The November 2002 Eruption of Piton De La Fournaise, Réunion: Tracking the Pre-Eruptive Thermal Evolution of Magma Using Melt Inclusions

Bull Volcanol (2009) 71:1077–1089 DOI 10.1007/s00445-009-0287-5

RESEARCH ARTICLE

The November 2002 eruption of Piton de la Fournaise, Réunion: tracking the pre-eruptive thermal evolution of magma using melt inclusions

Nathalie Vigouroux & A. E. Williams-Jones & Paul Wallace & Thomas Staudacher

Received: 19 November 2007 /Accepted: 29 April 2009 /Published online: 6 June 2009 # Springer-Verlag 2009

Abstract The November 2002 eruption of Piton de la estimate eruption temperatures. The November 2002 melt Fournaise in the Indian Ocean was typical of the activity of inclusion compositions suggest that they were at temper- the volcano from 1999 to 2006 in terms of duration and atures between 1,070°C and 1,133°C immediately before volume of magma ejected. The first magma erupted was a eruption and quenching. This relatively wide temperature basaltic liquid with a small proportion of olivine phenocrysts range may reflect the fact that most of the melt inclusions

(Fo81) that contain small numbers of melt inclusions. In were from olivine in lava samples and therefore likely subsequent flows, olivine crystals were more abundant and underwent minor but variable amounts of post-eruptive 2+ richer in Mg (Fo83–84). These crystals contain numerous melt crystallization and Fe loss by diffusion due to their and fluid inclusions, healed fractures, and dislocation relatively slow cooling on the surface. In contrast, melt features such as kink bands. The major element composition inclusions in tephra samples from past major eruptions of melt inclusions in this later olivine (Fo83–84)isoutof yielded a narrower range of higher eruption temperatures equilibrium with that of its host as a result of extensive post- (1,163–1,181°C). The melt inclusion data presented here and entrapment crystallization and Fe2+ loss by diffusion during in earlier publications are consistent with a model of magma cooling. Melt inclusions in Fo81 olivine are also chemically recharge from depth during major eruptions, followed by out of equilibrium with their hosts but to a lesser degree. storage, cooling, and crystallization at shallow levels prior to Using olivine–melt geothermometry, we determined that expulsion during events similar in magnitude to the relatively melt inclusions in Fo81 olivine were trapped at lower small November 2002 eruption. temperature (1,182±1°C) than inclusions in Fo83–84 olivine (1,199–1,227°C). This methodology was also used to Keywords Piton de la Fournaise . Volcanic plumbing system . Melt inclusions . Editorial responsibility: M. Clynne Olivine cumulates . Post-entrapment modifications . A. E. Williams-Jones Volatiles . Magma evolution Department of Earth and Planetary Sciences, McGill University, Montreal, QC, Canada Introduction P. Wallace Department of Geological Sciences, University of Oregon, Eugene, OR, USA The 530-ka Piton de la Fournaise shield volcano is the current manifestation of the Réunion hot spot, which T. Staudacher produced the Deccan Traps in India at about 65 Ma Observatoire Volcanologique du Piton de la Fournaise, IPGP, Réunion, France (Courtillot et al. 1986). It is located at the southern end of the Mascarene Basin, encompassing the islands of Maur- Present address: itius and Rodrigues. Piton de la Fournaise rests on the south * N. Vigouroux ( ) flank of the older and now extinct Piton des Neiges, which Department of Earth Sciences, Simon Fraser University, Burnaby, BC, Canada V5A 1S6 forms most of present day Réunion Island (Fig. 1), and is e-mail: [email protected] one of the world’s most active volcanoes. 1078 Bull Volcanol (2009) 71:1077–1089

The anatomy of Piton de la Fournaise is similar to that of the early phase (first day) of the November 2002 eruption Kilauea in that a central conduit links a shallow magma crystallized from the magma hosting it, which was more chamber with the summit craters, and flank eruptions are evolved and cooler than the magma erupted during past fed by lateral dyke intrusions connected to the central events studied by Bureau et al. (1998b). The bulk of the conduit (Peltier et al. 2007). Fissures from historic flank olivine, which was ejected in increasing proportions during eruptions are mostly distributed along an arcuate rift zone the later stages of the 2002 eruption (late phase) and is less extending from the northeast of the caldera to the southeast, evolved, appears to have been entrained in this lower passing through the central craters (Peltier et al. 2005). temperature magma. This caused it to experience a larger Recent models based on seismicity and deformation of the amount of pre-eruptive cooling than olivine of similar volcano prior to eruption suggest that Piton de la Fournaise composition ejected during the earlier events studied by has a magma chamber with a volume of ~300×106 m3 Bureau et al. (1998b). These results are consistent with the located between sea level and 500 m above sea level (a.s.l.; model of crystallization, accumulation, storage, and erup- Sapin et al. 1996; Nercessian et al. 1996; Peltier et al. tion proposed previously by Bureau et al. (1998b, 1999), in 2007). A deeper magma chamber, located at 5–6 km below which they used melt inclusions to reveal the crystallization sea level (b.s.l.), has also been postulated (Battaglia et al. and storage of olivine Fo83–85 at a variety of depths ranging 2005) and olivine crystallization has been shown to occur from ~5 km b.s.l. to the near surface, prior to eruption from as deep as 15 km b.s.l. (Bureau et al. 1998a, b). Finally, a the central conduit area of the volcano. They showed that crystal cumulate zone, evident as a high-density plug on the the volatile contents found in the melt inclusions from a basis of seismic data, is likely to be present within the single eruption could not be reconciled with a plausible central cone above sea level (Nercessian et al. 1996; Sapin degassing trend and therefore must represent trapping of et al. 1996) and to extend to the Moho at a depth of about individual melt batches with variable degassing histories. 12.5 km b.s.l. (Gallart et al. 1999). Information on the depth of crystallization and accumu- Background geology lation of magma within the volcano is crucial to further understanding the plumbing system of Piton de la Four- The summit area of Piton de la Fournaise volcano is naise. Melt inclusions provide an effective tool for this enclosed on three of its sides by the 4,500-year-old Enclos purpose because they trap small pockets of magma at Fouqué, the remnant of the youngest of three calderas, different stages during the crystallization of crystals in whereas its east flank is open to the Indian Ocean. The response either to undercooling (Faure et al. 2003)or summit rises to 2,631 m a.s.l. and is crowned by two changes in the chemical environment (e.g., magma mixing) intersecting craters, Bory and Dolomieu, the latter one (Sobolev 2007). Analysis of the chemical composition, being the larger and representing the current center of including volatile contents (H2OandCO2), of melt activity (Fig. 1). inclusions, yields information on the depth of crystalliza- Most of the historic eruptions have been from vents tion and storage, as well as trends in the chemical evolution located inside the walls of the current caldera, at various of the magma. At Piton de la Fournaise, melt inclusion elevations along the rift zone (summit to 1,000 m a.s.l.) but studies of olivine from both prehistoric (~3,000–50 ka) and three of the last four large-volume (>50×106 m3) deep- historic (1931, 1977, 1998) eruptions have shown that seated eruptions (1977, 1986, 1998) have had at least one olivine crystallization and accumulation occurs over a wide vent located outside of the caldera rim along an extension range of depths from the Moho to approximately sea level of the intracaldera rift zone. Prior to the collapse of the (Bureau et al. 1998b, 1999). The samples analyzed in these current caldera (Enclos Fouqué), the southeast and the studies were from large-volume, mostly olivine-rich mag- northeast rift zones were active along the margins of the old mas erupted from vents located outside the current caldera caldera rim (Plaine des Sables collapse; Fig. 1) and a few and rift system (prehistoric eruptions) and intracaldera vents vents/eruptive centers were preserved after the sector active in historical times (1931, 1998). collapse that created the current caldera (e.g., Piton For this study, we analyzed olivine-hosted melt inclu- Manapany; Bureau et al. 1998b). sions from the moderate-volume (8×106 m3) flank eruption The Piton de la Fournaise volcano is still in its shield of November 2002, a seismically shallow event (down to building stage, erupting lavas that range from aphyric 1.5 km b.s.l.) characteristic of the activity at Piton de la basalts to oceanites, which are olivine-rich basalts equiva- Fournaise from 1999 to 2006. Using the compositions of lent to the Hawaiian picrites (Upton and Wadsworth 1966). these inclusions after correction for post-entrapment mod- Most of the olivine erupted at Piton de la Fournaise ranges ifications, we have characterized the thermal history of the in composition from Fo81 to Fo87 (Albarède et al. 1997; magma from which the olivine crystallized and in which it Bureau et al. 1998a, b, 1999). However, basalts containing erupted. Results of this study suggest that olivine ejected in Fo83–84 olivine have been the dominant lava type during the Bull Volcanol (2009) 71:1077–1089 1079

