Conditions of Magma Storage, Degassing and Ascent at ...€¦ · RECEIVEDJUNE 16, 2009; ACCEPTED...

Conditions of Magma Storage, Degassing andAscent at Stromboli: New Insights into theVolcano Plumbing System with Inferenceson the Eruptive Dynamics

NICOLE ME¤ TRICH1,2�, ANTONELLA BERTAGNINI2 ANDANDREA DI MURO3,1

1LABORATOIRE PIERRE SU« E, UMR9956, CNRS-CEA, CE-SACLAY, FRANCE2ISTITUTO NAZIONALE DI GEOFISICA E VULCANOLOGIA, PISA, ITALY3LABORATOIRE GEOLOGIE DES SYSTEMES VOLCANIQUES, IPGP/UPMC, PARIS, FRANCE

RECEIVEDJUNE 16, 2009; ACCEPTED NOVEMBER 6, 2009ADVANCE ACCESS PUBLICATION DECEMBER 30, 2009

Stromboli is known for its persistent degassing and rhythmic strom-

bolian activity occasionally punctuated by paroxysmal eruptions.

The basaltic pumice and scoria emitted during paroxysms and

strombolian activity, respectively, differ in their textures, crystal con-

tents and glass matrix compositions, which testify to distinct condi-

tions of crystallization, degassing and magma ascent. We present

here an extensive dataset on major elements and volatiles (CO2,

H2O, S and Cl) in olivine-hosted melt inclusions and embayments

from pyroclasts emplaced during explosive eruptions of variable

magnitude. Magma saturation pressures were assessed from the dis-

solved amounts of H2Oand CO2 taking into account the melt compo-

sition evolution. Both pressures and melt inclusion compositions

indicate that (1) Ca-basaltic melts entrapped in high-Mg olivines

(Fo89^90) generate Stromboli basalts through crystal fractionation,

and (2) the Stromboli plumbing system can be imaged as a succession

of magma ponding zones connected by dikes.The 7^10 km interval,

where magmas are stored and differentiate, is periodically recharged

by new magma batches, possibly ranging from Ca-basalts to basalts,

with a CO2-rich gas phase.These deep recharges promote the forma-

tion of bubbly basalt blobs, which are able to intrude the shallow

plumbing system (2^4 km), where CO2 gas fluxing enhances H2O

loss, crystallization and generation of crystal-rich, dense, degassed

magma. Chlorine partitioning into the H2O^CO2-bearing gas

phase accounts for its efficient degassing (�69%) under the

open-system conditions of strombolian activity. Paroxysms, however,

are generated through predominantly closed-system ascent of basaltic

magma batches from the deep storage zone. In this situation crystalli-

zation is negligible and sulfur exsolution starts at �170MPa.

Chlorine remains dissolved in the melt until lower pressures, only

16% being lost upon eruption. Finally, we propose a continuum in

explosive eruption energy, from strombolian activity to large paroxys-

mal events, ultimately controlled by variable pressurization of the

deep feeding system associated with magma and gas recharges.

KEY WORDS: Stromboli; melt inclusions; magmatic volatiles;

CO2 fluxing; magma degassing

I NTRODUCTIONThe dynamics of magma ascent and degassing stronglycontrol the explosivity of basaltic magma eruptions,magma fragmentation, texture and dispersion of products.Slow magma ascent and pure open-system degassing oroutgassing have been associated with low-intensity strom-bolian activity because of differential transfer of coalescinggas bubbles (e.g.Wilson & Head,1981) or bubble accumula-tion, foam collapse and gas pocket (slug) migrationupwards in the conduits (e.g. Jaupart & Vergniolle, 1988).At the opposite end of the spectrum, high-intensity explo-sive activity is expected to result from rapid magmadecompression and ascent during which the melt remains

�Corresponding author. Present address: Laboratoire Pierre Su« e,UMR9956, CNRS-CEA, CE-Saclay, France.E-mail: [email protected]

� The Author 2009. Published by Oxford University Press. Allrights reserved. For Permissions, please e-mail: [email protected]

JOURNALOFPETROLOGY VOLUME 51 NUMBER 3 PAGES 603^626 2010 doi:10.1093/petrology/egp083

Downloaded from https://academic.oup.com/petrology/article-abstract/51/3/603/1531471/Conditions-of-Magma-Storage-Degassing-and-Ascentby gueston 16 September 2017

in equilibrium with its coexisting gas phase and the mix-ture (gasþmelt) expands as a closed system (Cashman,2004). Actually, the notion of closed-system degassing maybe ambiguous. It was recently proposed that basalticmagmas may be fluxed by CO2-rich bubbles of deep deri-vation during their ponding or storage at crustal depths, aprocess that may enhance magma dehydration (Spilliaertet al., 2006; Johnson et al., 2008; Collins et al., 2009). Thisprocess is supported by the high CO2 flux at basaltic volca-noes and ultimately should influence the crystallizationpath of magmas (for a review, see Me¤ trich & Wallace,2008).In the Aeolian island arc (Fig. 1), Stromboli represents

a suitable target volcano to address the question of the rela-tionship between the eruptive dynamics and conditions ofmagma ascent and crystallization with a focus on the spe-cific role of the CO2-rich gas phase. Stromboli is famousfor its rhythmic strombolian activity, which has persistedover 1800 years (Rosi et al., 2000) and is thought to bedriven by differential transfer of gas slugs (e.g. Chouetet al., 1974; Jaupart & Vergniolle, 1988; Allard et al., 1994;Burton et al., 2007a). Moreover, Stromboli is well knownfor its large persistent degassing (e.g. Francis et al., 1993;Allard et al., 1994), which contributes to �95% of the bulkdischarge of SO2 (on average 250�50 t/day), a relativelyhigh CO2 flux (on average 1·4�103 t/day) and lowmagma extrusion rate (Allard et al., 2008). Besides thedominant low-intensity explosive activity, more energeticexplosive episodes, so-called paroxysms, have periodicallyoccurred since the beginning of the present-day activity(Rosi et al., 2000). Paroxysms are impulsive, short-lived

events consisting of a sequence of blasts from differentcraters. The higher magma discharge rate during parox-ysms, relative to normal strombolian activity, results inthe emission of variable volumes of pumice (from �103 to�105 m3; Bertagnini et al., 2003). As documented byRittmann (1931), the most energetic and hazardous explo-sive events (large-scale paroxysms) generate convectiveplumes up to �10 km high and the largest tephra volumes.Pumice fallouts spread over the sea, pumice bombs andmeter-sized fluidal spatters coat the northern and the west-ern volcano flanks, and meter-sized ballistic blocks arelaunched a distance of up to 2 km. During the last two cen-turies only 25 large paroxysms have been documented(Barberi et al., 1993). During small-scale events (e.g. inAugust 1998 and 1999), both the volume and dispersion ofthe products are smaller. Decimeter-sized blocks andbombs are confined within a few hundred meters from thecraters and small transient convective columns producesparse fallout of light pumice lapilli on the volcano slopes.In the last 15 years, about 24 well-documented paroxysmshave occurred and are classified as small-scale paroxysms(Bertagnini et al., 2008); intermediate-scale events occurredon 5 April 2003 and 15 March 2007.The petrology and mineralogy of basalts of the last�1800 years have been widely described (Francalanciet al., 1999, 2004, 2005; Me¤ trich et al., 2001; Bertagniniet al., 2003, 2008; Landi et al., 2004, 2006). The basalticmagmas emitted during strombolian activity, lava flow epi-sodes, and paroxysmal eruptions have closely similarbulk compositions in major elements, irrespective of theamounts of crystals. The crystal-rich basalt emitted asdense, black scoria during strombolian activity and aslava flowsçthe so-called high porphyritic (HP) magmaor basalt (Francalanci et al., 2004)çis partially degassedand highly viscous (�104 Pa s). It contains on average45^55 vol. % of plagioclase, clinopyroxene, and olivine inequilibrium with a shoshonitic (SiO2 �52wt %) residualmelt. On the other hand, the crystal-poor basalt eruptedas ‘golden’ highly vesicular pumice during paroxysmsçso-called low porphyritic (LP) magma or basalt(Francalanci et al., 2004)çis rich in dissolved volatiles(43wt %) and has a low viscosity (7^15 Pa s; Me¤ trichet al., 2001; Pichavant et al., 2009). Because of their low con-tent of crystals (55 vol. % of olivine and clinopyroxene),the glassy matrices of the pumices are basaltic, as are thebulk-rocks. During paroxysmal eruptions, the HP and LPmagmas mingle in the upper parts of the volcano plumb-ing system (Bonaccorso et al., 1996; Bertagnini et al., 1999).Both mineralogical and melt inclusion studies have sug-gested a rather distinct transition between the deep-derived, volatile-rich LP magma and the shallow HPmagma. Such a process would require a mechanismallowing very efficient water degassing and a high crystal-lization rate.

Fig. 1. Map of Stromboli in the Aeolian arc (Italy). Stars refer to thesampling sites of pumices (ST531, 127, 82) produced during large-scaleparoxysms. Pumice lapilli emitted on 15 March 2007 were sampledoffshore Punta Lena. The samples of the August 1998 and August1999 small-scale eruptions were collected in the summit crater zone.

JOURNAL OF PETROLOGY VOLUME 51 NUMBER 3 MARCH 2010

604Downloaded from https://academic.oup.com/petrology/article-abstract/51/3/603/1531471/Conditions-of-Magma-Storage-Degassing-and-Ascentby gueston 16 September 2017

This study aims to elucidate the conditions of magmaascent, crystallization and degassing as a function of theeruptive dynamics, with emphasis on the possible effect ofCO2 fluxing on magma dehydration, the transitionbetween the LP and HP magmas, and the bulk degassingbudget of the volcano. Melt inclusion volatile and majorelement concentrations have previously failed to provide acomprehensive constraint on the conditions of magmadegassing at Stromboli because of the small size of theinclusions and their hosts. To overcome this limitation andfulfill our objectives we adopted new strategies of analysisand sampling. In particular, we have developed themicro-Raman analysis of water in basaltic glasses, whichhas been proved to be efficient and accurate (Di Muroet al., 2006; Mercier et al., 2009). The new analyses of water(�80) were then combined with measurements of C, Sand Cl in melt inclusions and embayments, using nuclearand electron microprobes. The selected samples are repre-sentative of the entire range of paroxysmal events, fromlarge-scale paroxysms to the intermediate (15 March2007) and small-scale (23 August 1998) events. Moreover,we have complemented our previous dataset on the shal-low degassed magma. Combining the new and previousdata on the major element geochemistry of melt inclusionsand the pressure-related behavior of volatiles (CO2, H2O,S, Cl) provides strong constraints on the conditions ofmagma storage, crystallization, degassing and eruption atStromboli.

ERUPT IVE PRODUCTS ANDSAMPLE SELECTIONPumiceLarge-scale paroxysms

We sampled the eruptive products of two of the most recentlarge-scale paroxysms (ST127 and ST531; Fig. 1). SampleST127 is made up of centimeter- to decimeter-sizedpumice bombs recovered at 540m above sea level (a.s.l.)on the NE volcano flank (38·797778N, 15·2199948E). Thisarea is regarded as the source region of one of the hot ava-lanches generated during the 1930 paroxysm by the slidingof the fall deposit accumulated on the steep sandy slopesof the volcano (Rittmann, 1931). On the basis of its strati-graphic position, we tentatively correlate this sample tothe 1930 eruption. The ST531 pumice was sampled in an�30 cm thick deposit emplaced by a paroxysmal event at550ma.s.l. along the path to the summit on the east flankof the Sciara del Fuoco (38·8000858N, 15·2158278E). Thedeposit comprises an incipiently welded spatter layer, con-sisting of decimeter-sized pumice and scoria bombs, overly-ing a 5 cm thick, coarse ash and fine lapilli basal layer.Paleomagnetic data indicate for this paroxysmal eventan age between the 15th and 17th century AD (Speranzaet al., 2004).

In addition we acquired new data on a previouslydescribed pumice sample (sample ST82) (Me¤ trich et al.,2001) in which we discovered high-Mg olivines in the¼^½ mm fraction that led us to update our dataset onthe most primitive melts. These pumice clasts were col-lected in a 1cm thick tephra layer encountered at a depthof 1·50m in a stratigraphic trench dug at nearly500ma.s.l. on the NE flank of the volcano (38·7955408N,15·2238668E; Trench 2 of Rosi et al., 2000).

Intermediate-scale paroxysms

The 15 March 2007 event occurred during an effusive crisislasting from 27 February to 2 April 2007 and was wellmonitored (Barberi et al., 2009). A short-lived convectivecolumn fed the fallout of ash and pumice lapilli on the vol-cano slopes as far as the sea, and meter-sized blocks wereejected to a distance of 1·5 km from the craters. We focushere on pumice lapilli floating on the sea offshore PuntaLena (Fig. 1) that were collected byT. Caltabiano (INGV,Catania) immediately after the paroxysm. These sampleshave provided the opportunity to obtain a complete data-set on olivine-hosted melt inclusions for this type of explo-sive event that greatly complements our previous resultsobtained on pumices from the 5 April 2003 paroxysm(Me¤ trich et al., 2005).

