Investigating Magma Plumbing Beneath Anak Krakatau Volcano, Indonesia: Evidence for Multiple Magma Storage Regions

Examensarbete vid Institutionen för geovetenskaper ISSN 1650-6553 Nr 193

Investigating Magma Plumbing Beneath Anak Krakatau Volcano,

Indonesia: Evidence for Multiple Magma Storage Regions

Börje Dahrén

Copyright © Börje Dahrén och institutionen för geovetenskaper, Berggrundsgeologi, Uppsala universitet.

Tryckt hos Institutionen för geovetenskaper Geotryckeriet, Uppsala universitet, Uppsala 2010

ii L

Referat

Att öka förståelsen för transport och lagring av magma är en stor utmaning för petrologer och vulkanologer. Detta gäller speciellt för explosiva vulkaner, där förståelsen av magma- lagringssystem är mycket viktig för att förutse dynamiska förändringar och därigenom också i riskförebyggande arbete. Denna studie syftar till att undersöka magma-lagringssystemet vid Anak Krakatau, den aktiva vulkanen som befinner sig på kanten av kalderan från Krakataus förödande utbrott år 1883. För detta ändamål tillämpas a.) klinopyroxen-smälta termobarometri (Putirka et al., 2003; Putirka, 2008), b.) plagioklas-smälta termobarometri (Putirka, 2005; Putirka, 2008), c.) klinopyroxen-barometri (Nimis & Ulmer, 1998; Nimis, 1999; Putirka, 2008) samt d.) olivin-smälta termometri (Putirka et al., 2007). Tidigare seismiska (Harjono et al., 1989) och petrologiska (Camus et al., 1987; Mandeville et al., 1996a; Gardner et al., under granskning, J. Petrol.) studier har undersökt denna frågeställning.

De petrologiska studierna påvisar ytligt förvar av magma, vid ett djup av 2-8 km. Den seismiska studien, å andra sidan, identifierade två områden med magmalagring, på djup av ~9 respektive ≥22 km.

Denna studie visar att klinopyroxen för närvarande kristalliserar i mitt i jordskorpan under Anak Krakatau (8-12 km), en nivå tidigare identifierad av seismiska undersökningar (Harjono et al., 1989). Plagioklas visar på ett ytligare förvar (4-6 km), vilket överenstämmer med tidigare petrologiska undersökningar (Camus et al., 1987; Mandeville et al., 1996a; Gardner et al., under granskning, J. Petrol.). Klinopyroxen äldre än 1981 uppvisar större kristallisationsdjup (8-22 km), vilket antyder att magma-förvaringssystemet närmat sig ytan under de senaste ~40 åren. Dessutom sammanfaller de identifierade djupen för lagring av magma med de dominerande litologiska gränserna i skorpan, vilket indikerar att lagringen styrs av diskontinuiteter och densitetsskillnader i skorpan. Denna studie visar således att petrologiska metoder är tillräckligt känsliga för att identifiera magma-lagringsnivåer, även där seismiska metoder misslyckas på grund av begränsningar i upplösning. Kombinerade seismiska och petrologiska studier har därför högre potential att uppnå en mer komplett karaktärisering av magma-lagringssystem vid aktiva vulkaniska komplex.

Nyckelord:

Anak Krakatau; termobarometri; magma-lagringssystem; klinopyroxen; plagioklas.

i Y

Abstract

Improving our understanding of magma plumbing and storage remains one of the major challenges for petrologists and volcanologists today. This is especially true for explosive volcanoes, where constraints on magma plumbing are essential for predicting dynamic changes in future activity and thus for hazard mitigation. This study aims to investigate the magma plumbing system at Anak Krakatau; the post-collapse cone situated on the rim of the 1883 Krakatau caldera. Since 1927, Anak Krakatau has been highly active, growing at a rate of ~8 cm/week. The methods employed are a.) clinopyroxene-melt thermo-barometry (Putirka et al., 2003; Putirka, 2008), b.) plagioclase-melt thermo-barometry (Putirka, 2005), c.) clinopyroxene composition barometry (Nimis & and Ulmer, 1998; Nimis, 1999; Putirka, 2008) and d.) olivine-melt thermometry (Putirka et al., 2007). Previously, both seismic (Harjono et al., 1989) and petrological studies (Camus et al., 1987; Mandeville et al., 1996a;

Gardner et al., in review, J. Petrol.) have addressed the magma plumbing beneath Anak Krakatau. Interestingly, petrological studies indicate shallow magma storage in the region of 2-8 km, while the seismic evidence points towards a mid-crustal and a deep storage, at 9 and 22 km respectively.

This study shows that clinopyroxene presently crystallizes in a mid-crustal storage region (8-12 km), a previously identified depth level for magma storage, using seismic methods (Harjono et al., 1989). Plagioclases, in turn, form at shallower depths (4-6 km), in concert with previous petrological studies (Camus et al., 1987; Mandeville et al., 1996a; Gardner et al., in review, J. Petrol.). Pre-1981 clinopyroxenes record deeper levels of storage (8-22 km), indicating that there may have been an overall shallowing of the plumbing system over the last ~40 years. The magma storage regions detected coincide with major lithological boundaries in the crust, implying that magma ascent and storage at Anak Krakatau is probably controlled by crustal discontinuities and/or density contrasts. Therefore, this study shows that petrology has the sensitivity to detect magma bodies in the crust where seismic surveys fail due to limited resolution. Combined geophysical and petrological surveys offer an increased potential for the thorough characterization of magma plumbing at active volcanic complexes.

Keywords

Anak Krakatau; thermobarometry; magma plumbing; clinopyroxene; plagioclase.

