Introduction

The textures of volcanic pyroclasts reflect complex histories of crystallization and vesiculation prior to and during magma ascent. Insight into the plumbing of the shallow magma system and history of magma ascent can be derived from groundmass glass compositions and crystal and vesicle textures of erupted pyroclasts as they respond to pressure, temperature, and composition changes in the system (e.g., Bain et al. 2019; Cashman 2004; Cassidy et al. 2015; Edmonds and Woods 2018; Giachetti et al. 2010; Shea et al. 2010; Swallow et al. 2018). Furthermore, microlite and vesicle textures record different stages of magma decompression and ascent because of differences in the kinetics of crystal and bubble nucleation and growth and the ability for bubbles to migrate out of the system. Microlite number densities (MNV) vary with decompression path and/or pre-eruptive equilibration depth (e.g., Befus and Andrews 2018; Cashman and McConnell 2005; Hammer et al. 1999; Hammer and Rutherford 2002; Klug and Cashman 1994; Shea et al. 2010; Toramaru et al. 2008) and duration of shallow storage (i.e., interevent repose time; Hammer et al. 2000). That is, microlite textures and phenocryst rim growth (Befus and Andrews 2018; Riker et al. 2015) record both pre-eruptive, or magma staging decompression, and syn-eruptive decompression paths. Vesicle textures also preserve information about pre-eruptive (Toramaru 2014) and syn-eruptive ascent, where small vesicles are attributed to syn-eruptive ascent of magma in the conduit and late-stage conduit processes (Klug et al. 2002; Klug and Cashman 1994; Mangan et al. 1993). Furthermore, vesicle number densities (VNV) can be used to estimate syn-eruptive magma decompression rates, if bubble nucleation is homogeneous (Cluzel et al. 2008; Toramaru 2006). Eruption explosivity and eruption style are intimately linked (e.g., Blundy and Cashman 2005; Cassidy et al. 2018; Degruyter et al. 2012; Jaupart and Allegre 1991; Melnik and Sparks 1999; Sparks 1997; Woods and Koyaguchi 1994) to the complex feedback loop of magma ascent, degassing, crystallization, and melt compositional evolution (Cashman 2004; Couch et al. 2003; Hammer et al. 1999, 2000; Kirkpatrick 1981; Sparks 1997). By combining groundmass glass composition and vesicle and microlite textural information, we can learn more about the shallow magma storage and ascent, which strongly control eruption style.

Past textural studies of eruptive products document variations in microlite and vesicle size, shape, and abundance that are attributed to differences in pressure/temperature pathways (e.g., Cashman 2004; Gardner et al. 1998; Hammer et al. 1999; Taddeucci et al. 2004; Wright et al. 2012), where bulk composition is held constant. However, subtle compositional heterogeneity is found in the intermediate-composition eruptive products of many stratovolcanoes, especially those produced during single short-lived explosions (e.g., Trident (Coombs et al. 2000), Augustine (Larsen et al. 2010; Roman et al. 2006), Redoubt (Coombs et al. 2013; Wolf and Eichelberger 1997), Ruapehu (Pardo et al. 2014), and Soufrière Hills (Humphreys et al. 2010)).

Despite the occurrence of modest compositional heterogeneity, its effects on microlite crystallization and vesiculation during decompression have not been previously investigated in detail on a suite of co-erupted natural samples. Here, we focus on explosive eruptive deposits that span the andesitic compositional field from two sequential vulcanian explosions of the 2006 eruption of Augustine Volcano. We use variations in componentry, groundmass glass and plagioclase compositions, temperature and oxygen fugacity from Fe-Ti oxide pairs, vesicle number density (VNV), and plagioclase microlite number density (MNV) and crystallinity (ϕ) to answer two questions: (1) how do pyroclast characteristics vary between low-silica and high-silica andesites erupted during the same explosive event? and (2) for two sequential explosions, can we use variations in textures and compositions to better understand magma decompression, shallow magma storage, and changes in the magmatic system that affect eruption explosivity and style? We show that small changes in andesite whole-rock and accompanying melt SiO2 content result in variations in pyroclast density, vesicle number density, and microlite content, even among pyroclasts from a single deposit. At Augustine, this compositional diversity provides evidence for spatially discrete shallow magma storage prior to eruption, and we use our data to investigate the complex decompression pathways of co-erupted magma from this network of storage regions.

The 2006 eruption of Augustine Volcano

Augustine Volcano is located in Alaska’s Cook Inlet and is part of the eastern Aleutian volcanic arc (Fig. 1). The volcano comprises a summit dome complex surrounded by an apron of pyroclastic density current (PDC) and debris-avalanche deposits (Waitt and Beget 2009). Augustine’s historical eruptions in 1883, 1935, 1964, 1976, 1986, and 2006 each produced andesitic to dacitic fall deposits, PDCs, and lava domes (Larsen et al. 2010; Roman et al. 2006; Waitt and Begét 2009).

Fig. 1
figure 1

Distribution of eruptive deposits from the 2006 Augustine eruption (modified from Coombs et al. 2010 and Wallace et al. 2010). Inset map of Alaska shows Augustine’s location, a red dot, in the Cook Inlet of the eastern Aleutian volcanic arc. Sample locations for textural and geochemical analyses in this paper are shown with large symbols and labeled by sample number and event number (e.g., 313 (E9)). Small circles represent sample locations for which additional density and componentry analyses were completed. Note that deposits of events 9 and 10 are plotted on top of the map but in reality these deposit areas lie below the continuous and effusive phase units. Blue and yellow dashed lines and arrows indicate the deposit boundary and direction of ash-fall for explosive events 9 and 10–13, respectively (Wallace et al. 2010). See Vallance et al. (2010) for a complete description of the depositional units

Full size image

\The 2006 eruption was documented through observation flights, remote automated camera imagery, and satellite images during the eruption sequence, which allowed for distinction between individual deposits (Coombs et al. 2010; Larsen et al. 2010; Vallance et al. 2010; Wallace et al. 2010). Post-eruption sampling (e.g., Coombs et al. 2010; Vallance et al. 2010; Wallace et al. 2010) and whole-rock analysis (Larsen et al. 2010) revealed that deposits consisted of four dominant lithologies:

  1. 1)

    Low-silica andesite (LSA): whole-rock compositions are between 56.4 and 59.3 wt% SiO2 (Larsen et al. 2010) and LSA includes two pyroclast subtypes, scoriaceous and dense. LSA scoria pyroclasts are moderately vesicular and occur in various shades of brownish-gray-olive green. Dense LSA pyroclasts are glassy, angular, and medium to light gray (Vallance et al. 2010; Wallace et al. 2010).

  2. 2)

    High-silica andesite (HSA): whole-rock compositions are between 62.2 and 63.3 wt% SiO2 (Larsen et al. 2010) and HSA pyroclasts are moderately to highly vesicular and light gray (rarely white), pink, to salt-and-pepper in color (Vallance et al. 2010; Wallace et al. 2010).

  3. 3)

    Intermediate andesite: whole-rock compositions are from 58.3 to 62.2 wt% SiO2 (Larsen et al. 2010) and intermediate andesite pyroclasts are dense, medium gray in color, poorly to non-vesicular (Vallance et al. 2010; Wallace et al. 2010).

  4. 4)

    Banded andesite: whole-rock compositions are from 57.3 to 62.2 wt% SiO2 (Larsen et al. 2010) and banded pyroclasts are a combination of the above with distinct to diffuse banding (Vallance et al. 2010; Wallace et al. 2010).

Each lithology is present in each deposit type throughout the entire eruptive sequence, but componentry varies temporally and spatially among deposits (Fig. 2; Coombs et al. 2010; Vallance et al. 2010).