Small-scale paroxysms

The 23 August 1998 explosive event generated a 1-km-highconvective column, lasting a few seconds, and produced ashower of spongy and fibrous lapilli on the ESE slope andthe fallout of spatter and bombs in the summit area(Bertagnini et al., 1999). The multi-clast sample (ST130),collected at a distance of a few hundred meters from theactive craters (Fig. 1), is formed of pumice bombs andfibrous pumice lapilli.

Scoria samplesThe selected samples represent the HP magma stored inthe upper part of the plumbing system that was expelledduring small-scale paroxysms of 23 August 1998 (ST133B)and 26 August 1999 (ST182, ST177). They were collectedin the summit crater area and include two centimeter-sizedspatter clasts (ST133 and ST177) and highly vesicular scor-iaceous bombs (ST182). These samples were analyzed tocomplement previously studied scoriae entrained duringlarge paroxysms (i.e. ST82, Me¤ trich et al., 2001) to builtup a comprehensive dataset on the HP basalt emittedduring explosive events of variable energy.

ANALYT ICAL PROCEDURESOlivine and clinopyroxene crystals were hand-pickedunder a binocular microscope from the 0·25^1mm sizefraction of gently crushed scoria and pumice. The mineral

ME¤ TRICH et al. STROMBOLI VOLCANIC PLUMBING SYSTEM


phases were embedded in epoxy resin, and polished toexpose their glass inclusions.Major element concentrations of melt inclusions were

measured using an SX50 CAMECA electron probe(Camparis, Jussieu, France), with a 10 nA defocusedbeam and a counting time varying from 10 to 25 s.Concentrations of minor elements (S, Cl, P) were deter-mined using a 30 nA defocused beam and long countingtime on peak (120 s). We took into account the Ka wave-length shift of sulfur. Accuracy and precision of the S, Cland P analyses were checked on reference glasses (basaltsAlV981R23 and Vg2, pantellerite KE12) as previouslydescribed (Me¤ trich et al., 2001). Composition profiles inolivine were measured at 30 kV acceleration potential,with a 300 nA focused beam, and 120 s counting time forCa, Mn, P and Ni. Olivines were also analyzed andmapped with a Philips XL30 scanning electron micro-scope equipped with EDAX DX 4 (Dpt Scienze dellaTerra, Pisa, Italy).Carbon concentrations were determined with the

Laboratoire Pierre Su« e nuclear microprobe (Saclay,France), using the 12C(d, p)13C nuclear reaction. Analyseswere made in the scanning mode, with a 5 mm beam.The integrated charge varied from 2 to 1·5 mC, implyinga minimum detection limit between 60 and 70 ppmof CO2, respectively. Carbon concentrations were cali-brated against a scapolite standard (6800 ppmC) and awater-bearing Etna basaltic glass (EtII-1) containing300�30 ppmC (Spilliaert et al., 2006). The latter glassconcentration is reproduced with 10% precision.Micro-Raman spectroscopic analyses performed on

glasses (glass inclusions, embayments, glassy matrices)from Stromboli allow us to check for the possible occur-rence of micro-crystallization and quantify the totalamount of dissolved water (H2OþOH). Analyses werecarried out at Laboratoire Pierre Su« e (Saclay, France)using a Renishaw inVia instrument. The spectrometer isfitted to a Leica microscope with a 100� objective toexcite and collect the signal from a very small area (about1 mm in diameter). Analyses were performed at the samplesurface in confocal mode. The excitation source was agreen diode laser (532 nm) filtered to deliver a low poweron the sample (c. 7 mW) during short counting times(5100 s). The selected analytical conditions minimize scat-tering from the embedding medium and prevent sampleheating and oxidation. External and internal analyticalprocedures (Di Muro et al., 2006) for the intensity of theH2OT band (3580 cm^1) and scaled intensity of the H2OT

band to that of the silicate network (980 cm^1), respectively,were calibrated on a set of basaltic glass standards fromEtna (H2OT¼ 0·40^3·04wt %). Reference glasses werepreviously characterized by Karl Fisher titration, Fouriertransform IR (FTIR) and micro-Raman spectroscopy(Mercier et al., 2009). Analyses affected by focusing errors

were easily detected and corrected by cross comparison ofthe two procedures. Cubic baseline correction was appliedbefore determination of band intensities. One to threeanalyses were performed for each inclusion and then aver-aged. Error propagated on the external calibration proce-dures ranges between 6 and 10% relative.

PUMICE: TEXTURES ANDCOMPOSIT IONS OF MINERALSAND OF THEIR INCLUSIONSPumice is produced exclusively during paroxysms(Bertagnini et al., 2008). The LP pumice samples studiedhere carry euhedral crystals entrained from the shallowHP magma at the time of eruption and inherited crystalswith patchy zoning that testify to successive events of disso-lution and crystallization as commonly described inStromboli pumice (e.g. Bertagnini et al., 2003; Francalanciet al., 2004).We focus here on the few per cent (53^5%) ofcrystals that are euhedral or have dynamic crystallizationtextures, with melt inclusions preserved as glass upon cool-ing, as confirmed by micro-Raman spectroscopy. Tinynuclei (51 mm) of Fe-oxides that were detected by the laserbeam only in some clear glassy inclusions were related tonucleation upon cooling. Clinopyroxene, the dominantmineral phase in pumice, usually contains small andpartly crystallized melt inclusions. Some of them were ten-tatively homogenized at 1150^11608C and 1 atm in a verti-cal furnace using ArþH2 gas circulation, quenched inwater, and analyzed. They systematically containFe-oxides and daughter crystals that make their composi-tions of questionable reliability. We thus interpret onlywell-preserved, glassy melt inclusions hosted in olivine.Representative compositions of olivines, melt inclusions,embayments, and matrix glasses are reported inTable 1.

Pumice of large-scale paroxysmsTheir olivine crystals span a limited range in compositionfrom Fo91^85 in the cores to Fo90^84 in the rims. OlivineFo89^90 that coexists with diopside (Mg-number¼ 0·91^0·92) is present in each of the three studiedpumice samples (ST531, ST127, ST82). This substantiatesthe previous finding (Bertagnini et al., 2003) that highMg-olivines are characteristic only of the large-scale par-oxysms. These olivines are euhedral crystals, occasionallyaggregated, in the 1^½ mm grain size of samples ST531and ST127 (Fig. 2a). In ST82 pumice, Fo89^90 olivines arefound in the ¼^½ mm grain-size fraction, as hopper-shaped crystals (Fig. 2b) that according to Donaldson(1976) are produced during dynamic crystallization athigh undercooling and/or a combination of undercoolingand isothermal crystallization (Kohut & Nielsen, 2004).Observation of several crystals illustrates the textural andcompositional evolution of olivine crystals during



Table1:

Selectedcompositionsofmeltinclusions,embaym

entsandmatricesofpumiceandscoriafrom

Strom

boli

Large-scaleeruptions

Med

ium-scale

eruption

Date:

15th–17thcentury

undated

undated

20th

century

15.3.2007

Sam

ple:

ST531pumice

ST82

pumice

ST81

Pumice

ST127pumice

ST150307A

pumice

M.I.,

Mtx,w.r.:

4b17b

39a

39b

Mtx

w.r.

133

102a

139a

139b

Mtx

gl.

w.r.

M.I.av.1

5a9a–d

M.I.av.

M.I.av.

55d

26e

14a

Mtx

w.r.2

n:

32

22

103

23

34

220

sd4

1050

sd14

sd3

73

7

SiO

2(w

t%)

47·41

48·87

48·07

47·75

50·47

49·66

48·37

49·00

47·49

48·35

49·27

48·16

50·9

48·24

(20)

47·92

47·93

47·95

(42)

47·72

(53)

47·33

47·07

47·55

49·84

49·2

TiO

20·87

0·91

0·85

0·97

0·90

0·83

0·83

0·84

0·87

0·83

0·94

1·00

0·9

0·97

(4)

0·87

0·91

0·91

(4)

0·88

(5)

0·86

0·92

0·99

0·92

0·92

Al 2O3

14·96

16·59

15·58

15·85

17·50

16·38

15·56

16·42

14·51

15·04

17·21

16·68

16·6

15·43

(39)

15·20

17·33

17·45

(29)

17·39

(27)

17·27

15·94

16·82

18·08

17·5

FeO

6·42

6·78

8·43

7·83

7·68

7·64

6·70

6·58

7·24

7·32

7·61

7·80

8·22

6·97

(34)

6·48

7·76

7·64

(19)

7·49

(17)

7·40

8·43

7·71

8·00

7·93

MnO

0·09

0·16

0·10

0 ·17

0·16

0·15

0·17

0·19

0·15

0·17

0·16

0·28

0·16

0·13

(3)

0·15

0·20

0·15

(4)

0·16

(6)

0·14

0·17

0·15

0·12

0·16

MgO

3·94

2·80

4·66

4·28

5·63

6·89

5·07

2·78

6·03

5·28

5·79

6·05

6·66

4·28

(56)

4·66

4·03

4·93

(40)

4·85

(36)

5·06

4·88

5·12

5·31

6·39

CaO

15·56

14·59

13·60

13·20

10·29

11·44

13· 84

13·98

13·70

13·96

12·26

13·04

10·9

15·19

(32)

14·78

12·82

12·10

(37)

12·05

(13)

12·16

13·07

12·41

11·55

11·6

Na 2O

2·20

2·32

2·08

2·35

2·79

2·47

2·30

2·44

2·39

2·16

2·78

2·35

2·39

2·46

(6)

2·37

2·33

2·31

(15)

2·28

(10)

2·34

2·13

2·07

2·48

2·53

K2O

1·60

1·76

1·61

1·71

2·31

2·08

1·70

1·88

1·58

1·56

2·00

2·07

1·82

1·71

(5)

1·79

1·48

1·63

(9)

1·65

(6)

1·48

1·34

1·49

1·93

1·88

P2O5

nd

0·53

0·56

0·56

0·58

0·47

0·55

nd

0·55

0·54

0·59

nd

0·4

0·59

(2·5)

0·58

0·51

0·58

(2)

0·56

(1)

0·57

0·55

0·63

0·58

0·56

Snd

0·218

0·207

0·191

0·030

nd

0·191

nd

0·199

0·199

0·058

nd

nd

0·255

(11)

0·233

0·207

0·160

(8)

0·166

(6)

0·171

0·186

0·166

0·031

nd

Cl

nd

0·211

0·188

0·180

0·118

nd

0·178

nd

0·175

0·181

0·129

nd

nd

0·215

(5)

0·188

0·174

0 ·155

(5)

0·155

(4)

0·152

0·160

0·143

0·136

nd

H2O

2·6

2·6

2·6

3·0

0·3

0·2

3·1

2·9

3·5

2·6

0·3

nd

nd

3·0

nd

nd

2·8

(1)

2·9

3·8

2·9

2·9

0·6

nd

CO2(ppm)

1482

1207

1094

1304

nd

nd

1377

1685

1704

1572

nd

nd

nd

1100

nd

nd

1629

(90)

1730

1478

1164

1579

nd

nd

Total

95·81

98·44

98·64

98·16

98·76

98·16

98·70

97·20

98·57

98·36

99·12

97·44

98·9

99·54

95·22

95·68

98·93

98·43

98·92

97·74

98·15

99·58

98·7

CaO

/Al 2O3

1·0

0·88

0·87

0·83

0·59

0·70

0·89

0·85

0·94

0·93

0·71

0·78

0·66

0·98

(2)

0·97

0·74

0·70

(2)

0·69

(1)

0·70

0·82

0·74

0·64

0·66

K2O/N

a 2O

0·7

0·76

0·77

0·73

0·83

0·84

0·74

0·77

0·66

0·72

0·72

0·88

0·76

0·70

(3)

0·76

0·63

0·70

(4)

0·72

(3)

0·63

0·61

0·72

0·78

0·74

S/C

l(w

t)nd

1·03

1·1

1·1

0·3

nd

1·1

nd

1·1

1·1

0·4

nd

nd

1·2

(51)

1·2

1·2

1·1

(1)

1·1

(2)

1·1

1·2

1·2

0·2

nd

H2O/K

2O

1·6

1·5

1·6

1·8

0·1

0·1

1·8

1·6

2·2

1·7

0·2

nd

nd

1·8

nd

nd

nd

1·7

1·8

(1)

2·6

2·2

1·9

0·3

nd

Fo(m

ol%)

89·5

87·5

85·3

85·2

89·3

88·3

87·3

86·5

88·4

(5)

89·7

86·2

85·8

(4)

85·3

86·0

85·0

85·5

SiO

240·45

40·46

40·23

40·07

40·96

40·18

39·96

40·51

40·79

(34)

40·67

40·36

40·26

(33)

39·95

40·58

40·55

40·24

MgO

47·88

46·57

44·87

44·66

48·04

47·58

47·04

46·10

47·48

(23)

48·81

46·00

45·47

(58)

45·05

45·47

44·71

45·07

MnO

nd

nd

0·23

0·27

0·19

nd

0·25

0·24

0·45

(6)

0·32

0·29

0·23

(2)

0·24

0·28

0·25

0·27

(continued

)



Table1:

Continued

Large-scaleeruptions

Med

ium-scale

eruption

Date:

15th–17thcentury

undated

undated

20th

century

15.3.2007

Sam

ple:

ST531pumice

ST82

pumice

ST81

Pumice

ST127pumice

ST150307A

pumice

M.I.,

Mtx,w.r.:

4b17b

39a

39b

Mtx

w.r.

133

102a

139a

139b

Mtx

gl.

w.r.

