Pyroclast textural variation as an indicator of eruption column steadiness in andesitic Plinian eruptions at Mt. Ruapehu
Between 27 and 11 cal. ka BP, a transition is observed in Plinian eruptions at Mt. Ruapehu, indicating evolution from non-collapsing (steady and oscillatory) eruption columns to partially collapsing columns (both wet and dry). To determine the causes of these variations over this eruptive interval, we examined lapilli fall deposits from four eruptions representing the climactic phases of each column type. All eruptions involve andesite to basaltic andesite magmas containing plagioclase, clinopyroxene, orthopyroxene and magnetite phenocrysts. Differences occur in the dominant pumice texture, the degree of bulk chemistry and textural variability, the average microcrystallinity and the composition of groundmass glass. In order to investigate the role of ascent and degassing processes on column stability, vesicle textures were quantified by gas volume pycnometry (porosity), X-ray synchrotron and computed microtomography (μ-CT) imagery from representative clasts from each eruption. These data were linked to groundmass crystallinity and glass geochemistry. Pumice textures were classified into six types (foamy, sheared, fibrous, microvesicular, microsheared and dense) according to the vesicle content, size and shape and microlite content. Bulk porosities vary from 19 to 95 % among all textural types. Melt-referenced vesicle number density ranges between 1.8 × 102 and 8.9 × 102 mm−3, except in fibrous textures, where it spans from 0.3 × 102 to 53 × 102 mm−3. Vesicle-free magnetite number density varies within an order of magnitude from 0.4 × 102 to 4.5 × 102 mm−3 in samples with dacitic groundmass glass and between 0.0 and 2.3 × 102 mm−3 in samples with rhyolitic groundmass. The data indicate that columns that collapsed to produce pyroclastic flows contained pumice with the greatest variation in bulk composition (which overlaps with but extends to slightly more silicic compositions than other eruptive products); textures indicating heterogeneous bubble nucleation, progressively more complex growth history and shear-localization; and the highest degrees of microlite crystallization, most evolved melt compositions and lowest relative temperatures. These findings suggest that collapsing columns in Ruapehu have been produced when strain localization is prominent, early bubble nucleation occurs and variation in decompression rate across the conduit is greatest. This study shows that examination of pumice from steady phases that precede column collapse may be used to predict subsequent column behaviour.
KeywordsComputed microtomography (μ-CT) Degassing Pyroclast Vesicle size distribution X-ray synchrotron
A robust understanding of the controls on Plinian column behaviour can be important for hazard mitigation. The behaviour of an explosive eruption plume determines the range and type of impact it has on the surrounding land and communities. On one hand, high, steady and sustained eruption plumes may impact wide areas through pyroclastic fallout, disrupt air traffic, or in the case of large eruptions, perturb normal climatic cycles. On the other hand, collapsing columns commonly produce deadly pyroclastic density currents (PDCs). Eruption hazards thus change with column stability, where production of eruption plumes and gravity-driven pyroclastic flows are not mutually exclusive.
Ruapehu is an active composite volcano located at the southern end of the Taupo Volcanic Zone. It comprises several overlapping craters that are progressively younger towards the south (Hackett and Houghton 1989). The most recent Plinian eruptions at Ruapehu produced tephras mapped as the ~27 to ~11 cal. ka BP Bullot Formation (cf. Donoghue et al. 1995). Pardo et al. (2012a) subdivided the Bullot Formation into six eruptive periods. Each period groups several eruption units, ranging from thin phreatomagmatic to thick sub-Plinian and Plinian fall deposits. Fluvial and aeolian deposits indicating breaks in the eruptive activity stratigraphically bound each eruption unit.
We assume that textures in the selected juvenile lapilli smaller than −4ϕ did not suffer significant post-fragmentation modifications (e.g. Klug and Cashman 1996; Mangan and Cashman 1996) and use them to infer magmatic physical processes within the conduit. By studying conduit processes in correlation with the variations in bulk and glass chemistry and in eruption column type, we propose a model to explain the variable degrees of Plinian column stability at Ruapehu.
Plinian eruptions from Ruapehu
Steady columns are exemplified by the Shawcroft (Sw) Eruptive Unit. All exposures of Sw (n = 158) exhibit a single, massive lapilli bed with deposit characteristics indicating sustained column conditions during a single climactic column phase, preceded and followed by smaller phreatomagmatic eruptions (Fig. 1). A distinctive feature of this unit is the predominance of finely porphyritic, dense to microvesicular pumice lapilli, abundant accidental lithic andesite clasts (31 vol% in the 3 to 2ϕ size fraction) with diverse textures and degrees of alteration. Subrounded, non-juvenile hypocrystalline andesitic lithics, microphaneritic diorites and accidental, metasedimentary-quartzitic siltstone clasts from the Tertiary basement (cf. Graham et al. 1990) are commonly embedded in the pumice lapilli. Although fine lithics are abundant and poorly vesicular shards with grooves are found, there is a lack of distinctive quenching cracks and stepped surfaces (cf. Büttner et al. 1999). This observation indicates excavation and erosion of the hydrothermally altered conduit walls probably promoted entry of geothermal fluids, perhaps aiding fragmentation, but a true fuel-coolant interaction did not occur. The ~13 cal. ka BP eruption discharged a bulk volume of 0.6 km3, with a column height of 29 km and a mass discharge rate (MDR) of ~108 kg s−1.
Oscillatory columns that varied in height without collapsing are represented by the ~22 cal. ka BP Mangatoetoenui (Mgt) Eruptive Unit. The lithofacies characteristics indicate four eruptive phases (Fig. 1). The Mgt eruption began with a conduit-clearing phreatic explosion, producing a thin, basal lithic bed. Subsequently, a sustained eruptive column arose, producing a fall deposit that gradually waned in intensity over time as indicated by normal grading of pumice lapilli. After a brief interval, a second buoyant and steady eruptive plume deposited another coarse-grained and massive fall bed. An uppermost fine-grained bed accumulated from the dissipating cloud. Except for the basal lithic bed, the lithic content is <22 vol% in the 3ϕ size fraction (Fig. 1). Distinction of the Mgt unit is the coarsely porphyritic, coarsely vesicular juvenile clasts with subspherical and ellipsoidal vesicles. Dense, platy-shaped to predominantly highly vesicular ash shards, exhibiting a range of vesicle shapes, indicate dry magmatic fragmentation. This eruption unit involved at least 0.3 km3 of total ejecta and a maximum column height of 22 km at an average MDR of maximum 107 kg s−1.