Fig. 2
figure 2

From 14 of the 17 sample locations, lithologic components (by mass) of the four main lapilli-size pyroclast types, low-silica andesite (LSA), intermediate andesite, banded andesite, and high-silica andesite (HSA), versus phase of eruption. Low-silica andesite is subdivided into two pyroclast types: dense (dense LSA) and scoria (LSA scoria). Augustine Volcano 2006 eruption begins with an explosive phase (events 1–13) and ends with an effusive phase. The timeline of the 2006 Augustine eruption (modified from Coombs et al. 2010) shows the 13 discrete explosive events (vertical lines) of the explosive phase, with events 9 and 10 highlighted in black; they are the focus of this study

Full size image

The 2006 eruption is separated into three distinct eruptive phases (Fig. 2; Power et al. 2006), similar in progression to other historical eruptions of Augustine Volcano (e.g., 1976, 1986; Roman et al. 2006; Waitt and Begét 2009). Phases are distinguished by variations in eruptive style, componentry of eruptive products (Coombs et al. 2010; Vallance et al. 2010; Wallace et al. 2010), and by corresponding deformation of the edifice (Cervelli et al. 2010).

Following several weeks of unrest, the explosive phase of the 2006 eruption began on January 11, 2006, lasted 17 days (January 11–28, 2006), and included 13 discrete explosions (events 1–13). The explosive phase began with two discrete explosions (events 1–2) on January 11 that produced snow-rich avalanches. On January 13–14, a series of six explosions (events 3–8) produced PDCs that traveled down the north, east, and south flanks of the volcano (Fig. 1). Between January 14 and 17, a LSA lava dome formed in the eastern half of the summit crater (Coombs et al. 2010).

On January 17, event 9 erupted through the LSA lava dome. The resulting PDC left a deposit rich in dense LSA on the southwest flank and the tephra fall deposits are distributed to the west (Fig. 1; Coombs et al. 2010; Vallance et al. 2010; Wallace et al. 2010). After event 9, a HSA lava dome is inferred to have effused from January 17 to 27 (Coombs et al. 2010; Larsen et al. 2010).

On January 27, explosive event 10 generated two PDCs. The event began with an early phase PDC that left a deposit rich in dense LSA pyroclasts on the north-northwest flank. A later phase of event 10 generated the largest volume PDC, named the Rocky Point pyroclastic flow (RPpf), that swept down the north flank with a 4.7-km runout distance (Fig. 1; Coombs et al. 2010; Vallance et al. 2010). The second PDC is more HSA dominant than the earlier PDC of event 10. This increase in HSA marked a transition in eruption style from intermittent explosions to a subsequent continuous phase, which was more HSA-dominant (Coombs et al. 2010). Following event 10, three short explosive events (events 11–13) on January 27 and 28 produced unidentified but likely minor on-island deposits. A cumulative tephra fall deposit from events 10–13 lies to the south (Fig. 1; Wallace et al. 2010).

Following the intermittent explosive phase characterized by vulcanian activity, the continuous phase of the eruption, January 28 through February 10, began with 4 days of nearly continuous block-and-ash flows to the north and northeast of the summit. The magma flux waned around February 2 and lava flow effusion lasted from February 2 to 10. Following a 23-day hiatus, the effusive phase began on March 3 with lava effusion and frequent but small block-and-ash flows. Active lava effusion ended around March 16, thus concluding the 2006 Augustine eruption (Coombs et al. 2010).

Methods

Sample collection for componentry and density measurements

In 2010, from 17 sites, we collected samples from deposits of the explosive to effusive phases of the 2006 Augustine eruption for componentry and density measurements (Figs. 1 and 2; Table S1 in supplementary information). PDC deposit samples were sieved in the field to select the 16–32 mm (lapilli) size range. For fall deposits, which are typically finer grained than PDC deposits, we hand-picked larger lapilli-size pyroclasts from the deposits. Sorted lapilli-size samples were rinsed under tap water to remove adhering matrix.

For each of the 17 samples, following the criteria of Vallance et al. (2010) and Wallace et al. (2010), the lapilli-size pyroclasts were visually sorted into one of the four lithologic components (LSA, HSA, intermediate andesite, or banded andesite). Low-silica andesite was further subdivided into scoria (LSA scoria) and dense (dense LSA) lapilli. For the remainder of the paper, we refer to the lapilli as pyroclasts.

We measured the density of 4112 pyroclasts in total for all lithological components from the 17 samples (Table S1 in supplementary information) using the density method based on Archimedes principle (Houghton and Wilson 1989). Within each lithologic component for a given sample, density measurements were made for all pyroclasts (for samples with < 100 pyroclasts in that component) or 100 pyroclasts (for samples with > 100 pyroclasts in that component; Fig. 3 and Table S1 in supplementary information).

Fig. 3
figure 3

Density histogram of 16–32 mm pyroclasts by lithologic components, sample location, and sample type for 14 of the 17 sample locations (see Table S1 in supplementary information; not plotted are samples 312, 300, and 406). Explosive phase pyroclasts have a weak to strong bimodal density distribution. Late event 10 (from explosive phase) PDC, continuous, and effusive phase pyroclasts have a unimodal density distribution

Full size image

Sample selection for textural and compositional analysis

In this study, we focus on five samples (three PDC and two tephra fall) relating to events 9 and 10 (see Fig. 1; Table 1) for textural and compositional analysis. Event 9 and 10 explosions both occurred in the presence of a pre-existing dome, are two of the larger vulcanian explosions, and produced distinguishable deposits formed, for the most part, by single explosions. The three PDC samples we examined include event 9 PDC (sample 401), early event 10 PDC (sample 405), and late event 10 PDC (sample 404; samples correspond to Expf, Expc, and RPpf, respectively; see Vallance et al. (2010) for more details). The two tephra fall samples represent event 9 (sample 313) and events 10–13 (sample 400). Sample 400 is likely dominated by event 10 compared to events 11–13 products based on the relatively longer seismic signal duration, airborne ash recorded from radar, and the largest measured lightning flashes of event 10 (Coombs et al. 2010; McNutt et al. 2010; Thomas et al. 2010). Pyroclasts from these five samples were selected from the compositional endmembers, LSA and HSA (Table 1), in order to characterize vesicle and plagioclase microlite textures and groundmass glass and plagioclase composition.

Table 1 List of samples, sample type, explosion date, and corresponding geochemical and textural data
Full size table

We selected pyroclasts for detailed study based upon measured clast density (cf. Shea et al. 2010). We assume that density modes provide information about the dominant eruptive processes of each compositional endmember, and density tails preserve information about end-member processes (Fig. 4). Throughout this paper, unless otherwise distinguished as density tails, samples are from the mode of the density distribution for each event. For the 30 selected pyroclasts (including density tails of LSA scoria and HSA for event 9), plagioclase microlite texture and groundmass glass composition were analyzed. Of these, eight pyroclasts were also analyzed for plagioclase and Fe-Ti oxide composition (event 9 and early event 10). For event 9, we quantified the vesicle texture of four pyroclasts (LSA scoria and HSA; Table 1).

Fig. 4
figure 4

Event 9 (left column) and event 10 (right column) density histograms of 16–32 mm pyroclasts by sample type and location (e.g., Fall (sample 313)). Symbols above the histogram indicate LSA scoria, dense LSA, and HSA pyroclasts selected from the mode (filled) or tail (half-filled symbols; LSA scoria and HSA only) of the density distribution for textural and chemical analysis. Symbols in the legend are arranged from low density, ρ, to high density where symbols for low-density tails have the bottom-half of the symbol filled and symbols for high-density tails have the top-half of the symbol filled. Values above symbols are the average peak value at the center of the 100 kg/m3 density bin. The symbols and color schemes are used consistently in the following figures

Full size image

Image acquisition and textures

Vesicle and plagioclase microlite textures were characterized using backscattered electron (BSE) scanning electron microscope (SEM) images. Images were obtained on a desktop ASPEX EXpress SEM at the U.S. Geological Survey (USGS) Cascades Volcano Observatory using a 20-keV electron beam. The working distance was 16–17 mm at × 139–555 magnification or 45–47 mm at × 27 magnification.