Fig. 1 Digital elevation model of the summit area of Piton de la Fournaise. Inset shows location on Reunion Island. The two summit craters (Bory and Dolomieu) are within the youngest caldera (L’Enclos Fouqué). The two older collapse features are also labeled (Rivière des Remparts and Plaine des Sables). The cones of the 1998 eruption (Hudson, Krafft, and Kapor), the November 2002 cone (Guanyin), and the extent of the 2002 lava flows are shown. The 2002 lava flowed down the steep slope of the Grandes Pentes before crossing the Grand Brûlé and entering the ocean. Also shown is the location of the prehistorically active Piton de Caille cone, and arrows point toward the locations of two other prehistoric cones referred to in this study. The map is modified from Longpré et al. (2007)

last ~400 ka. These lavas contain 6 to 8 wt.% MgO and eruption rates (20.8 m3/s; February 2005) and ejected show little chemical variability, which has led them to be olivine-rich basalts (Vlastélic et al. 2005, 2007). Vlastélic referred to as “steady-state basalts” (Albarède et al. 1997). et al. (2007) and Peltier et al. (2009) have identified the The restricted compositional variability of the erupted lavas 1998–2007 period as an eruptive cycle during which the is probably caused by the buffering effect of cumulate crystal eruption of steady-state basalts with little entrained olivine mush zones, which also serve to filter out more Mg-rich antecrysts (crystals foreign to the erupting melt but olivine that crystallized at higher pressures (Albarède et al. originating from the magmatic system) was progressively 1997;Bureauetal.1999). replaced, starting in June 2001, by the eruption of olivine- Since historical times, volcanic activity has been rich lavas containing olivine antecrysts of cumulate origin. dominated by small- to moderate-volume (small-scale) eruptions, which occur intermittently between major, The November 2002 eruption large-volume eruptions. Six major eruptions have occurred in the twentieth and twenty-first centuries: 1931, 1961, After 3 months of intermittent seismic activity, a large 1977, 1986, 1998, and 2007 and 21 small-scale eruptions seismic crisis involving several hundred earthquakes occurred between the major eruptions of 1998 and 2007. located above sea level and strong summit deformation The volumes of material expelled during these small-scale started at Piton de la Fournaise on November 15, 2002 eruptions are one to two orders of magnitude smaller than andlasted5h(Longpréetal.2007). The eruption tremor those expelled during the major eruptions. Major and small- began on November 16, with fissures opening on the east scale events can also be distinguished on the basis of flank of the volcano at elevations of 1,900 to 1,600 m, seismic data. The 1998 and 2007 major eruptions were followed by the flow of lava down the same flank. A small accompanied by earthquakes at depths >5 km b.s.l., cone, called Piton Guanyin, formed on one of the most active whereas small-scale events are associated with earthquake fissures at an elevation of ~1,600 m (Fig. 1). Hawaiian-type depths of ≤2 km b.s.l. (Observatoire Volcanologique du activity commenced immediately and was characterized by Piton De La Fournaise (OVPDLF), unpublished data). lava fountains reaching up to 80 m in height. On the There is no correlation between eruptive volume and following day (November 17), the fountains were smaller, effusion rate. Some major eruptions had low effusion rates only up to 30 m high, due partly to drag imposed by a small (3.5 m3/s; 1998) and erupted nearly aphyric basalt, whereas lava lake that had developed within the cone’s interior. By numerous small-scale eruptions since 1998 have had high then, the eruption tremor had decreased by a factor of four, 1080 Bull Volcanol (2009) 71:1077–1089 and the fissures located at ~1,850 and ~1,750 m elevation length in the early tephra to 4 mm in length in the latest lavas. had ceased to be active, leaving Piton Guanyin as the only Olivine crystals in the late-stage lavas contain Cr-spinel active center. inclusions with a Cr # of 54–63 (Cr #=100Cr/(Cr+Al)), have Activity remained generally constant until November visible healed fractures, and are cut by numerous secondary 29th, when the eruptive tremor, number of seismic events, melt and fluid inclusion planes (Fig. 2). Some display kink and the height of the lava fountains increased suddenly banding, a feature found in other olivine crystals from Piton while the summit began to deflate. Shallow seismic events de la Fournaise identified as cumulates (Albarède et al. 1997; beneath the summit were frequent and particularly intense Bureau et al. 1999). The lack of these features in the early- with up to 6,000 events per day until December 23, when a stage olivine suggests that this olivine is not of cumulate pit crater appeared within the Dolomieu crater followed by origin. a rapid decline of seismicity (Longpré et al. 2007). Lava Primary melt inclusions are variable in size, ranging emissions ended on December 3, after 18 days of eruption from ~25 to ~150 μm in diameter. They are isolated or that had produced 8×106 m3 of material at an average occur in clusters, whereas secondary melt inclusions are output rate of 5.1 m3/s (Vlastélic et al. 2007). generally smaller and occur along healed cracks. The dominant shape is ellipsoidal but round, highly elongate, and negative crystal shapes also occur. The primary melt Rock and melt inclusion descriptions inclusions display varying degrees of crystallization due either to cracking and volatile loss upon ascent or slow Samples were collected at various times during the 2 weeks cooling during flow of the host lava at the surface. Some spanned by the eruption and comprised spatter cone tephra, primary inclusions are not visibly crystallized and were lava quenched in a bucket of water and the quickly cooled targeted for analysis. However, there is evidence that they tops of ‘a’a lava flows collected along levees. Information have undergone post-entrapment crystallization of olivine on sample type, location, and time of emplacement is on their walls (see below). All primary melt inclusions provided in Table 1. Sample REU0211-161, which is contain a shrinkage bubble. spatter cone tephra released on the first day of the eruption, is referred to hereafter as the “early-stage” sample, and all other samples, which are from lavas erupted between the Methodology 25th of November and the 2nd of December, are referred to as the “late-stage” samples. This division corresponds to a X-ray fluorescence change in the composition and proportion of the olivine crystals in the samples. Whole-rock major element analyses were conducted on The groundmass of lava samples contains visible microlites fused beads prepared from ignited samples using a Phillips (<300 μm in diameter) of Cr-spinel, plagioclase, and PW2400 3-kW automated X-ray fluorescence spectrometer clinopyroxene, whereas the tephra groundmass is glassier system at the Geochemical Laboratories of McGill Univer- and contains significantly less Cr-spinel (early-stage sample sity. The accuracy for SiO2 was within 0.5 wt.% of the REU0211-161). All samples contain olivine crystals that amount reported and for all other major elements was 1% increased in size as the eruption progressed, from <0.5 mm in of the amount present; the overall precision was 0.5%.