v

Table of Contents

1. Introduction ... 1

2. Field work... 2

3. Geotectonic setting ... 5

4. Previous estimates of magma storage depth ... 9

5. Bulk rock geochemistry... 11

6. Analytical method ... 14

7. Petrography and mineral chemistry ... 15

7.1 Plagioclase phenocrysts ... 15

7.2 Clinopyroxene phenocrysts ... 18

7.3 Olivine ... 19

7.4 Groundmass ... 20

8. Estimates of bedrock density and pre-eruptive volatile content ... 21

9. Method... 22

9.1 Clinopyroxene-melt thermo-barometers ... 22

9.2 Clinopyroxene barometers ... 23

9.3 Plagioclase-melt thermobarometers ... 24

9.4 Olivine-melt thermometers ... 24

10. Results ... 26

10.1 Pressures and temperatures from clinopyroxene-melt thermobarometry ... 26

v L

10.2 Pressure estimates from clinopyroxene barometry ... 29

10.3 Pressures and temperatures from plagioclase-melt thermobarometry ... 30

10.4 Temperature estimates from olivine-melt thermometry ... 31

11. Discussion ... 34

12. Conclusions ... 39

References ... 41

Appendix 1 – Chemical composition of analysed clinopyroxene phenocrysts ... 46

Appendix 2 – Chemical composition of analysed plagioclase phenocrysts ... 50

Appendix 3 – Chemical composition of analysed olivine ... 53

Appendix 4 – Chemical composition of analysed orthopyroxene ... 54

Appendix 5 – Chemical composition of analysed titanomagnetite ... 55

Appendix 6 – Chemical composition of analysed glass and groundmass ... 56

1

1. Introduction

The Krakatau volcano complex, western Java (Indonesia), is one of the most infamous volcanoes worldwide due to the cataclysmic eruption of 1883, being the latest in a sequence of caldera forming events (van Bemmelen, 1949; Camus et al., 1987). In 1927, a new volcanic cone breached the ocean surface, earning the name Anak Krakatau, “child of Krakatau”. The aim of this investigation is to constrain the depth of magma storage region(s) beneath Anak Krakatau, which is approached by employing pressure and temperature modelling calculations that use measured mineral and rock composition data and calibrated thermodynamic formulations. The focus will be on a.) clinopyroxene-melt thermo-barometry (Putirka et al., 2003; Putirka, 2008), and b.) plagioclase-melt thermo-barometry (Putirka, 2005). This will be complemented by a.) clinopyroxene composition barometry (Nimis, 1999; Putirka, 2008) and b.) olivine-melt thermometry (Putirka et al., 2007). The mineral dataset consists of electron microprobe (EPMA) and X-ray fluorescence (XRF) analyses of minerals and rocks erupted between 1883 and 2002. The results will serve as an independent test of previous estimates of magma storage depths derived by geophysical means, plagioclase-melt geobarometry and in- situ isotope stratigraphy (Camus et al., 1987; Harjono et al., 1989; Mandeville et al., 1996a;

Gardner et al., in review, J. Petrol.). Improved knowledge of the magma plumbing system

beneath Anak Krakatau will allow for better understanding and prediction of future activity at

this highly dynamic volcanic complex.

2

2. Field work

In September and October 2008, field work was carried out on the Indonesian islands of Java and Bali, to sample rocks and fumarole gases at 16 active volcanoes. The expedition was part of a project funded by Vetenskapsrådet and Uppsala University. Also, my participation on the trip was made possible by additional funding from Otterborgs donationsfond.

See Fig. 1 map of the field area. According to the Global Volcanism Program, run by the Smithsonian Institute, Krakatau was more or less active between October 2007 and October 2009. Our visit, however, took place during a period of relative quiescence. Notably, a new crater formed in 2007, situated on the southern flank, just below the old summit crater (Fig.

2). Although no eruptions took place during our visit of Anak Krakatau, the evidence of recent eruptions were abundant. The southern flank below the newly formed crater was covered in volcanic ash, and was devoid of vegetation. Volcanic bombs of varying size, with fresh bomb sags (Fig. 3) were scattered around the volcanic cone, especially on the terrace on the eastern flank (Fig. 2). See Fig. 2-7 below for pictures taken during the expedition.

Figure 1. Map of the Sunda Straits, after Gardner et al. (in review, J. Petrol.). The location of the other volcanoes in the north-south trending volcanic lineament is marked with triangles. Inset (b) is a close up on the Krakatau complex, the dashed line indicating the outline of the pre-1883 Krakatau island.

3

Figure 2. The southern flank of Anak Krakatau. Visible is the 1960’s crater rim (black), the pre-2007 summit crater (blue), as well as the currently active crater (red).

Figure 3. Picture taken on the terrace on the eastern flank (visible in Fig. 2), just below the main active cone. This area was strewn with countless volcanic bombs, ranging in size from centimetres to several meters in diameter. Many of the bombs had fresh bomb sags implying that they were recently erupted. This particular bomb was erupted in late 2001/early 2002. In the bottom right corner, one can see destroyed solar panel, associated with the KrakMon surveillance system operating on the Krakatau islands.

Figure 4. A view from inside the active Anak Krakatau crater. Note the fumaroles on the far crater wall.

Other fumaroles, located in the summit crater, were sampled to be used in other studies (e.g. Blythe et al., 2009).

4

Figure 5. A view from the top of Anak Krakatau towards Rakata (Fig. 1), one of the three islets remaining after the collapse of the old Krakatau island in 1883.

Figure 6. A part of the 1883 caldera wall on the island of Rakata. In other words this is a view inside the pre-1883 Krakatau volcano (approximately in the centre of Fig. 5). Below the volcanic rocks, one can see the top of a sequence of what is likely sedimentary rocks, and/or pyroclastic deposits. Note also the dyke swarm cutting trough the rocks of the old Krakatau edifice.

Figure 7. View from beach of Rakata, were the expedition spent the night after the excursion to Anak Krakatau.

The boat in the middle left of the image was used for transportation from the mainland

5

3. Geotectonic setting

Anak Krakatau (Anak) is located in the Sunda Strait between the Indonesian islands of Sumatra and Java (Fig. 1). Geologically, Anak is a part of the Sunda arc where the Indo- Australian plate is subducted beneath the Eurasian plate. In west Java, this occurs at a rate of 67±7 mm per year (Tregoning et al., 1994). The Sunda arc is an active volcanic region with the Krakatau complex being one of the most active parts. Since 1927, Anak has had numerous eruptions, and has grown to a height of ~315m (Hoffmann-Rothe et al., 2006), which converts to an average of 8 cm per week. Anak is a part of a volcanic lineament (Nishimura &

Harjono, 1992), with Panaitan to the south and Sukadana to the north (Fig. 1). This lineament is related to a north-south trending fracture zone, manifested in a shallow seismic belt with foci depths predominantly in the range of 0-20 km (Harjono et al., 1989; Nishimura &

Harjono, 1992; Špičák et al., 2002). Furthermore, the Krakatau complex is located at the intersection of the volcanic lineament and a fault (Nishimura & Harjono, 1992; Deplus et al., 1995), both of which would contribute to a heavily fractured bedrock facilitating magma transport. The projection of this fault is seen in the bathymetric map (Deplus et al., 1995), as indicated by the depressions labeled B and C (Fig. 8). The whole of the Sunda Strait is subjected to extensive faulting and rifting, attributed to the clockwise rotation of Sumatra relative to Java by 20° during the late Cenozoic (Ninkovich, 1976; Nishimura et al., 1986;

Harjono et al., 1991). The angle of subduction changes from near perpendicular (13°) in front of Java to oblique (55°) in front of Sumatra (Jarrard, 1986). The Sumatran rotation has resulted in extension, as reported in Harjono et al. (1991) and associated thinning of the crust to ~20 km in the Sunda Strait, as compared to 25-30 km in Sumatra and west Java (Nishimura

& Harjono, 1992). The micro-seismic study by Harjono et al. (1989) estimated the crustal thickness directly below Anak Krakatau to be ~22 km. The magmatism in the Sunda Strait is thus not strictly subduction zone related, but must also be considered to be to some degree extensional. The influence of the rifting is manifested in Sukadana (Fig. 1), where an 0.8-1.2 Ma old MORB-type basalt is found (Nishimura & Harjono, 1992).

The bimodal nature of the Krakatau complex, with extended periods of basaltic and/or basaltic-andesitic eruptions culminating in colossal caldera forming ignimbrite eruptions (Fig.

9), was discussed by Van Bemmelen (1949), and has since been strengthened by findings of

other authors (Camus et al., 1987; Mandeville et al., 1996a).

6

The formation of the most recent caldera occurred on the 23

of August 1883, when the Krakatau island collapsed (Mandeville et al., 1996). This resulted in a submarine caldera ~100 m deeper than the surrounding sea floor (Fig. 8). Note also that Anak is situated on the north- eastern rim of this caldera. The volume of products of the 1883 eruption has been estimated to be 17-25 km

(Deplus et al., 1995), ~20 km

(Rampino & Self, 1982) or 12.5 km

(Mandeville et al., 1996a). There is evidence of at least two more large ignimbrite eruptions at Krakatau.