Vesicle number density

Vesicle number densities (VNV) for four event 9 pyroclasts were calculated using FOAMs software (Shea et al. 2010). For each pyroclast, the collected BSE images were comprised of an exponential nest of 4 overlapping magnifications that include (1) scan of thin section, (2) 3 BSE images at × 27, (3) 6 BSE images at × 139, and (4) 12 BSE images at × 555 (Shea et al. 2010). These BSE images were used in the FOAMS software to calculate VNV (Shea et al. 2010). Vesicle diameters range in size from 0.001 to 2 mm. At the lowest magnification images, we hand-traced vesicles and crystals larger than 0.2 mm and at least 10 vesicles were counted to ensure adequate statistical representation (Shea et al. 2010). For higher magnification images, crystals, vesicles, and redrawn broken vesicle walls were digitized by hand (Shea et al. 2010) and a minimum vesicle diameter threshold was set at 10 pixels (Liu et al. 2015; Shea et al. 2010). Vesicle number density, VNV, values are groundmass-referenced (discarding crystals > 50 μm) and vesicularity-corrected (see Shea et al. 2010 for method details).

Microlite number density

For plagioclase microlite textural characterization, 20 pyroclast thin sections were imaged representing textures from event 9, which include low- and high-density tails, and 10 thin sections of pyroclasts from event 10. The 30 thin sections were imaged either at × 555 with a resolution of 1024 × 1024 pixels or × 280 at a resolution of 2048 × 2048 pixels, resulting in pixel sizes of 0.16 μm. For the × 555 resolution, 9 images were collected and stitched together with 50% overlap using Fiji software (Preibisch et al. 2009; Schindelin et al. 2012) to produce a single image measuring ~ 1700 × 1900 pixels. In order to capture at least 500 microlites per sample (except for microlite-poor HSA samples; Table 1), plagioclase microlites and phenocrysts were hand-traced on 2 to 6 images (images > 1024 pixels) for LSA and 6 images for HSA. For samples 405 and 404, dense LSA thin-sections were digitized on only two 1024 × 1024 pixel images at 0.16 μm/pixel resolution as these samples were microlite rich.

Microlite number density (MNV) was quantified for plagioclase, the most abundant microlite phase in all samples, although subsidiary mafic microlite phases include pyroxene, trace amphibole, and Fe-Ti oxides. All plagioclase microlites and all phenocryst phases were hand-traced in each of the 2–6 SEM BSE images per pyroclast. The average length of the counted microlites range from a minimum of 1.55 μm, based on the statistical shape inaccuracies at < 10 pixels (Liu et al. 2015; Higgins 2000), to a maximum of 50 μm, based upon visual inspection of plagioclase crystals in the Augustine samples. We define the groundmass as comprising microlites and glass but excluding crystals > 50 μm and vesicles. Plagioclase microlite crystallinity, ɸ, is the fractional area of the groundmass covered by plagioclase microlites, including those cut by the image edge (Hammer et al. 1999; Riker et al. 2015). The number of plagioclase per groundmass area, MNA, were quantified by counting the hand-digitized microlites that were fully enclosed within each image and those that were cut by the left-hand and upper margin of the image. Using MNA and ɸ, we calculated the volumetric plagioclase number density MNV = (MNA)/S, where S is the characteristic crystal size, S = (ɸ /(MNA))0.5 (Hammer and Rutherford 2002).

Electron microprobe

Groundmass glass, plagioclase, and oxide pair compositions from eight event 9 and 10 pyroclasts (see Table 1) were analyzed on a JEOL JXA-8900 electron microprobe at a USGS facility in Denver, Colorado. A 15-keV electron beam at 20 nA with a 3- or 5-μm diameter beam was used to analyze matrix glass and plagioclase. Reported compositions are from single spot analyses and secondary standards were used (Table S2 in supplementary information). Alkalis were measured first to avoid significant loss by migration, but minor migration was corrected using a time dependent intensity correction with a log-quadratic (hyper-exponential) fit. Oxide compositions were analyzed with a 15-keV focused electron beam at 20 nA. All analyzed oxide pairs were touching ilmenite-magnetite pairs and analyzed spots lie within 10 μm of the contact.

Groundmass glass compositions of the remaining 22 pyroclasts from the 5 samples (includes event 9 density tails and additional density modes) were analyzed on the JEOL-JXA 8530F in the Mineral Sciences Department of the Smithsonian Institute National Museum of Natural History. A 15-keV electron beam at 10 nA with a 3- or 5-μm diameter beam was used to analyze matrix glass. Reported compositions are from single spot analyses and a secondary standard was used (Table S2 in supplementary information). In order to compare matrix glass results between the two electron microprobes, pyroclast 401LSAS9 was run on both. Replication of analyses between electron microprobes required using a log-linear fit to the time dependent intensity correction for the JEOL-JXA 8530F. For both electron microprobes, groundmass glass and plagioclase analyses with oxide totals outside 97.5–101.5 were excluded from the results.

Results

Componentry

The relative proportion of LSA, HSA, intermediate andesite, and banded andesite components varies between eruptive events and between eruptive phases (Fig. 2). The mass percentage of LSA remained at or above 49% through most of the explosive phase of the eruption. Event 9 and the early event 10 deposits have mass fractions of LSA > 49%; however, late event 10 PDC deposit has a sharp drop in the proportion of LSA to < 50% and HSA increases to > 20%. Deposits from the continuous and effusive phases contain > 50% LSA and < 20% HSA (Fig. 2). This general pattern is consistent with previously published values although specific proportions differ, as Vallance et al. (2010) found for the 8 mm size fraction a higher percentage of LSA for events 3–8 (72% versus our 49%) and during the effusive phase (85% versus our 30%). These variations in proportions are likely a combination of different analyzed size fractions, true variability within single deposits, mass fraction versus pyroclast point counts, and that visual sorting into components is subjective.

Pyroclast density

Pyroclast densities range from 440 to 2890 kg/m3 for all studied deposits of the 2006 eruption (Fig. 3 and Table S1 in supplementary information). For the 17 samples that span the entire eruptive sequence, deposits from the explosive phase generally have a bimodal density distribution and deposits from the continuous and effusive phase have a unimodal density distribution (Fig. 3). For events 9 and 10, we observe a similar overall range of densities, 444 to 2660 kg/m3, with distinct differences observed between fall and PDC deposits from the same explosion (Fig. 4(a–d)), among compositional components of the same deposit (Fig. 4(a–e)), and among different PDC deposits from the same event (Fig. 4(d, e)).

Event 9

Event 9 samples from both fall and PDC deposits exhibit weak bimodal density distributions with strong modes at fairly low densities (1200 and 1400 g/cm3) and smaller modes at higher densities (2500 and 2200 kg/m3; Fig. 4(a, c)). Event 9 deposits are dominated by LSA scoria pyroclasts. High-silica andesite pyroclasts form a unimodal density distribution and skew towards the low density peak of the cumulative bimodal density distribution, whereas LSA (scoria and dense) and the intermediate and banded andesite compositions form two distinct modes.

Whereas both fall and PDC samples from event 9 are dominated by low-density pyroclasts, fall samples have lower mode densities overall than PDC samples (1200 versus 1400 kg/m3). This is also true for both LSA scoria and HSA compositions where tephra fall and PDC modal densities are 1200 kg/m3 and 1400 kg/m3, respectively for LSA, and 700–800 kg/m3 and 1000 kg/m3, respectively for HSA (Fig. 4(a, c)).