Collapsing columns are represented by the Oruamatua (Oru) Eruptive Unit, which was emplaced between ~13 and ~12 cal. ka BP. The thinly bedded lithofacies with sharp internal contacts given by grain-size variation suggests unsteady conditions or accumulation from various eruptive pulses. The Plinian phases are represented by thick and widely exposed pumice lapilli fall beds, usually preceded by phreatomagmatic events that produced thin, fine-grained, accretionary lapilli-rich, platy ashfall beds. Matrix-supported deposits of variable thickness and limited distribution were accumulated from coeval concentrated PDCs and indicate partial column collapse. The high lithic content (up to 30 vol% in the 3ϕ size fraction; Fig. 1), including accidental, metasedimentary clasts embedded in the pumice lapilli, and the presence of poorly vesicular, blocky and mossy ash shards with abundant stepped surfaces, indicates that external water participated in magma fragmentation. The estimated erupted bulk volume is 0.5 km3, with an ~37 km high column at an MDR of maximum 108 kg s−1.
The youngest (11.6 cal. ka BP) Okupata-Pourahu (Okp) Eruptive Unit marks a clear change in vent position. This unit is similar to the preceding Oru (3) in that it consists of pyroclastic fall-and-flow deposits (cf. Donoghue et al. 1999). However, the scarce accidental lithic fragments in the fine ash (<23 vol%) and the predominance of highly vesicular shards with tube-shaped vesicles indicate dry, magmatic fragmentation. Okp produced a minimum erupted bulk volume of 0.4 km3 and a column height of 25 km at MDR of ~107 kg s−1.
We restrict our analyses to the Plinian fallout record of Ruapehu, in order to compare variations in eruptive column stability.
Pumice clasts were sampled throughout the fallout deposits from the four eruptive units described above. For textural analyses, a total of 30 juvenile clasts ranging in size from −3 to −4ϕ were randomly picked from a single stratigraphic position within the climactic phase of each type unit, at 10–15 km from source. Clasts larger than −4ϕ as well as any sample showing vesicularity and/or crystallinity gradients were excluded to avoid the effects of post-fragmentation expansion (Thomas and Sparks 1992). No evidence of welding or vesicle collapse upon deposition was noted. Due to the lapilli size, not all samples with textural analysis have corresponding bulk and glass geochemistry (ESM-1 and Table SM 3.1).
For bulk chemical analyses, a total of 68 samples were analysed for bulk chemistry (ESM-1), of which 64 have previous textural work with gas porosity (ESM-2). In order to complete 15 to 20 analyses for each of the four type units, additional pumice clasts for bulk chemistry were randomly picked from the same positions as the clasts sampled for textural work. Major element concentrations were determined with a Siemens SRS 3000 sequential X-ray spectrometer with an Rh tube, which uses Bruker SPECTRA-plus software (V1.51). The reproducibility is ±1 % (1σ). For trace elements, 34 international standards were used for calibration and samples were analysed using pressed powder pellet with theoretical detection limits of 1–2 ppm. La, Ce and Th were measured for overlap corrections, and reproducibility is 2σ < 5 % for all elements except Ba, La, Ce, Pb and Th (ESM-1). Groundmass glass and phenocryst rim compositions were analysed at the University of Auckland with a JEOL JXA-840 electron probe microanalyser (EPMA) interfaced with a Princeton Gamma Tech Prism 2000 Si (Li) EDS X-ray detector at an accelerating voltage of 15 kV, a beam current of 800 pA and a total count time of 100 s. A 2-μm focused beam was used to analyse crystal phases, whereas a defocused beam of 15 μm in diameter was used for glass to minimize alkali loss. The KN-18 rhyolitic glass standard was used as a monitor during glass analyses. After removing points and samples where anomalous high values of Na2O, K2O, MgO and FeO pointed the interception of microlites/nanoliths within the beam, a total of 76 analytical points were obtained within 10 lapilli throughout the eruption sequence (ESM-1).
Porosity and density (gas pycnometry)
Cylindrical cores (10–20 mm in diameter) of pumice lapilli (−4ϕ and −3ϕ) were drilled, and density bulk volumes were obtained with a Micromeritics GeoPyc 1360 envelope density analyser (±1.1 % reproducibility). Skeletal (δskel, which only considers the volume of connected pores across the sample) and solid (δsol = density which excludes all porosity in the sample) densities were obtained with a Quantachrome’s gas pycnometer, using pure nitrogen (N2) as the flowing gas (±0.2 % reproducibility of gas flow). Five to six measurements were made per sample. Total porosity was calculated by following Houghton and Wilson’s method (1989), with average standard error σ = 0.3–0.5 %. Connected (φconn = pores accessible to N2 gas through an interconnected pore network) and isolated (φiso = pores inaccessible to N2 gas) porosities were obtained following Klug and Cashman’s method (1996). Corresponding data are given in ESM-2.
In order to image the 3D bubble structure within pumice clasts, we obtained X-ray microtomographic images of representative pumice textural types. Due to large computing requirements, only one sample per textural type within each eruption unit was scanned. Representative cores were selected based on modal porosity values. A total of 13 samples were analysed to span the range of textures observed. Subcores of 10 mm in diameter were cut for coarsely vesicular samples (pixel size = 4 μm) and 5 mm for finely vesicular samples (pixel size = 2 μm). X-ray imaging was carried out with a synchrotron source at the Lawrence Berkeley National Laboratory (LBNL, Ca, USA), as well as with a phoenix nanotom 180 X-ray computed microtomography (μ-CT) at Institut des Sciences de la Terre d’Orléans (ISTO; CNRS-l’Universitéd’Orléans, France), giving comparable results. 2D attenuation radiographs were reconstructed into 3D volumes and corrected for ring artefacts using Background Normalize 0.3 v and Octopus v8 software (IIC UGent, Zwijnaarde, Belgium) at LBNL, and Phoenix datos/x CT software (General Electric) at ISTO. Total of 600 to 1,767 scans of each sample were collected at LBNL during 180° rotation, using a super bend magnet source (beamline 8.3.2) with a local magnetic field of 4.37 T, a 400 mA ring current and 1.9 GeV of ring energy. Total of 2,000–2,300 scans of each sample were collected at ISTO during 360° rotation, using a tungsten filament and molybdenum target. Operating voltages were in the range of 80 to 100 keV, with currents of 50–90 nA (ESM-3).
