Journal of Petrology Advance Access originally published online on March 24, 2008
Journal of Petrology 2008 49(5):857-884; doi:10.1093/petrology/egn008
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Magmatic Differentiation at an Island-arc Caldera: Okmok Volcano, Aleutian Islands, Alaska
1Department of Earth Sciences, Wills Memorial Building, University of Bristol, Bristol BS8 1RJ, UK
2Gemoc, Department of Earth and Planetary Sciences, Macquarie University, Sydney, NSW 2109, Australia
3Alaska Volcano Observatory, Geophysical Institute, University of Alaska, Fairbanks, AK 901775-7320, USA
4Department of Geological Sciences, 1272 University of Oregon, Eugene, OR 97403, USA
RECEIVED JUNE 25, 2007; ACCEPTED FEBRUARY 12, 2008
 |
ABSTRACT |
|---|
TOP
ABSTRACT
INTRODUCTIONOkmok volcano is situated on oceanic crust in the central Aleutian arc and experienced large (
15 km3) caldera-forming eruptions at 12 000 years BP and 2050 years BP. Each caldera-forming eruption began with a small Plinian rhyodacite event followed by the emplacement of a dominantly andesitic ash-flow unit, whereas effusive inter- and post-caldera lavas have been more basaltic. Phenocryst assemblages are composed of olivine + pyroxene + plagioclase ± Fe–Ti oxides and indicate crystallization at 1000–1100°C at 0·1–0·2 GPa in the presence of 0–4% H2O. The erupted products follow a tholeiitic evolutionary trend and calculated liquid compositions range from 52 to 68 wt % SiO2 with 0·8–3·3 wt % K2O. Major and trace element models suggest that the more evolved magmas were produced by 50–60% in situ fractional crystallization around the margins of the shallow magma chamber. Oxygen and strontium isotope data (
18O 4·4–4·9
, 87Sr/ 86Sr 0·7032–0·7034) indicate interaction with a hydrothermally altered crustal component, which led to elevated thorium isotope ratios in some caldera-forming magmas. This compromises the use of uranium–thorium disequilibria [(230Th/ 238U) = 0·849–0·964] to constrain the time scales of magma differentiation but instead suggests that the age of the hydrothermal system is 100 ka. Modelling of the diffusion of strontium in plagioclase indicates that many evolved crystal rims formed less than 200 years prior to eruption. This addition of rim material probably reflects the remobilization of crystals from the chamber margins following replenishment. Basaltic recharge led to the expansion of the magma chamber, which was responsible for the most recent caldera-forming event. KEY WORDS: Okmok; caldera; U-series isotopes; Sr-diffusion; time scales; Aleutian arc
|
INTRODUCTION |
|---|
Caldera-forming eruptions are amongst the most spectacular and hazardous geological phenomena, with the potential to cause significant damage and to affect global climate [see Self (2006
) for a recent review]. However, there remains little consensus as to whether such eruptions are cyclical events representing an inevitable consequence of the build-up over time of evolved and explosive magmas. Consequently, detailed studies are required to understand why these catastrophic eruptions occur, how and over what time scale the erupted magmas are generated, and whether pre-eruptive magmatic records give any warning that such large eruptions are about to occur. Unfortunately, there have been very few short-lived isotope studies that might constrain the time scales of magma residence prior to caldera-forming eruptions, largely because there are few systems young enough for such studies. Here we present the results of a study of Okmok volcano in the Aleutian island arc, which has experienced two large caldera-forming eruptions within the last 12 000 years and thus affords an opportunity to try to address some of these issues. Traditional petrological and geochemical techniques are integrated with a range of isotopic data and trace element diffusion modelling to investigate the liquid line of descent and to extract as much information as possible about the processes and time scales involved. Because the multiple datasets constrain different aspects of the system we have chosen to present each set of results and interpret them separately and to finish with a working model that summarizes our conclusions. |
GEOLOGICAL SETTING AND PREVIOUS WORK |
|---|
The Aleutian arc extends for almost 3000 km from the Alaskan mainland in the east to the Komandorski Islands just off Kamchatka in the west (Fig. 1a), and represents the surface expression of the subduction of the Pacific Plate as it moves northwards beneath the North American Plate (Fournelle et al., 1994
). The arc is built on continental crust to the east and oceanic crust to the west, with a transition defined by the break in slope on the continental shelf at 164°W (House & Jacob, 1983). Recent geochemical studies of the volcanic products from the arc have highlighted evidence for variable relative contributions from the mantle wedge, subducted sediments and both fluids and (in the west) partial melts released from the downgoing altered oceanic crust (George et al., 2003; Kelemen et al., 2003; Jicha et al., 2004). Additionally, a study contrasting the time scales of tholeiitic magma evolution at an oceanic volcano (Aktuan) with calc-alkaline magmatic evolution at a continental volcano (Aniakchak) was presented by George et al. (2004).
|
Okmok is situated just west of the ocean–continent transition on the NE of Umnak Island at around 168°W (Fig. 1a). Byers (1959
, 1961) conducted the first thorough mapping program on Umnak Island, a direct result of military interest in Okmok after its eruption in June 1945, which was observed from the US air base at Fort Glenn on the SE flank of the volcano. He surmised that the 8–10 km diameter caldera at Okmok was the product of at least two major ignimbrite eruptions, based on the presence of twin fault scarps in the north. Furthermore, it was concluded that the bulk of the summit region in each case must have collapsed inwards, on the basis that the observed fraction of lithic fragments in the ignimbrite sheets fails to account for the volume of the missing pre-caldera edifice. Miller & Smith (1975, 1987) first examined the twin caldera-forming events and estimated them to be 5500 years apart, although this has recently been refined to 9950 years (Beget et al., 2005). The presence of an ancient lake in the caldera was established by work on the flood deposits in the Crater Creek gorge (Wolfe & Béget, 2002)—the result of a catastrophic failure of the caldera wall at 1500–1000 years BP, which released the largest flood known from the last 10 000 years. Geophysical investigations indicate that a body of magma still exists beneath Okmok (Mann et al., 2002).
Previous work at Okmok has focused on parental magmas (Nye, 1983; Nye & Reid, 1986), evolutionary trends (Miller et al., 1992; Miller, 1995) and subduction inputs (Miller et al., 1994; Class et al., 2000). High-Mg basalts at Cape Idak to the NE of Okmok caldera (see Fig. 1b) require high mantle temperatures (as high as 1500–1600°C), which have been accounted for by upwelling in the mantle wedge and the subsequent interaction of this hotter material with slab-derived fluids (Nye & Reid, 1986; Johnston & Draper, 1992). The calculated liquid composition ID1* for one of the Cape Idak basalts is more primitive than any lava yet found in the Aleutian arc (MgO > 15 wt %; Cr = 1300 ppm; Ni > 400 ppm), and has a liquidus temperature of 1409°C (Nye & Reid, 1986). Experimental data on another magnesian basalt from the Idak plateau, sample ID16, suggest that it last equilibrated with the mantle at around 12 kbar (Johnston & Draper, 1992).
|
STRATIGRAPHY AND SAMPLING |
|---|
The stratigraphy of erupted products from Okmok has recently been described at length by Larsen et al. (2008) to which the reader is referred for details. For the purposes of this paper, the sequence can be subdivided into the
12 000 years BP (Okmok I) and 2050 years BP (Okmok II) caldera-forming events. These are temporally separated by the Crater Creek series of caldera-infilling lavas (herein termed inter-caldera) and since the 2050 years BP eruption, the caldera has seen the development of a series of intra-caldera basaltic cones (Fig. 1b).
The 12 000 years BP caldera-forming event emplaced a pyroclastic formation similar in volume and internal stratigraphy to the 2050 years BP caldera-forming unit (see below). It was preceded by a thin initial mixed rhyolite and andesite Plinian phase overlain by a dominant volume of andesitic pyroclastic deposits. This eruption also formed a caldera similar in diameter to the later event, although slightly offset to the NE where remnants of its walls and intra-caldera fill remain (e.g. the Crater Creek series). It should be noted that no pre-caldera samples are available for the 12 000 years BP event and that much of this study has focused instead on variations in intra-caldera, syn-caldera and post-caldera samples, with respect to the 2050 years BP event. Nevertheless, we present data for eight samples of the 12 000 years BP andesite and three from the rhyodacite.