M.I.av.1

5a9a–d

M.I.av.

M.I.av.

55d

26e

14a

Mtx

w.r.2

n:

32

22

103

23

34

220

sd4

1050

sd14

sd3

73

7

FeO

10·02

11·85

13·73

13·83

10·27

11·19

12·21

12·87

11·06

(46)

9·98

13·12

13·42

(3)

13·81

13·19

14·11

13·61

NiO

nd

nd

0·16

0·14

0·19

nd

0·17

0·14

nd

nd

nd

0·13

(1)

0·14

nd

0·10

0·13

CaO

0·29

0·29

0·30

0·30

0·30

0·30

0·32

0·28

0·24

(7)

0·22

0·23

0·26

(7)

0·26

0·28

0·30

0·28

Restoredco

mpositions

PEC

(%)

7·5

8·0

4·0

3·5

6·0

9·0

0·0

1·5

0·08

7·5

5·5

2·2

0·02

1·5

2·7

1·2

SiO

2(w

t%)

46·89

48·19

47·76

47·48

47·92

48·20

47·49

48·23

47·66

47·48

47·51

47·78

47·60

47·23

46·89

47·46

TiO

20·80

0·83

0·82

0·94

0·78

0·76

0·87

0·82

0·89

0·82

0·86

0·89

0·87

0·85

0·90

0·97

Al 2O3

13·84

15·26

14·96

15·29

14·62

14·94

14·51

14·82

14·23

14·29

16·38

17·06

17·13

17·01

15·51

16·62

FeO

tot

6·69

7·19

8·64

8·04

6·91

6·99

7·24

7·40

7·28

6·69

8·06

7·77

7·58

7·48

8·58

7·78

MnO

0·08

0·14

0·09

0·16

0·16

0·18

0·15

0·16

0·12

0·14

0·19

0·15

0·16

0·14

0·17

0·14

MgO

7·23

6·30

6·27

5·69

7·65

6·81

6·03

5·89

7·64

7·31

6·34

5·82

5·46

5·66

5·96

5·60

CaO

14·41

13·44

13·07

12·75

13·03

12·75

13·70

13·76

14·05

13·90

12·13

11·84

11·87

11·99

12·72

12·27

Na 2O

2·04

2·13

2·00

2·27

2·17

2·22

2·39

2·13

2·27

2·22

2·21

2·26

2·25

2·30

2·07

2·04

K2O

1·48

1·62

1·55

1·65

1·60

1·71

1·58

1·54

1·58

1·69

1·40

1·60

1·62

1·46

1·30

1·47

P2O5

nd

0·49

0·53

0·54

0·52

nd

0·55

0·54

0·54

0·54

0·48

0·56

0·55

0·56

0·53

0·62

Snd

0·201

0·199

0·184

0·179

nd

0·199

0·196

0·235

0·219

0·195

0·157

0·164

0·168

0·181

0·164

Cl

nd

0·194

0·180

0·173

0·168

nd

0·175

0·178

0·198

0·176

0·164

0·152

0·153

0·150

0·156

0·141

H2O

2·4

2·4

2·5

2·9

2·9

2·7

3·5

2·6

2·8

nd

nd

2·7

2·9

3·8

2·8

2·9

CO2(ppm)

1371

1207

1050

1258

1294

1533

1704

1548

1012

nd

nd

1593

1704

1455

1133

1560

P(M

Pa)

130

163

180

nd

202

220

252

208

175

nd

nd

211

228

235

176

212

Size(mm)

80100

100

100

80irreg.

4100

60nd

nd

nd

80–100

100

100

6050

Bubbles

one

one

one

one

no

one

one

one

nd

one

one

several

one

several

one

two

(continued

)



Table1:

Continued

Small-scaleeruption

Strombolianactivity

Date:

August

1998

August

1999

August

1999

5April2003

August

1999

Present

Sam

ple:

ST130pumice

ST133scoria

ST182scoria

VDLscoria

ST177scoria

scoria

M.I.,

Mtx,w.r.:

1511a–d

Mtx

w.r.

10a

10c

Mtx

10a

16c

1514a

1411a–d

56a–h3

4e6b

2e–o

Mtx

4

n:

210

132

25

22

23

4/Emb.

12sd

122

27

40

SiO

2(w

t%)

46·73

49·38

50·37

49·72

51·28

51·50

52·44

52·55

52·58

50·52

52·80

52·70

51·47

(33)

51·45

(28)

50·78

51·31

51·69

52·27

TiO

20·92

0·98

1·08

0·97

1·61

1·48

1·75

1·66

1·43

1·93

1·62

1·65

1·56

(3)

1·64

(5)

1·68

1·65

1·55

1·54

Al 2O3

17·49

17·60

17·61

16·71

15·23

15·17

15·55

15·62

15·21

15·06

15·46

15·61

15·74

(6)

15·55

(10)

16·08

14·98

15·91

15·92

FeO

7·81

7·17

8·15

8·49

10·94

10·58

9·77

11·09

10·80

12·17

10·29

10·19

10·53

(31)

10·80

(21)

11·13

11·19

10·53

9·92

MnO

0·20

0·13

0·16

0·17

0·28

0·17

0·22

0·32

0·27

0·18

0·16

0·17

0·23

(4)

0 ·21

(5)

0·24

0·20

0·12

0·19

MgO

5·20

3·30

5·51

6·62

3·39

3·35

3·43

3·27

3·36

2·84

3·37

3·38

3·21

(15)

3·28

(7)

2·28

2·77

2·60

3·53

CaO

12·27

11·81

11·17

11·46

7·09

7·17

7·22

7·41

7·13

8·42

7·26

7·40

7·25

(11)

7·11

(22)

8·31

8·21

7·68

7·67

Na 2O

2·37

2·67

2·69

2·53

3·63

3·56

3·5

3·70

3·48

3·49

3·33

3·26

3·46

(9)

3·45

(8)

3·36

3·29

3·48

3·32

K2O

1·57

2·09

2·16

1·81

4·00

4·21

4·34

4·19

3·81

3·44

4·22

4·20

4·30

(3)

4·17

(12)

4·05

3·60

4·25

4·09

P2O5

0·58

0·55

0·69

0·45

1·11

1·26

1·3

1·03

1·03

1·11

1·22

1·22

1·23

(2)

1·25

(2)

1·10

1·27

1·18

1·19

S0·167

0·119

0·019

0·58

0·074

0·072

0·006

0·073

0·040

0·099

0·009

0·009

0·010

(2)

0·010

(3)

0·095

0·081

0·019

0·007

Cl

0·154

0·176

0·112

nd

0·257

0·200

0·103

0·219

0·164

0·271

0·121

0·118

0·124

(4)

0·124�(5)

0·154

0·198

0·131

0·120

H2O

nd

nd

nd

nd

0·5

0·5

nd

0·4

1·1

nd

0·4

0·4

nd

nd

0·6

0·5

nd

nd

CO2(ppm)

nd

nd

nd

nd

nd

nd

nd

355

273

nd

nd

nd

nd

nd

473

nd

nd

nd

Total

95·44

95·97

99·72

99·51

99·42

99·25

99·63

101·57

100·45

99·53

100·27

100·30

99·10

99·04

99·90

99·27

99·12

99·77

CaO

/Al 2O3

0·70

0·67

0·63

0·69

0·47

0·47

0·46

0·47

0·47

0·56

0·47

0·47

0·47

(2)

0·46

(1·6)

0·52

0·55

0·48

0·48

K2O/N

a 2O

0·66

0·78

0·80

0·72

1·10

1·18

1·24

1·13

1·09

0·99

1·27

1·29

1·23

(3·5)

0·08

(2)

1·21

1·09

1·22

1·24

S/C

l(w

t)1·1

0·7

0·2

nd

0·3

0·4

0·1

0·3

0·2

0·4

0·1

0·1

0·1

nd

1·21

(5)

0·6

0·4

0·15

0·1

H2O/K

2O

nd

nd

nd

nd

0·1

0·1

nd

0·1

0·3

nd

0·1

0·1

nd

nd

0·1

0·1

nd

nd

Fo(m

ol%)

86·2

85·6

71·6

71·6

70·4

70·1

70·9

69·3

69·3

69·7

(2)

69·8

(1)

69·7

70·1

69·9

SiO

239·93

39·51

37·70

37·70

38·00

37·39

37·88

37·16

37·16

37·55

(18)

37·22

(39)

37·92

37·39

37·37

MgO

46·25

45·98

36·13

36·13

34·91

34·66

35·33

34·28

34·28

34·40

(32)

35·28

(17)

34·65

34·66

34·38

MnO

0·28

0·38

0·50

0·50

0·49

0·49

0·48

0·51

0·51

0·50

(51)

nd

0·61

0·49

0·49

FeO

13·25

13·78

25·33

25·33

26·16

26·39

25· 81

27·11

27·11

26·61

(15)

27·21

(35)

26·89

26·39

26·30

NiO

nd

nd

nd

nd

0·05

0·05

nd

0·05

0·05

0·05

(51)

nd

nd

0·05

0·05

CaO

0·29

0·36

0·34

0·34

0·31

0·31

0·50

0·31

0·31

0·31

(1)

0·34

(1)

0·39

0·31

0·33

(continued

)



Table1:

Continued

Small-scaleeruption

Strombolianactivity

Date:

August

1998

August

1999

August

1999

5April2003

August

1999

Present

Sam

ple:

ST130pumice

ST133scoria

ST182scoria

VDLscoria

ST177scoria

scoria

M.I.,

Mtx,w.r.:

1511a–d

Mtx

w.r.

10a

10c

Mtx

10a

16c

1514a

1411a–d

56a–h3

4e6b

2e–o

Mtx

4

n:

210

132

25

22

23

4/Emb.

12sd

122

27

40

Restoredco

mpositions

PEC

(%)

2·0

5·5

2·5

3·0

2·0

1·0

5·5

0·0

1·2

1·0

5·2

4·0

4·0

SiO

2(w

t%)

46·60

48·84

50·94

51·08

52·26

52·43

49·83

52·80

51·30

51·31

50·11

50·75

51·16

TiO

20·90

0·93

1·57

1·44

1·62

1·42

1·83

1·62

1·54

1·62

1·60

1·58

1·49

Al 2O3

17·14

16·63

14·85

14·72

15·31

15·06

14·23

15·46

15·55

15·39

15·24

14·38

15·33

FeO

tot

5·78

5·63

11·30

11·02

9·76

9·35

12·92

8·74

9·16

10·97

10·36

11·79

11·10

MnO

0·19

0·12

0·27

0·17

0·31

0·26

0·17

0·16

0·22

0·20

0·23

0·19

0·11

MgO

6·02

5·65

4·20

4·33

3·90

3·67

4·62

3·37

3·58

3·60

3·97

4·03

3·76

CaO

12·03

11·18

6·92

6·97

7·27

7·06

7·98

7·26

7·16

7·04

7·90

7·90

7·41

Na 2O

2·32

2·52

3·54

3·45

3·63

3·45

3·29

3·33

3·42

3·41

3·19

3·16

3·35

K2O

1·54

1·97

3·90

4·09

4·11

3·77

3·25

4·22

4·25

4·13

3·84

3·45

4·09

P2O5

0·57

0·52

1·08

1·23

1·01

1·02

1·05

1·22

1·22

1·24

1·04

1·22

1·14

S0·164

0·113

0·072

0·069

0·072

0·040

0·093

0·009

0·010

0·010

0·090

0·078

0·019

Cl

0·151

0·166

0·251

0·194

0·215

0·162

0·256

0·121

0·123

0·122

0·146

0·190

0·126

H2O

nd

nd

0·5

0·5

0·4

1·1

nd

0·4

nd

nd

0·6

0·5

nd

CO2(ppm)

nd

nd

nd

nd

355

273

nd

nd

nd

nd

473

nd

nd

P(M

Pa)

nd

nd

nd

nd

4043

nd

nd

nd

nd

68nd

nd

Size(mm)

2540

nd

nd

3030

3060

100–80

40–60

2530

25

Bubbles

one

one

one/FeS

one/FeS

one/FeS

no

nd

no

no

no

no

no

no

n,number

ofan

alyses;Mtx,matrixglass;w.r.,

whole

rock;gl,glass

surroundingcrystals.Fo¼[100�Mg/(MgþFe)]at.ratio.PEC

(%)ispost-entrap

men

tolivinecrystallization

calculatedwithaKDvalueof0·30

atNNOþ1an

dFMQ,respectively,forbasaltican

dshoshonitic

meltinclusions.

P(M

Pa),pressurescalculatedusingtheCO2–H

2O

solubility

model

ofPap

aleet

al.(2006).Size(mm)refers

tomeltinclusionsize.Bubblesrefers

tonumber

ofbubbles;

FeS

refers

tosulfideglobuleswhen

they

occur.

nd,notdetermined

.1Averaged

meltinclusionco

mposition,datafrom

Bertagniniet

al.(2003).

2Pumicewhole-rock

(STR150307A)from

Lan

diet

al.(2009).

3Averaged

meltinclusionco

mposition,datafrom

Metrich

etal.(2005).

4Averaged

electronmicroprobean

alysis

ofglassymatricesofHPbasaltfrom

Lan

diet

al.(2009).