Image rendering and processing was carried out with a combination of Avizo Fire 6.2 (Visualization Sciences Group (VSG), Zuse Institute Berlin) and 3D applications of ImageJ (National Institute of Health; Schneider et al. 2012). Subvolumes of interest were cropped from rendered volumes to produce the maximum possible dataset for image processing (600-pixel-sided cubes). A set of video files was produced (ESM-4) for qualitative analyses with VGA Studio Max 2.0 (Volume Graphics GmbH, Heidelberg, Germany), Avizo Fire 6.2 and ImageJ. In general, the heterogeneity of each sample was larger than the maximum possible computable size of the virtual subvolumes. Hence, three to four subvolumes (9–34 mm3 for Mgt; 11–18 mm3 for Sw; 6–11 mm3 for Oru; and 2–15 mm3 for Okp) were analysed, and individual results merged into a single dataset (38–68 mm3 for Mgt; 71–68 mm3 for Sw; 24–42 mm3 for Oru; and 8–226 mm3 for Okp). When individual subvolumes and merged volumes were not comparable, we describe each textural domain separately (e.g. Okp sample Ph16a-5a: 226 mm3).
Image processing and quantification was carried out using the BoneJ and 3D Particle Analyser plug-ins of ImageJ, with the following steps: (1) adjustment of brightness and contrast; (2) image filtering via median smoothing × 2.0; (3) segmentation, with greyscale threshold; (4) generation of binary images; (5) dilation, erosion and hole-filling; as well as (6) pore separation by a step-by-step watershed process to separate vesicles that were connected by fine apertures. Vesicle size distributions were calculated from whole vesicles only. Specific sample details are included in ESM-3.
For X-ray microtomography images, brightness is proportional to density and atomic number (Z); hence, pore spaces are black, plagioclase and glass are dark grey, pyroxene is light grey, and titanomagnetite is bright/white (cf. Ketcham and Carlson 2001; Degruyter et al. 2010). Clinopyroxene (Cpx), orthopyroxene (Opx) and titanomagnetite (Mt) were separable by greyscale thresholding; these were collectively defined as “mafic crystals”, and the calculated crystal number densities (Nx) were corrected for vesicularity.
Errors in vesicle calculations are introduced by the computed vesicle decoalescence process, use of small subvolumes, and the exclusion of the largest sized vesicle populations in order to improve image resolution (cf. Gualda et al. 2010a, b; Giachetti et al. 2011). Thus, 3D images were used in combination with gas pycnometry and observations in 2D backscattered electron (BSE) images.
Polished thin sections from three to six clasts representing all textural types within each unit (ESM-3.2) were imaged with 2D BSE under a scanning electron microscope (SEM) at Massey University, with a 20 kV accelerating voltage and a 55-μA beam current, at an 11.3-mm working distance. In addition, a Zeiss Sigma field emission SEM was used with an HKL INCA Premium Synergy Integrated ED/EBSD system (Oxford Instruments) at the University of Otago. Estimation of feldspar microlite crystallinity (χ) and vesicle-free microlite area number density (Nax) in 2D was completed using BSE images at × 1,000 magnification because they cannot be separated from glass using microtomography imagery. Due to the wide range in plagioclase shapes (swallowtail and acicular) and the low number of microlite counts obtained in this study (ESM-3), we followed Underwood’s method (1970) for stereological conversions based on probabilistic intersections (ESM-3.2 and Table ESM-3.3). Following the findings of Castro et al. (2003) and Mock and Jerram (2005) and considering that in this study, only χ is used to correct vesicularity, expected small deviations from real 3D data would not significantly alter our interpretations.
All datasets were merged to calculate the following: vesicle number densities (Nv), vesicle volume distributions (VVDs), vesicle size distributions (VSDs), cumulative vesicle volume distributions (CVVDs) and cumulative vesicle size distributions (CVSDs) (e.g. Blower et al. 2002; Klug et al. 2002) (Tables ESM-3.4 and ESM-3.5 in ESM-3). Nv values were corrected for groundmass crystallinity and are presented as melt-referenced Nv (e.g. Gurioli et al. 2005).
Magma compositions and magma temperature
Whole rock compositions of the studied eruptive units have overlapping SiO2 contents between 55.6 and 60.9 wt% (ESM-1). Although bulk composition is similar between units, the range of observed bulk composition varies. Clasts from the stable eruption column case (Sw) comprise the narrowest range of composition, whereas clasts from dry collapsing columns (Okp) contain the largest range of composition, extending to 3 wt% higher SiO2 than stable or oscillatory eruption column cases (Fig. 2a). By contrast, groundmass glass compositions vary greatly between the units (Fig. 2b). The steady column (Sw) lapilli bulk and glass compositions are the most uniform, with 56.5–57.9 wt% bulk SiO2 and 61.6–64.3 wt% SiO2 in the glass (anhydrous). In the oscillatory column case (Mgt), bulk composition varies from 55.7 to 57.8 % SiO2, and two groundmass compositional groups occur (Fig. 2b): (1) andesite-dacite (SiO2 59.8–65.4 wt%), for the microlite-poor textures, which includes components of mafic bulk chemistry and predominate in the first climactic phase (L-Mgt); and (2) dacite (SiO2 63.7–65.0 wt%), for the microlite-rich textures of restricted bulk composition (Fig. 2a), which predominate in the second climactic phase (U-Mgt). The collapsing columns (Oru and Okp) have the widest bulk compositional (SiO2% 56.9–59.7 in Oru and 55.6–60.9 in Okp) and textural range (Fig. 2a), including the most siliceous clasts. Samples analysed for Oru and Okp have the most siliceous groundmass glass (Fig. 2b), at 71.3–75.0 wt% (Oru) and 70.4–73.0 wt% (Okp) SiO2, which is consistent with the data reported by Donoghue et al. (1999).
