High-temperature oxidation of proximal basaltic pyroclasts, 1886 Tarawera, New Zealand

Microlite crystallization in erupting basalt can occur in the conduit, in flight, or in situ after deposition. Distinguishing the products of primary versus secondary (post-fragmentation) crystallization can be challenging in near-vent environments, but is vital for interpreting shallow conduit conditions from pyroclast textures. Here, we examine pyroclasts of the 1886 basaltic Plinian eruption of Tarawera volcano, New Zealand, to assess the roles of primary versus secondary crystallization of microlites. Lapilli and ash were selected from (a) an ultra-proximal section (T47), < 100 m from vent, which is dominated by pyroclasts derived from the Plinian jet and column margin, and (b) a medial fall deposit section (T43), 2.5 km from the fissure, which contains products from the umbrella cloud. Strong contrasts in pyroclast groundmass crystallinities exist between sections, from near-holocrystalline (90–97% void-free corrected; VFC) in T47 pyroclasts to highly crystalline (77–83% VFC) in T43 pyroclasts. Subhedral-euhedral Fe–Ti microlites (< 3 μm) are ubiquitous and abundant in T47 pyroclasts, whereas they are virtually absent in T43 pyroclasts. Olivine is present in both T47 and T43 pyroclasts but evidence of its subsolidus transformation is only seen in T47 clasts, whereas in T43 clasts olivine is fresh. Near-complete crystallization of the groundmass and subsolidus transformation of olivine in the ultraproximal T47 clasts are evidence that post-depositional modification of primary pyroclast textures occurred in ultra-proximal environments at Tarawera as a response to high residual temperatures and oxidizing conditions, aided by short transport times of relatively coarse ejecta at high accumulation rates, likely supplemented by intense vent-derived heat. Ultra-proximal basaltic eruption products can continue to crystallize after deposition and are thus unlikely to be faithful indictors of shallow conduit magmatic processes; we recommend the use of medial and distal products instead.


Introduction
Basaltic Plinian volcanic eruptions are rare but can be unexpected and therefore hazardous events, and hence better understanding the triggering mechanisms and conduit processes involved is critical for hazard assessment (e.g., Coltelli et al. 1998;Houghton et al. 2004a, b;Costantini et al. 2009). The proximal deposits of large Plinian explosive eruptions are only rarely preserved due to the destruction or burial of the vent area during or after the explosive phase of the eruption (e.g., Novarupta, Fierstein et al. 1997;Houghton et al. 2004a, b). An exceptionally well-preserved proximal near-vent deposit was produced from a basaltic Plinian eruption at Tarawera volcano, New Zealand, in 1886 (Sable et al. 2006b): the most recent well-documented basaltic Plinian eruption on Earth and the only one with quantitative constraints and visual observations. In addition to the near-vent (proximal) deposits at Tarawera, the medial-distal deposits are also preserved, which enables us to examine differences in pyroclast textures with distance from the vent.
Pyroclast shapes and textures from historical basaltic Plinian eruptions represent the best source of information about eruption and fragmentation dynamics for powerful end-members (e.g., Sable et al. 2006a, b;Sable et al. 2009;Costantini et al. 2010;Bamber et al. 2020), since eye-witness observations of such eruptions are rare and conduits and deeper sub-volcanic systems are not accessible. Pyroclasts can be generally categorized into two groups : (1) juvenile, derived from the fragmentation of newly erupted magma, or (2) lithic, derived from entrainment of the country rock hosting the volcanic system. Microtextures of juvenile pyroclasts reflect one or more of the following processes: (a) magma ascent to fragmentation, (b) cooling during transport, (c) post-depositional modification, and (d) recycling while in a hot state. Most work focuses on (a) (e.g., Sparks 1978;Houghton and Wilson 1989;Polacci et al. 2003;Cimarelli et al. 2010;Arzilli et al. 2019;Di Genova et al. 2020;Heinrich et al. 2020;Taddeucci et al. 2021), with fewer studies addressing (b) and (c) (e.g., Sparks and Wright 1979;Yamagishi and Feebrey 1994;Tait et al. 1998;Hort and Gardner 2000;Capaccioni and Cuccoli 2005;Del Moro et al. 2013;Moitra et al. 2018), and (d) (e.g., D'Oriano et al. 2013D'Oriano et al. , 2014Deardorff and Cashman 2017).