Table 1 Sample descriptions

Sample Date emplaced Date sampled Location Type

Early phase REU0211-161 Nov. 16, 2002 Nov. 16, 2002 S fissure, 1,600 m elevation Spatter cone tephra REU0211-171 Nov. 17, 2002 Nov. 17, 2002 N fissure, 1,550 m elevation Water-quenched lava Late phase REU0211-252 Nov. 25, 2002 Nov. 25, 2002 S channel levee, 700 m elevation Surficially cooled ‘a’a lava REU0211-254 Nov. 25, 2002 Nov. 25, 2002 S channel, 700 m elevation Water-quenched lava REU0212-014 Nov. 29, 2002 Dec. 1, 2002 S channel levee, 110 m elevation Surficially cooled ‘a’a lava REU0212-034 Nov. 29, 2002 Dec. 3, 2002 S channel levee, 120 m elevation Surficially cooled ‘a’a lava REU0212-021 Dec. 2, 2002 Dec. 2, 2002 N channel levee, 120 m elevation Surficially cooled ‘a’a lava

Only the last three digits of each sample name are used to identify the samples in subsequent tables Bull Volcanol (2009) 71:1077–1089 1081

Fourier-transform infrared spectroscopy

The water and carbon contents of primary melt inclusions were measured using a Nicolet Magna 560 Fourier- transform infrared spectrometer (FTIR) with a continuum IR microscope and OMNIC software at the University of Oregon. Doubly polished olivine wafers were prepared such that the melt inclusions were intersected on both sides. The average thickness of these wafers was 30 μm and the average aperture size used in the FTIR microscope was 50×30 μm. Transmission spectra were collected in the wavelength range of 6,000–650 cm−1 and then converted to

absorption values. For H2O absorption measurements, the peak height at 3,535 cm−1 was measured above the background. For carbonate measurements, an unpublished Fig. 2 Olivine crystal from sample REU0212-014 containing glassy peak fitting program supplied by S. Newman was used to primary melt inclusions, Cr-spinel inclusions, and trails of secondary melt inclusions obtain absorbance values at wavelengths of 1,430 and 1,515 cm−1. Owing to the thinness of the wafers, most spectra exhibited interference fringes, which in many cases 2− masked any possible signal of the CO3 doublet. For this Electron microprobe reason, only two inclusions yielded spectra in which the background was smooth enough to confidently analyze

Analyses of major elements in olivine and major and CO2. volatile elements (S and Cl) in the primary melt inclusions Absorbance was converted to concentration using the Beer– were carried out at McGill University using a Jeol JXA- Lambert law, a melt density of 2.8 g/cm3 and the following −1 −1 2− 8900L electron microprobe and the University of Oregon absorption coefficients: 394 L mol cm for CO3 using a Cameca SX-100, both with an atomic number, (calculated from the melt inclusion compositions using the absorption, and fluorescence correction procedure. Operat- method of Dixon and Pan 1995) and 63 L mol−1 cm−1 for OH −1 ing conditions for melt inclusion analysis at McGill and H2Oat3,535cm (Dixon et al. 1995). The values of University were as follows: 15 kV, 15 nA, 20 s peak these parameters are lower than those employed by Bureau et counting times for all elements except Fe and Mn (30 s), Ti, al. (1998b, 1999) but are more appropriate for the Piton de la Ni, and P (40 s), and Cl (50 s). The beam diameter was Fournaise basalts. Melt inclusion thickness was determined by defocused to 10 μm. A basaltic glass standard (BMAK mounting the wafers on a thin needle and immersing them in Smithsonian 113498) was used for major elements except oil of refractive index 1.657. The wafers were then viewed Na and K, for which an obsidian glass standard (KN9) was under the microscope at 100× magnification and tilted such used. Operating conditions for olivine analysis were as that the field of view was perpendicular to the width of the follows: 15 kV, 40 nA, 20 s peak counting times for all wafer. This allowed for a cross-sectional view of the wafer elements, and a beam diameter of 2 μm. An olivine and inclusion and measurement of the inclusion thickness standard was used for all elements. The average precision using the objective reticule. (2σ) is 6% or better for all elements except for MnO (30%),

P2O5 and S (20%), and Cl (35%). At the University of Oregon, 60 s counting times were used for all elements Results except Ti, Al, Mg, and P (40 s) and Ca, K (20 s). For melt inclusion analysis, the beam was set to 20 kV with a current Whole-rock chemistry of 30 nA and a diameter of 5 μm. Linear regression of count rate to time zero was performed on Na and K to Two water-quenched lava samples were collected by a team correct for the effects of alkali migration. The basaltic glass of scientists from the Observatoire Volcanologique du Piton standard VG-2 (USGS) was used for all elements. For de la Fournaise, one on November 17th (REU0211-171) olivine analysis, the beam conditions were as follows: and the other on November 25th (REU0211-254). Their 15 kV, 30 nA, and a focused beam. An olivine standard was chemical compositions are reported in Table 2. Sample used for all elements. The average precision (2σ)is5%or REU0211-254 is more magnesian than sample REU0211- better for all the elements analyzed except MnO for which 171, containing 8.7 wt.% MgO compared to 6.9 wt.%, and the precision is ≤25%. has slightly lower concentrations of most other oxides. The 1082 Bull Volcanol (2009) 71:1077–1089 data agree well with other published whole-rock composi- a 20 tions for the November 2002 eruption (Vlastélic et al. 2007) Fo80 and point to an increase in olivine content as the eruption 18 All whole rock Nov 2002 whole rock 82 16 progressed (increase in MgO content and dilution of most Nov 2002 olivine crystals 84 other oxides due to accumulation of olivine Fo84; Fig. 3a). 14

(wt.%) 86 The whole-rock data for the November 2002 eruption fall Olivine control line

total 12 along the compositional trend defined by other lavas Dec. 2 FeO 10 erupted from Piton de la Fournaise (Fig. 3; cf. Albarède et Nov. 25 Nov. 16-17 al. 1997; Bureau et al. 1999; Vlastélic et al. 2005, 2007). 8

6 Olivine and melt inclusion major element chemistry 0 10 20 30 40 50 MgO (wt.%) All olivine crystals have cores varying in composition from b Fo83 to Fo84, except in the early-stage sample (REU0211- 0.90 Clinopyroxene 161), which is from tephra released on the first day of the fractionation eruption; cores of olivine crystals from this sample have a Olivine control line composition of Fo . Most of the crystals have a compo- 3 81 O 0.80 sitionally distinct rim varying in thickness from 10–15 μm 2 and in composition from Fo76 to Fo82. CaO/Al Only primary melt inclusions in olivine were analyzed, 0.70 Early phase 2002 and their compositions are reported in Table 3. The most Late phase 2002 Past major eruptions conspicuous feature of the analyses is the highly variable 0.60 0102030 MgO (2.2–5 wt.%) and FeOtotal (5.5–10 wt.%, all iron expressed as FeO) content of the inclusions. If these MgO (wt.%) compositions were representative of the original magma, Fig. 3 a MgO vs. FeOtotal contents of whole-rock samples from Piton they would require that the host olivine have a composition de la Fournaise compiled from Albarède et al. (1997), Bureau et al. of Fo68–78, which is more Fe rich than the observed values (1999), and Vlastélic et al. (2005, 2007) and including the November 2002 eruption (data from Vlastélic et al. 2007 and this study; eruption (Fo81–84). This indicates that the melt inclusions are out of dates are indicated). The compositions of olivine from the 2002 equilibrium with their olivine hosts and implies that the eruption are also plotted (black squares) and their corresponding inclusions have undergone post-entrapment modification forsterite contents are indicated. The whole-rock data fall along an (Table 3; uncorrected compositions). olivine control line representing accumulation of olivine (dominantly After entrapment of a melt inclusion, olivine crystallizes Fo83–84) in the lavas. Lavas extruded after November 17, 2002 show evidence of olivine accumulation, the proportion of which increased along the walls of the inclusion as temperature decreases as the eruption progressed. b MgO vs. CaO/Al2O3 contents of whole- and there is diffusion of iron from the melt into the host rock samples and compositionally restored melt inclusions (see text olivine (Danyushevsky et al. 2000). To account for these for a description of the correction procedure) from the November 2002 changes, the compositions of the melt inclusions were eruption and past major eruptions (Bureau et al. 1998b). Both melt inclusions and whole-rock data show evidence of clinopyroxene fractionation. The lack of clinopyroxene phenocrysts in the November Table 2 Whole-rock compositions in weight percent 2002 lavas (only minor groundmass/microphenocrystic clinopyroxene is present) does not preclude clinopyroxene fractionation from the Sample REU0211-171 REU0211-254 liquid at depth. Albarède et al. (1997) argued that fractionation of Eruption date Nov. 17, 2002 Nov. 25, 2002 clinopyroxene in the deeper parts of the edifice accompanied early differentiation of the Piton de la Fournaise lavas, and this process is SiO2 48.70 48.32 likely responsible for the variability in the CaO/Al2O3 ratios of melt inclusions (Bureau et al. 1999) TiO2 2.69 2.56