The inter-caldera Crater Creek series comprise a sequence of lavas erupted prior to the 2050 years BP event, filling part of the earlier caldera and exposed only as a veneer on the present caldera wall (Fig. 1b). The 10 lavas analysed here are basaltic andesite to andesite in composition. A unit known as the Middle Scoria Unit was also erupted sometime between the two caldera-forming eruptions (Wong, 2004), and is composed of basaltic andesite scoria from which two samples were collected. This unit occurs across the south to east flanks of Okmok. Although no published eruption ages are available for the Crater Creek samples, they are thought to have been extruded shortly before the 2050 years caldera collapse for two reasons. First, there is no evidence of any lava-flow breaching the earlier caldera, as the most prominent unit between the two caldera-forming packages on the flanks is the Middle Scoria Unit. Because the uppermost flow of the Crater Creek sequence is close to the top of the old caldera wall, and as no other caldera-filling flows are found, it therefore seems likely that the top of Crater Creek represents the final stages of inter-caldera effusive activity. Second, a switch from effusive to pyroclastic flow regimes towards the top of the sequence is matched by an increase in silica contents, and may represent the early stages of more explosive activity just prior to the caldera-forming event.
The 2050 years BP caldera-forming unit is composed of an upper mafic pyroclastic fall and flow sequence and a basal rhyodacitic tephra fall. Two samples were taken from the basal rhyodacite that was erupted hours or days before the main caldera-forming event, which was accompanied by the emplacement of the mafic pyroclastic deposits. The conclusions of the study of Larsen et al. (2008) are that the basaltic andesite ash-flows that make up the final stages of the caldera-forming eruption are a hybrid of the resident andesite (which was erupted as an ash-fall, immediately prior to the ash-flows) and a freshly intruded basaltic magma, and that this mixing event was the trigger for the caldera-forming eruption. We also collected 13 samples from the mafic units, which form over 99% of the 15 km3 erupted volume and should therefore not be considered as truly ‘bimodal’. Deposits drape the flanks of Okmok, overlapping onto the deposits of the 12 000 years event outside the caldera as shown in Fig. 1b.
Post 2050 years BP intra-caldera cone-building lavas are subdivided on the basis of composition into those erupted in the NW and SE halves of the caldera. They are basaltic to andesitic in composition and those from the NW of the caldera (Cones B, E, G, H and ‘I’) are more mafic than those erupted in the SE (Cones A, C, D and F). Age constraints are relative only for all post-caldera samples, apart from dated historic flows. The cones were given relative ages by Byers (1959) and these have been reassessed by recent remapping of much of the caldera floor (C. Neal, personal communication, 2003). Historical eruptions at Okmok have occurred in 1820, 1931, 1938, 1945, 1958, 1960, 1981, 1983, 1988 and 1997 (Miller et al., 1998), as well as major explosive eruptions in 1817 and 1899. At the time of collecting the 14 samples presented here only the 1958 and 1997 eruptions had identifiable flow units. It had been thought that a satellite vent, Mount Tulik (Fig. 1b), was the origin of many of these eruptions, but that vent is deeply eroded and probably Pleistocene in age (Coats, 1950). Cone A is thought to be the site of most of the historical eruptions.
|
ANALYTICAL TECHNIQUES |
|---|
For the mineral analyses, thin sections were carbon-coated for analysis on a Cameca Camebax Micro, at the University of Bristol, controlled by SamX operating software. The accelerating voltage was 15 keV, producing a focused beam with a current of 15 nA. Standards used for calibration of the spectrometers were St John's olivine (Mg, Si), albite (Na, Si), sanidine (K), andradite (Ca), MnO (Mn), Fe (Fe) and SrTiO3 (Ti). A secondary standard, KK1 (kaersutite) was used to verify the calibration. Analyses were generally discarded if oxide totals fell below
98%. Sr and Ba analyses of plagioclases were determined on gold-coated samples by secondary ion mass spectrometry (SIMS) at the NERC Ion Microprobe Facility at the School of Geosciences, at the University of Edinburgh. A Cameca ims 4f ion microprobe was used to follow the traverses made across plagioclase crystals. The incident primary 16O– ion beam was accelerated at 14·5 keV, with a beam current of <10 nA. Resulting sputter pits were 25 µm diameter. Positive secondary ions were accelerated (at 4·5 keV) into a double focusing mass spectrometer. Elemental intensities were measured on the electron multiplier for between 3 and 10 s per peak (and for 10 cycles). An NIST silicate glass standard (SRM-610) was run at the start of each day, and this was used to correct any sample data obtained. As a secondary check, a feldspar megacryst (SHF1) with known Sr concentration of 505 ppm was run and corrected using the NIST standard. During analyses, mass 130·5 was used to check background noise levels. These ranged from around 0·1–0·4 c.p.s., and each analysis was corrected for background individually. SiO2 was used as a reference oxide, so where known from electron microprobe analysis data, appropriate values were also used for each analysis. Ion yield patterns were similar to NBS SRM-610 (see Hinton, 1990), but the ionization efficiency relative to Si was 10–25% lower, as observed in previous measurements of the reference crystals (R. Hinton, personal communication, 2002). Thus a Sr ppm value was obtained for each point, and then a final correction made to account for the known offsets between NBS 610 and the plagioclase standard, Lake County Plagioclase, to give final values. Typical error on a SIMS trace element analysis is 10% and is mainly related to counting statistics. Accuracy and precision are estimated to be 10% and 2%, respectively.
Whole-rock samples were analysed at Washington State University by X-ray fluorescence (XRF) for major elements (Johnson et al., 1999), reported here normalized to 100% on a volatile-free basis. Trace elements were analysed by inductively coupled plasma mass spectrometry (ICP-MS) at Washington State University (Knaack et al., 1994).
U and Th concentrations and isotope ratios were determined on samples that were spiked with a mixed 236U–229Th tracer and dissolved using an HF–HNO3–HCl mix in heated Teflon pressure bombs. U–Th isotopes were separated using TRU-Spec resin and analysed by multicollector (MC)-ICP-MS using a ThermoFinnigan Neptune at the University of Bristol following protocols described by Sims et al. (2008). Reproducibility and accuracy were assessed via multiple measurements of Table Mountain Latite (TML-3), which yielded (230Th/238U) = 0·998 ± 0·006 (2
).
For Sr isotopes, the Sr-cut was taken from the TRU-Spec columns and purified by means of a single pass through Sr-Spec resin. Samples were loaded onto single rhenium filaments in 1 µl 1N phosphoric acid (H3PO4) and analysed by thermal ionization mass spectrometry on a ThermoFinnigan Triton system at the University of Bristol. Measurements were made by a static method, with all cups separated by a single mass unit. Mass fractionation was measured and corrected for using the natural value of 88Sr/86Sr = 8·375209. SRM-987 was measured at regular intervals and samples normalized to an SRM-987 value of 87Sr/86Sr = 0·710248. Standard reproducibility was ± 0·0011% (2).
Selected samples were analysed for oxygen isotopes by CO2 laser fluorination at Caltech and the Universities of Wisconsin and Oregon, including data reported earlier by Bindeman et al. (2001a,b). A detailed description of analytical techniques, standards and reproducibility has been given by Bindeman et al. (2004). Magmatic 18O was obtained by measuring the 18O of 1–2 mg of one or more unaltered phenocryst phases and then applying experimentally and empirically derived mineral–melt or mineral–mineral isotopic fractionation relationships described by Bindeman et al. (2004). Errors for single (phenocryst or melt) measurements are of the order of ±0·1 (2). For derived magmatic 18O, the estimated error is ± 0·5 (2).
|
PETROGRAPHY AND MINERAL CHEMISTRY |
|---|
Mineral compositions are included as Supplementary Data (Appendix A) and modal proportions are shown for representative samples through the stratigraphy in Fig. 2.
|
The
12 000 years BP caldera-forming eventThe products of the 12 000 years BP caldera-forming event are close to aphyric with only rare phenocrysts of plagioclase and clinopyroxene (Larsen et al., 2008).
Intercaldera ‘Crater Creek’ series
Samples from the Crater Creek sequence are porphyritic to aphyric basaltic andesites to andesites, each with a groundmass ranging from glassy to microcrystalline (up to 50 µm grainsize) composed of plagioclase, clinopyroxene and Fe-oxides, as in the post-caldera suite.
Plagioclase is common to all samples and displays a wide range in modal abundance, from 10 to 31%. Grain size varies up to 2 mm, as in the post-caldera suite. Normal zoning between a calcic core and a more sodic rim is present in the majority of crystals. Whereas cores display the same range of anorthite contents (An93–67), rim compositions in this series are generally much more sodic than in the post-caldera products (as low as An45 in the most evolved sample, BF00CRP2) across a similar range of whole-rock compositions. Feldspar laths are present in the matrix in all cases, with compositions similar to the rims of the larger phenocrysts. Clinopyroxene often occurs in composite grains with plagioclase or olivine and, in general, has a similar composition to that in the post-caldera samples. A single analysis from CR4 seems to represent a more evolved composition, with higher Fe. Grains are often 500 µm in diameter. Olivine in the Crater Creek series is fairly Fe-rich (Fo72–45), especially in the more evolved (younger) whole-rock compositions. Spinel is restricted to the same evolved samples, where it is often found as inclusions in plagioclase or as part of a composite crystal, as well as a rare phenocryst phase.