Close to the vent, especially within the crater, newly deposited coarse pyroclasts may continue to cool slowly or even be reheated. While the pyroclast exteriors are likely quenched during transport, if the interiors are above the temperature of glass transition, they may experience post-depositional textural modifications. The degree of textural modifications is controlled by the temperature of the clasts and the rate of heat loss. Rapid deposition of juvenile pyroclasts at temperatures above the glass transition may permit: (1) microlite nucleation and growth (e.g., Szramek et al. 2010), (2) bubble nucleation, growth, collapse, and deformation (e.g., Rothery et al. 2007), (3) subsolidus transformation of minerals, e.g., Fe-Ti oxide and olivine (e.g., Haggerty and Baker 1967;Del Moro et al. 2013), and (4) ductile flattening, annealing and welding of pyroclasts (e.g., Giordano et al. 2005;Sumner et al. 2005;Bertagnini et al. 2011). Slow cooling of pyroclasts in situ is promoted by (i) larger clasts, as they cool more slowly, (ii) short transport times in the atmosphere, (iii) enhanced heat retention of the deposit due to high accumulation rates, and (iv) heat from the nearby vent, adjacent volcanic jets, or high temperature volatiles streaming through the vent and deposits (e.g., Sparks and Wright 1979;Sumner 1998;Sumner and Branney 2002;Woodcock et al. 2016). Pyroclasts deposited far from the vent, i.e., in medial and distal locations, undergo rapid cooling in the plume and are generally cold at deposition (Thomas and Sparks 1992). Groundmass textures of products from medial and distal deposits thus best preserve conduit processes, whereas near-vent deposits may be more equivocal.
Recycling of hot juvenile pyroclasts in the vent can also modify primary volcanic textures (e.g., Rosseel et al. 2006;D'Oriano et al. 2013D'Oriano et al. , 2014Lefebvre et al. 2013;Schipper and White 2016;Deardorff and Cashman 2017). Textures associated with recycling or reheating of pyroclasts in an intra-crater setting (e.g., at Stromboli) include changes in external morphology, extensive crystallization of the groundmass, precipitation of oxides on external clast surfaces, and subsolidus transformation of mineral phases (D'Oriano et al. 2013(D'Oriano et al. , 2014. Oxidized basaltic pyroclasts display a red or brown discoloration, chiefly controlled by the growth or alteration of iron-oxide mineral species (Houghton and Carey 2015). The significance of textures formed by high-temperature oxidation of basaltic pyroclastic deposits in the ultra-proximal environment has not previously, to our knowledge, been investigated in detail.
In this study, we focus on the 1886 Tarawera eruption to investigate post-depositional modifications of primary pyroclastic textures. The 1886 Tarawera eruption deposits are useful for this investigation due to the excellent preservation and comparability of the ultraproximal and medial deposits. This allows us to investigate textural differences between pyroclasts that experienced long transport times and rapid cooling, versus pyroclasts which experienced extended timescales for cooling at high temperatures in the deposit adjacent to the vent. We examine in detail the microlite, olivine and vesicle textures of lapilli and ash samples from unwelded horizons, comparing products from a proximal section dominated by ballistic and small clasts (median diameter −2.4 to -4.7 Φ for Plinian phase deposits; Carey et al. 2007) sedimented from jet margins and the lower convective column (Carey et al. 2007;Sable et al. 2009;T47), and a medial fall deposit section with ash and lapilli sedimented from the Plinian umbrella cloud (Sable et al. 2009;T43). We identify characteristics of 1886 pyroclasts that are attributed to post-depositional high-temperature oxidation.

Eruption chronology
The eruption of Tarawera began early on June 10, 1886, when an 8-km-long fissure opened along Mt Tarawera in the northeast, and basaltic magma was subsequently explosively erupted from over 50 discrete vents over a period of approximately 4 h. Visual observations of the eruption and its timing were recorded as eye-witness accounts that provide quantitative constraints on the dynamics and impacts of the eruption (Smith 1886a, b; Smith 1886a, b; Williams 1887). A detailed chronology was inferred from these accounts (Cole 1970;Nairn 1979;Keam 1988Keam , 2016Houghton and Wilson 1998). At ~ 12:30 a.m. on June 10 (all times are local), sustained ground shaking began (Smith 1886a, b). Earthquake activity increased in intensity until the eruption began at ~ 1:30 a.m. from a vent within the Wahanga dome, in the northeastern portion of Mt Tarawera (Keam 1988) (Fig. 1a). By 2:30 a.m., an 8-km-long line of craters hosted erupting vents along the Mt Tarawera segment (Williams 1887). Eyewitness accounts reported that these vents were in sustained eruption forming a convective column more than 10 km high (Williams 1887). At ~ 3:30 a.m., the fissure propagated across the Rotomahana swamps, lakes and geothermal system. Interaction of the basaltic magma with ground, surface and geothermal waters produced a 10-13-km-high eruption column that deposited wet fall and surge deposits, the "Rotomahana Mud" (Cole 1970). The entire 17-km rift, including the Rotomahana and Waimangu segments, was in eruption until ~ 6:00 a.m. (Hutton 1887;Keam 2016), with some weak explosions persisting for years at Waimangu (Cole 1970).