Al2O3 14.32 13.60 FeOa 11.09 11.32 corrected following the methods described in Danyushevsky MnO 0.17 0.17 et al. (2000). The extent of this post-entrapment crystalli- MgO 6.85 8.72 zation and iron loss for the melt inclusions of the CaO 11.31 10.77 November 2002 eruption is readily seen by comparing their compositions with those of the lavas (Fig. 4). If a melt Na2O 2.72 2.54 inclusion is in equilibrium with its host olivine, its K O 0.77 0.72 2 composition should plot at the intersection of the liquid P O 0.33 0.32 2 5 line of descent for olivine fractional crystallization (curves a Total iron reported as FeO labeled L.l.d.) and the line representing liquid compositions ulVlao 20)71:1077 (2009) Volcanol Bull Table 3 Analyzed melt inclusion compositions in weight percent

Sample 161-01 161-02 161-03 161-04 161-05 161-06 161-07 252-01 252-04 252-07 252-08 014-01 034-04 034-11 034-12 021-03 021-05 021-12 021-13

SiO2 50.28 50.43 50.34 50.03 50.02 49.63 50.15 48.72 50.08 50.47 48.96 49.69 51.47 51.28 50.38 51.06 50.45 53.08 52.24 TiO2 3.10 3.08 3.07 3.15 3.09 3.19 3.08 2.99 3.18 3.08 3.68 3.05 2.93 2.96 2.94 3.46 3.29 2.76 3.34 Al2O3 16.19 16.26 16.23 16.18 16.14 16.25 16.33 15.42 16.17 16.24 15.84 15.40 16.50 16.34 16.66 17.49 16.89 17.67 18.15 FeOa 9.99 9.73 10.07 9.81 9.96 9.80 9.75 10.09 8.39 9.05 9.18 9.44 8.15 8.07 8.39 5.52 7.56 5.90 5.50 MnO 0.16 0.18 0.17 0.16 0.17 0.17 0.15 0.19 0.11 0.13 0.14 0.12 0.12 0.14 0.11 0.08 0.13 0.08 0.07 MgO 3.44 3.24 3.18 3.60 3.50 3.20 3.08 4.04 3.36 3.45 3.27 4.99 2.56 2.93 3.00 2.59 3.01 2.29 2.23 CaO 13.31 13.17 13.42 13.16 13.14 13.03 13.15 13.21 12.63 12.34 12.83 11.27 14.13 13.69 13.88 14.82 13.90 13.00 12.56 – Na2O 2.71 2.78 2.77 2.73 2.78 2.74 2.75 2.61 2.91 3.17 3.02 3.06 2.87 2.84 2.91 3.05 2.69 2.93 2.95 1089 K2O 0.78 0.77 0.75 0.74 0.76 0.75 0.78 0.60 0.95 0.81 0.94 1.18 0.67 0.90 0.70 1.06 0.90 1.66 1.79 P2O5 0.35 0.39 0.32 0.38 0.32 0.34 0.34 0.71 0.53 0.46 0.83 0.56 0.29 0.44 0.36 0.49 0.39 0.08 0.08 S 0.087 0.091 0.083 0.085 0.090 0.090 0.091 0.094 0.120 0.095 0.064 0.128 0.079 0.161 0.103 0.061 0.115 0.040 0.040 Cl 0.028 0.032 0.029 0.021 0.025 0.029 0.032 0.024 0.025 0.026 0.024 0.042 0.028 0.037 0.030 0.033 0.027 0.014 – Host Fo 81 81 81 81 81 81 81 84 84 84 83 83 84 84 84 84 84 84 84 % p.e.cb 13 14 14 12 12 14 14 15 19 17 18 11 23 20 20 22 21 24 24 Tquench °C 1,095 1,092 1,089 1,099 1,097 1,092 1,088 1,109 1,093 1,093 1,092 1,133 1,073 1,082 1,084 1,072 1,084 1,070 1,072 Ttrapping °C 1,181 1,182 1,182 1,181 1,182 1,184 1,184 1,213 1,220 1,211 1,214 1,207 1,222 1,215 1,214 1,213 1,218 1,220 1,223

Corrected compositions in wt.%c SiO2 48.97 49.01 48.91 48.80 48.76 48.29 48.70 47.47 48.32 48.76 47.43 48.70 49.07 49.17 48.47 48.83 48.48 50.30 49.59 TiO2 2.75 2.71 2.70 2.81 2.75 2.81 2.70 2.60 2.68 2.63 3.12 2.76 2.38 2.46 2.46 2.84 2.73 2.23 2.69 Al2O3 14.38 14.33 14.27 14.44 14.36 14.29 14.30 13.42 13.61 13.84 13.44 13.94 13.39 13.58 13.92 14.34 14.02 14.28 14.63 Fe2O3 1.42 1.39 1.38 1.43 1.42 1.38 1.37 1.38 1.30 1.32 1.30 1.48 1.20 1.25 1.27 1.21 1.26 1.17 1.16 FeO 10.29 10.34 10.35 10.26 10.28 10.35 10.38 10.15 10.35 10.28 10.31 9.96 10.60 10.44 10.41 10.55 10.44 10.67 10.69 MnO 0.14 0.15 0.15 0.15 0.15 0.15 0.13 0.16 0.10 0.11 0.12 0.11 0.10 0.11 0.09 0.06 0.11 0.07 0.05 MgO 7.47 7.51 7.53 7.46 7.47 7.53 7.54 8.90 9.32 8.97 8.95 8.53 9.63 9.25 9.17 9.33 9.41 9.42 9.43 CaO 11.82 11.60 11.79 11.75 11.69 11.46 11.51 11.50 10.64 10.52 10.88 10.20 11.47 11.38 11.59 12.15 11.54 10.51 10.13 Na2O 2.40 2.45 2.43 2.44 2.47 2.41 2.41 2.27 2.45 2.70 2.57 2.77 2.33 2.36 2.43 2.50 2.24 2.37 2.38 K2O 0.69 0.68 0.65 0.66 0.68 0.66 0.68 0.52 0.80 0.69 0.80 1.07 0.54 0.75 0.58 0.87 0.75 1.34 1.45 P2O5 0.31 0.34 0.28 0.33 0.29 0.30 0.30 0.62 0.44 0.36 0.70 0.51 0.23 0.36 0.30 0.40 0.32 0.07 0.07 S 0.077 0.080 0.073 0.075 0.080 0.079 0.080 0.081 0.101 0.081 0.054 0.116 0.064 0.134 0.086 0.050 0.095 0.033 0.032 Cl 0.025 0.028 0.025 0.019 0.022 0.026 0.028 0.021 0.021 0.023 0.020 0.038 0.023 0.031 0.025 0.027 0.022 0.011 – a Total iron as FeO b Percentage of post-entrapment crystallization c Corrected for both iron loss by diffusion and post-entrapment crystallization 1083 1084 Bull Volcanol (2009) 71:1077–1089