Ninkovich, (1979) located two major dacitic ashfalls near the trench ~350 km south and southwest of Krakatau, which he associated with Krakatau and attributed to 60,000 BC and

“recent”. The “recent” ashfall has not been radiometrically dated. However, two estimates, from correlations with historical documents are suggested in the literature, namely 416 AD (Camus et al., 1987) and 535 AD (Wohletz, 2000).

Figure 1. Bathymetric map (Deplus et al., 1995) of the Krakatau complex. Isolines indicate 20 m contours. The depression labeled “A” is the

~240 m deep caldera from the 1883 eruption. Note the location of Anak Krakatau on the north-eastern rim of the caldera. The east-west trending fault intersecting the volcanic lineament is visible as the depressions labeled “B” and

“C”.

Figure 2. The bimodal cyclicity of Krakatau, as reported in van Bemmelen (1949). The composition of the present day eruption products are still dominated by basaltic-andesites (~55 wt% SiO2).

7

Drill cores obtained during hydrocarbon exploration by Pertamina-Aminoil provides information on the bedrock at depth in the Sunda Strait. The closest of these wells (C-1SX) is located ~30 km southeast of Anak (Fig. 1). The C-1SX well penetrated a continuous sedimentary sequence of Quaternary to upper Pliocene age. The lithology is dominated by marine clays and clay-dominated siliciclastic rocks interbedded with volcanoclastic material to a depth of at least 3000 m (Nishimura & Harjono, 1992; Mandeville et al., 1996a).

Findings by (Lelgemann et al., 2000) suggest that the extension and rapid subsidence of the Sunda Strait have created space for up to 6 km graben fill. Thus, the total depth of the sediments and sedimentary rocks below Krakatau can be constrained to between 3 to 6 km.

The Pertamina-Aminoil wells all failed to reach the basement below the sedimentary sequence, but other wells to the southeast of Sumatra and northwest of Java have drilled Cretaceous granites and quartz-monzonites (Hamilton, 1979). The assumption of a sedimentary sequence

underlain by a plutonic

basement below

Krakatau (Harjono et al., 1991) is supported by the findings of sedimentary

(Mandeville et al., 1996b; Gardner et al., in review, J. Petrol.), granitic (Oba et al., 1983) as well as dioritic, gabbroic and meta-basic (Oba et al., 1983;

Gardner et al., in review, J. Petrol.) xenoliths in Krakatau lavas and pyroclastic flows. The crustal velocity model used in the micro-seismic study

Figure 3. Stratigraphy of the bedrock below Anak Krakatau. The lithology is inferred from findings of xenoliths (Oba et al., 1983; Camus et al., 1987;

Mandeville et al., 1996b; Gardner et al., in review, J. Petrol.) and seismic studies (Harjono et al., 1989; Kopp et al., 2001).

8

by Harjono et al. (1989) identifies three boundaries in the crust below Krakatau with distinguishable crustal velocities, which could correspond with lithological boundaries. These boundaries would be at depths of roughly 4, 9 and 22 km respectively, with the lowermost boundary representing the Moho. The upper boundary (4 km) very likely represent the sedimentary-plutonic crustal boundary. The middle boundary (9 km) represent a density contrast, possibly caused by a change in lithology from a light density plutonic rock (e.g.

granite) to a higher density plutonic rock (e.g. diorite or gabbro). See Fig. 10 for a schematic

stratigraphy of the bedrock below Anak Krakatau.

9

4. Previous estimates of magma storage depth

The methods previously employed to estimate magma storage depth beneath Krakatau are a.) plagioclase-melt thermobarometry, b.) chlorine content in melt inclusions, c.) loci of seismic attenuation zones, and d.) in-situ crystal isotope stratigraphy. The results of these studies will be outlined below.

Mandeville et al. (1996a) employed plagioclase-melt thermobarometry (Housh & Luhr, 1991) to estimate the depth of the pre-1883 magma chamber. Their results indicate shallow depths of crystallization for plagioclase, with pressure estimates in the range of 1 to 2 kbar (~4 to 8 km). This was complemented by analysing chlorine content in melt inclusions (Metrich & Rutherford, 1992), resulting in an independent estimate of 1 kbar (~4 km).

Camus et al. (1987) estimated depth of crystallization for plagioclase in rocks erupted between 1883 and 1981, using

plagioclase-melt thermobarometry (Kudo & and Weill, 1970), resulting in estimates of 0.5 to 2 kbar (~2 to 8 km).

Gardner et al. (in review, J. Petrol.) have carried out in situ

Sr/

Sr analyses on plagioclase from the 2002 eruptions, employing LA-ICPMS, and conclude that crystallization of many plagioclase grains must have taken place during assimilation of sedimentary country- rock. This constrains the depth of final crystallization to within the upper three or four kilometres simply on stratigraphic grounds - from drill hole evidence - in agreement with the plagioclase-melt thermobarometry results discussed above.

Harjono et al. (1989) analysed the seismic signature from 14 earthquakes near Anak Krakatau during 1984, using data from analogue seismograms. Two seismic attenuation zones

Figure 11. The two seismic attenuation zones detected below Anak Krakatau, redrawn from Harjono et al. (1989). Circles represent earthquake foci with (open circles) and without (filled circles) S- wave attenuation. The inferred magma storage regions are represented by the red shapes. Note that the resolution in this study is too low to conclude whether or not the two attenuation zones are connected.

10

beneath the volcanic edifice were identified in that study, one a small and irregular zone at a depth of approximately 9 km, and another much larger one at 22 km (Fig. 11). Note that the seismic attenuation zones coincide with the lower boundary of the medium and lower crustal velocity zones discussed above. However, it was not possible to resolve whether the two attenuation zones are connected or not, nor if these represent large volume chambers or a plexus of smaller pockets and chambers.

The present study uses, for the first time, barometry based on clinopyroxene, to provide an

independent and complementary test to these isotopic, geophysical and geobarometric

constraints, with the aim to establish a model of the plumbing system beneath Anak Krakatau.

11

5. Bulk rock geochemistry

Bulk rock chemistry of flows, bombs and ash erupted from Krakatau and Anak is plotted in Fig. 12. The bulk rock data was supplied by Mairi Gardner, a final year PhD student at University College Cork, Ireland, who works with Prof. Troll on other aspects of Anak (Gardner et al., in review, J. Petrol.), or otherwise taken from the literature (Zen &

Hadikusumo, 1964; Self, 1982; Camus et al., 1987; Mandeville et al., 1996a). All oxides have been normalized to 100%, (volatile free) and iron content is reported as FeO

. Note that the bulk rock analyses of lava flows and bombs erupted between 1990 and 2002 carried out by Gardner et al. (in review, J. Petrol.), were done on the exact same samples that were used for the petrographic and microprobe analyses in this study. Analyses of rocks erupted during the 1883 eruption as well as the time period 1960-1981 are also plotted in Fig. 12. Note that, in the TAS diagram (Le Bas et al., 1986) below, all Krakatau rocks plot in the subalkaline field (Fig. 12).