Fig. 2. (a^d) Backscattered electron micrographs of olivines from large-scale paroxysm pumices. The numbers refer to olivine compositions(Fo mol %). (a) Aggregated euhedral crystals with weak normal zoning (ST531, ol11); (b) nearly euhedral crystal resulting from the incompletegrowth of a hopper olivine (ST531, ol8); (c) skeletal crystal with dendritic overgrowths and a high density of irregular melt inclusions in itsinner part; this texture results from the sealing of the hollow central cavity of a hopper olivine (ST82, ol110); (d) euhedral, homogeneous crystalwith rounded melt inclusions (ST531, ol2). (e, f) Dark-field images of olivines from pumice clasts of the 15 March 2007 intermediate-scale parox-ysm showing: (e) skeletal, embayed crystal hosting numerous melt inclusions of variable size and shape, and (f) hopper olivine.(g) Backscattered electron micrograph of a nearly euhedral, homogeneous olivine from pumice of the 23 August 1998 small-scale paroxysm(ST130, ol15). (h) Transmitted-light image of a homogeneous olivine from scoria of the 26 August 1999 small-scale paroxysm, with melt inclu-sions regularly aligned along crystal growth directions (ST177, ol6).



their growth. Initially, an internally hollow crystal(hopper olivine) forms with a magnesian compositionFo89^90 (Fig. 2b). Then, the central cavity becomes par-tially sealed by further growth of a less magnesian olivinethat entrapped several melt/glass inclusions. The final tex-ture of the crystal is that of a closed skeletal olivine (Fig.2c), with dendritic overgrowths on some crystals. Such aninterpretation is critical because it accounts for the vari-able quantity of melt/glass confined in the center of manygrains, and the apparent reverse zoning of some olivines(Fig. 2b and c).In all high-Mg olivines (Fo89^90), the melt inclusions pre-

serve primitive compositions of Ca-basalts with highCaO/Al2O3 ratios (0·8^1·0) irrespective to their size(Fig. 3a). These compositions are even more calcic thanthose commonly reported for the bulk pumice erupted atStromboli (on average CaO/Al2O3¼0·68�0·02), apartfrom some rare exceptions (CaO/Al2O3¼0·81, Pompilioet al., 2009). A further important observation is thatthe residual glasses wrapping the crystals have preservedvariable compositions with CaO/Al2O3 ratio up to 0·78(Table 1).Less magnesian olivines (Fo87^84) are commonly simple

euhedral crystals that contain melt inclusions (Fig. 1d)whose compositions (CaO/Al2O350·80) resemble those ofthe bulk pumice and matrix glasses (Fig. 3a). As a wholethe olivine-hosted melt inclusions analyzed in samples(ST531, 127 and 82) from large-scale eruptions record arange of the CaO/Al2O3 ratio (Fig. 3a) attributed to thepredominant effect of clinopyroxene crystallization. Thevariability in the K2O/Na2O ratio suggests an additionalprocess of mixing between magmas differing in theirK2O contents, in agreement with previous finding(Bertagnini et al., 2003). Inputs of slightly different magmabatches in the Stromboli plumbing system have also beenproposed on the basis of trace element geochemistry(Pompilio et al., 2009) and Sr isotopes (Francalanci et al.,1999, 2005).The total volatile content (CO2þH2OþSþCl) in

fully enclosed inclusions is 43wt %, with H2O varyingfrom 2·9 to 2·3wt % (one value at 3·5wt %) and theH2O/K2O ratio ranging from 2·2 to 1·6 (Fig. 3b). TheirCO2 concentrations decline from 2260 to �1100 ppm,with the highest value found in a 40 mm melt inclusion,with one small bubble, in a 250 mm olivine of Fo90(Electronic Appendix 1, available for downloading athttp://www.petrology.oxfordjournals.org). Interestingly,the S/Cl ratio averages 1·2^1·1� 0·1, whereas the S and Clconcentrations decline, respectively, from 2510 to 1600 andfrom 2060 to 1610 ppm (Fig. 4a^c). Such a feature canhardly be explained by fractional crystallization alone.Partial exsolution of sulfur into the vapor phase can occurat high pressure while Ca-clinopyroxene is crystallizing.Actually, the solubility of sulfur dissolved as sulfate species

possibly decreases as the melt becomes less calcic, as pre-viously discussed for Stromboli (Bertagnini et al., 2003).However, the contemporaneous decline of the S and Clamounts dissolved in the basaltic melt led to a suggestionof additional processes.We propose that the S and Cl evo-lution (Fig. 4c) is best accounted for by both high-pressurepartial exsolution of sulfur and mixing between magmasslightly differing in their initial S and Cl concentrationsrather than similar S and Cl partitioning in the vaporphase at high pressure.

The case of intermediate-scale eruptionsIn the pumice erupted during the March 2007 paroxysm,olivine Fo86^84 (Fig. 2e and f) coexists with clinopyroxene(Mg-number¼ 0·88^0·84). Despite systematic investi-gation, high Mg-olivines (Fo89^90) as described abovehave not been found. Clinopyroxene shows sectorzoning. Olivine crystals display both skeletal morphol-ogies and embayed structures. Skeletal olivines (Fig. 2e)and crystals with hollow cavities (Fig. 2f) were observedin the 1^¼ mm size fraction. We have focused onthe former olivines that systematically host large(80^100 mm) curvilinear, fully enclosed melt inclusionswith several bubbles, unsealed glass embayments andsmall size inclusions confined in their central part(Fig. 2e). The olivine textures of these pumice samples areequivalent to those described above for large-scaleparoxysms.Major element compositions of large curvilinear melt

inclusions resemble those of the glassy matrices and oftheir bulk-rocks (on average CaO/Al2O3¼0·70; Fig. 3a).Cooling was fast enough to occasionally preserve slightcompositional gradients at the interface between olivineand glass. Analytical profiles of MgO measured in theseinclusions reveal that the boundary layer extends for lessthan 10 mm and results in a maximum MgO decrease of�1wt % at the olivine contact (Table 1). The compositionsof unsealed glass embayments plot in the same composi-tional domain, testifying to their formation during equilib-rium olivine crystallization (Faure & Schiano, 2005).Thus, inclusions and embayments share the same composi-tion, their K2O/Na2O ratio clustering at �0·7 (Fig. 3a).This implies that crystallization upon magma ascent wasnegligible, as confirmed by the scarcity of microlites inthe groundmass.We also verified the possible effects of a boundary layer

on the melt inclusion compositions, owing to rapid crystalgrowth. In a few crystals we analyzed on average 10 inclu-sions per crystal (Electronic Appendix 2). The chemicalprofiles obtained on 50^60 mm inclusions in the centralpart of crystals reveal no systematic chemical variation(26e, 14a in Table 1). Additionally, no systematic chemicaldifference was found as a function of the melt inclusion



Fig. 3. Variation of K2O/Na2O (a) and H2O/K2O (b) vs CaO/Al2O3 in olivine-hosted melt inclusions, glass embayments and glassy matricesof Stromboli pumice and scoria. Pumice samples from large-scale paroxysms include ST531, ST127 and ST82. In (a) the domain of pumicebulk compositions is plotted for comparison. In (b) the microphotographs in the upper and lower corners illustrate, respectively, a fully enclosedmelt inclusion from ST82 pumice and an embayment in the 15 March 2007 pumice. Abbreviations: LSP, ISP, SSP, large-, intermediate- andsmall-scale paroxysms, respectively; MI, melt inclusions; Emb, unsealed embayments; Mtx, glassy matrices. [1] Averaged composition of typicalglassy matrices of present-day HP basalt from Landi et al. (2009).



size (Electronic Appendix 2). We did not observe sig-nificant enrichment in S with respect to Cl (S/Cl¼1·2against 1·1�0·1) as produced in experiments at highcrystallization rates (Baker, 2008). Only the smallestinclusions (20^25 mm) record boundary layer effects inhaving abnormally high concentrations in phosphorus(P2O54 0·65 wt % up to 1·06wt %, against 0·55wt %),an element known for its slow diffusivity (Baker, 2008).The fully enclosed melt inclusions in the 15 March 2007

pumice preserve a high total volatile content (on average:

H2O¼ 2·7�0·1wt %, except for one value at 3·8wt %;CO2¼1590� 90 ppm; S¼1570�80 ppm; Cl¼1520�50 ppm; Table 1). Their H2O/K2O ratios cluster at 1·7^2(Fig. 3b). Their averaged S/Cl ratio (1·0� 0·1; Fig. 4a and b)is similar to that measured in melt inclusions fromanother intermediate-scale paroxysm (5 April 2003;Me¤ trich et al., 2005). The glassy matrix contains 170 pmof S and �1270 ppm of Cl, similar to the values(S �250 ppm, Cl �1150 ppm) measured in the 5 April2003 pumice matrices.

1.5Basaltic melt

1.0S

/Cl

Change in redox state of sulfur

Shoshonitic melt

0.5

S

(a)

1.5 0.25

0.0

K2O wt%

( )

1.0

0.15

l wt%

S/C

l

Basaltic melt

0.0

0.5

0.4 0.6O

0.05

0.0 0.1 0.2 0.3

C

S

(b) (c)

P2O5 wt% S wt%

1 2 3 4 5

Fig. 4. (a) Variation of S/Cl vs K2O in olivine-hosted melt inclusions, glass embayments and glassy matrices of Stromboli pumice and scoria.Variation of the S/Cl ratio as a function of P2O5 (b) and of Cl vs S (c) upon decompression and ascent of the basaltic magma. Symbols as inFig. 3a. In (c) previous melt inclusion data for Stromboli pumice from large-scale eruptions (Me¤ trich et al., 2001; Bertagnini et al., 2003) are plot-ted for comparison (open diamonds).



To follow the behavior of each volatile species uponmagma decompression we have specifically analyzed someof the unsealed glass embayments (Fig. 5a^c). Theseformed during olivine crystallization in the ponding zoneand are commonly in contact with a large gas bubble(Fig. 5c). Compositional profiles from the innermost partto the outer matrix glass evidence an overall decrease ofH2O (from 2·9 to 0·6) and S (from 1550 to 170 ppm). TheCl concentration remains nearly constant (on average1550 ppm); the lowest values measured in the outermost

parts of embayments are equivalent to that of the matrixglass. Their major element composition, however, does notchange significantly (Figs 3b and 4) in agreement withlimited crystallization during magma ascent. We havenot found diffusion profiles covering the whole range ofvolatile concentrations within one single embayment.As an example, the concentrations in H2O and CO2 vary,respectively, from 2·5 to 1·7wt % and from 1372 to926 ppm in one embayment with a gas bubble at theoutlet (ST150307A-19, Fig. 5a). The lowest values havebeen found within the 10^20 mm boundary layer surround-ing the gas bubble. Sulfur behaves similarly, and variesfrom 1360 to 1160 ppm. We infer that the decompressionrate was high enough to prevent extensive diffusion of thevolatiles into the gas bubble.

Pumice of small-scale paroxysmsOne of the typical features of these pumice samples(ST130) is that olivine Fo85^86·2 is extremely rare andoccurs only in the 250 mm size fraction as nearly euhedralmicrocrysts (Fig. 2g) together with clinopyroxene(Mg-number¼ 0·88). Olivine-hosted inclusions share thesame CaO/Al2O3 (0·70) and S/Cl (1·1) ratios with thosemeasured in the March 2007 samples (Table 1).Water wasnot analyzed in these rare inclusions. However, a highwater content of �3wt % is expected from the totaloxide content of �96wt % of the electron microprobeanalyses (Table 1).