2+ 2+ in equilibrium with the host olivine (lines labeled Foxx). (Mg/Fe )melt/(Mg/Fe )olivine,wasassumedtobeequalto Any deviation from this point of intersection defines a 0.306 based on experimental studies of Reunion magmas vector, the magnitude and direction of which depend on the and other basalts (Roeder and Emslie 1970;Fisketal. extent and relative proportions of post-entrapment crystal- 1988). We used a melt FeO/FeOtotal ratioof0.86basedon lization (vector parallel to the liquid line of descent) and olivine–spinel-liquid equilibrium (Maurel and Maurel diffusive iron loss (vector parallel to the ordinate). From 1982) and Cr-spinel Fe2+/Fe3+ ratios reported by Bureau Fig. 4, it is evident that there were substantial post- et al. (1998b) for past summit eruptions at Piton de la entrapment changes in the compositions of the November Fournaise (1.81±0.03 for Dolomieu crater samples). The

2002 melt inclusions, particularly in the concentration of latter data imply fO2 conditions ~1 log unit above FMQ at iron for some inclusions. As mentioned above, these melt 1,200°C. However, this estimate, which is on the high end inclusions have analyzed compositions that are in equilib- of estimates for other oceanic hot spot magmas (e.g., rium with olivine of composition Fo68–78 (average Fo71) but Óskarsson 1994; Rhodes and Vollinger 2005)shouldbe are hosted in olivine of composition Fo81 or Fo83–84.By treated with some caution as Cr-spinel re-equilibrates comparison, the compositions of melt inclusions from rapidly with the melt and an incomplete analysis of minor major eruptions appear to have undergone relatively minor elements like Ti and V can significantly affect the change. They are in equilibrium with olivine of composi- calculated Fe2+/Fe3+ value (Clynne and Borg 1997). An tion Fo80–83 (average Fo81) but are hosted in olivine of error in fO2 of ±1 log unit could change the corrected composition Fo83–85. MgO content of a melt inclusion by up to 10% (relative). To correct for the post-entrapment changes, we added We modeled the melt inclusions as closed systems (Fe3+ 2+ Fe back into the inclusion until its composition reached behaves incompatibly), and consequently, the FeO/FeOtotal the liquid line of descent for olivine fractionation and then ratio of the melt increased slightly during incremental incrementally added olivine to the melt until the Mg/Fe2+ addition of equilibrium olivine. ratio of the latter reflected equilibrium with its host For all of the November 2002 melt inclusions, the 2+ olivine. For this purpose, the partition coefficient, KD= proportion of Fe lost to diffusion and olivine crystallization after melt entrapment is estimated to have been 14% to 56% and 11% to 24% respectively. The compositions of

melt inclusions hosted in olivine (Fo83–85) from past major Early phase 2002 eruptions (Bureau et al. 1998b) were corrected in the same 14 Late phase 2002 Corrected compositions manner; the iron loss and olivine crystallization were Past major eruptions 12 estimated to be 0% to 16% and 3% to 8%, respectively. L.l.d. Corrected major element compositions for the November 10 L.l.d. 2002 primary melt inclusions are presented in Table 3 and (wt.%) compared to those of melt inclusions from major eruptions

total 8 Fo85 (hosted by Fo83–85 olivine) in Fig. 4. FeO 6 L.l.d . (P.E.C.) – Fo83 The melt inclusions hosted in Fo83 84 olivine from the Fo71 Fo81 2002 eruption have corrected MgO contents from 8.5 to 4 Iron loss 9.6 wt.%, SiO contents from 47 to 50 wt.%, and total Foxx (host) 2 2 alkali (Na2O+K2O) contents from 3 to 4 wt.%. These 0 246810 12 compositions fall within the range of corrected composi- MgO (wt.%) tions of melt inclusions hosted in Fo83–84 olivine from

Fig. 4 Uncorrected FeOtotal vs. MgO contents of melt inclusions from major eruptions (Figs. 3b and 4; Bureau et al. 1998b). The the November 2002 eruption and from past major eruptions (Bureau et melt inclusions hosted in Fo81 olivine have a restricted al. 1998b). Also shown are the corresponding compositions after compositional range with a MgO content of ~7.5 wt.%, correction for post-entrapment crystallization and Fe2+ loss. The shaded field represents the whole-rock compositions shown in Fig. 3a. SiO2 contents from 48 to 49 wt.%, and total alkali contents A liquid in equilibrium with a given olivine composition will lie along of 3 wt.% and are more evolved than the average melt the line labeled with the olivine composition (Foxx) assuming an FeO/ inclusion hosted in Fo83–84 olivine (Figs. 3b and 4). FeOtotal ratio of 0.86 (Bureau et al. 1998b). These lines were calculated using a KD of 0.306 (Roeder and Emslie 1970; Fisk et al. 1988). Liquid lines of descent (L.l.d.) for olivine fractionation are shown starting with Volatile content liquids in equilibrium with olivine Fo84 and Fo81. A melt inclusion in equilibrium with its host olivine (Foxx) and with the appropriate liquid Owing to differences in the H2O and CO2 absorption line of descent (implying no Fe2+ loss) will plot at the intersection of 2+ coefficients used in this study and those used by Bureau et these two lines. The effects of post-entrapment crystallization and Fe loss by diffusion are illustrated in the inset. P.E.C.=post-entrapment al. (1998b, 1999), the H2O and CO2 contents of the melt crystallization along the inclusion–host interface inclusions in previous studies were underestimated; we Bull Volcanol (2009) 71:1077–1089 1085

Table 4 H2O and CO2 contents

a Corrected for post-entrapment crystallization. Melt inclusions with no associated major element analysis were corrected using the average amount of post-entrapment crystallization for November 2002 melt inclusions. Most CO2 values in the November 2002 melt inclusions are below detection limit (–) due to analytical difficulties discussed in the text b The difference between the absorption coefficients used by Jendrzejewski et al. (1996a, b) and Bureau et al. (1998b, 1999) and the coefficients from Dixon and Pan (1995) results in an error (underestimate) of 19% for H2O and 1% for CO2 have recalculated the published values using the more One of the inclusions (sample REU0211-252-3) contains a commonly accepted coefficients (Table 4). FTIR analysis of visible sulfide globule with appreciable proportions of Cu primary melt inclusions from samples spanning the Novem- and Ni (10 and 8 wt.%, respectively; data from electron ber 2002 eruption yielded H2O contents in the range 0.1 to microprobe energy dispersive spectrometry). However, no 1.8 wt.%; most values were between 0.2 and 0.8 wt.% sulfide globules are present in the host olivine crystal.

(Table 4). Only two melt inclusions yielded reliable CO2 Previous studies of Piton de la Fournaise have reported the values due to the analytical difficulties discussed above. presence of sulfide globules in both melt inclusions and the These values are 165 and 1,010 ppm and are for inclusions host olivine. The sulfur content of the melt inclusions from samples REU0212-014 and REU0211-252, respectively investigated in these studies averages 1,140±260 ppm

(sample REU0211-252 also has the highest H2Ocontent). (Bureau et al. 1998a, b). The chlorine contents of our The corresponding vapor saturation pressures are 425 and melt inclusions vary from below detection to 390 ppm and 2,630 bars, respectively (calculated with VolatileCalc; New- are comparable to those of melt inclusions from major man and Lowenstern 2002). Vapor saturation is inferred from eruptions, which are in the range 270±50 ppm (Bureau et the presence of fluid inclusions in these samples, as well as in al. 1998a, b). olivine (Fo83–87) from earlier eruptions studied by Bureau et al. (1998b, 1999). It is also consistent with the wide range of

H2OandCO2 contents reported by them for melt inclusions Discussion in these samples. Our calculated saturation pressures are considered minimum values because we did not analyze the Trapping and eruption temperatures of melt inclusions volatile content of the shrinkage bubbles. Contents of both