Eruption deposits
The first detailed physical volcanology study of the 1886 deposits was carried out by Walker et al. (1984), who defined five proximal units on crater walls within 400 m of the inferred vents, and mapped these units across Mt Tarawera. Houghton and Wilson (1998) later re-grouped the five units into three: phreatomagmatic Unit 1, magmatic Unit 2/3, and phreatomagmatic Unit 4/5. These three units can be followed along the walls through the entire 8-km-long fissure segment. The proximal deposits now exposed on the crater wall were not directly exposed to the vent or eruption jet during the climactic phase of the eruption because latestage phreatomagmatic activity subsequently excavated earlier deposits while widening the craters (Sable et al. 2006b) (Fig. 2). Houghton and Wilson (1998) correlated the medial-distal fall deposit (Figs. 1a and 3) with proximal Unit 2/3 deposits in the crater walls at Mt Tarawera (Figs. 1 and 2) and proposed that intense eruptive activity from the northeastern fissure segment was sustained for most of the eruption's duration. We define the current crater walls, about 50-150 m from the original vents, as 'ultra-proximal' (Fig. 1). The ultra-proximal deposits vary between black and red in color along the Mt Tarawera portion of the fissure (Fig. 1b) (Walker et al. 1984;Houghton and Wilson 1998;Sable et al. 2006b;Carey et al. 2007). Walker et al. (1984) attributed color differences to changes in eruption behavior and intensity. The color of the deposits was later suggested to have more complex origins, because color changes transect both bedding and individual clasts (Fig. 1c) (Houghton and Wilson 1998). Sable et al. (2006b) separated Unit 2/3 deposits on the crater walls into tephra "packages" of contrasting dispersal (Fig. 2). Linear half-thicknesses (t 1/2 ) of individual tephra packages were determined, with a higher t 1/2 equating to greater heights of particle release and hence derived from vents in higher intensity eruption; lower t 1/2 equates to lower particle-release heights, from vents erupting at lower intensity. Field measurements give t 1/2 values ranging from < 20 to ~ 280 m, which suggest that different vents erupted with variable intensities over the course of the eruption (Sable et al. 2006b). Four of the thirteen major craters produced widespread packages with t 1/2 > 150 m, most likely representing the vents erupting at high intensity that fed the convecting column and Plinian umbrella cloud (F, H, J, L, Fig. 1; Sable et al. 2006b).
Alternating non-welded and welded beds are observed within Unit 2/3 (Carey et al. 2007;Sable et al. 2009). Welding varies laterally and vertically within packages, both abruptly and gradationally and is strongest in locally dispersed packages, although zones of strong welding do occur in a few very widespread packages (Sable et al. 2006b). Similar degrees of welding found in packages with t 1/2 between 5 and 100 m demonstrate the complexity associated with simultaneous deposition of clasts from multiple vents of contrasting intensity (Sable et al. 2006b). Some non-welded packages have accumulation rates higher than suggested minimum rates for welding (e.g., Sparks and Wright 1979), justified by such packages either (i) including clasts with longer transport times in the upper levels of the plume; and/ or (ii) having a high content of cold lithic wall rock; and/or (iii) having some degree of magma-water interaction (Sable et al. 2006b). Clast flattening is also observed in some beds, although these are often the largest clasts (e.g., bombs; Carey et al. 2007).