A clot of crystals around 10 mm in diameter is present in sample BF00CRP2, displaying an interlocking cumulate texture. The mineral assemblage remains the same, with the addition of ilmenite and pigeonite as rare cumulate phases. Plagioclase defines the cumulate texture, with 500 µm long tabular crystals. Clinopyroxene occurs as large crystals, usually inter-cumulus and thus without good euhedral shape, whereas spinel is late-stage and generally <200 µm. Mineral compositions are distinct from those in the host rock and have more evolved compositions.
The 2050 years BP caldera-forming event
The 2050 years BP caldera-forming package is composed of three units: a basal rhyodacite ash-fall, an andesitic ash-fall, and an upper basaltic andesite ash-flow. The rhyodacite fall deposit is composed of crystal-poor pumice (<5% phenocrysts), with rare lithic fragments. The glass matrix is used as a proxy for the liquid composition of the magma, and an average of several microprobe analyses agrees well with the whole-rock composition for the unit with the exception of anomalies in Fe2O3 and Na2O, which may be attributed to the occurrence of sodic plagioclase and Fe-oxide phases. Plagioclase core compositions range from An44 to An53, with rim compositions remarkably uniform at An46–47, although the data are from a single sample and may not be representative of the whole unit. In a more detailed study of the same formation (J. F. Larsen, unpublished data), core compositions of An40–67 were reported. Phenocrysts range from 200 to 1000 µm, and are generally fresh and euhedral, with generally narrow rims. Disequilibrium features such as embayments or fritted edges are generally absent. Clinopyroxene (Wo39–46 En32–44 Fs17–25) occurs in composite grains or as single euhedral phenocrysts that are generally more Fe-rich than those from the more basic rock-types, suggesting that clinopyroxene is a liquidus phase, although not a common one. Spinel is also present, falling at the evolved end of the range for island arc basalts, defined on the basis of Cr–Fe3+–Al by Eggins (1993); it has lower Mg-number than crystals from more basic samples. The petrology of the andesitic ash-fall unit was not studied in detail but Larsen et al. (2008) have provided a description of phenocryst compositions and textures. Plagioclase phenocrysts range from An52 to An90, with rounded or slightly embayed forms suggesting limited dissolution. In general, the unit is crystal-poor.
The upper basaltic andesite ash-flow deposit is composed of crystal-poor pumice (7% phenocrysts) with 10% lithic fragments. As a result, the glass matrix has a slightly higher silica content than the whole-rock analysis. Plagioclase core compositions cluster at An45–50, with rims at An58–65 such that crystals are reversely zoned. A few, possibly xenocrystic cores are present with An82–87. J. F. Larsen (unpublished data) has isolated two populations of plagioclase in a more extensive study of this unit—one with a range of An47–93, and another concentrated around An62. The crystals reported here are probably from the former group, whereas the rim compositions, which match the second population in that study, most probably represent the equilibrium composition. Indeed, the population with a wider range of compositions is reported to have many crystals with fritted and resorbed, or ‘channelled’ edges, whereas the An60 group seems much less altered. Plagioclase crystals are around 200–1000 µm in size, although there is also a population of microlites (<100 µm) in the matrix glass with compositions around An60. The andesitic ash-fall unit contains plagioclase ranging from An52 to An90; these crystals are often rounded or slightly embayed, yet not to the extent of those in the ash-flow (J. F. Larsen, unpublished data). Clinopyroxene is a rare phenocryst phase and has the composition of augite (e.g. Wo36 En48 Fs16). Olivine is also a rare phase although at Fo70 it is out of equilibrium with the groundmass and therefore probably xenocrystic. Spinel occurs as a phenocryst phase and also as inclusions within olivine.
Post-caldera lavas
Post-caldera samples are porphyritic to aphyric basalts and basaltic andesites, each with a groundmass ranging from glassy to microcrystalline (up to 50 µm grain size) composed of plagioclase, clinopyroxene and Fe-oxides.
Plagioclase is ubiquitous, occurring at all scales up to 2 mm and in modal proportions of 7–32%. Rocks with a very high modal proportion of plagioclase may have accumulated the phase from other magmas. Simple normal zoning is evident in many crystals, with calcic cores (up to An94) surrounded by more sodic rims (as low as An65). This variation exists in crystals as small as 200 µm. Smaller, tabular or acicular grains lack a calcic core and are composed entirely of rim-composition material. Rims vary from 50 to 100 µm in width. Disequilibrium textures are rare, although crystals are commonly cracked or pitted. The observed range in anorthite content for the post-caldera plagioclases is An93–42.
Clinopyroxene occurs most commonly in large glomerocrysts along with plagioclase and sometimes olivine. Modal proportions vary (see Fig. 2), whereas the compositional range is rather restricted (Wo49–41 En48–43 Fs4–15); pigeonite is found in some samples. Olivine is present in all but the most evolved lavas, although always at modal proportions <1%; the average composition is Fo80. Grains are often large (up to 1 mm) and generally display good original euhedral shapes, but with deep embayments often leading to a skeletal appearance. Fe-oxides were identified as a phenocryst phase only in sample H1 of the post-caldera suite.
|
INTENSIVE PARAMETERS |
|---|
The relationship between the major-element composition of plagioclase and the liquid from which it has grown can provide quantitative information on the pressure of crystallization and magmatic volatile content, as well as showing whether a crystal population is really in equilibrium with its host (Sisson & Grove, 1993a
, 1993b; Heath et al., 1998). Figure 3 shows the variation of [Ca]/[Na] in plagioclase and in calculated parental liquids from Table 2 (see below). It is clear that the more mafic (higher Ca/Na) samples contain a wide range of plagioclase compositions at each inferred liquid composition. Core compositions in particular show marked disequilibrium (equilibrium is shown by the anhydrous line at Kd = 1) and although much of the range could be the result of changing water contents in the parental liquid, it is unlikely that the volatile content of a liquid would range from zero to 6% at constant [Ca]/[Na]. Instead, it is proposed that many of the cores crystallized from liquids with higher [Ca]/[Na]—such as the parental liquid ID1* (Nye, 1983), which has [Ca]/[Na] = 6·5 and could crystallize >An90 plagioclase at volatile contents <6%. It should also be noted that Kd plag/liquid increases approximately two-fold at pressures of 1·5–2 kbar, such that even lower H2O contents are required if crystallization occurred at depth from a primary melt. In contrast, rim compositions are much closer to being in equilibrium with their coexisting groundmass composition. In basaltic samples at higher [Ca]/[Na], rim compositions range from just below the anhydrous 0–8 kbar curve to just above the 2% H2O at 2–5 kbar curve (Fig. 3). The inference drawn is that the rim material crystallized from the coexisting interstitial melt (now groundmass) at the given volatile contents and pressures. In more evolved samples ([Ca]/[Na] <2) rim compositions show a greater range of Kd values, such that liquid volatile contents between zero and 4% H2O are inferred, and in one example (sample BF00-G2), up to 6%.
|
|
Electron microprobe analyses of the rhyodacite glass (98JLOK4b) in this study all gave low oxide totals, on average 3·02% below 100, which may be attributed to lost volatiles—the bulk of which are likely to be H2O. Additionally, the average of six melt inclusion analyses (J. F. Larsen, unpublished data) yields a volatile content of 3·15%—although this probably represents the state of the host liquid at an earlier stage in its evolution (i.e. when the inclusion was formed during crystal growth). The difference method has also been employed to estimate water contents in the basaltic andesite magma (99JLOK4g_5), using the average of several analyses of the matrix glass. This gives a volatile content of 1·75% for the andesite magma.