H2OandCO2 overlap with those obtained by Bureau et al. The initial composition of a primary melt inclusion hosted (1998b) for melt inclusions in Fo83–85 olivine from major by olivine may be assumed to be the same as that of the eruptions. corresponding magma and reflect equilibrium with the The concentrations of sulfur in melt inclusions analyzed olivine at the time of entrapment. When the host olivine in our study span a wide range, from 330 to 1,380 ppm. begins to cool, the melt inclusion will evolve by crystalliz- 1086 Bull Volcanol (2009) 71:1077–1089 ing olivine along the rim of the inclusion, diffusing Fe2+ major eruptions using the raw data of Bureau et al. (1998b), into the host and growing a shrinkage bubble due to corrected using the same procedure employed for the 2002 thermal contraction of the melt and the volume decrease inclusions and a pressure of 1,200 bars. The calculated caused by crystallization. This evolution will terminate trapping temperatures vary with the olivine composition once the inclusion cools fast enough (i.e., upon eruption) to and range from 1,199°C to 1,227°C. inhibit further diffusion of elements, not only from the Quenching temperatures were calculated by applying the inclusion but also within it (thereby preventing further Sugawara geothermometer to the uncorrected melt inclu- crystallization of olivine). The temperature at which this sion compositions using the same pressure and water occurs is referred to as the quenching temperature. If the content estimates as above. This assumes no water loss inclusion were to cool slowly during an eruption, such as from the melt inclusions due to pre-eruptive diffusion and might be the case in the interior of a lava flow, the no drop in pressure within the inclusions. The resulting quenching temperature would be significantly lower than temperatures are 1,088°C to 1,099°C for the early-stage the eruption temperature. However, if it cooled quickly, for melt inclusions and from 1,070°C to 1,133°C for the late- example because of its location in a small pyroclast, the two stage inclusions (Fig. 5). For the past major events, the temperatures could be quite similar. quenching temperatures are in a narrow range from 1,163°C To estimate temperatures of melt inclusion trapping and to 1,181°C (Fig. 5). quenching, we used the empirical MgO-in-liquid geother- Although the quenching temperatures for the November mometer of Sugawara (2000; Eq. 6a), which has been 2002 melt inclusions are all equivalent within the error of calibrated for a range of pressures, temperatures, and the geothermometer, there is a significant amount of scatter compositions appropriate for the Reunion magmas. This in the temperatures recorded by the late-stage melt geothermometer incorporates the effects of pressure and inclusions. If the melt inclusions cooled from the same

H2O content on the equilibrium temperature of olivine- magma at a similar rate, they should all record the same saturated liquids. For Hawaiian lavas, the model yields quenching temperature, and given rapid cooling (explosive- results similar to those obtained with the experimentally ly erupted), this temperature should be similar to the derived geothermometers of Montierth et al. (1995) and eruption temperature. In order to evaluate this cooling Helz and Thornber (1987); these lavas include composi- history, we plotted the amount of iron loss by diffusion tions (low MgO) similar to those analyzed in the November experienced by each inclusion (FeO in weight percent) 2002 inclusions. versus its calculated quenching temperature (Fig. 5). The The Sugawara geothermometer was applied to the amount of iron loss is a function of the time-averaged corrected melt inclusion compositions at a pressure of cooling rate and the size of the inclusion (Danyushevsky et 1,200 bars (average trapping pressure calculated from the al. 2000, 2002). Based on a comparison of the data for

H2O and CO2 contents of the 2002 melt inclusions). If the Reunion melt inclusions to cooling rate-dependent curves water content of a melt inclusion had not been determined, determined for an average inclusion radius of 25 μm, we the average analyzed water content of the 2002 melt conclude that the inclusions from past major eruptions and inclusions was assumed. This value (0.55 wt.%) is similar those from the early phase of the 2002 eruption, all from to the average value obtained for melt inclusions previously explosively erupted tephra samples, cooled at time- analyzed in Fo83–85 olivine by Bureau et al. (1998b; averaged rates of ≥100°C/day. Some of the late-stage 0.86 wt.%, recalculated in Table 4). The uncertainty in the inclusions from the 2002 eruption, which were hosted by calculated pressure and water content of the melt inclusions lava, also experienced similarly fast cooling. However, is less than the standard error of the geothermometer, which other late-stage inclusions cooled more slowly, i.e., at a is ±40ºC. The resulting calculated trapping temperatures are time-averaged rate of <100°C/day. Much of this slow given in Table 3. Melt inclusions from the early stage of the cooling likely occurred in the lava flow, but we cannot eruption (Fo81 olivine hosts) yield an average trapping rule out the possibility that some of it may have occurred temperature of 1,182±1°C, whereas the trapping temper- during prolonged storage in the plumbing system. atures of those from the late-stage (Fo83–84 olivine hosts) For the purpose of this study, we based our estimate of range from 1,207°C to 1,223°C. These temperatures are the eruption temperature for the November 2002 event on equivalent within the error of the geothermometer. the quenching temperatures of those inclusions that cooled However, given the systematic differences in calculated relatively quickly (≥100°C/day). Temperatures for the temperatures, olivine host compositions, and melt MgO early-stage melt inclusions (sample REU0211-161) and contents between the early and late-stage melt inclusions, those from sample REU0211-252, which was erupted on we conclude that the difference in trapping temperatures of the 25th of November, cluster around a value of 1,100°C. A melt inclusions sampled in the two stages is real. We also single inclusion from sample REU0212-014 also falls along calculated trapping temperatures for melt inclusions from the rapid cooling curve of 100°C/day and appears to record Bull Volcanol (2009) 71:1077–1089 1087 a higher eruption temperature than the other inclusions Comparison between small-scale and major eruptions (1,133°C). However, in the absence of supporting data, it is uncertain whether or not this lava, which was extruded on The depth of crystallization and accumulation of olivine

November 29th (the onset of renewed explosive activity (Fo83–84) sampled by major eruptions is estimated to have during the 2002 eruption), was indeed hotter. Therefore, ranged from 5 km b.s.l. to <1 km below the summit of the assuming that the magma erupted in the November 2002 volcano (Bureau et al. 1998b). The November 2002 melt event was at a temperature of 1,100°C, it follows that this inclusions were trapped at depths varying from 1.5 km a.s.l. magma was cooler (possibly by up to 70°C) than magma to 6 km b.s.l. (values obtained from partial pressures of erupted during past major eruptions, an interpretation that is H2O and CO2 recorded by melt inclusions). As has been the consistent with the presence of more evolved olivine case for major eruptions, the November 2002 event also phenocrysts in the products of the early phase of the sampled Fo83–84 olivine from a variety of levels within and November 2002 eruption. below the volcanic edifice. Although the November 2002 eruption may have Nov. 2002 Major eruptions sampled a similar population of olivine crystals to those 7 1 ˚C/day (150 days) sampled by major eruptions, there is seismic evidence for 6 brittle crustal deformation and magma movement possibly 81 84 having been initiated at shallower depths. The November 5 10 (15) 2002 eruption, as well as the three small eruptions 4 preceding it (March 2001, June 2001, and January 2002), Trapping Fo Trapping Fo had seismic event hypocenters (volcano-tectonic earth- 3 100 (1) quakes) located at depths varying between 1.5 and 0.5 km