Figure 12. TAS diagram (Le Bas et al., 1986) plotting bulk composition of rocks from Anak Krakatau (black circles) and Krakatau (red circles). The data was taken from the literature. (Zen & Hadikusumo, 1964; Self, 1982; Camus et al., 1987; Mandeville et al., 1996a; Gardner et al., in review, J. Petrol.). The Krakatau rocks include products from the ignimbrite eruption in 1883 as well as older basaltic dyke rocks. The vast majority of the rocks erupted from Anak Krakatau plot in a very narrow region in the basaltic-andesite field, with rocks from single events plotting in the basaltic and andesitic fields.

12

Using the classification of Peccerillo and Taylor (1976), the rock suite plots mostly within the medium-K part of the calc-alkaline series (Fig. 13), in concert with the findings of other authors (Camus et al., 1987). Note, however, that when using the definition of Peacock (1931), the rock suite would be classified as calcic rather than calc-alkaline, with an alkali lime index of 61.6.

Figure 43. K₂O vs SiO₂ plot (Peccerillo & Taylor, 1976) for rocks from Anak Krakatau (black circles) and Krakatau (red circles). Except for two of the basaltic dyke rocks from pre-1883, all rocks plot within the calc-alkaline series, and there almost exclusively within the medium-K part.

The rocks plot in two main groups on a TAS diagram. The pumices and obsidians of the 1883 eruption plot in the dacite-rhyolite field, while the lava flows and bombs from Anak belong to a rather homogenous suite of basaltic-andesites. The exceptions to this would be the 1960-1963 and 1981 eruptions (basalts and acidic-andesites, respectively), representing single events. Note also that several basaltic dyke rocks from the island of Rakata with a similar composition to the 1963 basaltic flows have been reported (Camus et al., 1987). The early history of Anak is not well documented, as very few analyses have been performed on rocks

0 1 2 3 4

48 53 58 63 68

Arc tholeiite series Calc-alkaline series

High-K

calc-alkaline series Shoshonite series

SiO ₂

K 2 O

13

erupted between 1927 and 1960. However, there are indications that the early Anak rocks did not differ significantly from the more recent ones, as silica content in ashes and bombs erupted in the period 1928-1935 has been reported to be in the range of 51.81 to 54.76 wt. % (van Bemmelen, 1949), overlapping with the SiO

content of the recent basaltic-andesites (Fig. 12). This corresponds well with the observation of Camus et al. (1987), that the composition of the tuff ring and lava flows on Anak appeared to belong invariably to the basaltic-andesite suite. Field observations in 2008 also support this as the lava bombs of the 2007-2008 eruptions appear to be virtually identical in composition to the 2002 bombs. Thus, all bulk rock analyses and field observations indicate that the bulk of Anak Krakatau island is made up of basaltic-andesites, with minor components of basalt plus sparse acidic-andesite.

This suggests the presence of a steady state magma storage system under the volcano,

presently producing basaltic-andesite from parental basalt with limited variation of the final

product.

14

6. Analytical method

Mineral chemistry as well as glass and groundmass composition was analysed at Uppsala University (Sweden) using a Cameca SX 50 Electron Probe Microanalyser (EPMA), equipped with three crystal spectrometers (WDS), secondary (SE) and backscattered electron detectors (BSE). The EPMA is an advanced microchemical instrument that is used to determine the chemical composition of e.g. mineral samples. The sample, prepared as a thin section, is beamed with high energy electrons with an accelerating voltage of 20 kV and a current of 15 nA, causing the sample to emit characteristic X-ray signatures, allowing the determination of chemical composition. The electron beam is focused using several electromagnetic lenses.

The diameter of the beam is commonly 1-2 µm, though a beam size of up to 25 µm was used for the analysis of groundmass composition. The wide beam analyses of groundmass included glass and microcrysts, but avoided phenocrysts. Glass compositions were analysed both in the groundmass and in melt inclusions. International reference materials were used for calibration and standardisation (e.g. Andersson, 1997).

Figure 14. Schematic illustration of an Electron Probe Microanalyser (EPMA).

Image adopted from Bundesanstalt für Geowissenschaften und Rohstoffe, Hannover (www.bgr.bund.de)

15

7. Petrography and mineral chemistry

The lavas examined are all highly porphyritic, dark, and partly vesicular. Plutonic as well as sedimentary xenoliths occur. The homogeneity of the bulk chemistry of the rocks is reflected in the petrographic features, as all thin sections examined share the same characteristics, outlined below. The analysed dataset consists of 152 clinopyroxene, 121 plagioclase, 19 olivine, 26 orthopyroxene and 4 titanomagnetite spot analyses where collected from 15 clinopyroxene, 12 plagioclase, 4 olivine and 4 titanomagnetite mineral grains. The relatively small number of analysed minerals was considered adequate due to the relatively homogenous mineral chemistry, especially regarding clinopyroxene. Also, 14 analyses of glass in groundmass and melt inclusions where collected as well as another 14 wide beam (~10x10 µm) analyses of groundmass, in order to be able to calculate an average groundmass composition including both glass and microcrystalline phases. Mineral chemistry of clinopyroxene from Mandeville et al. (1996) and Camus et al. (1987) will be included in the model calculations in order to increase the temporal resolution of the data set. The modal composition is, on average, 70% groundmass, 25% plagioclase, 4% clinopyroxene and less than 1% olivine crystals, as determined from point counting (4 thin sections, 144 points each).

These mineral phases will be outlined below. Representative microphotographs and electron backscatter images of the rocks analysed are displayed in Fig. 15.

7.1 Plagioclase phenocrysts

Plagioclase is the volumetrically dominant phenocryst phase, making up approximately one fourth of the total rock volume. Plagioclase phenocrysts are mostly subhedral, but a few are anhedral. Sieve-like textures are a common feature with numerous melt inclusions present.

The size of the plagioclase crystals are on the order of 0.5-2 mm. Numerous plagioclase

crystals, when viewed under polarized light, appear to have experienced stages of growth

and/or dissolution, as shown by cores that have been partially resorbed at some point before

growth re-commenced (Fig 13a, 13d). Normal as well as reverse zoning has been

documented.

16

Figure 15a. Subhedral plagioclase crystal with highly sieve-textured core and rim regions. This implies a dynamic magmatic system. Crossed polars.

Figure 15b. Sieve-textured plagioclase crystal with numerous melt inclusions. Crossed polars. Note also the very thin overgrowth rim.

Figure 15c. Several intergrown anhedral plagioclase crystals, all with a sieve-texture. Crossed polars.

Figure 15d. Plagioclase crystal. Note that the anomalously dark brown colour of the plagioclase is due to the thin section being thicker than normal (~100 µm).

The outer regions are sieve-textured with a very thin overgrowth of the rims, just like the plagioclase in Fig.

15b. Crossed polars

Figure 15e. A plagioclase crystal with sieve-like texture, having grown around several small clinopyroxenes.

Again, note the sieve-texture and the thin overgrowth on the rims. Crossed polars.

Figure 15f. A plagioclase crystal intergrown with several smaller clinopyroxene crystals. Crossed polars.

17

Figure 15g. Euhedral clinopyroxene crystal. Dark brown, vesicular groundmass consisting of acicular plagioclase, orthopyroxene, titanomagnetite and glass. Crossed polars.