Event 10

Similar to event 9, event 10–13 tephra fall and early event 10 PDC samples have bimodal density distributions (Fig. 4(b, d)). Based on the bimodal density distribution, similar proportion of LSA and HSA, and similar textures we assume event 10–13 tephra fall deposit is dominated by products of early event 10. Density modes for LSA scoria and dense LSA from event 10–13 fall sample are subequal at 1200 and 2400 kg/m3, respectively (Fig. 4(b)). The density modes are broadly similar to the event 9 fall modes, though LSA scoria and dense LSA modes are not subequal. Early event 10 PDC density modes are 1700 for LSA scoria and 2500 kg/m3 for dense LSA (Fig. 4(d)), which is denser than event 9 PDC (Fig. 4(c)). High-silica andesite pyroclasts in event 10–13 fall and early event 10 PDC samples are volumetrically minor and are all low density, similar to event 9.

The density distribution shifts towards a strong unimodal distribution during the later PDC of event 10 and contains a much higher fraction of HSA (Fig. 4(e)). The density mode for LSA scoria is 1600 kg/m3 and for dense LSA is 2000 kg/m3. The density mode for HSA is 1600 kg/m3. As with event 9, the LSA scoria and HSA mode of event 10–13 fall is lower than the respective modes for the two PDC deposits. Though late event 10 PDC has a unimodal distribution, the density mode for LSA scoria and HSA is similar to early event 10 PDC but the mode for dense LSA is lower (Fig. 4(d, e)).

Event 9 vesicle number densities

By compositional category, event 9 pyroclasts from the fall deposit have higher VNV than pyroclasts from the PDC deposit (Table 1 and Fig. 5). For HSA, the event 9 fall pyroclast has a VNV of 1.6 × 107 mm−3 whereas the PDC pyroclast has a lower VNV of 4.1 × 106 mm−3. For LSA scoria, VNV are overall lower than HSA for both fall, 3.1 × 106 mm−3, and PDC, 1.9 × 106 mm−3.

Fig. 5
figure 5

Vesicle number density (VNV) versus vesicularity of four pyroclasts from Augustine Volcano event 9 explosion. Red symbols are LSA scoria and blue symbols are HSA. Included for comparison are additional VNV data from two vulcanian explosions in 1997 from Soufrière Hills Volcano (Giachetti et al. 2010) and a single vulcanian explosion in 2010 from Soufrière Hills Volcano (Burgisser et al. 2019). On the right-hand side are example grayscale images digitized from BSE SEM images to calculate vesicle abundance

Full size image

Microlite number density and crystallinity

Pyroclasts from events 9 and 10 show broad ranges of plagioclase microlite number density, MNV, 10(4.6–6.7) mm−3, and microlite crystallinities, ɸ, 0.02 to 0.39 (Figs. 6 and 7). Although crystallinities of some LSA and HSA pyroclasts overlap, overall LSA trends to higher and HSA to lower crystallinities, consistent with previous qualitative observations (Larsen et al. 2010). LSA pyroclasts have plagioclase MNV of 10(5.5–6.7) mm−3 and ɸ of 0.16–0.39. High-silica andesite pyroclasts have plagioclase MNV of 10(4.6–6.2) mm−3 and ɸ of 0.02–0.24. We find that plagioclase MNV and ɸ of low- and high-density tails for pyroclasts from event 9 are more variable, with MNV of 10(5.5–6.4) mm−3 and ɸ of 0.18–0.28 for LSA and MNV of 10(4.6–6.2) mm−3 and ɸ of 0.01–0.35 for HSA (Fig. 6). Microlite MNV generally increase from event 9 to early event 10 and are highest within pyroclasts from late event 10 PDC. In contrast, there is no consistent change in crystallinity with time.

Fig. 6
figure 6

Plagioclase microlite number density (MNV) versus crystallinity for LSA and HSA pyroclasts. Generally MNV increases from event 9 to late event 10. Additional textures found in thin section are denoted with pentagon (silica phase) and angled line (broken plagioclase phenocrysts) symbols. Inset cartoon is a simplified illustration of previous experimental and textural work (e.g., Cashman and McConnell 2005; Clarke et al. 2007; Couch et al. 2003; Hammer and Rutherford 2002; Martel and Schmidt 2003; Miwa et al. 2009; Shea and Hammer 2013; Wright et al. 2007) where MNV increases with increasing decompression (dP/dt) or temperature (dT/dt) rates and crystallinity increases with duration of shallow storage

Full size image
Fig. 7
figure 7

Digitized BSE SEM images with hand-traced plagioclase microlites in gray, groundmass glass in white, and vesicles, excluded phenocrysts, and excluded broken plagioclase in black. Rows correspond to deposit type from eruptive events and columns correspond to high-silica andesite (HSA; blue outline), low-silica andesite scoria (LSA scoria; red outline), and dense low-silica andesite (dense LSA; orange outline)

Full size image

Pyroclasts from event 9 to late event 10 vary in groundmass texture in two additional ways: (1) the presence of a silica phase in three HSA pyroclasts (one a high-density-tail) and all dense LSA pyroclasts and (2) the abundance of broken crystals in two HSA and five LSA pyroclasts (Fig. 6). All dense LSA pyroclasts contain variable amounts of amorphous silica phase (inferred by the lack of crystalline habit in BSE-SEM images), in the groundmass. Euhedral silica phase (inferred by crystalline habit in BSE-SEM images) is present in the groundmass of one high-density-tail HSA from event 9 (Fig. S1 in supplementary information). Amorphous silica phases are present in HSA pyroclasts from the early phase of event 10 (Fig. S1 in supplementary information). The groundmass of event 10 pyroclast 400DLSA6 is sufficiently devitrified that glass analysis or microlite crystal counts were not possible. LSA and HSA pyroclasts with ɸ above 0.10 or MNV above 105.9 mm−3 are more likely to contain clearly broken plagioclase phenocrysts crystals that are < 50 μm (Fig. 6), based on truncation of concentric chemical zones (Fig. S1 in supplementary information). Early event 10 pyroclasts contain the most abundant broken plagioclase crystals (Table 1) in both LSA and HSA. Broken plagioclase crystals are excluded from the MNV and ɸ calculations.

Glass and plagioclase compositions

Individual electron microprobe analyses of groundmass glass compositions are presented (Fig. 8, Table S2.1 in supplementary information) such that each pyroclast is represented by 4 to 16 compositional probe analyses. A majority of the pyroclasts have a relatively restricted range of groundmass glass compositions. Taken as a whole, LSA pyroclast groundmass glass compositions range between 64.5 and 78 wt% SiO2; HSA compositions are less variable with a range between 72 and 79 wt% SiO2. In general, groundmass glass K2O contents increase as SiO2 increases, whereas Al2O3, CaO, MgO, FeO, and TiO2 contents decrease with increasing SiO2 (Fig. 8). However, groundmass compositions of LSA and HSA are not colinear, showing that they do not represent a single liquid line of descent. Furthermore, within each LSA and HSA compositional suite the groundmass glass compositions of pyroclasts are not colinear (Fig. 8(a–d)). For example, all HSA pyroclasts do not follow a single liquid line of descent as illustrated by the early event 10 HSA samples that widely vary in K2O at the highest SiO2 contents (with amorphous silica; Fig. 8(b)).