The 1886 medial-distal deposits comprise only material sedimented from the high parts of the Plinian plume, and are dominated by pyroclasts erupted from vents within the craters F, H, J, and L (Sable et al. 2006b). Admixed ash from erupting vents in the Rotomahana portion of the fissure (from 3:30 a.m. to 6:00 a.m. on June 10) can be found in the upper layers of the 1886 deposits. This ash is called the 'Rotomahana Mud' and is a distinctive gray, fine-grained, lithic-rich ash, which commonly coats larger clasts. The temporal relationship between the Rotomahana Mud and the Tarawera lapilli varies across the footprint of the combined deposit (Carey et al. 2007;Sable et al. 2009;Rowe et al. 2021).

Sample collection and density measurements
Two sampling locations were selected to represent markedly different pyroclast sedimentation histories and postdepositional thermal environments (Fig. 1a). We used previously collected samples from an ultra-proximal section adjacent to Crater I (T47) that features widespread packages (t 1/2 > 80 m), and a medial section (T43) 2.5 km to the NE of the Tarawera fissure ( Fig. 1a) (Carey et al. 2007;Sable et al. 2009). This collection includes (a) 72 samples of ~ 100 16-32 mm juvenile clasts (lapilli) from T47, and (b) 11 samples of ~ 100 16-32 mm lapilli from T43. These clasts were sampled from narrow vertical intervals no more than 10 cm thick.
The degree of welding varies through the T47 section (Carey et al. 2007). All lapilli were sampled from nonwelded beds, although adjacent beds were welded (Fig. 2). T47-ρ22 is from a non-welded bed which lies on top of a bed containing flattened bombs. T47-ρ20 and T47-ρ19 are from alternating welded and non-welded beds containing flattened bombs. T47-ρ12 is from a non-welded bed with flattened bombs.
Although deposits range in color from red to black through the ultra-proximal deposits, all T47 clasts analyzed are red and undeformed. The T43 deposits are black to gray, and all T43 pyroclasts analyzed are black.
The density of each lapilli was measured following Houghton and Wilson (1989). Bulk vesicularity was calculated using a bulk dense rock equivalent (DRE) density of 2900 kg m −3 , (Houghton and Wilson 1989). Density histograms for studied samples are shown in Figs. 2 and 3. Five modal density lapilli from beds in Unit 2/3 were selected for microtextural analysis from each section to represent the bulk of the erupted magma. In some cases, where there are broad principal modes, lapilli were taken from multiple bins (e.g., T43-ρ08; Fig. 4).
Three samples of ash (G16 from ρ22), in the size range 250 µm-1 mm, were selected for textural analysis from Unit 2/3 at T47. Ash was embedded within resin then polished for Backscattered Electron (BSE) imaging.

Vesicle analysis
Selected lapilli were thin sectioned and nested images were generated using the FEI MLA650 scanning electron microscope (SEM) for low magnifications and the Hitachi SU-70 field emission analytical SEM for higher magnifications. See Online Resource 1 and Fig. S1 for more details on imaging strategy. We used Adobe Photoshop CS6 software to select vesicles from crystals and groundmass. See Online Resource 2 for more details on image processing. Binarized images were analyzed using the FIJI software package to calculate individual vesicle areas. Two-dimensional size distributions (number per unit area, N A ) were then converted to volume distributions (number per volume, N V ) using the stereological method of Sahagian and Proussevitch (1998). The cumulative number density (N V total) referenced to the matrix of the clast records the rate and timing of nucleation of vesicles during the rise of a body of magma. See Adams et al. (2006); Shea et al. (2010); Stovall et al. (2011); Parcheta et al. (2013) for further details on the method. All vesicularity data are provided in Table 1.

Groundmass crystallinity
Groundmass crystallinity (area %) was calculated from high magnification BSE images, at 250 µm horizontal field width and a scale factor of 10,240 px/mm. Such high magnifications were used to image groundmass crystallinity and to exclude phenocrysts (generally > 100 µm), xenoliths, and large vesicles. Mineral phases, glass and vesicles were then manually selected using Photoshop with the aid of element maps. The abundance of each phase (crystals, matrix glass and vesicles) was measured using the FIJI software package. The crystallinity % of each groundmass phase was calculated on a vesicle-free basis. All crystallinity data are provided in Table 2.
We measured the area % of plagioclase, olivine, pyroxene, and Fe-Ti oxide microlites (< 100 µm). Area % of pyroxene and groundmass glass in T43 is only approximate as BSE grayscales are indistinguishable using our image analysis methods. See Online Resource 3 for more details on methods used to gather crystal area % data.