SCOR IA I SSUED FROM THESHALLOW PLUMBING SYSTEM:TEXTURES AND COMPOSIT IONSOF OLIV INES AND OF THEIRINCLUSIONSThe HP basaltic scoriae contain on average 50 vol. % ofplagioclase An60^70, clinopyroxene Fs12^15, and minorolivine Fo70^73 (Francalanci et al., 2004, 2005; Landi et al.,2004, 2006, 2009). Systematic compositional profiles wereobtained for millimeter-sized olivines in scoria samplesST182, ST177 (August 1999) and ST 133 (August 1998). Theanalyzed crystals display homogeneous compositions withaveraged Fo contents varying from 69·3�0·2 to70·4� 0·3 from one crystal to another one. These olivinescontain solid inclusions of plagioclase An67 (sampleST182-11; Table 1) that is typical of the HP basalt. Olivinecommonly entraps small melt inclusions (20^40 mm),aligned along crystal growth directions (Fig. 2h). Thesetextural features were experimentally reproduced forclosed hopper olivine crystals after successive cooling andreheating cycles (Faure & Schiano, 2004). Despite its chem-ical homogeneity, olivine could have a complex crystalliza-tion history, as extensively documented for Stromboli

0.16

0.20

0.04

0.08

0.12

S w

t%

( )0.00

0.16

0.18

(a)

0.12

0.14

Cl w

t%

2000

250019

57

61

0.10(b)

500

1000

1500

CO

2pp

m

61

00 2 4

H2O wt%

(c)

Fig. 5. Variation of S (a) and Cl (b) vs H2O in melt inclusions andembayments hosted in olivines of the 15 March 2007 pumices(ST15030A sample). (c) Details of the CO2 and H2O evolution in 3different embayments (#19, 57, 61). Microphotograph shows the ana-lytical spots (H2O and CO2) for an embayment entrapped duringskeletal growth of olivine Fo86 (ST15030A-19).



plagioclase and pyroxene (Landi et al. 2004, 2006;Francalanci et al., 2004, 2005). The shoshonitic compositionof the melt inclusions in Fo69^70 olivine matches that ofthe typical glassy matrix of the scoria (Fig. 3a) with ahigh SiO2 content (on average 52·5wt %), low CaO/Al2O3 (0·48) and high K2O/Na2O (1·2). The amounts ofdissolved H2O (50·5wt %; H2O/K2O �0·1, Fig. 3b) andS (80 ppm, at the detection limit) are low. The Cl contentof the melt inclusions (on average 1200 ppm) is identicalto that of the glassy matrix, as is their S/Cl ratio (Fig. 4a).These features, commonly observed at Stromboli (e.g. 5April 2003 sample, Table 1), indicate the exsolution of bothS and Cl into the gas phase within the conduits.We empha-size that sulfide globules are uncommon in the inclusions,crystals and matrix glasses of the scoria.However, some olivines show a wider variability of

compositions from Fo71 to Fo73^74·7, as also reported inlava and scoria samples (Landi et al., 2006; Cigolini et al.,2008). They include plagioclase (An78·8) more calcic thanthat reported above. Their melt inclusions display somecompositional variability (0·485CaO/Al2O350·58;Fig. 3a) and have intermediate H2O (�1·1^1·9wt %;H2O/K2O40·1, Fig. 3b) and S contents (800^1000 ppm).Their Cl concentrations vary from 2560 to 1500 ppm,with S/Cl (wt) ratios up to 0·6 (Fig. 4a).We cannot excludethe possibility that the low H2O content (50·5wt %)reported for some inclusions that still have �800 ppm of Swas the result of re-equilibration of H2O with the sur-rounding degassed shoshonitic melt after their entrapment.Quenched immiscible sulfide liquid is commonly, but notsystematically, present within melt inclusions in olivineFo71^73. Sulfide globules have not been found in the hostcrystals themselves. There is no systematic relationshipbetween the amount of S dissolved in the melt and thepresence of sulfide globules. Finally, we emphasize thatsimilar compositional ranges were also observed for meltinclusions and host olivines of the HP basalt entrainedduring large-scale eruptions (e.g. scoria ST82s; Me¤ trichet al., 2001).

MAGMA SATURATIONPRESSURESThe minimum melt saturation pressure (PCO2þPH2O)and gas phase composition (XCO2, XH2O) were derivedfrom the dissolved amounts of H2O and CO2 measured inmelt inclusions, using the H2O^CO2 solubility model ofPapale (1999), updated by Papale et al. (2006). This modelallows a better estimation of the effect of melt compositionon the volatile solubility for calcic and calc-alkaline basalts(Moore, 2008).We have computed the total fluid pressure for each melt

inclusion in Fo90^85, representative of both Ca-basalticand LP basaltic melts (Table 1 and Electronic Appendix

1), for a temperature varying from 1200 to 11408C, accord-ing to Bertagnini et al. (2003). The fO2 was fixed at�NNOþ1, a value slightly higher than that estimated byPichavant et al. (2009) but in accordance with the absenceof dissolved sulfide species. The abbreviation NNO refersto the Ni^NiO oxygen buffer; �NNO¼ (log10fO2sample ^log10fO2NNO). A difference of 0·5 log unit lowers thepressure by a few MPa in the range of interest. In pumice(ST82, ST531) erupted during large-scale paroxysms,the saturation pressures range from 280 to �190MPa(Fig. 6) and are calculated for melt inclusions having vari-able compositions (CaO/Al2O3 from 0·7 to 1). It shouldbe noted that these pressures are possibly underestimatedbecause of CO2 diffusion in the melt inclusion bubblesupon cooling.The March 2007 basaltic melt recorded satu-ration pressures between 235 and 200MPa, with an aver-age value of 210�20MPa. The latter value, retrievedfrom fully enclosed, sealed inclusions, is considered repre-sentative of melt entrapment and olivine crystallizationconditions in the storage zone. A careful examination ofthe dataset indicates that values �190MPa are determinedin unsealed glass embayments for which the pressure regu-larly decreases to �40MPa (Fig. 6).The CO2 solubility in shoshonitic melts in the

0^100MPa pressure range is not known. Thus the satura-tion pressures calculated using the solubility model ofPapale et al. (2006) must be considered with caution, allthe more so because the computed CO2 solubility ishighly sensitive to the redox state. In the present study,calculations were made with total Fe as FeO. The calcu-lated pressures vary from �90 to 520MPa (Table 1;Electronic Appendix 1).

DISCUSS IONHigh-pressure magma crystallizationThe highest melt saturation pressure (PH2OþPCO2)retrieved from Ca-rich melt inclusions is 280MPa (Fig. 6),equivalent to a maximum 11km lithostatic depth, using anaverage crustal density of 2700 kg/m3. It is obvious thatthe pressures (depths) deduced from melt inclusionsrecord those of olivine crystallization, whose textures indi-cate dynamic crystallization. This observation impliesthat the Ca-basalt magma batches rise from a deeperlevel, possibly from the base of the 17 km thick continentalcrust underlying Stromboli volcano (Panza et al., 2007),where a low Vp and Vs volume was seismically detectedand interpreted as due to ponded partial melt (Barberiet al., 2007).High-pressure equilibrium experiments have failed to

reproduce the crystallization of olivine Fo89^90 from aCaO-rich hydrous melt (H2O� 3wt %) in presence of afluid phase, whereas diopsidic pyroxene is the liquidusphase (Di Carlo et al., 2006; Pichavant et al., 2009). To clar-ify the olivine crystallization conditions we have computed



the magma crystallization paths using the Adiabat ther-modynamic model (Ghiorso & Sack, 1995; Asimow &Ghiorso, 1998; Smith & Asimow, 2005). As starting condi-tions we used the average compositions of the Ca-basalticmelt inclusions (CaO/Al2O3¼0·9^1·0; MgO �7·6wt %),adjusted for post entrapment olivine crystallization, andan H2O concentration of 2·9^2·8wt % (St82-133; ST81-11;Table 1). Simulations at 300MPa indicate that clinopyrox-ene would be the only liquidus phase under moderatelyhydrous, low-CO2 conditions as experimentally observed(Di Carlo et al., 2006; Pichavant et al., 2009). Increasingthe melt water concentration to 3·2^3·5wt % does notinduce major changes. Therefore, we computed theolivine^melt equilibrium by forcing olivine saturation(1% olivine added) because the MgO concentrations ofthe inclusions, after correction for olivine post-entrapmentcrystallization, could be underestimated to some extent.We found that olivine Fo89·8^88·7 joins the liquidus phasewith clinopyroxene (Mg-number¼ 0·91^0·90) in equilib-rium with a melt having �2·9wt % of H2O, and �8·1wt% of MgO, between 300 and 250MPa. These two phasesstart crystallizing at a temperature of 1190^11858C at�NNOþ1 (Table 2), but the olivine amount remains low(�1wt %) as observed in the natural products. Predictionof olivine saturation, melt MgO content and temperatureof crystallization could be unreliable because the

thermodynamic models do not treat the effect of CO2 onphase equilibria and Ca activity. To test the validity ofthese calculations, we have calculated the theoreticalvalue of the partitioning coefficient K

Mg�FeD Ol�Liq [(Mg/

Fe2þ)melt/(Mg/Fe)ol] according to Toplis (2005). The mea-sured and predicted KD values are consistent for the giventemperatures and fO2 (Table 2).The melt inclusion compositions, olivine textures and

Adiabat computations support the idea of deep rechargeof gas and hydrous, Ca-basalt magma batches with13^14wt % CaO (CaO/Al2O3¼0·9^1), during large-scaleparoxysms. Ne-normative, calcic melts are not exclusive toStromboli, and have previously been reported as meltinclusions hosted in high-Mg olivine at Vulcano (Schianoet al., 2004), and bulk-rocks at Mt Etna (Kamenetskyet al., 2007) in Italy and, more broadly, at arc volcanoes(Schiano et al., 2000). Calcic magmas may form by interac-tion between mantle-derived melts and clinopyroxene- oramphibole-bearing lithologies possibly at the mantle^crust interface (Schiano et al., 2000; Me¤ dard et al., 2006).At Stromboli Ca-basalts, rich in CO2, H2O, S, and Clhave been recognized as melt inclusions from everypumice sample erupted during five large-scale paroxysms.Both Ca-rich melt inclusions and their host olivine withhopper and skeletal textures demonstrate fast magma crys-tallization events associated with deep magma inputs.

3

4

%

LSP - MI [1]LSP - MILSP - EmbLSP - MtxISP - 15/03/07 MIISP - 15/03/07 EmbISP 5/04/03 MI/Emb

1

2

H2O

wt% ISP - 5/04/03 MI/Emb

ISP - MtxScoria Aug. 1999 MI

00 50 100 150 200 250 300

Pressure (MPa)Fig. 6. Pressure-related evolution of the H2O content (Raman data) of olivine-hosted melt inclusions and embayments from pumices and scor-iae. Also reported are published (FTIR) data for pumice ([1]; Me¤ trich et al., 2001; Bertagnini et al., 2003). Abbreviations as in Fig. 3.



Table 2: Computed crystallization path of Stromboli magma

ST82 ST81

Bulk Melt Bulk Melt Bulk Melt Bulk Melt Bulk Melt

T (8C): 1185–75 1160 1150 1140 1185

fO2 – �NNO: þ1 þ1 þ1 þ1 þ1 þ1 þ1 þ1 þ1 þ1

P (MPa): 300 300 300 300 300 300 300 300 300 300

SiO2 (wt %) 48·6 48·5 48·4 48·4 48·3 48·3 48·3 48·2 47·8 47·7

TiO2 0·8 0·8 0·8 0·8 0·8 0·8 0·8 0·8 0·9 0·9

Al2O3 14·7 14·9 15·5 15·8 16·1 16·4 16·7 16·9 13·9 14·2

Fe2O3 2·3 2·2 2·3 2·2 2·3 2·2 2·3 2·2 2·4 2·4

FeO 5·0 5·1 5·1 5·2 5·2 5·2 5·2 5·2 5·2 5·3

MnO 0·2 0·2 0·2 0·2 0·2 0·2 0·2 0·2 0·1 0·1

MgO 8·3 8·0 7·5 7·3 7·0 6·8 6·6 6·3 8·9 8·6

CaO 13·1 12·9 12·4 12·1 11·9 11·6 11·4 11·1 13·7 13·5

Na2O 2·2 2·2 2·3 2·4 2·5 2·5 2·6 2·6 2·2 2·3

K2O 1·6 1·6 1·7 1·8 1·8 1·9 1·9 2·0 1·5 1·6

P2O5 0·5 0·5 0·6 0·6 0·6 0·6 0·6 0·6 0·5 0·5

H2O 2·9 3·0 3·2 3·2 3·3 3·4 3·5 3·6 2·7 2·8

CaO/Al2O3 0·89 0·86 0·80 0·77 0·74 0·71 0·68 0·66 0·99 0·95

Phase %

Melt 96·7 89·2 84·9 81·0 96·6

Olivine 0·9 0·3 0·4 0·5 0·1

Cpx 2·3 9·6 13·8 17·6 3·1

Olivine composition

SiO2 40·7 40·5 40·5 40·4 40·8

FeO 10·4 11·1 11·5 12·0 10·0

MnO 0·2 0·2 0·2 0·2 0·1

MgO 48·3 47·8 47·4 47·0 48·6

CaO 0·5 0·4 0·4 0·4 0·5

Fo (mol %) 89·2 88·5 88·0 87·5 89·7

Clinopyroxene composition

SiO2 50·8 50·3 49·9 49·5 50·8

TiO2 0·4 0·5 0·6 0·7 0·4

Al2O3 4·2 4·6 4·9 5·1 4·1

Fe2O3 2·9 3·2 3·4 3·6 3·0

FeO 2·9 3·1 3·2 3·3 2·7

MgO 16·4 16·0 15·7 15·4 16·3

CaO 22·3 22·2 22·2 22·2 22·5

Na2O 0·1 0·1 0·1 0·1 0·1

Mg-no. 0·91 0·90 0·90 0·89 0·91

KD measured 0·338 0·326 0·306 0·309 0·334

KD predicted 0·312 0·306 0·306 0·306 0·304

Bulk indicates bulk system (crystalþ liquid). �NNO¼ [log10 fO2sample – log10 fO2(NNO)], where NNO refers to the Ni–NiOoxygen buffer. Fo¼ [100�Mg/(Mgþ Fe)] at. ratio. Mg-number¼ [Mg/(Mgþ Fe2þ)]. KD predicted is theoretical KD value(Toplis, 2005).