Coexisting pyroxenes can be used to constrain crystallization temperatures and a selection of analyses are projected for this purpose according to the scheme of Lindsley (1983) in Fig. 4. No three-pyroxene assemblages are found, but coexisting augite + pigeonite or augite + orthopyroxene can be plotted, the tie-lines giving a temperature in each case. It should be noted that not all tie lines perfectly describe a single isotherm. Temperatures derived apply to pressures close to 5 kbar (Fig. 4). Single augite phenocrysts are also plotted as an indication of general temperature trends. The temperatures derived range from 900 to 1125°C with the lowest temperatures in the 2050 years BP rhyodacite and the highest in the andesite of the same unit.
|
The only separate evidence of the depth of a possible magma chamber is provided by a geophysical investigation in the aftermath of the 1997 eruption (Mann et al., 2002
), which suggests that the source of that magma lay some 2·5–5·0 km beneath the centre of the present caldera. This depth also suggests that active magma bodies are present at pressures of 1–2 kbar. |
MAJOR AND TRACE ELEMENT BEHAVIOUR |
|---|
Results
Table 1 presents whole-rock major and trace element data for the 52 samples collected. The majority of the samples range in SiO2 from 51 to 60 wt %, excepting the rhyodacites from the two caldera-forming events, which have 67–72 wt % SiO2. However, none of the samples are primitive and both MgO (5·8–0·2 wt %) and CaO (12·2–1· 8 wt %) decrease with increasing SiO2 (not shown). Na2O and K2O both increase with increasing SiO2 whereas TiO2, Al2O3, Fe2O3* and P2O5 broadly increase until about 54 wt % SiO2, and then decrease towards the more evolved compositions (Fig. 5). The mid-stage Fe enrichment is a characteristic of tholeiitic differentiation trends (Sisson & Grove, 1993b
; Arculus, 2003) and was also noted by Miller et al. (1992) in their Okmok sample set. In Fig. 6, most of the samples fall in the tholeiitic field on the FeO*/MgO vs SiO2 diagram of Miyashiro (1974) and are medium-K according to the classification scheme of Gill (1981), excepting the high-silica fall deposits, which are high-K. Under the alternative classification of Arculus (2003) the Okmok set of samples would be considered as medium-Fe, becoming high-Fe only in caldera-forming samples, as shown in Fig. 6. The variation in SiO2 and cumulative volume with stratigraphic height is shown in Fig. 7.
|
|
|
The variation of selected trace elements with SiO2 is shown in Fig. 8, and data for the complete range of analysed incompatible and rare earth elements (REE) are plotted in normal mid-ocean ridge basalt (N-MORB) normalized multi-element diagrams in Fig. 9. Zr, Ba and Th all increase with SiO2 across the compositional gap to rhyodacitic compositions. For these and other incompatible elements, a clear division exists between the post-caldera samples and all other suites, as the post-caldera examples have consistently lower concentrations at the same SiO2 content. The transition elements Sc and V show more complex patterns, behaving in a compatible manner only after the introduction of an oxide phase to the crystallizing assemblage (much like TiO2). Sr appears to be moderately compatible across the whole range of SiO2. Figure 9 shows trace element patterns typical of convergent margin lavas (e.g. Hawkesworth et al., 1997
), with enrichment in the large ion lithophile elements (LILE) such as Cs, Rb, Ba, Sr and Pb as well as depletion of the high field strength elements (HFSE) Ta and Nb, with respect to La and Th. The most evolved samples in the suite display elevated concentrations of all elements apart from Eu and Sr, which have negative anomalies, implying the crystallization and removal of plagioclase. However, some samples also exhibit positive Eu anomalies, implying accumulation of plagioclase. Concentrations of the REE are close to those of N-MORB (Sun & McDonough, 1989), with relative enrichment in the light REE.
|
|
The calculated primitive liquid composition (ID1), derived from a subaerially erupted olivine-cumulate sample by correcting for olivine accumulation (Nye & Reid, 1986
), is also shown in Figs 5, 6 and 8. This has been successfully evaluated by least-squares analysis as a potential parental liquid for the Okmok samples of Nye (1983) and Miller (1995). It represents the most mafic liquid at Okmok, and one of the most mafic documented arc compositions globally (Nye & Reid, 1986). If this liquid composition is included with the rest of the data, a further inflection occurs in the TiO2, Al2O3 and Fe2O3 trends at around 50 wt % SiO2 (Fig. 5).
Liquid line of descent
Whole-rock compositions do not always provide realistic liquid proxies and the accumulation of mineral phases obscures the true fractionation path of magmas on whole-rock plots, producing scatter and distorting trends. At Okmok, some post-caldera samples have Eu/Eu* >1, and the addition of plagioclase may account for their high observed Al2O3 and Sr contents (Figs 5 and 8). Similarly, the accumulation and removal of clinopyroxene and magnetite may be responsible for the scatter in Sc and V, respectively (Fig. 8). Furthermore, volatile contents and oxygen fugacities can be estimated from the chemical disequilibrium between phenocrysts and their host liquid (e.g. Sisson & Grove, 1993a)—and poor estimates will be produced if whole-rock compositions are relied upon. Thus, it is useful to estimate liquid (groundmass) compositions to obtain a true picture of liquid evolution in the suite, without the scatter and bias imparted by crystal accumulation. By combining point-counting data with average phenocryst compositions, an estimation of the groundmass composition can be made by mass balance. Average crystal compositions multiplied by their model abundance were subtracted from the whole-rock major element composition and the resultant composition re-normalized to 100%. Table 2 lists the modal proportions and average phase compositions used and the resulting calculated liquid composition for a representative selection of samples from each of the stratigraphic units investigated. Possible liquid lines of descent are shown in Fig. 10 for the post-caldera and Crater Creek series and the coherence of these schematic lines of evolution suggests that the technique described for calculating liquid compositions is a valid one. Therefore, these compositions are used as the basis for all subsequent modelling of the liquid line of descent. It appears that a similar transition occurs in each series, leading to depletion in Fe2O3 (and TiO2; not shown) at different SiO2 contents. Fe2O3 inflects later in the liquid trend of the Crater Creek series, which probably reflects the later appearance of Fe–Ti oxides in that stratigraphic group. This could account for the high TiO2 contents in the whole-rock data for the same samples (Fig. 5). The apparent enrichment of Al2O3 in the low SiO2 samples is reversed in the liquid composition data, suggesting that it may be an artefact of plagioclase accumulation (Fig. 10).
|
|
Very few primitive lavas are present in the modern Okmok suite (none >5·8 wt % MgO), and so the calculated liquid composition, ID1* (Nye & Reid, 1986
) was used to represent the flux of liquid from the mantle. We have used the MELTS algorithm (e.g. Ghiorso & Sack, 1995) to model possible evolutionary paths from this primitive liquid to the most mafic members of the modern suite. It should be noted that the term ‘evolution’ is applied to a shift towards more silicic compositions, and does not necessarily relate to time, which is addressed subsequently using U–Th disequilibria and Sr diffusion modelling. The results of a MELTS model for in situ crystallization (Langmuir, 1989) of a starting composition of ID1* (at 3 kbar and 0·2% H2O based on inferences drawn from the mineral equilibria data) are shown in Fig. 11. Olivine (Fo85) is the first phase to crystallize from the ID1* liquid, followed by clinopyroxene and finally plagioclase (An80). The residual liquids follow a path leading to the mafic end of the calculated liquid array for all oxides except Fe2O3* and Al2O3 (Fig. 11).
|
The discrepancy in Fe2O3* is problematic, whereas the Al2O3 misfit is most probably due to the suppression of plagioclase in the MELTS model until the residual liquid has reached >18 wt % Al2O3. Moreover, the plagioclase produced in the model is not as calcic as most observed core compositions, although it appears to match rim compositions well. Increasing the initial water content of the model to 2% does force the crystallization of suitably calcic plagioclase crystals, but they do not appear on the liquidus until the residual liquid has attained 23 wt % Al2O3, which seems unrealistic in view of the observed liquid compositions. Although the 0·2% H2O MELTS model does intersect the high-Al2O3, low-SiO2 end of the whole-rock array (because plagioclase crystallization is suppressed until the liquid reaches
18·5 wt % Al2O3) it is more likely that these high Al2O3 contents reflect plagioclase accumulation. If plagioclase is introduced at around 15 wt % Al2O3, observed liquid compositions with <17 wt % Al2O3 (e.g. BF00-1997f and BF00CR7) can be modelled. Figure 11 shows that MELTS models that use the calculated liquid compositions BF00-1997f and BF00CR7 can, with the exception of MnO, provide a good simulation of the apparent liquid line of descent. Additionally, it is clear from the evolution of P2O5 (Fig. 11) that a further phase must crystallize during the generation of the rhyodacite, in which P is compatible. Apatite is the most probable candidate, yet none is found in the rhyodacite or in any other sample in this study, and although it does appear as a liquidus phase in the MELTS modelling presented above, its appearance is too late to allow the liquid to evolve as suggested by the sample data. However, the in situ model means that intermediate and evolved products result from a return of residual liquid from a marginal zone where temperatures are lower and apatite stable. This may account for the lack of apatite in the erupted magmas.