Fig. 8
figure 8

(a–d) Groundmass glass compositions of Augustine 2006 pyroclasts such that each pyroclast is represented by 4 to 16 compositional probe analyses. Error bars in each panel represent standard deviation of secondary standards. (e) Whole rock (small symbols; Larsen et al. 2010) and groundmass glass composition (this study) of 2006 pyroclasts. Note that the glass compositions cannot be the products of fractionation of a single crustal assemblage (dashed ellipses show composition space of Augustine phenocrysts). (a–e) Pyroclast 401LSAS9 (highlighted in (a)) was analyzed on both the U.S. Geological Survey (open red square) and the Smithsonian Institution (closed red square) electron microprobe. (b) Outline of groundmass glass compositions corresponding to event 9, early event 10, and late event 10, with a particular focus on the trends of modal density pyroclasts (density tail compositions are partially transparent)

Full size image

Groundmass glass compositions of LSA and HSA pyroclasts vary between consecutive eruptive events (Fig. 8(b)). Groundmass glass compositions become more homogeneous through the studied eruptive sequence, from event 9, the most compositionally diverse, to early event 10 to late event 10. This change is most pronounced at the low silica end of the compositional range (minimum SiO2 contents vary from 64.5 to 71 to 75 wt% SiO2 for events 9, early 10, and late 10, respectively). In contrast, HSA glass compositions for event 9 and late event 10 have no clear progression in major element composition through the eruption sequence, but early event 10 HSA is the most evolved and contains the minimum and maximum wt% K2O compositions (Fig. 8(b)).

Microlite and phenocryst plagioclase compositions (Fig. 9; Table S2.2 in supplementary information) overlap for event 9 and early event 10 (late event 10 PDC sample was not analyzed). Plagioclase phenocrysts range in anorthite composition from An48–84 and An44–84 for LSA scoria and HSA pyroclasts, respectively. Event 9 LSA scoria and HSA and early event 10 LSA scoria have similar microlite and phenocryst rim compositions of An48–68; however, early event 10 HSA microlite and phenocryst rim composition is more albitic with An20–56. Early event 10 fall HSA pyroclast, which has the highest groundmass K2O (Fig. 8), contains the most albitic plagioclase from An20–44 (Fig. 9).

Fig. 9
figure 9

Anorthite composition from electron microprobe analysis (a) plagioclase microlites and phenocryst rims with An contents ranging from An20–68 and (b) plagioclase phenocrysts and cores with An contents ranging from An44–84. Histogram bins are 4 An (mol%). (a) A single HSA pyroclast from sample 400 (Fall) has low An (mol%) plagioclases that are high in SiO2, K2O and low Al2O3, and also contains a silica phase and high K2O groundmass glass. Note the colors in this figure vary slightly from the colors used for the symbols in previous figures

Full size image

Fe-Ti oxide compositions

Compositions of touching Fe-Ti oxide pairs in the groundmass and as inclusions in phenocrysts were analyzed and used to calculate temperature and oxygen fugacity (fO2). Pairs that contained exsolution lamellae or were zoned were not analyzed, as was the case for HSA pyroclasts from PDC deposits. Thus, we present analyses of only LSA scoria pyroclasts from PDC deposits and LSA scoria and HSA pyroclasts from fall deposits. For the 42 pairs that passed the Mg/Mn partitioning test for equilibrium (Bacon and Hirschmann 1988), temperature and fO2 were calculated using the geothermometer of Ghiorso and Evans (2008). Temperatures of 38 oxide pairs range from 868 to 1118 °C (Fig. 10; Table S2.3 in supplementary information), whereas 4 oxide pairs with anomalously high temperatures (> 1150 °C) were marked as outliers. Oxygen fugacity is consistently between 0.96 and 1.45 log units above the Ni-NiO buffer with one outlier at NNO + 0.58.

Fig. 10
figure 10

Oxygen fugacity (fO2) versus temperature for touching Fe-Ti oxide pairs from 2006 Augustine pyroclasts, as determined by the geothermometer from Ghiorso and Evans (2008). Data are from event 9 and early event 10 (this study) and previous studies of Augustine 2006 erupted products. Coombs and Vazquez (2014) pairs are from HSA pyroclasts from late event 10 PDC and the effusive phase and the pairs from Larsen et al. (2010) encompass all eruptive phases and all lithologies except HSA. Larsen et al. (2010) values are newly calculated using Ghiorso and Evans (2008) and differ from temperatures and fO2 originally published, which were calculated using QUILF (Andersen et al. 1993). Four outlier temperature estimates for LSA (> 1150 °C) are not plotted

Full size image

Fe-Ti oxide temperatures for event 9 are consistent with previously published Augustine 2006 Fe-Ti oxide data (Fig. 10; Coombs and Vazquez 2014; recalculated Larsen et al. 2010). For HSA, event 9 fall temperatures are similar to and within error of previously published HSA temperatures, 877 ± 8 °C (pyroclasts from late event 10 PDC deposit and effusive phase deposit; Coombs and Vazquez 2014). Although our results are consistent with the recalculated temperatures of Larsen et al. (2010) using the Ghiorso and Evans (2008) geothermometer, our results indicate a more reduced system, NNO + 1.4, than the published results of Larsen et al. (2010) which are based off the QUILF geothermometer (Andersen et al. 1993). Our results for HSA from event 10 fall yield higher temperatures (937 ± 13 °C) and lower fO2 (NNO + 1.2; Table 1; Fig. 10) than event 9.

With the exception of two LSA scoria oxide pairs that overlap with HSA temperatures, LSA scoria oxides yield overall higher temperatures than HSA, ranging from 990 to 1120 °C (average 1044 ± 40 °C without the two low-T pairs) and are consistent with the recalculated temperature data of Larsen et al. (2010) who examined all eruptive phases and all lithologies except HSA (Fig. 10). Individual LSA scoria pyroclasts have substantial temperature ranges of at least 50 °C. Temperatures overlap between event 9 and early event 10; however, early event 10 pyroclasts are on average more reduced than those from event 9 (∆NNO + 0.99 and ∆NNO + 1.25, respectively).

Discussion

Magma storage

Previous work sheds light on crustal storage and magma mixing prior to the 2006 Augustine eruption. Compositional endmembers LSA and HSA originate from at least two distinct resident magma sources (De Angelis et al. 2013) and are unrelated by simple crystal fractionation (Larsen et al. 2010). The rare occurrence of quenched mafic inclusions (51.8–57.1 wt% SiO2) in the 2006 deposits indicates intrusion of basalt magma prior to the eruption (Steiner et al. 2012). Previously, the origin of LSA was proposed to be a hybrid magma of intruding basalt and HSA (Larsen et al. 2010). However, the prevalent tschermakitic amphibole of the intruding mafic magma is rarely found in LSA (De Angelis et al. 2013), prompting the interpretation that LSA magma likely resided in a shallow reservoir for decades prior to eruption, similar to HSA magma (De Angelis et al. 2013; Thompson 2011).

The different oxygen fugacities and temperatures of LSA and HSA (Fig. 10) provide further evidence for separate shallow reservoirs of LSA and HSA. The average LSA equilibrium temperature is 1028 ± 44 °C (n = 26; disregarding n = 4 > 1150 °C) and the average HSA equilibrium temperature from this study and Coombs and Vazquez (2014) is 897 ± 31 °C (n = 24). Additionally, HSA oxygen fugacity is about 0.2 log units above LSA.

Based on geophysical data and melt inclusion pressures, LSA and HSA likely originate from reactivated shallow crustal reservoirs at depths of 4–6 km (Cervelli et al. 2010; Power and Lalla 2010; Webster et al. 2010). Geophysical data constrain the top of a crustal magma body at 4 km beneath the surface based on earthquake locations (Power and Lalla 2010) and at 5 km below the surface based on a modeled geodetic deflation signal (Cervelli et al. 2010). Melt inclusions from Augustine 2006 LSA and HSA pyroclasts indicate maximum entrapment pressures around 150 MPa (approximately 6 km) for 4.1–4.6 wt% H2O, in agreement with geophysical depths (Webster et al. 2010). Recent phase-equilibria experiments using HSA pyroclasts from late event 10 PDC (RPpf) deposit suggest crystallization pressures of 120–170 MPa (corresponding to a storage depth of 4.6–6.6 km), temperatures of 840–870 °C consistent with event 9 oxide-pair geothermometry results (Fig. 10), and microlite and phenocryst compositions of An50–70 that are within the An range observed for event 9 HSA plagioclase (Fig. 9; De Angelis et al. 2020).