Crystal size distributions and crystal number densities could not be calculated due to the difficulty in outlining individual crystals in the extremely crystal-rich groundmass made of overlapping crystals, especially in pyroclasts from the ultra-proximal deposits.

Juvenile clast density and bulk vesicularity
For this study, T47 ρ samples were chosen from beds within the most widespread packages and were therefore most likely dominated by clasts erupted from high intensity vents (following Sable et al. 2006a, b). The T43 ρ samples were chosen such that they represent Unit 2/3 below the major influx of Rotomahana Mud. T47 and T43 ρ samples are relatively similar in terms of their density ranges (800 to 2000 kg m −3 on average; Figs. 2 and 3) and their modal densities (1000-1200 kg m −3 on average), although T43 has a higher standard deviation (240-320 kg m −3 ) compared to T47 (190-230 kg m −3 ) due to the tails of dense clasts. Almost all selected T43 ρ samples have multiple modes, versus only one weakly bi-modal sample from T47 (Fig. 4). Red and black ultra-proximal deposits; coloration is not confined to beds or clasts of specific types or sizes, indicating that red oxidation is likely largely a post-depositional process. (c) Zoom in of (b). Yellow arrows point to a horizon of red-colored deposits with overlying black deposits. White arrows point to thin layer of red deposits underlain and overlain by black deposits ◂ Selected clast vesicularity ranges are slightly higher for T47 clasts (59.8-63.8%) than T43 clasts (56.3-61.8%).

2D observations of vesicle textures from both lapilli and ash
The vesicle textures in T47 lapilli are nearly identical to those in T43 (Table 1). Lapilli from both sections contain highly irregular, complex and polylobate-shaped vesicles with a wide range of sizes (mean size interval from selected lapilli are 0.02 to 2.5 mm; Fig. 5). Vesicles < 0.1 mm are generally more circular or lobate. Vesicle number densities fall in a narrow range of 2-5 × 10 6 cm −3 (with one outlier at 5 × 10 5 cm −3 ) for the five T47 lapilli and 1-6 × 10 6 cm −3 for the five T43 lapilli (Table 1), identical to vesicle number densities calculated by Sable et al. (2009). In both T47 and T43 lapilli, vesicles are typically smaller at the rims and larger in the cores, although this feature is more evident in T47 lapilli. The vesicle size distributions of T47 and T43 lapilli are broadly the same (Fig. S2).
Vesicle sizes in the ash particles from T47 ultra-proximal ejecta are extremely similar to those of its lapilli, although larger vesicles (> 0.3 mm) are limited in the ash. The T47 ash contains circular and lobate vesicles identical in size to the lower range of the T47 lapilli (mean size interval of 0.02 to 0.3 mm).
A subpopulation of small (< 30 µm) voids are common in all Tarawera lapilli (Fig. 5i,j,k,l), although their abundance generally correlates with groundmass crystallinity. We refer to these features as voids because they have separate sizefrequency distributions decoupled from the vesicle population > 30 µm (Fig. S2) and often contrasting morphologies and dispersal habits from larger populations of vesicles.
Although only modal density clasts from T47 and T43 samples were studied in detail, we also briefly examined textures of dense and vesicular clasts from the investigated samples. Dense and vesicular clasts from each sample have similar number densities to the modal clasts, but dense clasts The roughness of vesicle walls correlates weakly with groundmass crystallinity. Clasts with near-holocrystalline groundmass (T47) have highly irregular and angular vesicle walls due to the protrusions of microlites toward the inner portions of vesicles (Fig. 5i,j), whereas clasts with lower groundmass crystallinities (T43) have smoother inner walls with fewer protrusions of microlites (e.g., Fig. 5k,l).
To summarize, in general, the vesicle textures in T47 pyroclasts are nearly identical to those in T43, but there are subtle qualitative differences related to protrusions of microlites into vesicles that are not captured in quantitative data.

Microlites
Groundmass textures of the 10 modal density juvenile lapilli (5 from T47, 5 from T43; Fig. 4) and 3 ash samples across two size intervals (0 and 2 ϕ) from T47-G16 (equivalent stratigraphic height to ρ22) have been qualitatively and quantitatively characterized ( Table 2). We group the groundmass textures into two categories (near-holocrystalline and highly crystalline), based on the areal ratio of microlites to glass ( Table 2).