The pyroxene-melt geothermometer of Putirka (2008), using phenocryst rims and groundmass glass, was used to approximate magma temperatures. Results indicate that magmatic temperatures (in °C) slightly decrease with decreasing stratigraphic position from Mgt (997 to 1,137 °C), to Sw (978 to 1,075 °C), to Oru (953 to 1,008 °C) and Okp (931 to 1,031 °C).
Pumice textures and correlation with bulk chemistry
Relative proportions of each texture in each unit (and subunit), with the corresponding average bulk, (φbulk), connected (φconn) and isolated (φiso) porosities
microlite-poor (2–4 % ± 4) foamy clasts consisting of highly to extremely vesicular pumice, with subspherical vesicles grading from (1a) fine to (1b) coarse. Plateau borders (i.e. interfaces between glass films surrounding vesicles) are smooth or ruptured, dominantly curved or “Y”-shaped and vary from a few μm to 100 μm (commonly >25 μm). Pl microlites vary from subhedral with swallowtail or spiky edges giving acicular texture (2 to 15 μm) to euhedral, equant and rare with a resorbed core (20 to 31 μm). Average bulk porosity (φbulk) ranges from 68 to 91 % and average isolated porosity (φiso) from 0.4 to 6 %. Foamy bulk SiO2 ranges from 55.9 to 57.5 % (±0.7), including the most mafic components of all units (Fig. 2c).
Microlite-poor (2–4 % ± 4) sheared clasts consisting of highly vesicular lapilli with ellipsoidal, aligned coarse vesicles. Plateau borders are straight, slightly curved, or wrinkled, typically <15 μm. Pl microlites vary from subhedral with swallowtail or spiky edges giving acicular texture (2 to 14 μm) to euhedral, equant, some with swallowtail edges and rare resorbed core (10 to 23 μm). Average φbulk ranges from 72 to 79 % and φiso from 1 to 10 %. Bulk composition mirrors those in foamy textures.
Microlite-bearing (10–12 % ± 5), microvesicular, highly vesicular lapilli with vesicles varying from subspherical, concave, refolded or flattened with pinched terminations. Plateau borders are thick (~20–25), and Pl microlites are subhedral and spiky, giving an acicular texture (1–15 μm). Average φbulk ranges from 65 to 76 % and φiso from 1 to 10 %. Microvesicular bulk composition is restricted to between 56.9 and 58.0 % (±0.3).
Microlite-bearing (10–12 % ± 5), microsheared, incipiently to highly vesicular pumice clasts where vesicles are elongated, locally aligned with microlites and deformed. Plateau borders are ~10–20-μm thick, and microlites are subhedral to spiky (<33 μm). Average φbulk ranges from 62 to 63 % and φiso from 2 to 4 %. Microsheared composition is the most restricted, only ranging 57.0–57.8 % (±0.3).
Microlite-rich (18–22 % ± 6), fibrous clasts consisting of predominantly finely vesicular and commonly colour-banded lapilli. Vesicles are stretched, distorted and separated by very thin (<3 μm) glass fibres, commonly wrinkled. Microlites vary from euhedral to subhedral (1–21 μm). Some clasts are colour-banded, where differences in colour correspond to variations in vesicle shape and size, crystal size and microlite content, but the mineral assemblage is the same in each band. Average φbulk ranges from 60 to 69 % and φiso from 2 to 3 %. Grinded, non-juvenile andesitic lithics are commonly embedded. Fibrous clasts bulk SiO2 varies from 56.9 to 59.7 % (±1 %) in the wet collapsing columns and from 57.6 to 60.9 % (±1 %) in dry collapsing columns (Fig. 2c).
Microlite-rich (20–22 % ± 6), dense clasts consisting of incipiently to moderately vesicular lapilli, sometimes colour-banded, where vesicles are distorted, refolded and terminate in a tear-shaped or V-form. Plateau borders range from 5 to 100 μm (commonly >25 μm). Pl microlites vary from subhedral (1–10 μm), with swallowtail or spiky edges giving acicular texture, to euhedral prismatic (15–30 μm). Average φbulk ranges from 28 to 52 % and φiso from 2 to 8 %. Subrounded, non-juvenile hypocrystalline, fine-grained andesitic lithics and microphaneritic diorites with rounded outlines are commonly embedded. In Oruamatua dense clasts, metasedimentary quartzitic siltstones detached from the Tertiary basement occur. Bulk SiO2 of dense clasts displays different variation between units (Fig. 2c), the greatest is shown in dry collapsing column cases, from 55.6 to 59.0 % (±1.4). Some microvesicular and microsheared clasts grade into microlite-rich dense textures, so we applied a threshold of 60 % gas pycnometry measured vesicularity to distinguish them (cf. Houghton and Wilson 1989).
With the available data, there are no clear correlations between bulk chemistry variations and textural differences within each unit (e.g. fibrous clasts show the highest and lowest bulk SiO2 in Oru; Fig. 2a). Glass compositions (Fig. 2b) reflect differences in groundmass crystallinity between units (mainly between non-collapsing and collapsing cases), with increasing K2O and decreasing CaO in more microlite-rich samples.
When plotting bulk SiO2 of individual textures, irrespective of the eruptive unit (Fig. 2c), all samples overlap, except for the fibrous clasts, which are slightly offset (>2 %) towards more siliceous bulk compositions.
Texture and porosity in relation to eruption column type
The relative proportions of each texture within each unit are shown in Table 1. Bulk porosity distributions of each deposit are unimodal and skewed, except in the steady column case (Sw), where it is nearly Gaussian (Fig. 1), with a peak at 71 ± 8 %. Gas pycnometry analyses (ESM-2) reveal that φconn correlates broadly with φbulk and that φiso is relatively low in most units (0.2 to 12 %), reaching the highest values (up to 31 %) in Mgt.