After magma injection into the deep ponding zone, olivineand clinopyroxene crystallization may be enhanced by dif-ferent processes. Decompression, accounting for the micro-cryst textures in pumice produced during large-scaleparoxysms, is one possibility. Another process involvesthermal and chemical gradients that are able to promotecrystallization; these develop at the interface between theuprising, hot Ca-basaltic melt and the basaltic magmaresiding at 7^10 km depth. The two magmas differ in theirmajor element contents (CaO, MgO), their dissolved vola-tile amounts and temperature (�T� 308C).We also verified using Adiabat that hydrous Ca-basalts

(H2O¼ 2·9wt %) are able to generate LP basalts byremoval of 15^20wt % of solids (Table 2). Ca-basalts areregarded as the most likely parental magmas of theStromboli products erupted during the last 1800 years.Hence, we cannot exclude the possibility that the pondingzone located between 7 and 10 km depth is fed by basalticmagma batches that originated from parental Ca-basaltsthrough fractional crystallization at various crustal levels.Assuming that crystal fractionation is the dominant pro-cess that controls magma evolution, the dissolved amountof H2O in the melt is expected to increase. This feature isrecorded only in a few melt inclusions (H2O43wt %;Fig. 6) whereas the H2O content of the 15 March 2007inclusions averages at 2·7�0·1wt %. Hence, anon-negligible fraction of H2O could have been lost in thefluid phase at such high pressure (depth). Melt waterdepletion is possibly induced by increasing the XCO2/XH2O ratio in the vapor phase coexisting with the melt, asdiscussed further below. Magma CO2 fluxing is thoughtto possibly have a drastic effect on the water degassingpath of basaltic magma (Spilliaert et al., 2006; Johnsonet al., 2008; Me¤ trich & Wallace, 2008; Collins et al., 2009).We propose that such a process is active in the deepestparts of Stromboli plumbing system.

Conditions of magma ascent duringparoxysmsThe conditions of magma ascent have been retrieved fromthe fully enclosed inclusions in olivine from large- andmedium-scale explosive eruptions. We have computed thepressure-related evolution of H2O and CO2 in a basalticmelt under closed-system ascent and decompression(CSAD) conditions, using Papale’s updated model(Papale et al., 2006). CSAD computations were performedassuming 3·5 wt% of H2O in the system and that themelt was coexisting with variable amounts of gas phase at�250MPa. The CSAD curves without crystallizationplotted in Fig. 7a show the pressure-related evolution ofthe H2O and CO2 contents of the melt, with CO2 rangingfrom50·01 to 10wt %. Our data cannot be fitted whenconsidering no or a negligible fraction of gas phase in thesystem, which would produce a magma still containing a

high dissolved H2O content at low pressure (curve 1;Fig. 7a). A large part of our data points can be reproducedwhen considering decompression of an H2O-rich basalticmagma coexisting with 2·4wt % of CO2-rich gas phase(curve 2),490% of the CO2 being exsolved in the vaporphase. This proportion of exsolved CO2 is consistent withthe initial CO2 content of the Stromboli parental meltthat was inferred from the melt inclusion S content andthe time-averaged CO2/SO2 ratio in the gas emissions(Allard et al., 2008). A higher proportion of CO2 (4wt %)in the system was proposed by Scaillet & Pichavant (2005)close to our curve 3. Actually, the CSAD curves 2 and 3bracket the H2O and CO2 evolution recorded by meltinclusions from the pumice of five paroxysms. Altogetherour results demonstrate quantitatively and for thefirst time that the LP basaltic magma that sustains theparoxysmal eruptions rises in equilibrium with its gasphase, under prevalently closed-system conditions, fromthe storage zone lying at �7^10 km depth, until close tothe surface. More specifically, the volatile concentrationsof the 15 March 2007 samples indicate that the magmarose with its gas from conditions of 210�20MPa (equiva-lent to 8 km in depth). CSAD conditions for gas-richmagma are fully consistent with the phenomenology com-monly reported for paroxysms during which vertical jetsof gas and pyroclasts develop to feed a convective column(e.g. 5 April 2003 eruption; Rosi et al., 2006), related toincreasing acceleration of the melt and gas mixture upondecompression and final magma fragmentation (e.g.Parfitt & Wilson, 1999). On the other hand, the pumicematrices contain only scarce microlites and their majorelement compositions resemble that of the bulk rocks aspreviously observed for large-scale paroxysms (Bertagniniet al., 2003) and demonstrated for the 15 March 2007pumice (Fig. 8). Decompression, direct ascent, anderuption of LP basalt are fast enough to inhibit crystalnucleation. Such features exclude low-pressure stagnationof the ascending magma batch, which could havepromoted H2O loss from the melt and crystallization.

Tracking the effects of CO2 on magmadehydration and crystallizationThe pressure-related behavior of H2O and CO2 duringmagma ascent recorded by the unsealed glass embaymentsprovides other information (Fig. 7a). It reveals preferentialmelt depletion with respect to H2O that drives the datapoints on the left-hand side of curves 2 and 3. It is well illu-strated by olivine embayments in contact with a gasbubble in the 15 March 2007 pumice (microphotographin Fig. 5c). Prevalent depletion of H2O relative to CO2

could be expected during diffusion-controlled bubblegrowth, because of the slower diffusivity of CO2 withrespect to H2O (Watson, 1994). Following this line of rea-soning, sulfur, whose diffusivity resembles that of CO2 in



hydrous basaltic melt (H2O¼ 3·5wt %; Freda et al., 2005),would be significantly enriched compared with H2O, as isCl (Alletti et al., 2007). This hypothesis is not verified, aswe observed a coupled decrease of H2O and S whereas Clconcentrations remain nearly constant (Fig. 5a and b).Therefore, we have considered an alternative processduring which the fluid^melt partitioning of volatile speciesis controlled by the proportion and the composition of thevapor phase coexisting with the basaltic melt at eachdecompression step. H2O concentrations 52·5wt % at�100MPa require up to 5^10wt % of CO2 in vaporphase (curves 3 and 4, Fig. 7a). Preferential magma deple-tion in H2O by a CO2-fluxing process basically implies ahigh gas/melt ratio. Actually, evidence of heterogeneoustrapping of melt and gas (variable gas bubble/melt ratio)during olivine growth was previously reported for the 5April 2003 and historical pumices (Me¤ trich et al., 2001,2005; Bertagnini et al., 2003). Similar observations werealso made for Etna samples in which magma dehydrationby CO2-rich gas is best recorded in olivine-hosted meltinclusions showing a highly variable bubble/glass ratio(Spilliaert et al., 2006).

At Stromboli, the time-averaged CO2 flux of1·4�103 t/day coupled with a low magma extrusion ratehas led to the idea that the contribution (continuous or dis-continous) of streaming CO2-rich bubbles has a decisiveinfluence on the crystallization path of the basalticmagma (Allard et al., 2008). An important issue addressedhere is thus the possible influence of CO2 on the transitionbetween the LP and HP basalts at Stromboli.As illustrated in Fig. 7a, the basaltic melt still contains

between �2 and 1wt % of dissolved H2O between 100and 50MPa, where it coexists with a high proportion ofgas phase. The conditions are thus fulfilled for plagioclasecrystallization (Landi et al., 2004) and extensive crystalli-zation (H2O�1wt %; Di Carlo et al., 2006) when the LPmagma is ponding in this pressure interval, in agreementwith the few pressures determined for shoshonitic meltinclusions hosted in Fo69^72 (Fig. 6).We emphasize that nogradual change in the chemistry of minerals or bulk-rockswas observed in the HP basaltic magma, either erupted asscoria during normal strombolian activity and lava flowsor entrained by the uprising magma during large-scaleparoxysms (Me¤ trich et al., 2001; Bertagnini et al., 2003;

2 1

1500

2000

2500

pm4 3

2 1

0.2

0.4

0.6

0.8

Pressure

XC

O2

100 MPa

(b)

500

1000

1500

CO

2pp Pressure

0

500

0 1 2 3 4H2O wt%

(a)

Fig. 7. (a) Variation of CO2 and H2O dissolved in melt inclusions, embayments and glassy matrices of LP basalt erupted as pumice during par-oxysms. Isobars and the curves showing the pressure-related CO2^H2O evolution are calculated for the average basaltic composition of meltinclusions (ST150309A; Table 1) using Papale’s updated model (Papale et al., 2006). The curves show the CO2^H2O evolution duringclosed-system ascent and decompression (CSAD) of magma without crystallization. The starting conditions are: 3·5wt % of H2O and variableamounts of CO2 in the bulk system: curve 1 for 0·25wt % CO2; curve 2 for 2·4 wt % CO2; curves 3 and 4 for 5 and 10wt % CO2, respectively.It should be noted that 90% of the CO2 is exsolved into the gas phase at pressure �300MPa, except in the former case (0·25wt % CO2). Thedata points for the melt inclusions are bracketed by curves 2 and 3, which track closed-system conditions of magma decompression and ascent.Deviation to the left-hand side of curves 2 and 3 indicates the preferential effect of CO2 on magma dehydration, whereas data points that ploton the right-hand side suggest a higher initial H2O content in the melt. (b) Evolution of the gas phase composition (XCO2 in mole fraction)computed, using Papale’s updated model (Papale et al., 2006), with magma decompression, for total CO2 contents of 2·4, 5 and 10wt %, respec-tively. Symbols as in Fig. 6.



Francalanci et al., 2004, 2005; Landi et al., 2004). We pro-pose that the HP magma is generated by the effective crys-tallization of bubbly (LP) basaltic magma blobs fluxed bya high proportion of CO2-rich gas (45wt %) in the4^2 km depth range (Figs 7 and 8). Hence, CO2, by enhan-cing magma crystallization, indirectly affects the densityand viscosity of the Stromboli basalts and more broadlythose of water-rich magmas.

Sulfur and Cl degassingBy combining melt entrapment pressures, deduced fromthe H2O and CO2 measurements and the S concentrationsin melt inclusions, we have assessed the initial exsolutionpressure of S during ascent of basaltic melt to be�170MPa (Fig. 9a). A high pressure of sulfur partitioninginto a gas phase is in agreement with previous pressureestimates for water-rich, oxidized basaltic magmas(140MPa) as observed at Mt Etna (Spilliaert et al., 2006)and other volcanoes (e.g. Irazu' ; Benjamin et al., 2007).Therefore, a large proportion (475%) of the S is exsolvedbetween 170 and 50MPa (6^2 km) as its concentrationin the melt declines from 1570 to �500 ppm. In the samepressure range, Cl exsolution is almost negligible (Fig. 9b).As a whole, by comparing the S and Cl concentrationsin basaltic melt inclusions (in ppm, S¼1507�80;Cl¼1520�50) and those in the glassy matrix (in ppm,

S �250^170; Cl � 1270^1150) of the 15 March 2007 basalt,we assess a total S loss of 90% but only 16% for Cl. Theseestimates are also valid for the 5 April 2003 paroxysm ofintermediate magnitude as the range of S and Cl concen-trations prior and after eruption are similar (Me¤ trichet al., 2005).The bulk degassing budget of these two volatile species is

much more important, in particular for Cl, duringsteady-state activity (persistent degassing, strombolianactivity). It achieves 99% and �69% for S and Cl, respec-tively, as deduced from the S and Cl concentrations inmelt inclusions (in ppm, S¼ 980 to �80; Cl ¼ �2700^2900 to 1200) and glassy shoshonitic matrix (in ppm,S¼ 80� 40; Cl¼1200�50), and the proportion of crystals(Xc¼1 ^ f, where f¼K2O0/K2O¼ 0·43) of the HPmagma.The bulk loss of Cl calculated for typical strombo-lian activity at Stromboli is by far larger than that com-monly reported at basaltic volcanoes (10^50%; Me¤ trich &Wallace, 2008) but most probably applies to open-conduitbasaltic volcanoes producing H2O-rich basalts and exhibit-ing excessive degassing.The common occurrence of sulfide immiscibility in

shoshonitic melt inclusions suggests that the magmaachieves or is close to sulfide saturation, whereas sulfur ismainly dissolved as sulfate in the water-rich magma(Bertagnini et al., 2003). Transition from sulfate to sulfide

300 11.3

200re

(MP

a)7.6

Depth

(

Deep seatedvolatile-richLP basalt

100

Pre

ssu

3.8

km)

Shallow plagioclase-bearing, HP magma

0

0.2 0.6 1.0 1.4K2O/Na2O

0

Fig. 8. Pressure-related variation of the Na2O/K2O ratio in olivine-hosted melt inclusions, embayments and glassy matrices of pumice andscoria. This illustrates (1) prevalently closed-system decompression and LP basalt ascent, without crystallization, during paroxysmal eruptions,and (2) LP^HP basalt transition in the 100^50MPa pressure range. The photomicrograph on the lower left shows the glassy matrix of the15 March 2007 pumice (scale bar 100 mm). The shoshonitic compositions typical of glassy matrices (crosses) of the present-day HP basalt aretaken from Landi et al. (2009). Symbols as in Fig. 6.



speciation of sulfur in the melt requires a change in theredox conditions of the basaltic magma that is crystallizingand degassing in the 100^50MPa range. To test thishypothesis we verified the olivine-melt equilibria. For theshoshonitic melt hosted in Fo69·5 (ST182-11; Table 1) themeasured K

Mg�FeD Ol�Liq value (0·302) matches well that pre-

dicted (0·294) by the model of Toplis (2005), for an fO2 buf-fered at FMQ and a temperature of 11208C. Similarresults were obtained for melt inclusions having a smallsulfide globule hosted in Fo71·6 (i.e. ST133-10a; Table 1). For15 March 2007 pumice, the measured (0·32^0·34) and pre-dicted (0·31) KD values indicate equilibrium between oli-vine (Fo86) and a more oxidized melt (NNO betweenþ0·2 and þ1). Crystallization and degassing would drivethe melt towards more reduced conditions while S(470%) is significantly fractionated into the gas phase, sul-fide having a minor effect on the bulk S degassing. Sulfuris thus prevalently removed from the magma pondingbetween 4 and 2 km below the vents (Fig. 9a). Such a

feature is in agreement with the idea of a dominant recy-cling of degassed magma beneath the volcanic pile(2·8 km) (e.g. Francis et al., 1993; Allard et al., 1994, 2008).One of the striking observations is that Cl also seems to

be significantly removed from the melt, whereas it remainsdissolved until very low pressure during closed-systemmagma ascent (Fig. 9b). Degassed magma drains backinto the conduit and recycling could explain the low Clconcentrations. However, S should also be depleted in themelt, which is not the case as high S contents up to730^950 ppm are measured for the same inclusions(Fig. 9a). More broadly, the solubility of Cl depends onthe melt composition, water content and pressure, withopposite effects of pressure and crystallization upon cool-ing (Webster et al., 1999; Aiuppa et al., 2009a). AtStromboli, the efficient shift of melt composition frombasalt to shoshonite may affect the Cl solubility whereasthe [(AlþNaþCaþMg)/Si] molar ratio in the meltdeclines from 0·70 to 0·49. Alternatively, Cl depletion inthe more Si-rich shoshonite is possibly accounted for by itsefficient partitioning into an H2O-rich vapor (e.g.Websteret al., 1999; Alletti et al., 2009). The XH2O in the vaporphase is 0·3^0·5mol %, within the 100^50MPa pressurerange (Fig. 7b). We propose that preferential Cl removalfrom the shoshonitic melt by uprising CO2^H2O-bearinggas bubbles within this pressure range is a suitable pro-cess to account for excessive Cl degassing at Stromboli.This suggests a rather high pressure of Cl exsolutionduring refilling episodes of the shallow (�4 km) plumbingsystem. Following this line of reasoning, Stromboli basalticmagma is probably strongly degassed with respect to Cl(and S) prior to ascending through the shallow dike-likesystem (Chouet et al., 2008) where gas bubble percolationprevails (Burton et al., 2007b).