Heterogeneous groundmass glass composition is consistent with LSA and HSA magmas originating from separate crustal reservoirs; furthermore, the groundmass glass heterogeneity within LSA and HSA magmatic end members suggests more than two separate crustal reservoirs. The lack of a single liquid line of descent for either LSA or HSA groundmass glasses suggest that melt within each respective endmember is not related by crystallization alone but experienced different pressure-temperature histories in separate reservoirs (cf. Cashman and Edmonds 2019). For example, LSA scoria groundmass glass compositions lie along a single crystallization trend (Fig. 8(a–d)) but are not colinear with dense LSA (Fig. 8(a–d)) compositions from event 9 and early event 10. A subset of dense LSA have high-Cr whole-rock contents (De Angelis et al. 2013; Larsen et al. 2010) and a wide amphibole compositional range that may have resulted from HSA mixing with a mafic magma, whereas low-Cr dense LSA are most likely a product of more extensive degassing of a remnant LSA magma body (De Angelis et al. 2013). The lack of a single LSA liquid line of descent and bimodal Cr compositions suggest the 2006 eruption erupted multiple LSA magmas, and that degassing-induced crystallization alone did not drive melt evolution.

Additionally, HSA groundmass glass lacks a single linear compositional trend between SiO2 and K2O (Fig. 8(b)) pointing to eruption from more than one HSA magma body. We hypothesize that K2O variation results from different HSA parent compositions, where low-K2O HSA (Fig. 8(b)) could represent older shallowly stored magma or mobilized cumulate material (Larsen et al. 2010; Loewen et al. 2018). Ongoing tephra chronology of Augustine Volcano has found that incompatible element concentrations, including K2O, increased through time over the past 7 ka (Loewen et al. 2018). Nearby Redoubt Volcano has a similar temporal progression where incompatible element concentrations increased at specific SiO2 concentrations (Beget and Nye 1994). The explosive phase HSA 2006 eruptive products were previously found to contain trace elements, particularly LREE, similar to compositions of prehistoric Holocene lavas (Larsen et al. 2010). Larsen et al. (2010) suggested this divergence of composition from pre-2006 historical lavas could indicate some HSA is remobilized from the mid-crustal reservoir and is at least hundreds of years old. Furthermore, Larsen et al. (2010) also found an explosive phase low-K2O HSA pyroclast they suggest could represent older cumulate material. Experiments on 2006 HSA that replicate low-K groundmass glass show depletion can be caused by small amounts of biotite crystallization (De Angelis et al. 2020). Regardless of whether variation is a result of shallow-level crystallization or deeper source processes that appear to change through time, the presence of at least three K2O-SiO2 trends in groundmass glass data leads us to suggest there were multiple resident HSA reservoirs of various ages that were tapped and co-erupted in the 2006 Augustine eruption.

Heterogeneous whole rock, groundmass glass, and amphibole compositions provide evidence of multiple compositionally isolated reservoirs of LSA and HSA magma of various ages at Augustine Volcano (De Angelis et al. 2013; Larsen et al. 2006; Loewen et al. 2018; Harris 1994; Fig. 11). Building upon recent conceptual models for the 1986 (Roman et al. 2006) and 2006 eruptions (De Angelis et al. 2013; Larsen et al. 2010), physical separation of these bodies preserves compositional heterogeneity (Roman et al. 2006; De Angelis et al. 2013; Fig. 11) and prohibits homogenization of the magma through mixing within the eruptive sequence, which can occur as an eruption progresses (cf. Redoubt Volcano (Wolf and Eichelberger 1997)). The multiple magma reservoirs are proposed as a network of complex and interconnected dikes through which the heterogeneous magma ascends and continues to mix towards the vent (De Angelis et al. 2013; Roman et al. 2006; Fig. 11). Furthermore, we add that the discrete explosions of the explosive phase may have tapped different dikes that connect multiple magma reservoirs with the single shallow conduit and vent (Fig. 11).

Fig. 11
figure 11

Conceptual model of the plumbing system beneath Augustine Volcano and magma staging during the 2006 eruption. Event 9, early event 10, and late event 10 each tap different shallowly stored resident magma bodies of HSA and LSA. Shallow resident magma bodies vary in age from old to young for both HSA and LSA. Event 9 and early event 10 explosions have slightly different decompression pathways to staging magma in the upper conduit and products are erupted from the upper 2.5 km of the conduit. Late event 10 has little pre-eruptive staging and the majority of the eruptive volume is from the shallow magma storage zone. Note that the color of the various aged magma bodies does not necessarily correspond to the symbol colors used in the previous figures

Full size image

Magma staging prior to eruption

Pyroclast density distributions may provide proxies for the pre-eruptive staging depth distribution of magma prior to eruption (cf. Wright et al. 2007; we refer to all pre-eruptive magma decompression and residence above the storage zone as “staging”), and groundmass glass compositions provide information about decompression path (e.g., Blundy and Cashman 2001; Geschwind and Rutherford 1995; Hammer and Rutherford 2002). Density of event 9 and 10 pyroclasts correlates with the groundmass glass silica content (Fig. S2a in supplementary information), where denser pyroclasts are more evolved and experienced a greater degree of crystallization and gas loss (e.g., Bain et al. 2019; Cashman and Blundy 2000; Gardner et al. 1998; Wright et al. 2007). If degassing is the main driver of groundmass crystallization, then this correlation indicates that the most degassed samples have also experienced the greatest degree of outgassing. That is, the most crystalline samples represent the most degassed, shallowest (dome or shallow plug) material (Giachetti et al. 2010; Wright et al. 2007). Vulcanian or dome-forming eruptions often have bimodal density distributions where the bimodality reflects varying degrees of pre-eruptive outgassing across a range of staging depths, and strombolian, subplinian, or phreatomagmatic eruption styles tend to produce pyroclasts with more unimodal distributions (Mueller et al. 2011).

Like other vulcanian eruptions, event 9 and early event 10 produced PDC and tephra fall deposits with bimodal density distributions (Fig. 4; discussion of differences between PDC and tephra fall density distributions from the same event is beyond the scope of this paper). Denser pyroclasts were staged near or at the surface, consistent with the presence of surface domes prior to explosions. Additionally, dense LSA and a few HSA pyroclasts have silica phase both in the groundmass or as a vapor-phase precipitate that is consistent with residence at shallow depths (e.g., Blundy and Cashman 2001; Cashman and Blundy 2000; Couch et al. 2003; Preece et al. 2016). Lower-density pyroclasts originated deeper in the conduit or upper storage system. Early event 10 has the highest proportion of dense LSA, suggesting that the relative erupted volume of outgassed magma in the dome and upper conduit was greater than in event 9. Densification of the upper conduit may have decreased permeability and inhibited efficient outgassing deeper in the conduit (Cashman and McConnell 2005; Clarke et al. 2015; Giachetti et al. 2010), which allowed degassing to build pressure beneath event 9 and event 10 domes until it overcame the structural strength of the dense cap to initiate explosion (Clarke et al. 2015).