The steady column pumice lapilli (Sw) are finely porphyritic, mainly microvesicular, microsheared and dense (Table 1). Lapilli from oscillating columns (Mgt) have coarsely porphyritic textures and φiso. Sheared lapilli predominate within the normally graded deposit of the first Mgt climactic phase, with microvesicular and foamy clasts present at all stratigraphic levels. Calculated average φbulk values range from 76 ± 5 % (microvesicular) to 80 ± 6 % (foamy), with a weighted average for the unit of 79 ± 2 % (Fig. 3; Table 1; ESM-2). In the massive deposit of the second Mgt climactic phase, finely vesicular lapilli predominate, with microsheared, microvesicular and foamy pumice clasts found together (Table 1). Calculated average φbulk values within the second climactic phase range from 28 ± 18 % (dense texture) to 76 ± 6 % (foamy), with a weighted average φbulk of 66 ± 12 % for the unit (Fig. 3; ESM-2).
Pumice lapilli from collapsing columns (Oru and Okp) show the widest range of vesicle shapes and sizes. Nearly all pumice textural types are present at any stratigraphic level, including distinctive fibrous, some colour-banded, pumice clasts (up to ~40 % of Okp deposit), some with colour banding related to microlite content and vesicularity. Oru pumice lapilli are predominantly finely vesicular, showing average φbulk values between 51 ± 7 % (dense textures) and 65 ± 5 % (microvesicular textures) (Fig. 1). Weighted average bulk porosities vary stratigraphically from 57 ± 13 % (L-Oru) to 60 ± 15 % (M-and-U-Oru). Okp pumice lapilli average φbulk range from 49 ± 11 % (dense textures) to 90 ± 13 % (foamy textures), with a weighted average of 63 ± 18 % in the first climactic phase and 69 ± 17 % in the second climactic phase (ESM-2).
Textural quantification (3D image analysis)
Summary of textural parameters, by combining 2D groundmass data and 3D X-ray microtomography results on merged subvolumes of each sample
For the case of oscillating column (lower Mgt) pumice, VVDs are positively skewed (Fig. 4c, d), varying from unimodal (foamy) to polymodal (sheared), with main modes at ~800 μm. VSDs (Fig. 5c, d) show long tails of reduced slope at larger vesicle sizes. The dominantly microsheared lapilli of the late climactic phase (upper Mgt) have a unimodal and Gaussian VVD (Fig. 4e), with mode at 50 μm and a linear VSD (Fig. 5e).
All VVDs quantified for the phreatomagmatic collapsing column case (Oru) (Fig. 4f–h) are nearly unimodal and Gaussian, with modes clustered at 40–50 μm and exhibit VSD tails at larger sizes (Fig. 5f–h). For the magmatic collapsing column case (Okp; Fig. 5i–k), VVDs of microvesicular samples show variable vesicle volume fractions (Fig. 5i, j), including the lowest modal value of all units (Fig. 5i) and internal heterogeneities represented by subvolumes with variable and polymodal VVDs (Table ESM-3.4). Modal sizes are clustered around 40 μm.
Except for the Oru case, where all CVSDs are exponential, both exponential and power-law CVSDs were found within each of the studied units (Table ESM-3.5). Melt-referenced vesicle number densities (Nv-Corr) are nearly constant in all samples (1.8–8.9 × 102 mm−3), except in fibrous textures, where Nv-Corr varies from ~0.3 × 102 to 53 × 102 mm−3 (Table 2 and ESM-3.4). By multiplying the relative proportion (vol%) of each textural type within each eruption unit by its corresponding Nv-Corr, average Nv-Corr were obtained for each unit (Table 2; ESM-2). The obtained average Nv-Corr increases from 3.5 × 102 mm−3 in the deposits of oscillatory columns (Mgt) to 14 × 102 mm−3 in the fall deposits of collapsing columns (Okp) (Table 2). Highly heterogeneous fibrous textures in Oru and Okp (Table 2) result in greater diversity in Nv-Corr.
VSDs deviate from linearity at equivalent diameters (L) <10 μm (Fig. 5). However, vesicles <10 μm were not observed in thin sections of any samples, except those with fibrous textures. Average melt-referenced vesicularities measured with micro-CT are ~ equal or lower than average bulk porosities obtained by gas pycnometry (0–10 % lower). The largest deviation was found in sheared and fibrous textures. This highlights the limitation of micro-CT technique where vesicle walls are thinner (or vesicle size is smaller) than the voxel size and stereological assumptions of highly elongated vesicles. Regardless of these limitations, quantitative results presented above are consistent with qualitative and hand sample observations.
Microlite phases (<30 μm) are similar to phenocrysts, comprising plagioclase, pyroxene and titanomagnetite in all samples (Fig. 3). In lapilli from steady (Sw) to oscillating (Mgt) column types, plagioclase microlites are distinctively convolute in shape, with swallowtail morphology or acicular texture. In contrast, plagioclase microlites are euhedral to subhedral in the pumice from collapsing columns (Oru and Okp). Vesicle-free groundmass crystallinity varies between textures, increasing from foamy and sheared textures to microvesicular and dense clasts (Table 2; Table ESM-3.3).
Corrected magnetite content (including microlites and phenocrysts) varies between 0.2 and 10 vol% (Table 2; ESM-3). Vesicle-free magnetite number densities (NMt-Corr) vary from 0.4 to 4.5 × 102 mm−3 in samples with dacitic groundmass and between 0.0 and 2.3 × 102 mm−3 in those with rhyolitic groundmass (Table 2).