Inference on the plumbing system andvolcanic activityWe propose here an interpretative model of the plumbingsystem of Stromboli (Fig. 10) by combining the melt satura-tion pressures retrieved from the CO2 and H2O contentsof melt inclusions, olivine textures and the new and previ-ous mineralogical data.The pressures calculated using the Papale et al. (2006)

model confirm the presence of a magma ponding zoneextending vertically from 7 to 10 km, in accordance withprevious findings (Bertagnini et al., 2003) and availableexperimental petrology (Di Carlo et al., 2006). This deepsystem is refilled either directly by Ca-basalts as illustratedduring large-scale paroxysms or by magmas whose compo-sitions possibly span a range from Ca-basalt to slightlymore evolved basalts, depending of the magma rechargerates. System reloading with magma and CO2-rich gaspromotes the development of bubble plumes (Philips &Woods, 2001) and the formation of low-density LP basaltblobs, coexisting with variable proportions of CO2-rich

0.2

0.1

S w

t%

0.00.3

(a)

0.2

t%

0 0

0.1

Clw

(b)0.0

0 50 100 150 200 250

Pressure (MPa)

Fig. 9. Pressure-related evolution of S (a) and Cl (b) in sealed meltinclusions and unsealed embayments hosted in olivines and glassymatrices of Stromboli pumice and scoria samples. Sulfur is exsolvedfrom the basaltic melt into the gas phase at �170MPa. A noteworthyfeature is the contrasting behavior of Cl during closed-system ascentand decompression of LP basalt (no crystallization) with respect toconditions of much slower magma motion (steady-state conditions)and basalt crystallization (see text for discussion). Symbols as in Fig. 6



gas, which are able to move up, intrude and refill the shal-low system. Efficient crystallization of LP basalts wouldbe enhanced by CO2 fluxing, inducing water loss in thedepth interval between 54 and 2 km (below the vents).We thus highlight here a complex structural transitionfrom the deep (7^10 km) to the shallow (2^4 km) magmastorage systems, possibly related to the cumulative body orcrystal mush that was built up beneath Stromboli duringits past and present-day history (e.g. Francalanci et al.,2004, 2005). The existence of this body is evidenced by thepresence of inherited crystals (xenocrysts), recording

complex zoning and dissolution surfaces related torepeated interactions with the ascending LP magma,which are systematically found in pumices (e.g.Bertagnini et al., 2003; Francalanci et al., 2004, 2005). Thisbody or mush, to which is contributed the recycled,degassed HP basalt, possibly acts as a filter press. Thisinterpretation accounts for (1) the efficiency of CO2-richgas to drive, in open conduit conditions, the LP magma toextensively crystallize within the shallow plumbing systemroots, and (2) the small size in pumice of olivine and clino-pyroxene, which cluster at �0·5^0·8mm.

Fig. 10. Interpretative model of the plumbing system of Stromboli based on (1) the melt compositions and melt saturation pressures recorded bysealed melt inclusions and unsealed embayments trapped during olivine crystallization, (2) the crystal textures and (3) the chemistry ofStromboli pumice and scoria samples. The fluid saturation pressures are calculated using Papale et al. (2006) and the corresponding lithostaticdepth assuming an average density of the surrounding rocks of 2700 kg/m3. The geometry of the conduits within the uppermost 1kmçimagedas two complex dike systems underlying the summit craters with a sharp change in fracture dip at 80 and 280m, respectively, below sea level(Chouet et al., 2008)çis oversimplified. LP and HP basalts refer to low- and high-porphyritic magmas, respectively, according to Francalanciet al. (2004).



Moreover, we provide new constraints on the triggeringmechanisms of the large- to small-scale paroxysms, giventhat the persistent activity and the sustained flux of gas atStromboli imply a continuous (or discontinous, but perma-nent) refilling of the deep plumbing system. The occur-rence of deep recharges of magma with either its own gasphase or deeper-derived CO2, and the development of abubbly layer, leads us to propose that paroxysms are trig-gered by overpressure in the storage zone lying at 7^10 kmdepth. The variable energy of the paroxysms is regardedas a response to different rates of magma and gas reloadingand system pressurization, which in turn controls thevolume of the gas and magma discharged from the deepponding zone. Magma withdrawal through lava outpour-ing in the summit zone (i.e. 2003 and 2007 effusive crises)possibly amplifies the effect of decompression but is not acondition required for paroxysmal eruptions regardless oftheir magnitude. This conclusion is further supported bythe observation that paroxysms are not systematically pre-ceded by effusive activity (Bertagnini et al., 2008). Finally,we stress that the gradual increase of the number of explo-sion quakes, the tremor amplitude (Ripepe et al., 2009),and the CO2/SO2 ratio in the gas plume (Aiuppa et al.,2009b) recorded prior to the 15 March 2007 paroxysm aremost probably the direct consequences of the pressuriza-tion of the deep system.

CONCLUSIONSOur new observations on the textures and compositions ofboth olivines and their melt inclusions in basaltic pumiceand scoria of Stromboli shed light on the conditions ofmagma storage, crystallization, degassing and extrusion.The Stromboli plumbing system can be regarded as asuccession of magma ponding zones connected by dikes.The 7^10 km depth interval, at which magmas are storedand differentiate, is periodically recharged by newmagma batches, possibly ranging from Ca-basalt to basaltin composition, with a CO2-rich gas phase.The transition from the LP basalts towards HP,

plagioclase-bearing basalts is well-defined and based upona combination of several features: (1) arrival of magma ofdeep derivation with a gas phase that promotes the forma-tion of bubbly LP basalt blobs; (2) the existence of a filterpress (i.e. cumulative body or crystal mush) throughwhich the bubbly blobs travel prior to intruding the rootsof the shallow system (�4^2 km); (3) the ability of theCO2-rich gas to drive the LP basalt towards the domainof plagioclase crystallization in response to H2O loss, pro-moting extensive crystallization and the generation of HPmagma; (4) the low magma extrusion rate.The eruptive dynamics of the paroxysms are controlled

by uprising of magma that essentially remains in equilib-rium with its gas (prevalent closed-system conditions) asopposed to open-system strombolian activity dominated

by slower magma ascent rates and differential transfer ofgas bubbles.We emphasize that the triggering mechanismsof the paroxysmal eruptions probably reflect a subtle inter-play between magma supply rate, residence time and theviscoelastic response of the deep reservoir system in accor-dance with larger time spans between the largest explosiveevents.

ACKNOWLEDGEMENTSWe thank A. Aiuppa, T. W. Sisson and J. Corte¤ s for theirpositive and constructive comments, P. Papale for discus-sion on melt CO2 solubility, M. Pompilio for fruitfuldiscussions and his critical reading of an early version ofthe manuscript, D. Massare for heating experiments,O. Belhadj for her kind help in sample preparation, andP. Pantani for graphical assistance.

FUNDINGThis work was supported by the projects V2-Monitoringand research activity at Stromboli and Panarea (2004-2006 INGV-DPC agreement) and V2-Paroxysm(2007-2009 INGV-DPC agreement) and in part by ANR-MIME, contract ANR-07-BLAN-0130-02.

SUPPLEMENTARY DATASupplementary data for this paper are available at Journalof Petrology online.

REFERENCESAiuppa, A., Baker, D. R. & Webster, J. D. (2009a). Halogens in

volcanic systems. Chemical Geology 263, 1^18.Aiuppa, A., Federico, C., Giudice, G., Giuffrida, G., Guida, R.,

Gurrieri, S., Liuzzo, M., Moretti, R. & Papale, P. (2009b). The2007 eruption of Stromboli volcano: Insights from real-time mea-surements of the volcanic gas plume CO2/SO2 ratio. Journal of

Volcanology and Geothermal Research 182, 221^230.Allard, P., Carbonnelle, J., Me¤ trich, N., Loyer, H. & Zettwoog, P.

(1994). Sulphur output and magma degassing budget of Strombolivolcano. Nature 368, 326^330.

Allard, P., Aiuppa, A., Burton, M., Caltabiano, T., Federico, C.,Salerno, G. & La Spina, A. (2008). crater gas emissions and themagma feeding system of Stromboli volcano. In: Calvari, S.,Inguaggiato, S., Puglisi, G., Ripepe, M. & Rosi, M. (eds) Learningfrom Stromboli. American Geophysical Union, Geophysical Monograph 182,65^80.

Alletti, M., Baker, D. R. & Freda, C. (2007). Halogen diffusion in abasaltic melt. Geochimica et Cosmochimica Acta 71, 3570^3580.

Alletti, M., Baker, D. R., Scaillet, B., Aiuppa, A., Moretti, R. &Ottolini, L. (2009). Chlorine partitioning between a basaltic meltand H2O^CO2 fluids from Mt Etna. Chemical Geology 263, 37^50.

Asimow, P. D. & Ghiorso, M. S. (1998). Algorithmic modificationsextending MELTS to calculate subsolidus phase relations.American Mineralogist 83, 1127^1132.



Baker, D. R. (2008). The fidelity of melt inclusions as records of meltcomposition. Contributions to Mineralogy and Petrology doi:10.1007/s00410-008-0291-3.

Barberi, F., Rosi, M. & Sodi, A. (1993).Volcanic hazard assessment atStromboli based on review of historical data. Acta Vulcanologica 3,173^187.

Barberi, F., Civetta, L., Rosi, M. & Scandone, R. (2009). Chronologyof the 2007 eruption of Stromboli and the activity of the ScientificSynthesis Group. Journal of Volcanology and Geothermal Research 182,doi:10.1016/j.jvolgeores.2008.09.019.

Barberi, G., Zhang, H., Scarfi, L., Cocina, O., Castellano, M.,Chiarabba, C. & Patane' , D. (2007). Crustal evidence of a low velo-city Vp and Vs volume beneath Stromboli, Italy. Geophysical

Research Abstracts 9, 02621.Benjamin, E. R., Plank,T.,Wade, J. A., Kelley, K. A., Hauri, E. H. &

Alvarado, G. E. (2007). High water contents in basaltic magmasfrom Irazu' volcano, Costa Rica. Journal of Volcanology and

Geothermal Research 168, 68^92.Bertagnini, A., Coltelli, M., Landi, P., Pompilio, M. & Rosi, M.

(1999). Violent explosions yield new insights into dynamics ofStromboli volcano. EOS Transactions, American Geophysical Union 80,633^636.

Bertagnini, A., Me¤ trich, N., Landi, P. & Rosi, M. (2003). Stromboli anopen window on the deep feeding system of a steady state volcano.Journal of Geophysical Research 108, 2336, doi:10.1029/2002JB002146.

Bertagnini, A., Me¤ trich, N., Francalanci, L., Landi, P., Tommasini, S.& Conticelli, S. (2008). Volcanology and magma geochemistry ofthe present-day activity: constraints on the feeding system. In:Calvari, S., Inguaggiato, S., Puglisi, G., Ripepe, M. & Rosi, M.(eds) Learning from Stromboli. American Geophysical Union, Geophysical

Monograph 182, 19^38.Bonaccorso, A., Cardaci, C., Coltelli, M., Del Carlo, P., Falsaperla, S.,

Panucci, S., Pompilio, M. & Villari, L. (1996). Annual report ofthe world volcanic eruptions in 1993, Stromboli. Supplement to

Bulletin of Volcanic Eruptions 35, 8^14.Burton, M. R., Allard, P., Mure¤ , F. & La Spina, A. (2007a).

Magmatic gas composition reveals the source depth of slug-drivenstrombolian explosive activity. Science 37, 227^230.