Figure 15h. Euhedral clinopyroxene crystal. Crossed polars.

Figure 15i. Euhedral, clinopyroxene crystal. Crossed polars.

Figure 15j. Euhedral, twinned clinopyroxene crystal. In the bottom left corner, there is a tiny, partly resorbed olivine with very high colours. Note that the olivine rim is covered in discontinuous phases, as can also be seen in Fig. 15k below. Crossed polars.

Figure 15k. BSE image of a partly resorbed olivine (center of image), covered in discontinuous growth of orthopyroxene and titanomagnetite. The olivine crystal is surrounded by the groundmass composed of acicular plagioclase (dark grey laths), anhedral orthopyroxene (light grey), titanomagnetite (white crystals) and glass (irregular dark grey fields).

Figure 15l. BSE image. Intergrown plagioclase (bottom) and clinopyroxene (top). Note the abundance of melt inclusions and vesicles in the plagioclase. The clinopyroxene have several inclusions of opaque minerals, likely titanomagnetite.

18 The lowest anorthite

concentrations (An

) as well as the highest (An

) were found in plagioclase cores. The rims do also vary widely in their composition, with extremes of An

and An

. It is noteworthy that the An% variation between different grains is often greater than

within individual grains. Although the number of analysed individual grains (n=12) is insufficient to distinguish between different plagioclase generations, there seems to be two main groupings, in the range of An

and An

respectively. Other authors have independently reached similar conclusions, finding two plagioclase populations in Anak Krakatau lavas of An

and An

(Gardner et al., in review, J. Petrol.) and An

and An

(Camus et al., 1987). The composition of all plagioclase datapoints analysed for this study is illustrated in Fig. 16.

7.2 Clinopyroxene phenocrysts

Clinopyroxene is the second most abundant mineral phase, though markedly less common than plagioclase, making up approximately 4% of the total rock volume. The clinopyroxene crystals are, with few exceptions, euhedral. Most grains have melt inclusions, though considerably fewer than found in plagioclase. The clinopyroxene crystals are slightly smaller than plagioclase, in the region of 0.2-1.0 mm. Also, it is common to find plagioclase that has grown around clinopyroxene, implying that the main phase of plagioclase growth occurred after clinopyroxene crystallization, the two mineral phases may thus possibly record different levels of magma storage and crystallisation.

The overall composition of pyroxenes, in terms of the endmember mineral components as defined by (Morimoto et al., 1988), is plotted in Fig. 17. It is apparent that all the

Figure 16. Composition of all analyzed plagioclase crystals (n=121).

Composition of plagioclase varies between An45 and An80.

19

clinopyroxenes belong to the same compositional family. Therefore, the average composition of each individual clinopyroxene grain was calculated, which will be used in the thermobarometric model calculations. Note also that the core-rim variations are minor and unsystematic, though a tendency of normal zoning towards slightly Fe-richer rims is often observed. However, the opposite has also been found in a few grains. This would indicate that the clinopyroxenes are close to equilibrium with the basaltic-andesite host rock. Interestingly, Camus et al. (1987) noted that the variation in composition of the clinopyroxenes found in the 1981 acid-andesites relative to the ones in the earlier basaltic-andesites does not differ significantly, implying a common source region for the clinopyroxenes, despite the differences in bulk chemistry between the eruptions. Moreover, in terms of major elements, the composition of the recent (1990-2002) and the old (1883-1981) clinopyroxenes is very similar. There is, however, one notable distinction between the two, namely the Na

O component. The clinopyroxenes from the recent eruptions have Na

O contents of 0.20 ± 0.06 (n=152, range = 0.042-0.40, 1 std), which contrasts the older clinopyroxenes that fall between 0.31-0.50 % (n=15). Although

Na

O is a minor component in clinopyroxene, this is an interesting distinction, as the jadeite (NaAlSi

O

) component of clinopyroxene is supposed to increase (under equilibrium) with increasing pressure under which the melt has crystallised (Putirka et al., 1996; 2003;

Putirka, 2008). This implies that the older rocks may have formed at a deeper level than the recent ones.

7.3 Olivine

Olivine is rather uncommon (<1 wt. %) and has not been identified in all thin sections. The olivines found are very small, with diameters in the order of 0.02 to 0.15 mm, and are best

Figure 17. Composition of clinopyroxenes (filled circles, n=168) and orthopyroxenes (open circles, n=26). The clinopyroxene dataset include composition of minerals analyzed for this study (n=153) and older clinopyroxenes (n=15) reported in the literature (Camus et al., 1987; Mandeville et al., 1996a). All clinopyroxenes plot in a narrow region in the augite field, implying a common source. The composition of the orthopyroxenes found in the groundmass (n=26) is slightly more heterogeneous, spreading over the enstatite and pigeonite fields.

20

identified using electron backscatter images and EPMA analyses. All olivines observed are anhedral (resorbed), frequently with rims covered by over-growths of orthopyroxene and occasional titanomagnetite (Appendix A, image 11). There is a gradual normal zoning in all the olivines investigated towards more Fe-rich rims. Forsterite content is in the range of Fo

, with most olivine rims below Fo

.

7.4 Groundmass

The groundmass consists of 35% glass, 40% microcrystalline laths of plagioclase, 20%

orthopyroxene and 5% opaque phases, as determined by point counting on high magnification

electron backscatter images (7 images, 88 points each). The glass is dark brown to black, and

not transparent under plain-polarized light. The plagioclase in the groundmass, present as

anhedral laths and needles (An

), is chemically very similar to the larger plagioclase

phenocrysts. The orthopyroxenes occur in two modes: as a discontinuous phase on the rims of

partly resorbed olivine, and as microcrysts in the groundmass. No clinopyroxene has been

identified in the groundmass. Only four EPMA analyses were performed on the opaque

phases, all of which were identified as titanomagnetite. Although this is a small number of

analyses, our data coincide with the findings of Camus et al. (1987), who identified

titanomagnetite as the only opaque phase in basaltic-andesites from Anak.

21

8. Estimates of bedrock density and pre-eruptive volatile content

The H

O content is a very influential parameter in a number of the thermobarometers that will be used. The pre-eruptive volatile content of a rock can be approximated from the mass deficiency in EPMA analyses of groundmass glass and melt inclusions („the difference method‟), as described in Devine et al. (1995), provided that the volatiles make up >1%. The mass deficiency in the glass inclusions ranges from 1.2 % to 4.9 % (average = 2.4 %).

However, the precision of estimating volatile concentrations using the difference method is not very high, reported to be ±0.5 % (Devine et al., 1995). Mandeville et al. (1996a) estimated the pre-eruptive volatile content in the 1883 eruptive products to be 4 ± 0.5 wt. %. Due to the enrichment of H

O in magmas during fractional crystallisation, it would be reasonable to assume that the water content in the recently erupted basaltic-andesites are lower than in the considerably more felsic magma of the 1883 eruption. Therefore, the pre-eruptive H

O content will be approximated to be in the range of 2 to 4 wt. % for the thermobarometric calculations to follow. Furthermore, for the reasons stated above, the higher end of that range (3-4 wt. %) will be considered for basaltic-andesite bulk rock compositions, while the lower end (2 to 3 wt. %) will be considered for basaltic bulk compositions. Note that the H

O estimates henceforth will be labeled as X

, were X is the estimate in weight percent.