Bubble nucleation and growth
Vesicle textures in pyroclasts record bubble nucleation and growth processes that are influenced by magma storage conditions and decompression path (cf. Shea et al. 2010). By comparing textural information with experimental data and well-constrained natural samples, we can better constrain ascent conditions at Ruapehu. The mechanism of bubble nucleation (homogeneous or heterogeneous) affects the timing of bubble formation and thereby plays a first-order control on the degree of volatile supersaturation that magma attains during ascent. A comparison of VSDs and number densities in Ruapehu samples with compositionally similar experimental samples may be a place to start towards this end. However, the VSDs of pyroclasts are affected by combined processes of expansion, coalescence, deformation and collapse in response to the local dP/dt (cf. Cashman and Mangan 1994; Mangan and Cashman 1996; Burgisser and Gardner 2005; Polacci et al. 2008; Hamada et al. 2010). Such growth processes can mask original bubble size distributions and number densities. Therefore, we compare Nv with magnetite number densities to discriminate between homogeneous and heterogeneous bubble nucleation in Ruapehu samples with rhyolitic melts, because of the high efficiency of oxides to serve as nucleation sites at low degrees of volatile supersaturation (cf. Hurwitz and Navon 1994; Gardner and Denis 2004). Magnetite number densities are on the same order as vesicle number densities for the collapsing column type units, characterized by rhyolitic glass (Oru and Okp; Fig. 2), where the calculated NMt (Table 2) is of the order 0.2–1.4 × 102 mm−3, such that heterogeneous bubble nucleation is possible (cf. Cluzel et al. 2008). Further, Nv and Nv-Corr values are lower than the typical 1011–1016 mm−3 reported for homogeneous bubble nucleation in rhyolitic melts, implying that volatile supersaturation was not necessary for the magma to vesiculate (Hurwitz and Navon 1994; Cashman et al. 2000). These results suggest that for the 27–11 ka andesitic Plinian eruptions of Ruapehu involving clasts with rhyolitic melts (to date, only found in collapsing columns), magnetite crystals could provide effective sites for heterogeneous bubble nucleation.
This same mechanism is not likely; however, in the studied units indicating non-collapsing columns where rhyolitic melts were not found. Mangan et al. (2004) demonstrate that dacitic melts, such as the ones predominantly involved in steady and oscillatory columns, strongly wet mafic crystals so they do not serve as bubble nucleation sites, and homogeneous nucleation is more likely.
The nucleation depth and saturation pressure remain unknown because glass inclusions have only been found in antecrysts (Pardo 2012). If inclusions in antecrysts reequilibrated with their host melt (cf. Qin et al. 1992; Hauri et al. 2002), H2O contents, measured by the author using Infrared Spectroscopy (FTIR), between 3.4 and 5.4 wt% suggest minimum storage pressures of 100 to 200 MPa (Pardo 2012). Similar storage pressures and water contents have been found for recent Plinian eruptions at Ruapehu (cf. Kilgour et al. 2013).
The variable VVD and VSD reflect the complex bubble growth history of the erupting magmas. For example, the observed skewed VVD in foamy, microvesicular, sheared and fibrous textures (Fig. 4a, c, d), as well as the tails in VSD of most samples (Fig. 5a, c, d, f, h, j), are likely produced by bubble coalescence (cf. Shea et al. 2010). VVD with multiple modes, typical of shearing, some microvesicular and fibrous clasts, may be caused by shear-induced bubble deformation (Fig. 4d, i, k; cf. Larsen and Gardner 2000). Low vesicularities, typical of fibrous and dense clasts with deformed bubbles, may indicate bubble collapse (Fig. 3 dense textural type; cf. Moitra et al. 2013). In addition, textures with power-law CVSD (Table SM-3.4 in ESM-3), which include Sw-dense, Mgt-sheared, Oru-microvesicular, Okp-microvesicular and fibrous, may reflect multi-phase growth of vesicles during non-equilibrium degassing (cf. Blower et al. 2002).
The above mechanisms are less complex in steady eruption columns (Sw), where typically unimodal VVDs indicate a single growth stage and uniform degassing history. Subtle differences in VVD width reflect the local effect of coalescence. Irregular vesicle walls indicate incomplete vesicle relaxation, where local permeable paths developed forming dense textures.
In the case of oscillatory eruption columns (Mgt), the bimodal VVD and variable pumice textures reflect density gradients within the erupting magma (Figs. 2 and 3). These gradient and multiple bubble growth processes are also indicated by the coexistence of coarsely vesicular fragments with skewed and polymodal VVD and finely vesicular fragments with Gaussian VVD.
In the case of phreatomagmatic collapsing columns (Oru), bubble nucleation and growth were highly restricted by the interaction with external water. Unimodal VVDs of microsheared and fibrous textures indicate a single growth stage, whereas denser clasts reflect coalescence.
In the case of dry collapsing columns (Okp), multiple stages of bubble growth are indicated by polymodal VVDs and bipartite VSDs with steep slopes at large vesicle class sizes. The broad VVD in the Okp microvesicular and fibrous textures (Fig. 4i, k compared to Fig. 4a–h) also suggests advanced stages of vesicle coalescence and collapse.
A distinctive feature common to all fall deposits from collapsing columns is the presence of fibrous textures indicative of magma shear (cf. Stasiuk et al. 1996; Rust et al. 2003; Rust and Cashman 2004). Shear indicators along conduit walls in Ruapehu fibrous clasts include the following: stretched and aligned vesicles parallel to microlite’s longest axis orientation, abraded crystal fragments, shear bands and the presence of lithic inclusions (also present in dense clasts).
It is noteworthy that decompression rate calculations using the equation of Toramaru (2006) must be taken with care because this method was developed based on the data from 2D images. Additionally, although average decompression rates are similar between units, the inferred density gradient in Mgt and more extensively developed within Oru and Okp by the coexistence of multiple textures with progressively more variable bulk chemistry would imply the ejection of magma portions that have undergone different decompression paths and residence times in the conduit.
Factors governing column behaviour and conduit dynamics
Generally, eruption units related to collapsing columns are distinguished by the following: (1) the presence of abundant fibrous, siliceous clasts within lithofacies related to pre- or syn-collapse fallout phases and showing textures indicating multiple bubble growth stages and magma shear. (2) The greatest variability of bulk chemistry and pumice textures, where fibrous (and dense, in the dry eruption case) pumice clasts are the most compositionally heterogeneous; (3) the presence of dense and microvesicular clasts with nearly identical rhyolitic groundmass glass; (4) higher unit averaged microlite number densities (ESM-3) indicating longer shallow stalling than the steady column counterparts; and (5) potentially lower magma temperatures relative to steady column cases.