Trace element modelling was undertaken using various approaches (Finney, 2004). Our preferred model is the in situ crystallization model of Langmuir (1989), the results of which are summarized in Fig. 12. Because liquid compositions could be accurately calculated only for major elements, the trace element modelling utilized the whole-rock data and the partition coefficients that are listed in Table 3. The model simulated sequential evolution from the ID1* composition to a plagioclase supersaturated basaltic liquid (mimicking the plagioclase accumulated samples) to an andesitic liquid and finally a rhyodacite using the phase proportions listed in Table 4 (based on the MELTS results). In Table 4, the value f represents the fraction of residual liquid returned to the evolving chamber interior, from the crystallizing solidification zone; the total amount of crystallization (X) is also listed for each step.
|
|
|
The models shown in Fig. 12 suggest that significant (10%) magnetite crystallization occurs much earlier in the post-caldera suite than the inter-caldera suite. Much lower proportions of magnetite appear at this stage in the inter- and syn-caldera groups (all other symbols), allowing Ti to continue increasing. Particularly low proportions of magnetite allow the generation of the high-Fe–Ti group (see also Nye, 1983
), whereas around 1·5% magnetite generates the dominant intermediate compositions. Subsequently, the proportion of magnetite continues to rise, and the rhyodacitic, low-Ti liquids are eventually formed. The changing proportions of plagioclase crystallization (see Table 4) exert the dominant control on Sr content. In terms of K/Rb, Fig. 12 shows that model curves tend to flatten out and do not take into account samples with low K/Rb (marked with an ‘X’). This suggests that there may be a role for assimilation of a high-Rb crustal component, which we now explore in conjunction with the new isotope data. |
ISOTOPE RESULTS AND IMPLICATIONS |
|---|
In Table 5 we present new U–Th, O and Sr isotope data for representative samples from Okmok. U and Th concentrations are in the range of 0·5–3·8 and 1·0–7·4 ppm, respectively. For the majority of samples (234U/238U) is within analytical error (
1%) of unity consistent with a lack of alteration. However, some of the caldera-forming samples have (234U/238U) >1 and one (99JLOK42Dd) from the 12 ka BP eruption has (234U/238U) = 1·11 indicating non-magmatic U-series fractionation. (230Th/232Th) ranges from 1·29 to 1·37 with the exception of sample 00JLOK29K (fall), from the 12 ka BP eruption, which has (230Th/232Th) = 1·47. All of the samples have 238U excesses, with (230Th/238U) ranging from 0·80 to 0·96. Calculated (see Table 5 for details) magmatic 18O values range from 4·9 to 4·5 and ID16 yielded 18O = 5·6 ± 0·2 (I. N. Bindeman, unpublished data). Sr isotope results show a relatively restricted range of 87Sr/86Sr = 0·7032–0·7034 similar to previously reported data from Okmok (e.g. Class et al., 2000).
|
U–Th disequilibria
In a U–Th equiline diagram (Fig. 13), the Okmok samples lie within the field for other Aleutian arc samples and are very similar to samples from Akutan volcano located on oceanic crust just to the east. George et al. (2003
) have argued that the U–Th disequilibria of the Aleutian arc lavas are dominated by fluid addition of U to a mantle wedge whose (230Th/232Th) ratio was 1·3 as a result of sediment addition [see Turner et al. (2003a) for a recent review]. If this is correct, then some of the post-caldera samples, which have (230Th/232Th) = 1·3, may have erupted less than a few thousand years after fluid addition. In contrast, the time elapsed since fluid addition for most of the remaining samples would be 20 000–40 000 years, based on this model (see Fig. 13). Some or most of this inferred time could have been spent in the magma chamber beneath Okmok and, thus, provides an important constraint upon the time scale of differentiation. However, some of the samples with (230Th/232Th) >1·3 also have (234U/238U) >1 with the implication that their magmatic Th isotope signatures may have been modified (see further discussion below). Moreover, U-series studies of Akutan and Aniakchak have inferred that magma differentiation occurred over time scales of <10 000 years (George et al., 2004). Accordingly, we are hesitant to attribute significance to these U–Th model ages.
|
O and Sr isotopes
Although the most primitive lava ID16 has
18O = 5·6 (I. N. Bindeman, unpublished data) typical of pristine mantle (5·1–5·7; Taylor, 1980; Eiler, 2001), eight out of 10 of the Okmok samples have 18O lower than 4·9. The majority of low 18O magmas have been erupted at high latitudes, where precipitation can have 18O = –25, and so it has been inferred that such magmas have interacted with hydrothermally altered rocks (e.g. Muehlenbachs et al., 1974; Bindeman et al., 2001a,b, 2004). Importantly, Fig. 14 shows that there is a negative correlation between 87Sr/86Sr and 18O, which suggests that the variation in Sr isotopes is also due to interaction with hydrothermally altered material rather than variations in the amount of sediment addition to the source of the lavas. Hydrothermal fluids are also typically characterized by (234U/238U) >1 as a result of recoil effects and the enhanced mobility of 234U out of radiation-damaged lattice sites. 230Th is the daughter product of 234U and so time-integrated interaction with hydrothermally altered rocks will, potentially, also result in coupled increases in 18O and (230Th/232Th). Although there is considerable scatter, Fig. 15 does suggest that those rocks with the highest (230Th/232Th) ratios are also characterized by elevated 87Sr/86Sr and low 18O.
|
Assimilation models
In Figs 14 and 15 we show the results of a number of illustrative models that simulate the effects on a mafic Okmok magma (ID1* or BF00CR5) of assimilation of a partial melt of pre-existing solidified magma (i.e. magma chamber rind) of 100 ka age (see below) that had interacted with a hydrothermal system, which lowered its
18O from initial mantle-like values. Because the in situ crystallization model was required to account for the suppression of compatible element depletion during differentiation of the most evolved magmas, the decoupled fractional crystallization and crustal assimilation (FCA) model of Cribb & Barton (1996) was employed using the end-member compositions listed in Table 6.
|
|
In the simplest single-stage model, the ID1* magma, with an inferred 87Sr/86Sr ratio of 0·7030, undergoes bulk assimilation of wall-rock, which is assumed to have a trace element composition similar to the andesitic inter-caldera sample BF00CRP2 and
18O = –5. To match the data in Fig. 14, the assimilant needs to have a 87Sr/86Sr ratio of 0·707, and r, the rate of assimilation to fractional crystallization, is 0·1. Alternatively, if the 87Sr/86Sr ratio of the mantle end-member is 0·7032, a two-stage model can be invoked with an early stage 1 during which only O isotopes are affected, followed by a later stage 2 when 18O continues to decrease and 87Sr/86Sr begins to increase (see Fig. 14). The total amount of FCA is similar to that in the single-stage model. However, Fig. 14 also shows that Ba/Th decreases as 87Sr/86Sr increases (although the rate of this decrease is variable) and that the bulk assimilation model curve is too steep to account for most of the data. Instead, an assimilant with lower Ba/Th ratio is required. Ba/Th might be lowered in the assimilant by mobilization by the hydrothermal activity thought to be responsible for the low 18O. However, even if Ba is almost completely removed (such that Ba decreases from 1752 to 1 ppm), the observed range of values cannot be accounted for. Alternatively, Ba/Th in the assimilant may be lowered during partial melting of the wall-rock. Model curves are plotted in Fig. 14 showing that 10–50% batch melting can describe the data well. In particular, samples from the 2050 years BP eruption and the preceding Crater Creek series fall between the 10% and 20% melting curves, suggesting that, during this period, assimilation was by 15% batch melting.
The single-stage model is also shown on plots of 87Sr/86Sr and 18O vs (230Th/232Th) in Fig. 15. It is assumed that liquid supplied from the mantle has (230Th/232Th) 1·30 whereas the assimilant is inferred to be in secular equilibrium, with a value of (230Th/232Th) 1·54. This model passes through much of the data. However, the curve is not as steep as that of the data, and crucially, the model does not come close to the most mafic sample, which lies at a much lower value of (230Th/232Th). In light of this discrepancy, the two-stage model is also plotted in Fig. 15 and is able to describe the available data more accurately. During the first stage, the Th-isotope composition remains unchanged—as the assimilant at this time is inferred to have (230Th/230Th) = 1·30, identical to that of mantle liquids. Only 18O decreases, generating compositions close to those of the most mafic erupted magmas. From this point, the second stage of the model can pass through the remaining data with the correct slope if the (230Th/232Th) ratio of the assimilant is 1· 45. If the assimilant is assumed to be previously solidified magma (e.g. Bindeman et al., 2006) that had an initial (230Th/232Th) ratio of 1·3, then this model constrains the age of the assimilant to be 100 ka. However, the corollary is that the Th isotope ratios of at least some of the Okmok rocks have been raised by assimilation and thus cannot be used to constrain differentiation time scales.