Eruption

Vesicle number density

The use of density as a proxy for staging depth and time suggests that vesicles record syn-eruptive ascent path. We quantified VNV for four event 9 pyroclasts and can only infer syn-eruption decompression rates for this event. LSA and HSA pyroclasts have VNV of 106 to 107 mm−3, within one order of magnitude difference, and are consistent with nucleation and growth due to similar syn-eruptive conduit decompression rates. Augustine vulcanian event 9 VNV are similar to andesite pyroclasts from three Soufrière Hills Volcano vulcanian eruptions, two in 1997 and one in 2010 (Burgisser et al. 2019; Giachetti et al. 2010; Fig. 5). Previous calculations of  Soufrière Hills VNV correspond to decompression rates roughly between 1 and 5 MPa/s (Cassidy et al. 2018; Shea 2017). We assume based on the similar composition and VNV that Augustine event 9 syn-eruptive decompression rates are similar.

Constraints from groundmass textures

Magma decompression and ascent history are linked with eruption style and explosivity (e.g., Cashman and McConnell 2005; Hammer et al. 2000; Scandone and Malone 1985); hence, one might expect products of similar eruptions to have similar textural characteristics. As determined by experimental studies (e.g., Brugger and Hammer 2010; Cichy et al. 2011; Couch et al. 2003; Geschwind and Rutherford 1995; Shea and Hammer 2013) and natural volcanic samples (e.g., Cashman and McConnell 2005; Hammer et al. 1999, 2000; Martel et al. 2000), plagioclase MNV values correlate with water exsolution rate and therefore provide a proxy for magma decompression rates. For plagioclase MNV versus microlite crystallinity, there is a clear distinction in log MNV values between andesitic vulcanian and plinian/subplinian eruptions such that the more explosive, faster decompression eruptions produce pyroclasts with higher MNV values, especially at ϕ > 0.10 (Fig. 12). Vulcanian explosions from the Augustine 2006 eruption have plagioclase MNV values similar to those from other andesite vulcanian eruptions and are an order of magnitude or more below most products of subplinian to plinian andesite eruptions (Fig. 12).

Fig. 12
figure 12

Plagioclase MNV (mm−3) versus crystallinity for Augustine 2006 event 9 and event 10 pyroclasts compared to plagioclase microlite data sets of andesite vulcanian to plinian eruptions from around the world. Bulk andesite compositions show a strong textural correlation to eruption style where plagioclase MNV and crystallinity for vulcanian explosions cluster together and subplinian to plinian eruptions cluster together at higher MNV values

Full size image

In addition to textural differences linked to eruption style, quantitative crystallinity differences exist between HSA and LSA pyroclasts from Augustine Volcano, with HSA exhibiting overall lower values of crystallinity (Fig. 6). This is consistent with observations of Augustine 1986 and 2006 eruptive products where higher silica content pyroclasts are less crystalline (Gardner et al. 2006; Larsen et al. 2010; Roman et al. 2006). To understand what produces these textural differences between HSA and LSA, we step through two possibilities: (1) LSA and HSA experienced the same ascent histories from their respective discrete magma storage regions; therefore, there is a melt composition and/or temperature response to decompression that caused textural differences or (2) LSA and HSA experienced different ascent histories from their respective discrete magma storage regions that caused the textural differences. We assume that the late-stage, shallowest ascent history through the conduit was shared by pyroclasts that were emplaced during a single eruptive event. Here, we explore these two possibilities.

High-silica andesite versus low-silica andesite

First, we focus on the role of composition and/or temperature on groundmass crystallization. Late event 10 PDC samples have the most homogeneous groundmass compositions, a unimodal density distribution (Figs. 8(b) and 3(e)), and a large erupted volume (Coombs et al. 2010), which we interpret to reflect eruption from a relatively homogeneous storage depth range (at approximately 4 km depth) with little to no inter-eruptive staging. We argue that both HSA and LSA from late event 10 PDC experienced the same decompression path because of the lack of pause between the early phase and late phase of event 10 (Coombs et al. 2010). Based upon the lower plagioclase crystallinity and slightly lower MNV in HSA than in LSA (Fig. 6), we conclude there is a compositional and/or temperature control on plagioclase microlite crystallization in response to a single decompression path. This is consistent with experimental work that has shown viscosity and effective temperature constrain the relative rates of feldspar nucleation and growth (Hammer and Rutherford 2002), and more specifically, mafic melts have faster plagioclase growth rates than felsic melts (Hammer 2008).

Role of complex decompression history

We have established that there is a compositional and/or temperature control on groundmass crystallization. However, in order to examine the additional role of decompression path, we compare the more texturally and geochemically complex products of event 9 and early event 10, which we propose had multilevel staging histories compared to the more uniform products of late event 10 (Fig. 11). First, products of event 9 and early event 10 have lower and more heterogeneous plagioclase MNV and crystallinity than late event 10 (Fig. 6). If we use MNV as a proxy for decompression rate (Toramaru et al. 2008), then the lower and more heterogeneous MNV for event 9 may be the result of multiple pre-eruptive staging events before eruption (Fig. 11). We interpret the MVV range observed in event 9 (Fig. 6) to reflect syn-eruptive withdrawal from a range of depths that had variable pre-eruptive stalling times within the conduit and variable decompression paths towards the upper conduit. We attribute the differences between early and late event 10 pyroclasts to a lower average decompression rate for early event 10 because of pre-eruptive staging in the conduit, whereas magma decompressed more rapidly and uniformly from a storage region at depth during late event 10.

The greater range in crystallinity for event 9, 0.02–0.28, and early event 10, 0.11–0.39, products compared to late event 10 PDC, 0.04–0.24, is consistent with more widely varying P-T paths below the liquidus for a duration of days (up to 3 days repose for event 9 and 10 days repose for early event 10), during which magma decompressed and was staged in the shallow feeding system. The final distinct difference between events is the prevalence of broken crystals in early event 10 pyroclasts that we interpret to be the result of fairly rapid disruption and eruption of a brittle plug that had formed over the 10-day repose period. These variations all show that textural differences are a result of both compositional heterogeneity and decompression path before eruption (staging) and during eruption.

Groundmass glass composition and crystallinity

We find that the groundmass glass composition does not correlate with plagioclase microlite crystallinity (Fig. S2b in supplementary information) as would be expected for a single decompression-induced crystallization path (cf. Mount St. Helens, Cashman and McConnell 2005). Decompression crystallization as rim growth on pre-existing crystals was not quantified here and may account for some of the lack of correlation. However, qualitatively we note that LSA pyroclasts with > 72 wt% SiO2 in the melt do not appear to have more zoned phenocryst plagioclase than pyroclasts with < 72 wt% SiO2 groundmass glass. We suggest the lack of consistent correlation between groundmass glass composition and microlite crystallinity indicates that melt composition and crystallization record magma evolution and decompression of multiple separate magma bodies across shallow crustal storage levels (cf. Cashman and Edmonds 2019). We note however that pyroclasts from late event 10 PDC show the most restricted range in groundmass glass composition, consistent with textural characteristics, indicating a more uniform ascent history from a more restricted pressure interval of the heterogeneous magma storage region (Figs. 6, 8(b), and 11).

Eruption model

The geochemical and textural data presented here can be combined with previous work to reconstruct the upper-crustal magma storage and decompression processes active during events 9 and 10, and more generally, during the 2006 Augustine eruption (Fig. 11).

Six vulcanian explosions occurred on January 13–14 and erupted pyroclasts that span the andesite compositional range, producing a dense-rock-equivalent (DRE) volume of 12 × 106 m3 (Coombs et al. 2010). If the volume of the six events is divided equally and we assume a 30-by-45 m conduit (Coombs et al. 2010), each explosion tapped an evacuation depth of < 2.5 km beneath the summit and following each explosion, magma was refilled/staged in these shallow pathways.

From January 14–17, explosions paused between events 8 and 9 when LSA, intermediate and banded andesite, and only minor amounts of HSA ascended in the conduit and effused aseismically as a smooth-surfaced lava dome. The dome was first observed on January 16 and satellite data confirm the dome actively growing up until event 9, during which time magma continued to slowly ascend. The estimated DRE volume of this dome is 2.5 × 106 m3 (Coombs et al. 2010).