Column stability and MDRs are not directly related in Ruapehu Plinian events (Fig. 7a, b). The highest eruption columns are both steady (Sw) and collapsing (Oru) (Fig. 7b), and both contain the highest contents of dense pumice and country-rock lithics among all the studied lithofacies (Pardo 2012). Therefore, there must be additional factors controlling column stability. Internal factors controlling stability in Ruapehu include the following: (1) bubble nucleation mechanism (and timing); (2) magma parcels with high degree of shear and complex ascent history, including multiple bubble growth stages; (3) viscosity changes due to volatile loss and degassing-induced crystallization during stalling, as suggested by the coexisting microlite-rich fibrous, dense and microvesicular textures with highest groundmass crystallinity (Fig. 7c) in collapsing column cases; and (4) variable crystal fractionation within the erupting andesitic magmas, prompting an ~2–5 wt% range in bulk SiO2 and nearly identical, high-silica glass compositions in coexisting textures within each unit. This last argument (4) has to be taken with care because of the limited number of samples for which bulk and glass chemistry information is known. A compositional link to texture of pyroclasts in fall deposits is not clearly determined as has been found at Redoubt volcano, Alaska, in the pumiceous fall deposits from 2009 (Coombs et al. 2013). Although weighted average SiO2 (bulk) of Ruapehu collapsing column cases has more fibrous clasts at the high-SiO2 end of the range (Fig. 2b), the differences are incremental.
Based on the facies analyses of Pardo et al. (2012a), the transition from steady to collapsing Plinian columns occurs over an ~10-ka time period of frequent subPlinian and Plinian eruptions at Ruapehu (with time breaks of the order of hundreds of years). Here, we speculate that the recurrent eruption of magmas from the same reservoir, or a cyclically recharged reservoir at nearly the same depth, could promote the development of a physically and chemically heterogeneous conduit system with time. For example, unerupted, degassed residual magma volumes from previous injections in the upper crust could plug previous paths or progressively reduce/change the existing conduit geometry. The ejection of both juvenile and residual magmas, and the effects of shear-localization (cf. Shea et al. 2014), could explain the observed whole-rock chemistry variations (Fig. 2). Within such a cyclically recharged reservoir, viscous flow could trigger strain localization, and magma parcels with higher phenocryst content could segregate from high-shear to low-viscosity zones. Strain localization and its influence on magma rheology have been documented in other natural samples by Wright and Weinberg (2009). Similarly, it has been recently increasingly documented in Plinian pumice clasts at other stratovolcanoes (cf. Schipper et al. 2013; Shea et al. 2014).
Another possibility to explain the variability in bulk chemistry is the activation of multiple storage areas for the youngest collapsing column eruptions, which has been proposed for texturally and compositionally similar magmas at Redoubt (Coombs et al. 2013). The present understanding of Ruapehu plumbing system, based on long-term (250 ka) lava geochemistry (Price et al. 2012), comprises a plexus of dispersed reservoirs where each magma batch is “likely to have had a unique history with different sized magma storages evolving on varying time scales with a specific combination of AFC and mixing processes” (op cit). The variation observed in the ~10-ka window generating Plinian tephras is negligible in comparison to the ~coeval, highly chemically heterogeneous lavas (which range from basalt to dacite and display well-discriminated compositional fields) reported by Price et al. (2012). Variations observed here could be explained by variable phenocryst (fractionated + antecrysts) content of a texturally heterogeneous magma erupted from a cyclically recharged reservoir within the postulated complex plumbing system.
External factors likely affected column behaviour: (1) Magma-water interaction in Oru and (2) variation in conduit geometry may have occurred during eruption of Okp where a change in vent position is known. The predominance of fibrous and dense (some colour-banded) clasts in Okp may also reflect a change in geometry towards a shear-zone, dyke-like feeding system to the South.
A feedback mechanism between internal and external processes could occur in the Ruapehu system, altering the velocity profile in the conduit system, favouring complex viscous flow before fragmentation. In the case of Ruapehu andesitic Plinian eruptions, column collapse has been favoured not only by the incorporation of abundant dense fragments from the conduit margins (cf. Sable et al. 2006; Shea et al. 2011) but also by the development of extreme heterogeneity in the strain experienced by the fragmenting magma (Fig. 7d), possibly related to its higher heterogeneity after cyclical reservoir recharge, the incorporation of more fractionated magmas and the ultimate development of a dyke-feeding system under the South Crater.
If the clasts accumulating at the same time in a fall bed were simultaneously fragmented in the conduit, variable textures would represent heterogeneous physical conditions (i.e. vesicularity and crystallinity) across the conduit. A horizontal gradient in local decompression rate could produce more degassed, dense, microlite-rich magma close to conduit walls where ascent rate is slower. In the centre of the conduit, a lower shear rate (cf. Sable et al. 2006) would promote a foamy magma where vesicles grow freely and microlite crystallization could be inhibited. A rheologically heterogeneous rising magma would likely also fragment at different depths according to the local porosity/density conditions, enhancing column collapse as shown through experimental observations reported by Spieler et al. (2004).
Cross-conduit variation in ascent rate has been inferred in other andesitic eruptions (Rust and Manga 2002; Rosi et al. 2004), and for the Plinian phase of the trachytic Campanian Ignimbrite eruption (Polacci et al. 2003), Mt. Etna basaltic Plinian eruptions (Sable et al. 2006), the ~161 ka BP, rhyolitic Kos Plateau Tuff (Bouvet de Maisonneuve et al. 2009), and the 79 AD eruption of Vesuvius (Shea et al. 2012). Such a model would help explain the seemingly paradoxical discrepancy between the observation of Klug et al. (2002) that pyroclastic flow samples experienced higher dP/dt than fall samples, and Shea et al. (2012) results showing the opposite. Indeed, both may be true, where it is really the range in dP/dt that is greater (at both the high and low ends) in eruption columns that are about to collapse or are currently collapsing. Such a mechanism may help explain qualitative observations that pyroclastic flow deposits contain both isotropic, inflated pumice and a higher proportion of tube pumice than their Plinian fallout counterparts at many volcanoes worldwide (e.g. Sparks and Brazier 1982; Klug et al. 2002; Shea et al. 2012).