The total groundmass crystallinities of modal lapilli contrast markedly between T47 and T43 (Figs. 6, 7, 8, Table 2): T47 lapilli are near-holocrystalline (90-96 area %, e.g., Fig. 6a,b) while T43 lapilli have a relatively lower crystallinity but are still highly crystalline (77-83 area %, e.g., Fig. 7). Although only modal density lapilli were chosen to be studied, we observed that dense and vesicular clasts have identical crystallinities to modal clasts. Based on examination of many other grains in BSE, there is some subtle variability in groundmass texture of the products from each investigated sample, but we believe that the textures in selected clasts are representative of the bulk of the magma. Phenocryst content is very low (< 0.5 vol.% of the clasts; Sable et al. 2009) with little variability throughout each sample.
The groundmass of the lapilli and ash in both sections consists of differing proportions of microlites of plagioclase, pyroxene and olivine (Figs. 6 and 7; Table 2). Residual glass is present in varying proportions in T43  Fig. 4 Clast density distribution histograms for Unit 2/3 of the ultra-proximal (T47) samples and medial (T43) samples used in this study, drawn in stratigraphical order (see Fig. 2      there are more microlite protrusions on vesicle inner walls in T47 compared to T43. Vesicles in black, groundmass in gray, and microlites in gray and white but is virtually absent in T47. Fe-Ti oxide microlites of all sizes are only present in T47 lapilli and ash and are virtually absent in T43. Microlites are randomly distributed through the groundmass in both T47 and T43 ash and lapilli, and are chiefly randomly oriented, except near vesicles, where long axes are generally parallel to the vesicle margins. There are no rim-to-core variations in microlite crystallinity in either T47 or T43 pyroclasts. Size ranges of crystal phases between T47 and T43 are generally similar regardless of pyroclast size (e.g., ash versus lapilli for T47).

Olivine subsolidus transformation
Olivine is rare in all studied clasts (≤ 8%). In clasts from T47, olivine crystals show varying degrees of oxidation, reflecting subsolidus transformations (e.g., Haggerty and Baker 1967;Del Moro et al. 2013;D'Oriano et al. 2014). All olivine crystals observed (~ 250) in T47 clasts displayed subsolidus transformations (e.g., Fig. 9a,b,c; Online Resource 5). In contrast, all T43 olivines are fresh (e.g., Fig. 9d). In BSE images of T47 pyroclasts, olivines commonly show exsolution of Fe oxides as symplectitic intergrowths or patches, either along crystal rims, along fracture lines, or throughout the crystals (Fig. 9a,b,c). Although minerals associated with the subsolidus transformation of olivine are too small to be analyzed on the SEM by Energy Dispersive Spectroscopy (SEM-EDS) or by Electron Probe Micro Analysis (EPMA), we have tentatively identified the minerals by comparison of microtextures with those reported from experimental (Haggerty and Baker 1967) and lapilli. Olivine crystals in T47 pyroclasts display symplectitic texture, whereas olivine crystals in T43 pyrclasts remain fresh. Orange bor-ders enclose T47 pyroclasts, green border encloses a T43 pyroclast. Pl: Plagioclase, Px: pyroxene, Ox: Fe oxide, G: groundmass glass, Ol: olivine iron in the system (i.e., strictly related to the fO 2 conditions which control the subsolidus transformation). Mg-ferrite is the product of substitution of Fe 2+ by Mg 2+ during olivine oxidation (Del Moro et al. 2013). The mineral intergrown with the Mg-ferrite is likely forsterite, which, together with orthopyroxene, is a common product of subsolidus transformation of olivine (Haggerty and Baker 1967).

Groundmass and olivine textures
Pyroclasts deposited in the ultra-proximal environment (e.g., T47) are particularly subject to post-depositional textural changes because they are coarse (up to bomb-sized), experienced short transport times, and have accumulated quickly in a hot environment close to the erupting vents. Welded horizons within the T47 strata (Fig. 3) indicate that the deposits could reach a bulk temperature in excess of that of glass transition, permitting annealing, ductile deformation and welding processes. For high temperature basaltic magmas, post-fragmentation textural changes in lapilli and larger pyroclasts are difficult to avoid without extremely rapid cooling rates (Gardner et al. 1996).