On January 17, event 9 lasted 4 min and 11 s, and produced an eruption plume up to 13.5 km asl (Webley et al. 2010) and a PDC down the southwest flank (Coombs et al. 2010; Fig. 1). Low-silica andesite dominated the products of this event, with lesser intermediate and banded andesite and minor HSA (Fig. 2). Event 9 pyroclasts have a weak bimodal density distribution with the strongest mode at the LSA scoria range (Fig. 4). The weak mode for dense LSA likely represents fragments of the disrupted lava dome. The dominant LSA scoria mode represents andesite that resided in the conduit and/or shallow storage region and rapidly depressurized and vesiculated during the event. The impulsive and emergent acoustic waveform observed for this event (as well as the onset of event 10) can be explained as an abrupt outward acceleration of gases and volcanic material from the vent, which may be initiated by a large overpressure, followed by extended release of a pyroclast-gas mixture (McNutt et al. 2010). This is consistent with rapid disruption of the pressurized dome or plug, followed by depressurization, vesiculation, fragmentation, and eruption of the magma from deeper in the conduit. Erupted DRE volume of event 9 deposits is 8 × 105 m3, implying evacuation of the upper 750 m of the conduit.

Event 9 textures and compositions indicate complicated upper conduit staging through multiple decompression paths. Dominant LSA exhibits a restricted crystallinity range but wider ranges in MNV (Fig. 6) and melt composition (Fig. 8(b)) that suggest variable staging depths and decompression paths. HSA pyroclasts exhibit a wide range of microlite crystallinity, MNV, and melt compositions (Figs. 2 and 3). This heterogeneity within both compositional end members (Fig. S2b-c in supplementary information) is because of the wide range of pre-eruptive staging depths. Such depth variation could be expected from a system with multiple shallow magma batches, where early eruptions may have triggered several pre-eruptive decompression pulses on unerupted magma below. The differences between HSA and LSA microlite crystallinities are a result of the compositional response to these complex decompression paths.

Following event 9, the volcano was relatively seismically quiet with only brief periods of drumbeat earthquakes, which are associated with lava effusion (Power and Lalla 2010). An overflight on January 24 showed no changes at the summit; however, in the last few days of January, satellite data showed a new lava dome had effused that was lighter in color and blockier than the previous dome and later sampling indicated it was also more silicic (Coombs et al. 2010).

On January 27, 2006, after a 10-day eruption repose, event 10 lasted 9 min and formed two distinct, sequential PDCs that are separated into early and late event 10 deposits (Coombs et al. 2010). Early event 10 produced pyroclasts with a bimodal density distribution, but unlike event 9, this distribution is dominated by dense LSA (Fig. 4), suggesting a higher proportion of degassed plug material relative to magma sourced from deeper within the conduit/magma storage system. Additionally, early event 10 is dominated by LSA with subordinate intermediate and banded andesite and minor HSA. Although limited observations suggest the dome was dominated by HSA composition, there are plausible explanations for the low HSA and high dense LSA proportions for early event 10. Perhaps the dome was more heterogenous than noted previously or early event 10 destroyed more of the dome that formed prior to event 9 than the smooth dome that formed before event 10. Alternatively, there may have been multiple vents and/or shallow conduits. Similar to event 9, an impulsive and emergent acoustic waveform was observed for event 10 (McNutt et al. 2010), consistent with rapid disruption of the pressurized dome or plug.

Early event 10 textures indicate a simpler eruption path than for event 9. LSA and HSA exhibit a similar range in crystallinity compared to event 9 but tend towards the higher end of MNV. Melt compositions for early event 10 are more restricted in silica content than event 9 but there are at least two populations of HSA magma (in K2O content, Fig. 8(b)). We interpret the more homogeneous MNV values and melt compositions to indicate a simpler, syn-eruptive microlite crystallization decompression path. The longest repose interval of 10 days may have created the highest pre-eruptive overpressure of the sequence if overpressure increased through time as the shallow conduit plug continued to degas and densify. The higher abundance of broken phenocrysts in early event 10 products can be explained by the rapid depressurization of an over-pressurized system that creates sufficient strain rates to increase brittle fragmentation of melt and crystals.

During the latter part of event 10, apparently without pause, the more voluminous second PDC erupted (Coombs et al. 2010). Pyroclasts from this unit have a higher proportion of HSA and a unimodal density distribution (Figs. 2 and 3) that we interpret to reflect deeper pre-eruptive origin where low groundmass microlite crystallinities indicate there was little magma staging in the conduit. The larger DRE volume of the erupted deposits for late event 10 PDC, 1.25 × 107 m3, supports a syn-eruptive withdrawal from magma bodies around 5 km depth and indicates a shift from explosions sourced in the shallow conduit. The relatively high MNV suggest rapid ascent rates during this interval. We interpret that the earlier explosive events and the early phase of event 10 explosion connected the dike system beneath Augustine and provided an open pathway to eruption and rapid depressurization to tap the deeper and more homogenous LSA and HSA residual magma bodies (Fig. 11). After event 10, the Augustine system remained open and transitioned to less explosive, continuous, and effusive phases of the eruption.

Conclusions

Textural and geochemical constraints on deposits from discrete vulcanian explosive events at Augustine Volcano indicate that prior to its 2006 eruption, multiple shallow magma bodies of heterogeneous andesite resided under the volcano (Fig. 11). Low- to high-silica andesites repeatedly ascended through shallow dikes, mixing, and mingling prior to and during each vulcanian explosion. Resultant products of these vulcanian explosions are both compositionally and texturally diverse. Early vulcanian explosions, January 11–17, tapped approximately the upper 2.5 km of the staged magma beneath the volcano. Following a 10-day hiatus, January 17–27, explosive event 10, January 27, evacuated the conduit and reached deeper into the shallow storage system tapping a more homogeneous magma body dominated by high-silica andesite (HSA). This deeper tapping of magma presaged the transition from punctuated vulcanian to more continuous eruptive behavior (Coombs et al. 2010). Our data supports an Augustine conceptual eruption model that erupts heterogeneous magma because the geometry of the system allows for multiple decompression paths and heterogeneous chemical evolution (cf. Cashman and Edmonds 2019). We propose that such an upper-crustal magmatic system is common for stratovolcanoes that erupt heterogeneous, intermediate-composition, modest-volume products in vulcanian and subplinian styles. The main conclusions and implications from our study are as follows:

  • Augustine erupted multiple compositions in each explosive event by tapping multiple isolated reservoirs.

  • For the Augustine system, composition affects microlite formation; thus, for the same decompression path, high-silica andesite (HSA) typically has lower microlite crystallinity compared to low-silica andesite (LSA).

  • For the Augustine heterogeneous andesite system, plagioclase microlite crystallinity does not correlate with groundmass glass SiO2 because of a variation in bulk composition between multiple shallow magma batches and pressure-temperature pathways (cf. Cashman and Edmonds 2019)

  • Combining Augustine textural data with a range of other bulk andesite composition volcanoes, there exists a strong textural correlation to eruption style, where plagioclase MNV and crystallinity for vulcanian explosions group together and subplinian to plinian eruptions group together at higher MNV values (Fig. 12) supporting the knowledge that magma decompression rates influence eruption style (e.g., Cashman and McConnell 2005; Hammer et al. 2000; Scandone and Malone 1985).

  • At Augustine, a shift from heterogeneous groundmass glass compositions to more homogeneous products correlates with an interconnection of ascent pathways within the shallow plumbing system. Pathway interconnection allowed for tapping of a larger magma body and a shift from punctuated, more closed system behavior towards open-system behavior fed by this deeper body. This change in connection led to a rapid explosive eruption (event 10) but also led to a subsequent shift to more effusive behavior as overpressure subsided.