For the conversion of pressure estimates (kbar) to depth (km), the approximate densities of the respective stratigraphic units below Krakatau need to be established. In the seismic study by Kopp et al. (2001), a seismic line over the Java trench, ending just ~10 km south of Krakatau, was investigated. The stratigraphy proposed for the area close to the Krakatau complex by Kopp et al. (2001) will be used as density constraint applicable for the bedrock directly below Krakatau also (table 1). In Kopp et al. (2001), two different densities are suggested for different parts of the sedimentary succession, 2.23 and 2.4 g cm

, respectively.

For our purpose, an average of the two will be used.

Table 1. Densities of country rock below Anak Krakatau.

Inferred rock types Depth (km)

Density (g cm

) Sedimentary succession 0-4 2.32

Granitoids 4-9 2.8

Diorite/gabbro 9-22 2.95

Mantle >22 3.37

22

9. Method

The available rock forming mineral phases in the rocks are, plagioclase > clinopyroxene >

orthopyroxene > titanomagnetite > olivine. This allows the use of a number of igneous thermometers and barometers. The focus will be on the clinopyroxene-melt thermobarometry (Putirka et al., 1996; 2003; Putirka, 2008) and plagioclase-melt thermobarometry (Putirka, 2005) but will be complemented by other appropriate methods outlined below.

9.1 Clinopyroxene-melt thermo-barometers

Two models based on clinopyroxene-melt equilibria have been applied. The first and most established model is a thermobarometer developed by Putirka et al. (1996), and later calibrated for a wider selection of compositions and P-T conditions (Putirka, 1999; Putirka et al., 2003), including the application to hydrous magmas. This clinopyroxene-melt thermobarometer is an experimental regression model based on the jadeite- diopside/hedenbergite exchange equilibria between clinopyroxene and co-existing melt. The model has proved to be able to recreate P-T conditions for a wide range of magma compositions, within a reasonable margin of error, and has been widely used in the last decade (Shaw & Klügel, 2002; Putirka & Condit, 2003; Schwarz et al., 2004; Caprarelli &

Riedel, 2005; Klügel et al., 2005; Galipp et al., 2006; Mordick & Glazner, 2006; Longpré et

al., 2008; Barker et al., 2009). The Putirka et al. (2003) thermobarometer will henceforth be

termed PTB03. The standard errors of estimate (SEE) for PTB03 are ± 33 °C and ± 1.7 kbar

(Putirka et al., 2003). The second clinopyroxene-melt model to be used for comparison is a

barometer based on the Al partitioning between melt and clinopyroxene, and was recently

presented by Putirka (2008, eqn. 32c). That model is noteworthy as it is especially calibrated

for hydrous systems, requiring the input of a specific H

O estimate. This model will be named

PTB08. PTB08 also requires the input of a temperature estimate, which will be provided by

the PTB03 model. Note that PTB08 is not as firmly tested as PTB03, and is thus not

considered quite as reliable, even though the reported SEE of ±1.5 kbar (Putirka, 2008) is

even better than for PTB03. As PTB03 and PTB08 are based on different clinopyroxene-melt

exchange equilibria (Na and Al, respectively), any overlap of the two models would strongly

imply that the results are reliable.

23

Both PTB03 and PTB08 require the input of a mineral composition data and that of a co- existing melt. The importance of finding a suitable nominal melt, representing the equilibrium conditions of clinopyroxene formation, needs to be stressed, as it is the single largest source of error in mineral-melt equilibria models. This is especially true as there is no definite right or wrong when it comes to choosing a nominal melt, meaning that testing whether the nominal melt of choice represents an equilibrium melt is exceedingly important. Tests of equilibrium are often performed (Klügel & Klein, 2005; Longpré et al., 2008; Barker et al., 2009) using the Fe-Mg exchange coefficients, Kd[FeMg], between clinopyroxene and liquid (Duke, 1976). The Kd[FeMg] expected for a clinopyroxene-melt system in equilibrium would be 0.28 ± 0.08 (Putirka, 2008), and clinopyroxene-melt pairs that fall outside these boundaries will not be considered. As a further equilibrium test, it is useful to take into account the exchange equilibria of other components, such as Na-Al and Ca-Al, as initially suggested by Rhodes et al. (1979) and later expanded on by Putirka (1999). The Putirka (1999) model predicts the relative amounts of different clinopyroxene mineral components that would crystallize from a given nominal melt at the estimated P-T conditions. The predicted mineral components (PMC) can then be compared to the observed mineral components (OMC). If the clinopyroxene and melt compositions are approaching an equilibrium pair, the PMC should closely match the OMC. Here, the focus will be on the dioipside+hedenbergite component (DiHd), as it is the main component of the analysed clinopyroxenes, and it will provide a good complement to the other equilibrium tests that will be performed (Putirka, personal communication Aug. 2009).

9.2 Clinopyroxene barometers

To test the results of the Putirka clinopyroxene-melt thermobarometry (PTB03 and PTB08), a

clinopyroxene barometer not requiring the input of a coexisting melt would be ideal. The

clinopyroxene composition barometer developed by Nimis (1995; 1999) and Nimis & Ulmer

(1998) is widely used, despite having a tendency of systematically underestimating pressures

when applied on hydrous systems (Putirka, 2008). To eliminate the systematic error, this

barometer was re-calibrated for hydrous systems by Putirka (2008, eqn. 32b), with the added

requirement of an H

O estimate in addition to the temperature estimate already needed. This

barometer will be called NimCal08. The SEE for NimCal08 is at 2.6 kbar (Putirka, 2008).

24 9.3 Plagioclase-melt thermobarometers

Plagioclase-melt thermobarometry has been one of the preferred methods for petrologists to estimate pressures and temperatures of igneous systems, likely due to the abundance of plagioclase phenocryst in igneous rocks of varying composition and tectonic setting. Since the first thermometer was formulated (Kudo & and Weill, 1970), the approach has been developed further by various authors and a geobarometer has been incorporated (Housh &

Luhr, 1991; Sugawara, 2001; Ghiorso et al., 2002; Putirka, 2005; Putirka, 2008). Putirka (2005) calibrated the plagioclase-melt thermobarometer for hydrous systems, requiring the input of a H

O estimate in the modelling calculations. The thermometer in Putirka (2005) was later improved slightly (Putirka, 2008, eqn. 24a). Despite all this, the accuracy of plagioclase- melt geobarometry remains underwhelming, reproducing pressures within <3 kbar in most cases and, occasionally and apparently randomly, produces very poor results from some data sets. SEE for the plagioclase-melt thermometer is ± 36 °C and ± 2.47 kbar (Putirka, 2008).

The results of the plagioclase-melt thermobarometer will therefore be evaluated in reference to the findings of Gardner et al. (in review, J. Petrol.), who argue for shallow crustal plagioclase growth in the Anak magma plumbing system.