FeO loss (wt.%) 1000 (0.17) 2 10000 (0.02) b.s.l. prior to eruption, suggesting a buildup of magma pressure at those depths (Staudacher and Cheminée 2001; 1 100 (1) Staudacher and OVPDLF 2002; OVPDLF unpublished 0 report 2002). By comparison, the last major eruption of 1000 1050 1100 1150 1200 1250 1998 was accompanied by pre-eruptive volcano-tectonic ° Quenching Temperature ( C) earthquakes originating from 6 km b.s.l., suggesting magma Fig. 5 Calculated quenching temperature vs. FeO loss by diffusion replenishment from deep levels (Staudacher et al. 1998; (calculated initial FeO minus analyzed FeO assuming FeO/FeOtotal= Battaglia et al. 2005). 0.86) for each inclusion of the November 2002 eruption and the The late-stage olivine sampled by the November 2002 inclusions from past major eruptions (Bureau et al. 1998b). Symbols as eruption appears to be of cumulate origin based on the in Fig. 3. The solid time-averaged cooling curves are from Danyushevsky et al. (2002) for a melt inclusion radius of 25 μm presence of numerous healed fractures filled with secondary and a cooling interval of 150°C, starting at 1,220°C, which is the fluid and melt inclusions, kink banding, and disequilibrium trapping temperature of the melt inclusions hosted by Fo – olivine 83 84 between olivine (Fo83–84) and whole-rock compositions crystals. The number next to each curve represents the time-averaged (calculated to be in equilibrium with olivine of composition cooling rate (°C/day) and the number in brackets is the time in days that it would take the inclusions to cool to 1,070°C. The dashed curve Fo80–81). Olivine of similar composition expelled during represents a time-averaged cooling rate of 100°C/day for an inclusion past major eruptions has also been interpreted to be of radius of 25 μm, a starting temperature of 1,180°C and a cooling cumulate origin (Bureau et al. 1998a, b, 1999). It thus interval of 90°C, appropriate for melt inclusions hosted by olivine of appears that olivine crystals not expelled during a major composition Fo81 from the early stage of the 2002 eruption. Trapping temperatures for melt inclusions hosted in olivine of compositions eruption remain in the shallow plumbing system as Fo81 and Fo83–84 are indicated by the vertical lines (the lower end of cumulate piles until they are expelled by a later magma, the range is shown for Fo83–84 olivine-hosted melt inclusions). High and consequently, each eruption carries with it phenocrysts time-averaged cooling rates and limited pre-eruptive cooling charac- and antecrysts of varying proportions (cf. Bureau et al. terize inclusions from past major eruptions which quenched rapidly upon eruption (tephra samples). The gray shaded field encompassing 1998a, b, 1999; Vlastélic et al. 2007). This process has also these inclusions represents the range of eruption temperatures for past been documented at Etna, Sicily, by Armienti et al. (1994), major events. A portion of the November 2002 inclusions, including who showed that each erupted batch of magma contained all of the early phase inclusions, appears to have quenched at similarly new phenocrysts as well as crystals left behind during the high rates, suggesting limited post-eruptive modification. The gray shaded field encompassing these inclusions provides the best preceding eruption. Vlastélic et al. (2007) used the Pb representation of the range of eruption temperatures for the November isotopic signature of the recent lavas emitted from Piton de 2002 event. The remaining inclusions appear to have experienced a la Fournaise to suggest that the November 2002 eruption, lower time-averaged cooling rate. These inclusions were probably along with other small- to modest-scale eruptions since affected by slow cooling in the lava flow. There is also one late-stage melt inclusion that records an anomalously high eruption temperature 1998, were the result of extended magma storage at depth, (see text for discussion) whereas the major eruption of 1998 was the result of deep 1088 Bull Volcanol (2009) 71:1077–1089

magma injection into the shallow plumbing system. The composition of the host olivine (Fo83–84), although the results of our study suggest that magma ejected during eruption temperature calculated for major eruptions is these shallow modest eruptions is also of lower temperature ~1,170°C, i.e., 70°C higher than that of the November compared to magma ejected during major eruptions. 2002 eruption. The eruption of cumulate olivine with a

Cumulate olivine is entrained in this lower temperature, composition similar to most of the 1998 olivine (Fo83–84) more evolved magma and carried to the surface together and the eruption of olivine with a composition in with other olivine phenocrysts. Interestingly, the 1998 equilibrium with the whole rock of the 1998 eruption at magma, which erupted from the Kapor vent and contains the Kapor vent (Fo81) are consistent with the model for minor amounts of Fo84 olivine (Bureau et al. 1999), is magma evolution at Piton de la Fournaise proposed by chemically similar to melt trapped as inclusions by Fo81 Bureau et al. (1998a, b, 1999) and Vlastélic et al. (2007). olivine crystals from the early stage of the November 2002 Evidence from this and previous studies indicates a system eruption. This supports the idea that the November 2002 in which primitive magma is supplied from depth to the magma comprised a mix of residual 1998 melt, olivine shallow parts of the system and evolves through interac- phenocrysts possibly crystallized from the initial batch of tion with olivine cumulate piles and further crystallization 1998 magma and olivine antecrysts incorporated from a of olivine. Small-scale eruptions, such as the November zone of cumulate mush. 2002 event, expel magma which cools and evolves following the last major replenishment and entrains variable amounts of cumulate olivine depending on the Summary eruptive flux.

The November 2002 eruption was typical of the summit Acknowledgments We would like to thank Eric Delcher (Université eruptions that characterized volcanic activity at Piton de la de la Réunion), Jean-Louis Cheminée (Observatoire Volcanologique du Piton de la Fournaise), and Marc-Antoine Longpré (McGill Fournaise between 1999 and 2006. The composition of the University) for their help in the field and in sample collection. Patrick olivine crystals varied from Fo81 in the early phase of the Bachèlery (Université de la Réunion) provided important guidance and support. We are very grateful to Jim Clark (McGill University) eruption to Fo83–84 thereafter. Major element analysis has revealed that melt inclusions from the November 2002 and John Donovan (University of Oregon) for performing the microprobe analyses. Discussions with Don Baker and David Dolejs eruption are in chemical disequilibrium with their host (McGill University) helped clarify issues of olivine crystallization. olivine. Some melt inclusions from lava flows experienced Comments by Hélene Bureau and formal reviews by Keith Putirka, substantially greater degrees of post-entrapment modifica- Ivan Vlastélic, and Peter Michael greatly improved the final tion than melt inclusions from tephra samples that cooled manuscript. The project was supported by a Natural Sciences and Engineering very quickly. This is consistent with the observation that Research Council of Canada Undergraduate Student Research Award these melt inclusions record the lowest quenching temper- to NV and Discovery grant to AEW-J as well as travel funds provided atures. By comparing the amounts of diffusive iron loss by the Office Franco-Québécois de la Jeunesse awarded to NV. experienced by melt inclusions to those predicted for different time-averaged cooling rates, we classified our melt inclusions into slower and faster cooled populations References (i.e., with time-averaged cooling rates of <100°C/day and >100°C/day). Based on the compositions of the latter Albarède F, Luais B, Fitton G, Semet MP, Kaminski E, Upton BGJ, population of inclusions and the MgO-in-liquid geother- Bachèlery P, Cheminée JL (1997) The geochemical regimes of Piton de la Fournaise Volcano (Reunion) during the last mometer of Sugawara (2000), we estimate that the magma 530,000 years. J Petrol 38:171–201 erupted during the November 2002 event was at a Armienti P, Pareschi MT, Innocenti F, Pompilio M (1994) Effects of temperature of ~1,100°C. magma storage and ascent on the kinetics of crystal growth; the – Trapping temperatures of melt inclusions were calculated case of the 1991 93 Mt. Etna eruption. Contrib Mineral Petrol 115:402–414. doi:10.1007/BF00320974 by applying the Sugawara geothermometer to compositions Battaglia J, Ferrazzini V, Staudacher T, Aki K, Cheminée J-L (2005) that were corrected for post-entrapment crystallization of Pre-eruptive migration of earthquakes at Piton de la Fournaise olivine and diffusive iron loss. Inclusions from the early volcano (Réunion Island). Geophys J Int 161:549–558. stage of the November 2002 eruption, which are hosted by doi:10.1016/j.jvolgeores.2005.04.005 Bureau H, Métrich N, Pineau F, Semet MP (1998a) Magma–conduit Fo81 olivine, were trapped at an average temperature of interaction at Piton de la Fournaise volcano (Reunion Island): a 1,182°C, whereas those from the later stage of the eruption, melt and fluid inclusion study. J Volcanol Geotherm Res 84:39– 60. doi:10.1016/S0377-0273(98)00029-8 which occur in Fo83–84 olivine, were trapped at temperatures of 1,207°C to 1,223°C. Significantly, the latter Bureau H, Pineau F, Métrich N, Semet MP, Javoy M (1998b) A melt and fluid inclusion study of the gas phase at Piton de la Fournaise temperature range is very similar to that for melt inclusions volcano (Reunion Island). Chem Geol 147:115–130. from major eruptions (1,199°C to 1,227°C), as is the doi:10.1016/S0009-2541(97)00176-9 Bull Volcanol (2009) 71:1077–1089 1089

