59 Page 14 of 15 Bull Volcanol (2015) 77:59
Pyle DM (2000) Sizes of volcanic eruptions. In: Houghton B, Rymer H,
Stix J, McNutt SR, Sigurdsson H (eds) Encyclopedia of volcanoes.
Academic, San Diego, pp 263–269
Rampino MR, Self S (1982) Historic eruptions of Tambora (1815)
Krakatau (1883), and Agung (1963), their stratospheric aerosols,
and climatic impact. Quat Res 18:127–143
Reimer PJ, Bard E, Bayliss A, Beck JW, Blackwell PG, Bronk Ramsey C,
Buck CE, Cheng H, Edwards RL, Friedrich M, Grootes PM,
Guilderson TP, Haflidason H, Hajdas I, Hatté C, Heaton TJ,
Hoffmann DL, Hogg AG, Hughen KA, Kaiser KF, Kromer B,
Manning SW, Niu M, Reimer RW, Richards DA, Scott EM,
Southon JR, Staff RA, Turney CSM, van der Plicht J (2013)
IntCal13 and Marine13 radiocarbon age calibration curves 0–50,
000 years cal BP. Radiocarbon 55:1869–1887
Reubi O, Nicholls IA (2004a) Variability in eruptive dynamics associated
with caldera collapse: an example from two successive eruptions at
Batur volcanic field, Bali, Indonesia. Bull Volcanol 66:134–148
Reubi O, Nicholls IA (2004b) Magmatic evolution at Batur volcanic
field, Bali, Indonesia: petrological evidence for polybaric fractional
crystallization and implications for caldera-forming eruptions. J
Volcanol Geotherm Res 138:345–369
Reubi O, Nicholls IA (2005) Structure and dynamics of a silicic magmat-
ic system associated with caldera-forming eruptions at Batur
Volcanic Field, Bali, Indonesia. J Petrol 46:1367–1391
Ryu S, Kitagawa H, Nakamura E, Itaya T, Watanabe K (2013) K–Ar
analyses of the post-caldera lavas of Bratan volcano in Bali Island,
Indonesia—Ar isotope mass fractionation to light isotope enrich-
ment. J Volcanol Geotherm Res 264:107–116
Self S, King AJ (1996) Petrology and sulphur and chlorine emissions of
the 1963 eruption of Gunung Agung, Bali, Indonesia. Bull Volcanol
58:263–285
Self S, Rampino M (2012) The 1963–1964 eruption of Agung volcano
(Bali, Indonesia). Bull Volcanol 74:1521–1536
Self S, Rampino MR, Barbera JJ (1981) The possible effects of large 19th
and 20th century volcanic eruptions on zonal and hemispheric sur-
face temperatures. J Volcanol Geotherm Res 11:41–60
Siebert L, Simkin T, Kimberly P (2010) Volcanoes of the world, 3rd edn.
University of California Press, Berkeley
Sparks RSJ, Murphy MD, Lejeune AM, Watts RB, Barclay J, Young SR
(2000) Control on the emplacement of the andesite lava dome of the
Soufriere Hills volcano, Montserrat by degassing-induced crystalli-
zation. Terra Nov. 12:14–20
Sutawidjaja I (2009) Ignimbrite analyses of Batur Caldera, Bali, based on
14C dating. J Geol Indones 4:189–202
Tanguy J-C, Ribière C, Scarth A, Tjeptjep WS (1998) Victims from
volcanic eruptions: a revised database. Bull Volcanol 60:137–144
Turner MB, Cronin SJ, Stewart RB, Bebbington M, Smith IEM (2008)
Using titanomagnetite textures to elucidate volcanic eruption histo-
ries. Geology 36:31–34
Van Daele M, Moernaut J, Silversmit G, Schmidt S, Fontijn K,
Heirman K, Vandoorne W, De Clercq M, Van Acker J, Wolff
C, Pino M, Urrutia R, Roberts SJ, Vincze L, De Batist M
(2014) The 600 yr eruptive history of Villarrica Volcano
(Chile) revealed by annually laminated lake sediments. Geol
Soc Am Bull 126:481–498
Voight B, Constantine EK, Siswowidjoyo S, Torley R (2000) Historical
eruptions of Merapi Volcano, Central Java, Indonesia,1768–1998. J
Volcanol Geotherm Res 100:69–138
Wigley TML, Ammann CM, Santer BD, Raper SCB (2005) Effect of
climate sensitivity on the response to volcanic forcing. J Geophys
Res Atmos 110, D09107
Zen MT, Hadikusumo D (1964) Preliminary report on the 1963 eruption
of Mt. Agung in Bali (Indonesia). Bull Volcanol 27:269–299
Zollinger H (1845) Een uitstapje naar het eiland Bali. Tijdschrift voor
Nederlands Indie, jaargang 7, IV, p 43 (in Dutch)
Bull Volcanol (2015) 77:59 Page 15 of 15 59