The recommended equilibrium test for plagioclase-melt thermobarometry uses the ratio of the partitioning coefficients of the anorthite and albite components, Kd[An-Ab]. This is expected to be 0.10 ± 0.05 at low temperatures (T < 1050 °C), or 0.27 ± 0.11 at high temperatures (T > 1050 °C) (Putirka, 2008). Due to the highly variable composition, only the plagioclase datapoints closest to equilibrium with the selected nominal melt will be considered reliable. As a further test for equilibrium, the temperature estimate will be compared to a plagioclase saturation surface temperature calculated for the nominal melt (Putirka, 2008, eqn. 26). The plagioclase saturation surface temperature would be the lowest possible temperature for the nominal melt before plagioclase would start crystallizing. A close match between the temperature estimates of the plagioclase-melt and plagioclase saturation thermometers is expected for equilibrium conditions (Putirka, 2008).

9.4 Olivine-melt thermometers

Olivine-melt thermometers (Beattie 1993; Putirka 2007, eqn. 4) will also be tested on the few

olivines found. To test for olivine-melt equilibrium, the test proposed by Roeder and Emslie

25

(1970) has been used, where the partitioning coefficient of Fe and Mg between olivine and

liquid (Kd[FeMg]) should approach 0.30 ± 0.03. Though this value has since been shown to

vary with pressure and silica- and alkali-content, it remains generally valid at pressures below

20 to 30 kbar (Putirka, 2008). As reported by Putirka (2008), the two models that most

successfully manage to recreate temperatures for olivine-melt equilibria are the Beattie (1993)

and Putirka (2007, eqn. 4) models, henceforth labeled BO93 and PO07, respectively. Though

BO93 is the overall more successful olivine-melt thermometer, it has a tendency of

systematically overestimating temperatures for hydrous systems, a problem the PO07

thermometer is calibrated to avoid (Putirka et al., 2007; Putirka, 2008). PO07 may therefore

be considered the most suitable model (Putirka, 2008). The SEE of PO07 is ± 29 °C (Putirka,

2008). In this study, the main reason for employing olivine-melt thermometry is to provide an

independent test for the reliability of the clinopyroxene-melt thermobarometry. The

temperature estimates from clinopyroxene-melt and olivine-melt thermometry will be

compared and a close match would indicate a high reliability of the results (Longpré et al.,

2008; Putirka, personal communication Aug. 2009). This approach assumes that the

clinopyroxenes and olivines are coeval, and will only be applicable if both clinopyroxene and

olivine perform adequate equilibrium tests using the same nominal melt.

26

10. Results

In this section, the results of the thermobarometric models described above are presented. . The methods employed are a.) clinopyroxene-melt thermo-barometry (Putirka et al., 2003;

Putirka, 2008) b.) clinopyroxene composition barometry (Nimis & and Ulmer, 1998; Nimis, 1999; Putirka, 2008) c.) plagioclase-melt thermo-barometry (Putirka, 2005) and d.) olivine- melt thermometry (Putirka et al., 2007). The clinopyroxene-melt thermobarometry (PTB03 and PTB08) as well as the plagioclase-melt thermobarometry is considered most reliable (Putirka, 2008). The other models mentioned will mainly be used for reference, as overlap in results between different models is a very strong indication of reliable results.

10.1 Pressures and temperatures from clinopyroxene-melt thermobarometry

The first, and arguably most important step, when employing a mineral-melt equilibrium model is to find a suitable nominal melt. As mentioned above, the clinopyroxenes exhibit no obvious signs of being out of equilibrium with the host melt, as they are euhedral and lack any major compositional zoning. However, euhedral habitus and the lack of zoning does not by itself verify that the clinopyroxene was in equilibrium with the host melt, especially considering the sluggishness of clinopyroxene re-equilibration, that is on the order of months to years for a 5-10µm rim (Cashman, 1990), and the rather short repose

time at Anak Krakatau volcano, often with tens of smaller eruptions during a single year.

The most commonly used nominal melts are: a.) bulk rock composition (Caprarelli &

Riedel, 2005; Putirka & Condit, 2003; Putirka et al., 2003) and b.) groundmass or groundmass glass (Shaw & Klügel, 2002; Klügel et al., 2005; Longpré et al., 2008). In

Figure 18. Test for equilibrium using the Kd[FeMg]

between clinopyroxene and melt. The 1963 basalt and 2002 bulk rock both result in Kd[FeMg] values close to the ideal of 0.28 (Putirka, 2008), and are selected as the two viable nominal melt options.

60 65 70 75 80

0 20 40 60

100x Mg# c px

100xMg# melt

1963 Basalt Melt inclusions 2002 bulk rock Groundmass Glass

Out of equilibrium Out of equilibrium

27 addition to these options, two

more nominal melts will be evaluated here. These are melt inclusions (in plagioclase and clinopyroxene) and a more primitive bulk rock from a lava flow from the eruptive period 1960-1963 (henceforth labeled 1963 basalt) reported in Zen and Hadikusumo (1964). In Fig. 18, the Kd[FeMg] of the five nominal melt options outlined above are compared. It is apparent that groundmass and glass are far from equilibrium

conditions with the

clinopyroxene.

Of the remaining three options, the 1963 basalt and the bulk rock fit well within the expected boundaries, while inclusions appear to be only slightly out of equilibrium. In Fig. 19, the observed and predicted DiHd components are plotted, using the (a) 1963 basalt and (b) 2002 bulk rock. Both result in a very good match. It is thus not possible, using only

these methods, to determine whether the 1963 basalt or 2002 bulk rock is the better nominal melt. Both of those nominal melts will be used therefore, resulting in six sets of P-T estimates using the PTB03 and PTB08 models with different input of nominal melts and H

O estimates.

The results of these model calculations are summarised in Fig. 20a-b. The average P estimate of -0.57 kbar (range = -1.81 to 0.43) gained from PTB03 using the 2002 bulk rock as nominal

Figure 19. The predicted vs. observed mineral components of diopside+hedenbergite, using nominal melts (a) 1963 basalt and (b) 2002 bulk rock. Both nominal melts result in very similar results, in terms of predicted mineral compositions, indicating that both need to be considered viable nominal melts. Clinopyroxenes analyzed for this thesis and older clinopyroxenes (erupted 1883-1981) are represented by black and red circles respectively.

0,60 0,80 1,00

Obs erv ed c px C om ponents

Predicted px Components

Nominal melt: 2002 bulk rock

0,60 0,70 0,80 0,90 1,00

Obs erv ed c px C omp onents

Predicted cpx Components

Nominal melt: 1963 basalt

28

melt is an impossible result. Also, using the 2002 bulk rock as nominal melt, there is no overlap of the results from the PTB03 and PTB08 calibrations. This indicate that the 2002 bulk rock is not suitable as a nominal melt composition from which the clinopyroxene has crystallized, and will therefore not be considered further. The three sets of calculated pressures using the 1963 basalt as nominal melt are in a narrow range (0.59 to 3.77 kbar) with a high degree of overlap between PTB03 and PTB08 results. PTB03 and PTB08 were also employed using representative clinopyroxene compositions reported in the literature (Camus et al., 1987; Mandeville et al., 1996a). These clinopyroxenes come from rocks erupted between 1883 and 1981. The results of the old clinopyroxenes are plotted in Fig. 20c, spreading over a much larger P-T interval, and the majority of them record higher pressures and temperatures as compared to the more recent ones. This indicates that the bulk of the clinopyroxenes prior to the acidic-andesite eruption of 1981 crystallized at a greater depth compared to the recently erupted clinopyroxenes. Note that the old clinopyroxenes did not perform quite as well in the equilibrium test as the more recent clinopyroxenes. However, the difference in this equilibrium test between old and recent clinopyroxenes is very slight, and all datapoints are within the allowed 5% deviation. All results of clinopyroxene-melt thermobarometry, using the 1963 basalt as nominal melt, are reported in table 2.