Although the mixing models described above are necessarily non-unique they do serve to show that assimilation of old, hydrothermally altered magmatic rocks can lead to co-variations in Th–Sr–O isotopes similar to those observed in the Okmok lavas. We suggest that similar processes may be common in high-latitude systems, especially where voluminous caldera-filling lakes and the effects of the end of Pleistocene glaciation have promoted the formation of active, low 18O hydrothermal systems around magma chambers. An example is the Fisher eruption (Bindeman et al., 2001a, 2001b) although identification of such processes in lower latitudes might prove harder to detect in O isotopes.
|
DIFFUSION IN PLAGIOCLASE |
|---|
One method for constraining the time scales of magmatic processes is to utilize the tendency of fast-diffusing elements to redistribute to an equilibrium state while a crystal is at magmatic temperatures; the extent of this ‘re-equilibration’ can be used to constrain the time since the initial growth. Such an approach has been used in a number of case studies of magmatic olivine (Nakamura, 1995
; Pan & Batiza, 2002), plagioclase (Zellmer et al., 1999, 2003; Costa et al., 2003) and clinopyroxene (Morgan et al., 2004). Here we use the model developed for Sr and Ba diffusion in plagioclase by Zellmer et al. (1999, 2003). Sr has a high diffusion rate in plagioclase, yet its initial concentration upon crystal growth appears to vary little, as a result of a low dependence of DSr on temperature or melt composition. However, after crystal growth Sr will diffuse to be in equilibrium with the anorthite content, which does have a high dependence on composition, temperature and water content and is thus often more clearly zoned across a crystal. The slow, coupled inter-diffusion of CaAl–NaSi (Grove et al., 1984) ensures that the anorthite profile and thus also the ‘equilibrium’ trace element profile do not change over the time scales of interest. Therefore a comparison of Sr and anorthite contents across a crystal reveals the extent of re-equilibration, which is then combined with known diffusion rates to give the time since crystal growth.
Spatially resolved anorthite contents and Sr and Ba concentrations, as well as calculated equilibrium Sr and Ba concentrations [see Zellmer et al. (1999, 2003) for details of these calculations] for traverses across seven plagioclase phenocrysts from the various stratigraphic units are available as Supplementary Data (Appendix B). Figure 16 illustrates the results for one of the post-caldera samples. The equilibrium Sr profile is essentially a mirror image of the XAn profile and the local equilibrium profile [d(SrObs/SrEQ)/dx] is derived from the slope in SrObs/SrEQ between each pair of points in the profile. It should be noted that the significant deviation between the observed and calculated Sr profiles occurs within a few 100 µm of the rim of the crystal. In contrast, the core has essentially reached equilibrium. A similar pattern was found in all of the crystals analysed, suggesting that the cores had resided at magmatic temperatures for sufficient time to attain equilibrium whereas the rims had not. Ba profiles were also determined but because the covariance of Ba and anorthite content is high, we chose to use Sr for all age calculations.
|
Table 7 summarizes the results for the seven crystals and gives rim ages calculated using the Sr diffusion coefficient obtained by Giletti & Casserly (1994
). The diffusion coefficient for Sr (
Sr) is strongly dependent on temperature and composition, and the temperatures assumed were derived from the MELTS modelling above. Liquid compositions from each MELTS model were recalculated on an anhydrous basis and plotted against the anorthite content of the crystallizing plagioclase in 10°C cooling steps. The best estimate of temperature was assumed to occur when (1) the model liquid composition (SiO2 was used) best matched that of the observed matrix composition, and (2) the crystallizing plagioclase composition most closely corresponded to the observed anorthite content of the rim. We also calculated minimum growth rates using measured rim widths; these range from 10–10 to 10–11 cm/s (Table 7).
|
The calculated time scales range from 4 to 168 years and are given in stratigraphic order in Table 7, showing that the ages increase upward from 16–33 to 102 years through the Crater Creek series to a maximum of 168 years in the 2050 a BP caldera-forming event and then decrease to 4–53 years in the post-caldera samples. It should be noted that these do not indicate the time since crystal growth, but instead the time since the addition of new (i.e. rim) material. Moreover, this is common to all effusive samples throughout the 10 000 years history studied here, as all samples (excepting those produced during the recent caldera-forming event) show a similar rimwards increase in local disequilibrium, and give similar diffusion time scales. Thus, some crystallization event occurs <200 years prior to the eruption of each sample, when the host magma is at a temperature between 1045 and 1185°C, with water contents between 0·5 and 2·2%. Plagioclase of An50–75 crystallizes in each case, forming rims that rarely exceed 100 µm in width. This is especially intriguing given the variety of lava and tephra compositions and eruption styles, and we envisage two possible scenarios in which young rims might be formed.
In the first, crystallization of rim material takes place upon either cooling or decompression of the same magma that produced, and that hosts, the core of each crystal. The nature of the sudden jump from high to low anorthite content effectively rules out closed-system differentiation through isobaric cooling, as one would expect such a process to produce continuous normal zoning. However, changes in pressure and water activity also affect the anorthite content in plagioclase. For example, crystallization from hydrous melts produces more anorthitic plagioclase than from dry melts of a similar composition, yet whereas isothermal decompression of dry melts increases the anorthite content of the crystallizing plagioclase, similar decompression of a hydrous magma triggers the crystallization of less anorthitic plagioclase (Blundy & Shimizu, 1991). Thus, a hydrous melt may crystallize highly anorthitic plagioclase at depth, followed by low anorthite rims during decompression (normal zonation), whereas a dry melt will produce less anorthitic cores, with increasing anorthite during isothermal decompressive rim growth (reverse zonation).
The second, preferred, scenario involves mixing and, whereas large-scale mixing (hybridization) is responsible for only some of the whole-rock trends seen in the Okmok products, crystal–liquid mixing may be ubiquitous (e.g. Blundy & Shimizu, 1991). The transfer of crystals from one liquid to another will be encouraged if density contrasts exist between the crystals and their host liquids, and also by convective currents. In this model, the rims’ ages might reflect the time elapsed between entrainment of the cores from a marginal crystal mush zone and eruption (e.g. Turner et al., 2003b).
|
A WORKING MODEL FOR OKMOK |
|---|
The evolution of selected geochemical and isotopic parameters in the magmatic system at Okmok leading up to and following the 2050 years BP caldera-forming eruption is shown in Fig. 17 and illustrated schematically in Fig. 18. In the six oldest flows of the Crater Creek series, SiO2 and Zr decreased slightly with time, as do (230Th/232Th) and 87Sr/86Sr. Conversely, (238U/232Th) increases over the same interval of time (see Fig. 17). It seems probable that an influx of fresh magma was added to the chamber during this period, which was clearly more mafic than the resident magma, and was characterized by low (230Th/232Th), low 87Sr/86Sr and high (238U/232Th). We speculate that this recharge occurred immediately prior to caldera formation (Fig. 18), and that the injection of a large volume of basalt into the crust may even have acted as a trigger for that event (e.g. Sparks et al., 1977
). In the upper five flows of the series, evidence of recharge is replaced by evidence of fractional crystallization and assimilation. A sharp rise in SiO2 and Zr is matched by similar increases in 87Sr/86Sr and (230Th/232Th), and a drop in 18O (Fig. 17). Clearly, assimilation controls the isotopic shifts during this period, and large increases in SiO2 and Zr suggest that crystallization was driving at least part of the magma body towards much more evolved compositions. This period of assimilation and fractional crystallization could well account for the generation of the rhyodacitic component of the 2050 years BP event.
|
|
The 2050 years BP eruption was composed of the early rhyodacite and andesite ash-fall layers, and the later basaltic andesite ash-flow package, which makes up over 99% of the eruption by volume. It is likely that the first-erupted rhyodacite air-fall unit was generated by FCA processes in the upper crust, recorded by the uppermost flows of the Crater Creek inter-caldera sequence. The andesite fall unit was erupted next, with (238U/232Th) similar to the rhyodacite but with lower (230Th/232Th) and 87Sr/86Sr, suggesting that it may be the product of the same parental magma batch as the rhyodacite, at a less advanced stage of FCA differentiation. The isotopic data are, therefore, consistent with a model whereby the parental magmas of both fall units were resident in the Crater Creek magma chamber, prior to their eruption. The basaltic andesite flow unit represents a hybrid of at least two magma types, one of which is thought to represent a different batch of magma from the mantle with high (238U/232Th) and high (230Th/232Th).
The oldest post-caldera products, inferred to have erupted not long after the 2050 years BP event, appear to share some of the characteristics of magmas from the 2050 years BP event. (230Th/232Th) remains high, suggesting that these represent the remnants of the mixed basaltic andesite magma, the bulk of which was erupted as the 2050 years BP ash-flow unit. However, in the younger post-caldera samples (230Th/232Th) falls to even lower levels than those observed at the base of the Crater Creek series (Fig. 17) as less differentiated magmas began to erupt, reflecting the supply of magma from depth, and suggesting that the magmas did not stall for as long and experienced lower degrees of assimilation. These younger magmas also have higher 18O, indicating that they have not been modified by the assimilation of old crustal material. This suggests that the magma chamber was smaller during these effusive episodes, such that it did not contact older parts of the edifice (Fig. 18). Finally, evolved crystal rims formed in many of the products less than 200 years prior to eruption, which arguably reflects the remobilization of crystals from the chamber margins following replenishment.
|
SUPPLEMENTARY DATA |
|---|
Supplementary data for this paper are available at Journal of Petrology online.
|
ACKNOWLEDGEMENTS |
|---|
We thank John Gamble for editorial expertise, Steve Sparks and Tim Elliott for many helpful discussions, and Marlina Elburg, Jun-Ichi Kimura and Georg Zellmer for their very detailed comments, which greatly improved the final manuscript. Simon Turner is currently funded by an ARC Federation Fellowship. This is a BIG (Bristol Isotope Group) publication.