Lastly, variation in clast texture and the degree of pore anisotropy between pyroclastic fall versus flow deposits has been noted by others (Sparks and Brazier 1982; Klug et al. 2002; Shea et al. 2012). Our study indicates that textural variation is also observed in Ruapehu among the Plinian fallout phases of all column behaviour cases. In this context, the rheological heterogeneity across/along the conduit, correlated with coexisting juvenile and residual magmas, and shear-localization during steady phases, may contribute to the column unsteadiness and enhance its potential for subsequent collapse.
Here, we observe a correlation between the lithofacies associations indicative of eruption column behaviour during Plinian eruptions and specific textural parameters of pumice clasts from the resultant fallout, where an increase in textural heterogeneity also correlates with the bulk compositional range. Pumice from Ruapehu Plinian eruptions demonstrates evidence for variation in vesicle growth rate and distribution of shear, and spatial variability of decompression rate and degassing within the conduit during different phases of its eruptive history. This study indicates that over an ~10-ka time span, the shallow accumulation of residual magmas from previous events could favour the development of bulk magma chemical variability by differential crystal fractionation, highly heterogeneous flow and strong shear. These conditions, accompanied by the involvement of progressively more fractionated and potential lower temperature magmas, increased textural gradients across the conduit, maximum groundmass glass silica content (from 62 to 73 wt%) and groundmass crystallinity (from 4 to 20 vol%), appear to favour column collapse. We postulate that shear localization by viscous flow results in variable magma density and rheology, promoting variable decompression rates across the conduit and with time. Although not isolated processes, simultaneous eruptions of magmas varying ~5 % in bulk SiO2, magma shear and extreme gradients in vesicle size and shape during magma ascent result from processes (i.e. strain, variable decompression rates and bubble nucleation and growth mechanisms) that must be considered in combination to the factors commonly known to trigger column collapse.
This study was financed by the New Zealand Foundation for Research Science and Technology (NZ Natural Hazards Research Platform) programme, “Living with Volcanic Risk” and by a Massey University Doctoral Research Scholarship. This study would not have been possible without the help of Dr. Kate Arentsen with all the logistics behind. We thank Alastair McDowell for the assistance with CT-scanning at Lawrence Berkeley National Laboratory (California), along with Dr. Alain Burgisser and ERC grant 202844 under the European FP7, for providing the facilities at ISTO. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. Drs. Katharine Cashman, Ben Kennedy, Jim Gardner, Alan Palmer and the staff of Volcanic Risk Solutions at Massey University are gratefully thanked for the discussions during this project. Doug Hopcroft, Ritchie Sims, Ian Furkert, Bob Toes and Mike Bretherton assisted in the laboratory. The discussion here benefited greatly from reviews on an earlier version of the manuscript by Raffaello Cioni, Margaret Mangan and an anonymous reviewer. Cynthia Gardner, Thomas Giachetti and Michelle Coombs are extensively thanked for their detailed reviews.
Video showing (1) Renders of some of the textures, followed by (2) scrolls across: (a)sheared pumice type from the oscillatory column case (Lower-Mangatoetoenui). Vesicle shapes vary from subspherical to ellipsoidal and walls are smooth and thin compared to other pumice types. The field of view equals 800 × 300 pixels;(b) Microsheared texture from the steady phase of the Upper-Mangatoetoenui unit, with highly elongated and oriented vesicles. Field of view equals 800 × 370 pixels;(c) Microvesicular pumice clastfrom steady columns (Shawcroft). Large vesicles are distorted and have thick vesicle walls compared to other pumice types. Field of view equals 600 × 350 pixels;(d) Microvesicular pumice clast from phreatomagmatic, collapsing columns (Oruamatua). Vesicles are distorted and have thick walls. The field view is 600 × 400 pixels. (e) Fibrous pumice clasts from the magmatic, collapsing columns (Okupata-Poruahu). Vesicles are distorted, locally oriented and have the thinnest walls of all pumice types. (MOV 54235 kb)
- Cashman KV, Mangan MT (1994) Physical aspects of magmatic degassing II. Constraints on vesiculation processes from textural studies of eruptive products. In: Carroll MR, Holloway JR (eds) Volatiles in magmas. Rev Mineral 30:447–478Google Scholar
- Cashman KV, Sturtevant B, Papale P, Narvon O (2000) Magmatic fragmentation. In: Sigurdsson H, Houghton BF, McNutt SR, Rymer H, Stix J (eds) Encyclopedia of volcanoes. Academic Press, San Diego, pp 421–430Google Scholar
- Castro JM, Cashman KV, Manga M (2003) A technique for measuring 3D crystal-size distributions of prismatic microlites in obsidian. Am Mineral 88:1230–1240Google Scholar
- Donoghue SL, Neall VE, Palmer AS (1995) Stratigraphy and chronology of late Quaternary andesitic tephra deposits, Tongariro Volcanic Centre, New Zealand. Roy Soc NZ 25(2):112–206Google Scholar
- Mangan MT, Sisson TW, Hankins WB (2004) Decompression experiments identify kinetic controls on explosive silicic eruptions. Geophys Res Lett 31: L08605. doi:10.1029/2004GL019509
- Pardo N (2012) Andesitic Plinian eruptions at Mt. Ruapehu (New Zealand): from lithofacies to eruption dynamics. PhD-dissertation. Massey University, Palmerston NorthGoogle Scholar
- Pistone M, Caricchi L, Ulmer P, Burlini L, Ardia P, Reusser F, Marone F, Arbaret L (2012) Deformation experiments of bubble-and crystal-bearing magmas: rheological and microstructural analysis. J Geophys Res 117: B05208. doi:10.1029/2011JB008986
- Qin Z, Lu F, Anderson AT (1992) Diffuse reequilibration of melt and fluid inclusions. Am Mineral 77:565–576Google Scholar
- Underwood EE (1970) Quantitative stereology. Addison-Wesley, Cambridge, 274pGoogle Scholar