Figure 20. Results of PTB03 (filled circles) and PTB08 using 2H2O (filled triangles) and 3H2O (open triangles).

Recent and old clinopyroxenes are displayed in black symbols and red symbols respectively, Note that, when using the 1963 basalt (a), results of the PTB03 and PTB08 models are overlapping. The 2002 bulk rock (b) does not produce an overlap. This is a strong indication that the 1963 basalt is the most suitable nominal melt. The pressures and temperatures calculated for the old clinopyroxenes (c) are consistently higher than for the recent clinopyroxenes. SEE for PTB03 are ± 33 °C and ± 1.7 kbar. SEE for PTB08 is ± 1.5 kbar

0 1 2 3 4 5

1090 1100 1110 1120 1130

P (kba r)

T (° C)

(a)

Nominal melt:

1963 basalt

0 1 2 3 4 5 6 7 8

1100 1120 1140 1160

P (kba r)

T (° C)

(c)

Nominal melt:

1963 basalt

-2 -1 0 1 2 3 4 5

1080 1090 1100 1110 1120

P (kba r)

T (°C)

(b)

Nominal melt:

2002 bulk rock

29

10.2 Pressure estimates from clinopyroxene barometry

The NimCal08 barometer (Putirka, 2008, eqn. 32b), using temperature estimates calculated using the PTB03 model, yields pressures slightly lower than the PTB03 and PTB08 models.

Estimates of 2

and 3

result in average pressures of 1.03 kbar (-0.57 to 2.14) and 1.48 kbar (-0.11 to 2.59) respectively. As mentioned above, this model is not deemed very precise nor accurate, with a tendency of systematically underestimating pressure (Putirka, 2008).

Despite this, the overall overlap with model results from PTB03 and PTB08 lends further credence to these results. The results of the NimCal08 barometer is plotted in Fig. 21, and table 2, for comparison with the PTB03 and PTB08 models.

Figure 21. Results of the NimCal08 barometer.

Results using 2H2O and 3_H2O are represented by filled and open squares respectively. The partial overlap with results from the PTB03 and PTB08 models lends further credence to these results.

SEE for NimCal08 is at 2.6 kbar.

Table 2. Clinopyroxene-melt thermobarometry and clinopyroxene barometry.

Model Nominal melt

XH2O (%) Recent clinopyroxenes (1990-2002)

Old clinopyroxenes (1883-1981)

P (kbar) T (°C) P (kbar) T (°C)

PTB03 1963 basalt N/A 2.74

(1.46 to 3.77)

1116 (1102 to 1123)

4.85 (2.64 to 7.52)

1131 (1113 to 1154)

PTB08 1963 basalt 2 1.92

(0.59 to 2.83)

N/A 2.78

(1.15 to 5.61)

N/A

3 2.59

(1.26 to 3.50)

3.45 (1.82 to 6.28)

NimCal08 N/A 2 1.03

(-0.57 to 2.14)

N/A 2.41

(-1.72 to 5.66)

N/A

3 1.48

(-0.11 to 2.59)

2.86

(-1.27 to 6.11)

-1

0

1

2

3

1100 1110 1120 1130

P (kb ar )

T (°C)

30

10.3 Pressures and temperatures from plagioclase-melt thermobarometry

Of all the potential nominal melts, the 2002 bulk rock performed best in the Kd[ab-an]

equilibrium test with the plagioclase (Fig. 22), with the majority of the datapoints falling within the field of equilibrium.

The plagioclase-melt pairs outside the field of equilibrium will not be considered for the model calculations. However, at

>3.5

the temperature estimates calculated are all below 1050, requiring equilibrium conditions of Kd[Ab-An] that differ from 0.27 ± 0.11 (Putirka,

2008), i.e. the plagioclase is not in equilibrium with the bulk rock at >3.5

. This effectively constrains pre-eruptive H

O content in the basaltic-andesite to ≤3.5

. A further indication that the 2002 bulk rock is a suitable nominal melt for plagioclase-melt thermobarometry is the fact that the plagioclase saturation surface temperatures calculated are only on average ~10 °C higher than the temperature estimated using the plagioclase-melt thermometer. The results of plagioclase-melt thermobarometry, using 2002 bulk rock, 3

and 3.5

, is displayed in table 3 and Fig. 23. Note that there is no systematic difference between P-T estimates for plagioclase cores and rims. However, there is a strong correlation with An content and P-T estimates. High An contents, resulting in low Kd[An-Ab], correspond to low pressure estimates and vice versa. Plagioclase with medium An content appears to be closest to equilibrium with the bulk rock, with a range of An

yielding Kd[An-Ab] values very close to the ideal of 0.27 (0.24-0.30) determined by Putirka (2008). These “best fit” plagioclases result in a very tight range of temperature and pressure estimates, where T = 1065 to 1071 °C

Figure 22. Equilibrium test for plagioclase and three nominal melt options. The 2002 bulk rock result in the best fit, and will be used in the plagioclase-melt thermobarometry.

0 10 20 30 40 50

0 5 10 15

1000 x A n x A b l iq

1000 x Ab x An liq

2002 Bulk rock Groundmass 1963 Basalt

31

and P = 1.24 to 1.81 kbar (assuming 3

), or 1049 to 1055 °C and 0.72 to 1.24 kbar (assuming 3.5

). These two sets of P-T estimates are considered the most reliable.

Table 3. Results from plagioclase-melt thermobarometry.

T (°C) Saturation surface T (°C)

P (kbar) Kd[An-Ab] = 0.16-0.38

P (kbar) Kd[An-Ab] = 0.24-0.30

Nominal melt X_H2O (%) 1051

(1044 to 1060)

1061 1.06 (0.33 to 1.75) 1.01 (0.72 to 1.24) 2002 Bulk rock 3.5 1067

(1059 to 1076)

1076 1.61 (0.80 to 2.35) 1.56 (1.24 to 1.81) 2002 Bulk rock 3

10.4 Temperature estimates from olivine-melt thermometry

As the olivine phase is clearly not stable in the basaltic-andesite host rock, neither bulk rock chemistry nor groundmass can be considered as feasible nominal melts for olivine- melt thermometry. However, the more primitive bulk rock of the 1963 basalt seems to better represent the magma that gave rise to initial olivine crystallization. Three bulk rock compositions from the 1963 lava flow (Zen & Hadikusumo, 1964) will be compared:

analysis No. 4, 5, and an average of the two. Analysis No 5 has the highest magnesium

Figure 23. Results of plagioclase- melt thermobarometry, using 3.5_H2O (red triangles) and 3_H2O (blue circles). Only the plagioclase compositions closest to equilibrium with the 2002 bulk rock have been used in this plot.

The dotted lines indicate the respective estimated saturation surface temperatures (Putirka, 2008), very close to the calculated temperatures. Note also that there is no systematic difference between P-T estimates for plagioclase cores and rims. SEE for the plagioclase-melt thermobarometer are ± 36 °C and ± 2.47 kbar

0

0,5

1

1,5

2

1040 1050 1060 1070 1080

P (kbar )

T (° C)

3

3.5