Burton, M. R., Mader, H. M. & Polacci, M. (2007b). The role ofgas percolation in quiescent degassing of persistently activebasaltic volcanoes. Earth and Planetary Science Letters doi:10.1016/j.epsl.2007.08.028.

Cashman, K.V. (2004). Volatile controls on magma ascent and erup-tion. In: Sparks, R. S. J. & Hawkesworth, C. J. (eds) The State of

the Planet: Frontiers and Challenges in Geophysics. American Geophysical

Union, Geophysical Monograph 150, 109^124.Chouet, B., Hamisevicz, N. & McGetchin,T. R. (1974). Photoballistics

of volcanic jet activity at Stromboli, Italy. Journal of Geophysical

Research 79, 4961^4976.Chouet, B., Dawson, P. & Martini, M. (2008). Upper conduit struc-

ture and explosion dynamics at Stromboli. In: Calvari, S.,Inguaggiato, S., Puglisi, G., Ripepe, M. & Rosi, M. (eds) Learningfrom Stromboli. American Geophysical Union, Geophysical Monograph 182,81^92.

Cigolini, C., Laiolo, M. & Bertolino, S. (2008). Probing Stromboli vol-cano from the mantle to paroxysmal eruptions. In: Annen, C. &Zellmer, G. F. (eds) Dynamics of crustal magma transfer, storage and dif-

ferentiation. Geological Society, London, Special Publications 304, 33^70.Collins, S. J., Pyle, D. M. & Maclennan, J. (2009). Melt inclusions

track pre-eruption storage and dehydration of magmas at Etna.Geology 37, 571^574, doi:10.1130/G30040A.1.

Di Carlo, I., Pichavant, M., Rotolo, S. & Scaillet, B. (2006).Experimental crystallization of a high-K arc basalt: the Golden

Pumice, Stromboli Volcano (Italy). Journal of Petrology 47,1317^1343, doi:10.1093/petrology/egl011.

Di Muro, A.,Villemant, B., Montagnac, G., Scaillet, B. & Reynard, B.(2006). Quantification of water content and speciation in naturalsilicic glasses (phonolite, dacite, rhyolite) by confocal Micro-Raman spectroscopy. Geochimica et Cosmochimica Acta 70, 2868^2884.

Donaldson, C. H. (1976). An experimental investigation of olivinemorphology. Contributions to Mineralogy and Petrology 57, 187^213.

Faure, F. & Schiano, P. (2004). Crystal morphologies in pillow basalts:implications for mid-ocean ridge processes. Earth and Planetary

Science Letters 220, 331^344.Faure, F. & Schiano, P. (2005). Experimental investigation of

equilibration conditions during forsterite growth and meltinclusion formation. Earth and Planetary Science Letters 236, 882^898.

Francalanci, L., Tommasini, S., Conticelli, S. & Davies, G. R. (1999).Sr isotope evidence for short magma residence time for the 20thcentury activity at Stromboli volcano, Italy. Earth and Planetary

Science Letters 167, 61^69.Francalanci, L., Tommasini, S. & Conticelli, S. (2004). The volcanic

activity of Stromboli in the 1906^1998 period: mineralogical,geochemical and isotope data relevant to the understanding ofStrombolian activity. Journal of Volcanology and Geothermal Research

131, 179^211.Francalanci, L., Davies, G. R., Lustenmhower, W., Tommasini, S.,

Mason, P. R D. & Conticelli, S. (2005). Old crystal re-cycle andmultiple magma reservoirs in the plumbing system of the presentday activity at Stromboli volcano, South Italy: Sr-isotope in situ

microanalyses. Journal of Petrology 46, 1997^2021, doi:10.1093/petrol-ogy/egi045.

Francis, P., Oppenheimer, C. & Stevenson, D. (1993). Endogenousgrowth of persistent active volcanoes. Nature 366, 554^557.

Freda, C., Baker, D. R. & Scarlato, P. (2005). Sulfur diffusion inbasaltic melts. Geochimica et Cosmochimica Acta 69, 5061^5069.

Ghiorso, M. S. & Sack, R. O. (1995). Chemical transfer in magmaticprocesses IV. A revised and internally consistent thermodynamicmodel for the interpolation and extrapolation of liquid^solid equili-bria in magmatic system at elevated temperatures and pressures.Contributions to Mineralogy and Petrology 119, 197^212.

Jaupart, C. & Vergniolle, S. (1988). Laboratory models of Hawaiianand Strombolian eruptions. Nature 331, 58^60.

Johnson, E. J.,Wallace, P. J., Cashman, K., Delgado Granados, V. &Kent, A. (2008). Magmatic volatile contents and degassing-inducedcrystallization at Volcan Jorullo, Mexico: Implications for meltevolution and the plumbing systems of monogenetic volcanoes.Earth and Planetary Science Letters 269, 477^486.

Kamenetsky, V., Pompilio, M., Me¤ trich, N., Sobolev, A. V.,Kuzmin, D.V. & Thomas, R. (2007). Arrival of extremely volatile-rich high-Mg magmas changes explosivity of Mount Etna. Geology35, 255^258, doi:10.1130/G23163A.1.

Kohut, E. & Nielsen, R. L. (2004). Melt inclusion formation mechan-isms and compositional effects in high-An feldspar and hogh-Foolivine in anhydrous madic silicate liquids. Contributions to

Mineralogy and Petrology 147, 684^704.Landi, P., Me¤ trich, N., Bertagnini, A. & Rosi, M. (2004). Dynamics of

magma mixing and degassing recorded in plagioclase atStromboli (Aeolian Archipelago, Italy). Contributions to Mineralogy

and Petrology 147, 213^227.Landi, P., Francalanci, L., Pompilio, M., Rosi, M., Corsaro, R. A.,

Petrone, C. M., Nardini, I. & Miraglia, L. (2006). The December2002^July 2003 effusive event at Stromboli volcano, Italy: insightsinto the shallow plumbing system by petrochemical studies.Journal of Volcanology and Geothermal Research 155, 263^284.



Landi, P., Corsaro, R. A., Francalanci, L., Civetta, L., Miraglia, L.Pompilio & Tesoro, R. (2009). Magma dynamics during the 2007Stromboli eruption (Aeolian islands, Italy). Mineralogy, geochem-istry and isotope data. Journal of Volcanology and Geothermal Research

182, 255^268.Me¤ dard, E., Schmidt, M. W., Schiano, P. & Ottolini, L. (2006).

Melting of amphibole-nearing wehrlites: an experimental study ofthe origin of ultra-calcic ne-normative melts. Journal of Petrology 47,481^504.

Mercier, M., Di Muro, A., Giordano, D., Me¤ trich, N., Lesne, P.,Pichavant, M., Scaillet, B., Clocchiatti, R. & Montagnac, G.(2009). Influence of glass polymerisation and oxidation on micro-Raman water analysis in alumino-silicate glasses. Geochimica et

Cosmochimica Acta 73, 197^217, doi:10.1016/j.gca.2008.09.030.Me¤ trich, N. & Wallace, P. (2008). Volatile abundances in basaltic

magmas and their degassing paths tracked by melt inclusions. In:Putirka, K. & Tepley, F. (eds) Minerals, Inclusions and Volcanic

Processes. Mineralogical Society of America, Reviews in Mineralogy and

Geochemistry 69, 363^402.Me¤ trich, N., Bertagnini, A., Landi, P. & Rosi, M. (2001).

Crystallisation driven by decompression and water loss atStromboli volcano (Aeolian Islands). Journal of Petrology 42,1471^1490.

Me¤ trich, N., Bertagnini, A., Landi, P., Rosi, M. & Belhadj, O. (2005).Triggering mechanism at the origin of paroxysms at Stromboli(Aeolian archipelago, Italy): the 5 April 2003 eruption. Journal ofGeophysical Research 32, L103056, doi:10.1029.

Moore, G. (2008). Interpreting H2O and CO2 contents in melt inclu-sions: constraints from solubility experiments and modeling. In:Putirka, K. & Tepley, F. (eds) Minerals, Inclusions and Volcanic

Processes. Mineralogical Society of America, Reviews in Mineralogy and

Geochemistry 69, 333^361.Panza, G. F., Peccerillo, A., Aoudia, A. & Farina, B (2007).

Geophysical and petrological modelling of the structure and com-position of the crust and upper mantle in complex geodynamic set-tings: The Tyrrhenian sea and surroundings. Earth-Science Reviews80, 1^46.

Papale, P. (1999). Modeling of the solubility of two-component H2O^CO2 fluid in silicate liquids. American Mineralogist 84, 477^492.

Papale, P., Moretti, R. & Barbato, D. (2006). The compositionaldependence of the saturation surface of H2O þ CO2 fluids in sili-cate melts. Chemical Geology 29, 78^95.

Parfitt, E. A. & Wilson, L. (1999). Plinian treatment of fallout fromHawaiian lava fountains. Journal of Volcanology and Geothermal

Research 88, 67^75.Phillips, J. C. & Woods, A.W. (2001). Bubble plumes generated during

recharge of basaltic magma reservoirs. Earth and Planetary Science

Letters 18, 297^309.Pichavant, M., Di Carlo, I., Le Gac, Y., Rotolo, S. G. & Scaillet, B.

(2009). Experimental constraints on the deep magma feedingsystem at Stromboli volcano, Italy. Journal of Petrology 50, 601^624,doi:10.1093/petrology/egp014.

Pompilio, M., Bertagnini, A., Me¤ trich, N. & Belhadj, O. (2009).Geochemical decoupling between shallow and deep-seatedmagmas at Stromboli. Inference on the plumbing system.

Proceedings of Rittmann conference, 11^13 June 2009, Catania,Italy, p. 39.

Ripepe, M., Delle Donne, D., Lacanna, G., Marchetti, E. &Ulivieri, G. (2009). The onset of the 2007 Stromboli effusive erup-tion recorded by an integrated geophysical network. Journal of

Volcanology and Geothermal Research 182, 131^136.Rittmann, A. (1931). Der Ausbruch des Stromboli am 11 September

1930. Zeitschrift fu« rVulkanologie 14, 47^77.Rosi, M., Bertagnini, A. & Landi, P. (2000). Onset of the persistent

activity at Stromboli volcano (Italy). Bulletin of Volcanology 62,294^300.

Rosi, M., Bertagnini, A., Harris, A. J. L., Pioli, L. & Ripepe, M.(2006). A case history of paroxysmal explosion at Stromboli:timing and dynamics of the April 5, 2003 event. Earth and PlanetaryScience Letters 243, 594^606.

Scaillet, B. & Pichavant, M. (2005). A model of sulphur solubility forhydrous mafic melts: application to the determination of magmaticfluid compositions of Italian volcanoes. Annals of Geophysics 48,671^698.

Schiano, P., Eiler, J. M., Hutcheon, I. D. & Stolper, E. M. (2000).Primitive CaO-rich, silica undersaturated melts in island arcs:Evidence for the involvement of clinopyroxene-rich lithologies inthe petrogenesis of arc magmas. Geochemistry, Geophysics, Geosystems1, 1^36.

Schiano, P., Clocchiatti, R., Ottolini, L. & Sbrana, A. (2004). Therelationship between potassic, calc-alkaline and Na-alkaline mag-matism in South Italy volcanoes: a melt inclusion approach. Earthand Planetary Science Letters 220, 121^137.

Smith, P. M. & Asimow, P.W. (2005). Adiabat_1ph: A new public front-end to the MELTS, p-Melts, and pHMELTS models. Geochemistry,Geophysics, Geosystems 6, Q02004. doi:10.1029/2004GC000816.

Speranza, F., Pompilio, M. & Sagnotti, L. (2004). Paleomagnetism ofspatter lavas from Stromboli volcano (Aeolian Islands, Italy):Implications for the age of paroxysmal eruptions. Geophysical

Research Letters 31, L0260, doi:10.1029/2003GL018944.Spilliaert, N., Allard, P., Me¤ trich, N. & Sobolev, A. (2006). Melt inclu-

sion record of the conditions of ascent, degassing and extrusion ofvolatile-rich alkali basalt during the powerful 2002 flank eruptionof Mount Etna (Italy). Journal of Geophysical Research 111, B04203,doi:10.1029/2005/JB003934.

Toplis, M. J. (2005). The thermodynamics of iron magnesium parti-tioning between olivine and liquid: criteria for assessing and pre-dicting equilibrium in natural and experimental systems.Contributions to Mineralogy and Petrology 149, 22^39, doi:10.1007/s00410-004-0629-4.

Watson, E. B. (1994). Diffusion in volatile-bearing magmas. In:Carroll, M. & Holloway, J. R. (eds) Volatiles in Magmas.

Mineralogical Society of America, Reviews in Mineralogy and Geochemistry

30, 371^411.Webster, J. D., Kinzler, R. J. & Mathez, E. A. (1999). Chloride and

water solubility in basalt and andesite melts and implication formagmatic degassing. Geochimica et Cosmochimica Acta 63, 729^738.

Wilson, L. & Head, J. W. (1981). Ascent and eruption of basalticmagma on the Earth and Moon. Journal of Geophysical Research 86,2971^3001.



Conditions of Magma Storage, Degassing and Ascent at ...€¦ · RECEIVEDJUNE 16, 2009; ACCEPTED...

Documents

Transcript of Conditions of Magma Storage, Degassing and Ascent at ...€¦ · RECEIVEDJUNE 16, 2009; ACCEPTED...