*Corresponding author. E-mail: sturner{at}els.mq.edu.au
|
REFERENCES |
|---|
Arculus RJ. Use and abuse of the terms calcalkaline and calcalkalic. Journal of Petrology (2003) 44:929–935.
Asimow PD, Ghiorso MS. Algorithmic modifications extending MELTS to calculate subsolidus phase relations. American Mineralogist (1998) 83:1127–1131.[Abstract]
Bacon CR, Druitt TH. Compositional evolution of the zoned calcalkaline magma chamber of Mount Mazama, Crater Lake, Oregon. Contributions to Mineralogy and Petrology (1988) 98:224–256.[CrossRef][Web of Science]
Beget JE, Larsen JF, Neal CA, Nye CJ, Schaefer J. Preliminary volcano-hazard assessment for Okmok volcano, Umnak Island, Alaska. Alaska Division of Geological & Geophysical Surveys Report of Investigation 2004-3 (2005) 32.
Bindeman IN, Fournelle JH, Valley JW. 9100 bp eruption of Fisher Caldera, Umnak, Aleutians: low low-18O zoned tuff, a tephronological marker. Journal of Volcanology and Geothermal Research (2001a) 111:35–53.[CrossRef][Web of Science]
Bindeman IN, Fournelle JH, Valley JW. Low- 18O tephra from a compositionally zoned magma body: Fisher Caldera, Unimak Island, Aleutians. Journal of Petrology (2001b) 42:1491–1517.
Bindeman IN, Pomonareva VV, Bailey JC, Valley JW. Volcanic arc of Kamchatka: a province with high- 18O magma sources and large-scale 18O/16O depletion of the upper crust. Geochimica et Cosmochimica Acta (2004) 68:841–865.[CrossRef][Web of Science]
Bindeman IN, Schmitt AK, Valley JW. U–Pb zircon geochronology of silicic tuffs from the Timber Mountain/Oasis Valley caldera complex, Nevada: rapid generation of large volume magmas by shallow-level remelting. Contributions to Mineralogy and Petrology (2006) 152:649–665.[CrossRef][Web of Science]
Blundy JD, Shimizu N. Trace element evidence for plagioclase recycling in calc-alkaline magmas. Earth and Planetary Science Letters (1991) 102:178–197.[CrossRef][Web of Science]
Byers FM Jr. Geology of Umnak and Bogoslof Islands, Aleutian Islands, Alaska. US Geological Survey Bulletin (1959) 1028-L:267–369.
Byers FM Jr. Petrology of three volcanic suites, Umnak and Bogoslof Islands, Aleutian Islands, Alaska. Geological Society of America Bulletin (1961) 72:93–128.
Class C, Miller DM, Goldstein SL, Langmuir CH. Distinguishing melt and fluid subduction components in Umnak Volcanics, Aleutian Arc. Geochemistry, Geophysics, Geosystems (2000) 1. paper number 1999GC000010.
Coats RR. Volcanic activity in the Aleutian Arc. US Geological Survey Bulletin (1950) 974-B:47.
Costa F, Chakraborty S, Dohmen R. Diffusion coupling between trace and major elements and a model for calculation of magma residence times using plagioclase. Geochimica et Cosmochimica Acta (2003) 67:2189–2200.[CrossRef][Web of Science]
Cribb JW, Barton M. Geochemical effects of decoupled fractional crystallisation and crustal assimilation. Lithos (1996) 37:293–307.[CrossRef][Web of Science]
Eggins SM. Origin and differentiation of picritic arc magmas, Ambae (Aoba), Vanuatu. Contributions to Mineralogy and Petrology (1993) 114:79–100.[CrossRef][Web of Science]
Eiler JM. Oxygen isotope variations of basaltic lavas and upper mantle rocks. In: Stable Isotope Geochemistry. Mineralogical Society of America, Reviews in Mineralogy and Geochemistry—Valley JW, Cole DR, eds. (2001) 43:319–364.
Ewart A, Griffin WL. Application of proton-microprobe data to trace-element partitioning in volcanic rocks. Chemical Geology (1994) 117:251–284.[CrossRef][Web of Science]
Finney BM. Magmatic differentiation at an island-arc caldera: A stratigraphically constrained multi-isotope study of Okmok volcano, Aleutian islands, Alaska. In: Ph.D. thesis (2004) University of Bristol. 296.
Fournelle JH, Marsh BD, Myers JD. Age, character and significance of Aleutian arc volcanism. In: The Geology of Alaska. The Geology of North America, Volume G-1—Pfafker G, Berg HC, eds. (1994) Boulder, CO: Geological Society of America. 723–758.
Fujikama H, Tatsumoto M. Partition coefficients of Hf, Zr and REE between phenocrysts and groundmass. Journal of Geophysical Research (1984) 89:662–672.[CrossRef]
George R, Turner S, Hawkesworth C, Morris J, Nye C, Ryan J, Zheng S.-H. Melting processes and fluid and sediment transport rates along the Alaska–Aleutian arc from an integrated U–Th–Ra–Be isotope study. Journal of Geophysical Research (2003) 108. doi:10.1029/2002JB001916.
George R, Turner S, Hawkesworth C, Bacon CR, Nye C, Stelling P, Dreher S. Chemical versus temporal controls on the evolution of tholeiitic and calc-alkaline magmas at two volcanoes in the Alaska–Aleutian arc. Journal of Petrology (2004) 44:1–17.[CrossRef]
Ghiorso MS, Sack RO. Chemical mass transfer in magmatic processes. IV. A revised and internally consistent thermodynamic model for the interpolation and extrapolation of liquid–solid equilibria in magmatic systems at elevated temperatures and pressures. Contributions to Mineralogy and Petrology (1995) 119:197–212.[Web of Science]
Giletti BJ, Casserly JED. Strontium diffusion kinetics in plagioclase feldspars. Geochimica et Cosmochimica Acta (1994) 58:3785–3793.[CrossRef][Web of Science]
Gill J. Orogenic Andesites and Plate Tectonics (1981) Berlin: Springer. 390.
Grey DM, Eichelberger JC, George RM, Finney BM. Post-caldera eruption history of Okmok volcano, Umnak Island, Aleutians, Alaska. (2000) American Geophysical Union Fall Meeting.
Grove TL, Baker MB, Kinzler RJ. Coupled CaAl–NaSi diffusion in plagioclase feldspar: experiments and applications to cooling rate speedometry. Geochimica et Cosmochimica Acta (1984) 48:2113–2121.[CrossRef][Web of Science]
Hauri EH, Wagner TP, Grove TL. Experimental and natural partitioning of Th, U, Pb and other trace elements between garnet, clinopyroxene and basaltic melts. Chemical Geology (1994) 117:149–166.[CrossRef][Web of Science]
Hawkesworth CJ, Turner SP, McDermott F, Peate DW, van Calsteren P. U–Th isotopes in arc magmas: implications for element transfer from the subducted crust. Science (1997) 276:551–555.
Heath E, Macdonald R, Belkin H, Hawkesworth C, Sigurdsson H. Magmagenesis at Soufrière Volcano, St Vincent, Lesser Antilles Arc. Journal of Petrology (1998) 39:1721–1764.[CrossRef][Web of Science]
Hinton RW. Ion microprobe trace-element analysis of silicates: Measurement of multi-element glasses. Chemical Geology (1990) 83:11–25.[CrossRef][Web of Science]
House LS, Jacob KH. Earthquakes, plate subduction, and stress reversals in the eastern Aleutian arc. Journal of Geophysical Research (1983) 88:9347–9373.
Jicha BR, Singer BS, Brophy JG, Fournelle JH, Johnson CM, Beard BL, Lapen TJ, Mahlen NJ. Variable impact of the subducted slab on Aleutian island arc magma sources: evidence from Sr, Nd, Pb, and Hf isotopes and trace element abundances. Journal of Petrology (2004) 45:1845–1875.
Johnson DM, Hooper PR, Conrey RM. XRF analysis of rocks and minerals for major and trace element analysis on a single low dilution Li-tetraborate fused bead. Advances in X-ray Analysis (1999) 41:843–867.