Bureau H, Métrich N, Semet MP, Staudacher T (1999) Fluid–magma Nercessian A, Hirn A, Lepine JC, Sapin M (1996) Internal structure of decoupling in a hot-spot volcano. Geophys Res Lett 26 Piton de la Fournaise volcano from seismic wave propagation (23):3501–3504 and earthquake distribution. J Volcanol Geotherm Res 70:123– Clynne MA, Borg LE (1997) Olivine and chromian spinel in primitive 143. doi:10.1016/0377-0273(95)00042-9 calc-alkaline and tholeiitic lavas from the southernmost Cascade Newman S, Lowenstern JB (2002) VOLATILECALC: a silicate melt- Range, California: a reflection of relative fertility of the source. H2O-CO2 solution model written in Visual Basic for excel. Can Mineral 35:453–472 Comput Geosci 28:597–604. doi:10.1016/S0098-3004(01) Courtillot V, Besse J, Vandamme D, Montigny R, Jaeger JJ, Capetta H 00081-4 (1986) Deccan flood basalts at the Cretaceous/Tertiary boundary? Óskarsson N (1994) Oxidation state of iron in mantle-derived magmas Earth Planet Sci Lett 80:361–374 of the Icelandic rift zone. Hyper Interact 91:733–737 Danyushevsky LV, Della-Pasqua FN, Sokolov S (2000) Re- Peltier A, Ferrazzini V, Staudacher T, Bachèlery P (2005) Imaging the equilibration of melt inclusions trapped by magnesian olivine dynamics of dyke propagation prior to the 2000–2003 flank phenocrysts from subduction-related magmas: petrological impli- eruptions at Piton de la Fournaise, Reunion Island. Geophys Res cations. Contrib Mineral Petrol 138:68–83. doi:10.1007/ Lett 32:L22302. doi:10.1029/2005GL023720 PL00007664 Peltier A, Staudacher T, Bachèlery P (2007) Constraints on magma Danyushevsky LV, Sokolov S, Falloon TJ (2002) Melt inclusions in transfers and structures involved in the 2003 activity at Piton de olivine phenocrysts: using diffusive re-equilibration to determine la Fournaise from displacement data. J Geophys Res 112: the cooling history of a crystal, with implications for the origin of B03207. doi:10.1029/2006JB004379 olivine-phyric volcanic rocks. J Petrol 43:1651–1671 Peltier A, Bachèlery P, Staudacher T (2009) Magma transport and Dixon JE, Pan V (1995) Determination of molar absorptivity of storage at Piton de la Fournaise (La Réunion) between 1972 and dissolved carbonate in basanitic glass. Am Mineral 80:1339– 2007: a review of geophysical and geochemical data. J Volcanol 1342 Geotherm Res. doi:10.1016/j.jvolgeores.2008.12.008 Dixon JE, Stolper EM, Holloway JR (1995) An experimental study of Rhodes JM, Vollinger MJ (2005) Ferric/ferrous ratios in 1984 Mauna the H2O and carbon dioxide solubilities in mid-ocean ridge Loa lavas: a contribution to understanding the oxidation state of basaltic liquids. Part I: calibration and solubility results. J Petrol Hawaiian magmas. Contrib Mineral Petrol 149:666–674. 36:1607–1631 doi:10.1007/s00410-005-0662-y Faure F, Trolliard G, Nicollet C, Montel JM (2003) A developmental Roeder PL, Emslie RF (1970) Olivine-liquid equilibrium. Contrib model of olivine morphology as a function of the cooling rate Mineral Petrol 29:275–289. doi:10.1007/BF00371276 and the degree of undercooling. Contrib Mineral Petrol 145 Sapin M, Hirn A, Lepine JC, Nercessian A (1996) Stress, failure and (2):251–263. doi:10.1007/s00410-003-0449-y fluid flow deduced from earthquakes accompanying eruptions at Fisk MR, Upton BGJ, Ford CE (1988) Geochemical and experimental Piton de la Fournaise volcano. J Volcanol Geotherm Res 70:145– study of the genesis of magmas of Réunion Island, Indian Ocean. 147. doi:10.1016/0377-0273(95)00043-7 J Geophys Res 93:4933–4950 Sobolev AV (2007) Melt inclusions and host olivines: what do they Gallart J, Driad L, Charvis P, Sapin M, Hirn A, Diaz J, de Vogt B, tell about mantle processes and sources? Geoch Cosmo Acta 71 Sachpazi M (1999) Perturbation to the lithosphere along the (15):A951 Suppl hotspot track of La Reunion, from an offshore–onshore seismic Staudacher T, Cheminée JL (2001) Eruptions during late March 2001 transect. J Geophys Res 104:2895–2908 and on 11 June 2001. Bull Glob Volcanism Netw 26(5) Helz RT, Thornber CR (1987) Geothermometry of Kilauea Iki lava Staudacher T, OVPDLF (2002) Fissure eruption 16 November–3 lake. Bull Volcanol 49:651–658. doi:10.1007/BF01080357 December sent lava to the sea. Bull Glob Volcanism Netw Jendrzejewski N, Javoy M, Trull T (1996a) Mesures quantitatives de 27(11) carbone et d’eau dans les verres basaltiques naturels par Staudacher T, Bachèlery P, Semet MP, Cheminée JL (1998) spectroscopie infrarouge, I. Le carbone. CR Acad Sci Paris Geophysical portrayal of the March fissure eruptions. Bull Glob 322:645–652 Volcanism Netw 23(3) Jendrzejewski N, Javoy M, Trull T (1996b) Mesures quantitatives de Sugawara T (2000) Empirical relationships between temperature, carbone et d’eau dans les verres basaltiques naturels par pressure, and MgO content in olivine and pyroxene saturated spectroscopie infrarouge, II. L’eau. CR Acad Sci Paris liquid. J Geophys Res 105:8457–8472 322:735–742 Upton BGJ, Wadsworth WJ (1966) The basalts of Reunion Island, Longpré MA, Staudacher T, Stix J (2007) The November 2002 Indian Ocean. Bull Volcanol 29:7–22. doi:10.1007/ eruption at Piton de la Fournaise volcano, La Reunion Island: BF02597136 ground deformation, seismicity, and pit crater collapse. Bull Vlastélic I, Staudacher T, Semet M (2005) Rapid change of lava Volcanol 69(5):511–525. doi:10.1007/s00445-006-0087-0 composition from 1998 to 2002 at Piton de la Fournaise Maurel C, Maurel P (1982) Etude expérimentale de l’équilibre Fe2+- (Réunion) inferred from Pb isotopes and trace elements: evidence Fe3+ dans les spinelles chromifères et les liquids silicates for variable crustal contamination. J Petrol 46:79–107. basiques coexistants à 1 atm. CR Acad Sci Paris 295:209–212 doi:10.1093/petrology/egh062 Montierth C, Johnston AD, Cashman KV (1995) An empirical glass- Vlastélic I, Peltier A, Staudacher T (2007) Short term (1998–2006) composition-based geothermometer for Mauna Loa lavas. In: fluctuations of Pb isotopes at Piton de la Fournaise Volcano Rhodes JM, Lockwood JP (eds) Mauna Loa revealed: structure, (Reunion Island); origins and constraints on the size and shape of composition, history, and hazards. Geophysical monograph, vol the magma reservoir. Chem Geol 244:202–220. doi:10.1016/j. 92. AGU, Washington, DC, pp 207–217 chemgeo.2007.06.015