Johnston AD, Draper DS. Near-liquidus phase relations of an anhydrous high-magnesia basalt from the Aleutian Islands: implications for arc magma genesis and ascent. Journal of Volcanology and Geothermermal Research (1992) 52:27–41.[CrossRef]
Kelemen PB, Yogodzinski GM, Scholl DW. Along-stike variation in lavas of the Aleutian island arc: implications for the genesis of high-Mg# andesite and the continental crust. In: Inside the Subduction Factory. Geophysical Monograph, American Geophysical Union—Eiler J, ed. (2003) 138:223–274.
Knaack C, Cornelius S, Hooper P. Trace element analyses of rocks and minerals by ICP-MS. (1994) GeoAnalytical Laboratory, Department of Geology, Washington State University. http://www.wsu.edu/~geology/Pages/Services/ICP.html.
Langmuir CH. Geochemical consequences of in situ crystallisation. Nature (1989) 340:199–205.[CrossRef]
Larsen JF, Neal CA, Schaefer JG, Beget J, Nye C. Late Pleistocene and Holocene caldera-forming eruptions of Okmok caldera, Aleutian Islands, Alaska. Volcanism and Subduction: The Kamchatka region. Geophysical Monograph, American Geophysical Union (2007) 172:324–364.
Lindsley DH. Pyroxene thermometry. American Mineralogist (1983) 68:477–493.[Abstract]
Mann D, Freymueller J, Lu Z. Deformation associated with the 1997 eruption of Okmok volcano, Alaska. Journal of Geophysical Research (2002) 107. doi:10.1029/2001JB000163.
Miller DM. Petrogenesis of adjacent calc-alkaline and tholeiitic volcanoes on Umnak Island, Aleutian Islands, Alaska. In: Ph.D. thesis (1995) New York: Columbia University.
Miller DM, Langmuir CH, Goldstein SL, Franks AL. The importance of parental magma composition to calc-alkaline and tholeiitic evolution: evidence from Umnak Island in the Aleutians. Journal of Geophysical Research (1992) 97:321–343.
Miller DM, Goldstein SL, Langmuir CH. Cerium/lead and lead isotope ratios in arc magmas and the enrichment of lead in the continents. Nature (1994) 368:514–520.[CrossRef]
Miller TP, Smith RL. Two caldera-forming eruptions on Umnak Island, eastern Aleutian Islands. US Geological Survey Circular (1975) 733:45.
Miller TP, Smith RL. Late Quaternary caldera-forming eruptions in the eastern Aleutian arc, Alaska. Geology (1987) 15:434–438.
Miller TP, McGimsey RG, Richter DH, Riehle JR, Nye CJ, Young ME, Dumoulin J. Catalog of the Historically Active Volcanoes of Alaska. US Geological Survey Open-File Report 98-582 (1998) 104.
Miyashiro A. Volcanic rock series in island arcs and active continental margins. American Journal of Science (1974) 274:321–355.[Abstract]
Morgan DJ, Blake S, Rogers NW, Devivo B, Roland G, Macdonald R, Hawkesworth CJ. Timescales of crystal residence and magma chamber volume from modelling of diffusion profiles in phenocrysts: Vesuvius 1944. Earth and Planetary Science Letters (2004) 222:933–946.[CrossRef][Web of Science]
Muehlenbachs K, Anderson AT, Sigvaldsson G. Low- 18O basalts from Iceland. (1974) 38:577–588. Geochimica et Cosmochimica Acta.
Nakamura M. Residence time and crystallisation history of nickeliferous olivine phenocrysts from the northern Yatsugatake volcanoes, Central Japan: Application of a growth and diffusion model in the system Mg–Fe–Ni. Journal of Volcanology and Geothermal Research (1995) 66:21–100.
Nye CJ. Petrology and geochemistry of Okmok and Wrangell volcanoes, Alaska. In: Ph.D. thesis (1983) University of California Santa Cruz.
Nye CJ, Reid MR. Geochemistry of primary and least fractionated lavas from Okmok volcano, Central Aleutians: Implications for arc magmagenesis. Journal of Geophysical Research (1986) 91:10271–10287.[CrossRef]
Pan Y, Batiza R. Mid-ocean ridge magma chamber processes: constraints from olivine zonation in lavas from the East Pacific Rise at 9°30'N and 10°30'N. Journal of Geophysical Research (2002) 107. doi:10.1029/2001JB000435.
Pearce JA, Norry MJ. Petrogenetic implications of Ti, Zr, Y and Nb variations in volcanic rocks. Contributions to Mineralogy and Petrology (1979) 69:33–47.[CrossRef][Web of Science]
Rollinson H. Using Geochemical Data: Evaluation, Presentation, Interpretation (1993) Harlow: Longman Press. 352.
Self S. The effects and consequences of very large explosive volcanic eruptions. Philosophical Transactions of the Royal Society (2006) 364:2073–2097.
Sims KWW, Gill J, Dosseto A, Hoffmann DL, Lundstrom C, Williams R, Ball L, Tollstrup D, Turner S, Prytulak J, Glessner J, Standish J. An inter-laboratory assessment of the Th isotopic composition of synthetic and rock standards. Geostandards and Geoanalytical Research (2008) 32:65–91.[CrossRef][Web of Science]
Sisson TW, Grove TL. Experimental investigations of the role of H2O in calc-alkaline differentiation and subduction zone magmatism. Contributions to Mineralogy and Petrology (1993a) 113:143–166.[CrossRef][Web of Science]
Sisson TW, Grove TL. Temperatures and H2O contents of low-MgO high-alumina basalts. Contributions to Mineralogy and Petrology (1993b) 113:167–184.[CrossRef][Web of Science]
Sparks RSJ, Sigurdsson H, Wilson L. Magma mixing: a mechanism for triggering acid explosive eruptions. Nature (1977) 267:315–318.[CrossRef][Web of Science]
Sun S.-S, McDonough WF. Chemical and isotopic systematics of oceanic basalts: implications for mantle composition and processes. In: Magmatism in the Ocean Basins. Geological Society—Saunders AD, Norry MJ, eds. (1989) 42. London: Special Publications. 313–345.
Taylor HP Jr. The effects of assimilation of country rocks by magmas on 18O/16O and 87Sr/86Sr systematics in igneous rocks. Earth and Planetary Science Letters (1980) 47:243–254.[CrossRef][Web of Science]
Turner S, Bourdon B. Insights into magma genesis at convergent margins from U-series isotopes. In. Uranium-series Geochemistry. Mineralogical Society of America, Reviews in Mineralogy and Geochemistry—Bourdon B, Henderson GM, Lundstrom CC, Turner SP, eds. (2003a) 52:255–315.
Turner S, George R, Jerram DA, Carpenter N, Hawkesworth C. Case studies of plagioclase growth and residence times in island arc lavas from Tonga and the Lesser Antilles, and a model to reconcile discordant age information. Earth and Planetary Science Letters (2003b) 214:279–294.[CrossRef][Web of Science]
Villemant B, Jaffrezic H, Joran J.-L, Treuil M. Distribution coefficients of major and trace elements:fractional crystallization in the alkali basalt series of Chaîne des Puys (Massif Central, France). Geochimica et Cosmochimica Acta (1981) 45:1997–2016.[CrossRef][Web of Science]
Watson EB, Green TH. Apatite/liquid partition coefficients for the rare earth elements and strontium. Earth and Planetary Science Letters (1981) 56:405–421.[CrossRef][Web of Science]
Watson EB, Ryerson FJ. Partitioning of zirconium between clinopyroxene and magmatic liquids of intermediate composition. Geochimica et Cosmochimica Acta (1986) 50:2523–2526.[CrossRef][Web of Science]
Wolfe BA, Begét JE. Destruction of an Aleut village by a catastrophic flood release from Okmok Caldera, Umnak Island, Alaska. Geological Society of America, Abstracts with Programs (2002) 34:58–2.
Wong LJ. Physical volcanology of a sub-plinian and phreatomagmatic eruption at Okmok volcano, Alaska: implications for explosive mafic volcanism. In: M.S. thesis (2004) University of Alaska Fairbanks. 117.
Zellmer GF, Blake S, Vance D, Hawkesworth C, Turner S. Plagioclase residence times at two island arc volcanoes (Kameni islands, Santorini, and Soufrière, St Vincent) determined by Sr diffusion systematics. Contributions to Mineralogy and Petrology (1999) 136:35–357.
Zellmer GF, Sparks RSJ, Hawkesworth CJ, Wiedenbeck M. Magma emplacement and remobilization time scales beneath Montserrat: Insights from Sr and Ba zonation in plagioclase phenocrysts. Journal of Petrology (2003) 44:1413–1431.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  H. S. Cunningham, S. P. Turner, A. Dosseto, H. Patia, S. M. Eggins, and R. J. Arculus Temporal Variations in U-series Disequilibria in an Active Caldera, Rabaul, Papua New Guinea J. Petrology, March 12, 2009; (2009) egp009v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||