Pyroclasts with the highest microlite abundance are commonly inferred to represent cooler and slower ascending or stalled magmas along the margins of the upper conduit compared to their lower crystallinity counterparts inferred to ascend along the central axis of the conduit (e.g., Taddeucci et al. 2004;Sable et al. 2006a;Cimarelli et al. 2010). Such contrasts in crystal textures are commonly coupled with differences in vesicle textures; however, they are not coupled in our samples. Interpreting the near-holocrystalline (90-97%) T47 lapilli and ash as representative of syn-eruptive conduit conditions is not plausible because (i) the magma would have been too crystalline to flow (e.g., Caricchi et al. 2007), and (ii) a significant volume of this microlite-rich magma would need to be erupted at very high magma discharge rates and replenished quickly throughout this high intensity phase. This scenario is highly unlikely. We suggest that the discrepancy of groundmass crystallinity between T47 and T43 pyroclasts is due to post-depositional crystallization of dominantly plagioclase and Fe-Ti oxides, due to hot emplacement and slow cooling rates of the ultra-proximal deposits.
Post-depositional crystallization in T47 pyroclasts is also indicated by subhedral-euhedral Fe-Ti oxides < 3 µm which are completely absent in T43 clasts. Primary subhedraleuhedral morphologies of Fe-Ti oxides are often considered indicative of steady growth from low to moderate undercooling conditions (Hammer et al. 1999). These features have also, however, been attributed to magma oxidation before fragmentation due to convective circulation that allowed exposure to the atmosphere before recirculation in the upper part of the conduit, or due to recycling of juvenile pyroclasts (Genareau et al. 2010;D'Oriano et al. 2013D'Oriano et al. , 2014Deardorff and Cashman 2017). Convective circulation is highly unlikely due to the sustained nature of this eruption and the rapid ascent rates of the magma in the conduit (Sable et al. 2009). We suggest that the abundant and dispersed < 3 µm Fe-Ti oxides in T47 pyroclast textures are most consistent with post-depositional crystallization, rather than recycling, because other diagnostic textures of hot-state recycling such as cored lapilli are rare to absent (D'Oriano et al. 2013(D'Oriano et al. , 2014Deardorff and Cashman 2017). Further, recycling is not consistent with the sustained and intense eruption style during this phase.
We propose that the ultra-proximal deposits of the 1886 Tarawera eruption were emplaced at a bulk temperature that was above or close to the glass transition, and cooled relatively slowly due to high proximal accumulation rates. Further, these deposits could have been exposed to vent-derived heat moving upward and outward through the interstitial space in the deposits. We cannot constrain the timescales required for textural modification of the T47 pyroclasts in the ultraproximal setting because the exact behavior of the bulk deposit temperature through time is unknown, and uncertainties remain regarding crystalization kinetics of Fe-Ti oxides at temperatures close to the glass transition. However, thick m-scale densely welded horizons within the T47 section certainly suggest that bulk temperatures above the glass transition could have lasted many hours. This interpretation implies that the assumption that all ultraproximal 1886 pyroclast textures represent conduit conditions at fragmentation is incorrect.
At T47, both the lapilli and ash have extremely similar groundmass textures, including evenly distributed granular Fe-Ti oxides (< 3 µm) and near-complete groundmass crystallization. The similarities between these different grainsizes are surprising given the expected differences in particle cooling rates. These textures are ubiquitous in T47 products (approximately 100 ash particles in 0-2 Φ size range and all modal juvenile lapilli analyzed (13), see Online Resource 7). Further, lapilli and ash from the same stratigraphic horizon (T47-p22/T47-G16), have extremely similar groundmass crystallinities (Table 2). This is interesting because ash < 2 mm should have very rapid cooling rates in the atmosphere (a reasonable quench rate is 10-20 K s −1 ; Xu and Zhang 2002) and should have been deposited cold in the ultra-proximal environment (Moitra et al. 2018). Reheating ash above the glass transition is thermally very energy intensive. There are four options to explain the similar groundmass textures between T47 ash and lapilli: (1) all the ash is secondary, derived from lapilli and bombs by abrasion during sampling and particularly sieving; (2) the ash was rapidly deposited at bulk temperatures above the glass transition such that it could continue to crystallize the groundmass; (3) the ash was re-reheated after deposition; or (4) the ash and the lapilli have the same crystallinity and primary magmatic phases, but these textures were produced in substantially different ways; the ash is the result of more efficient brittle fragmentation of highly crystallized magma at the margins of the conduit, whereas the lapilli have textures reflecting extensive postdepositional crystallization. Juvenile grainsize histograms of T47 samples presented in Carey et al. (2007) do not show an increase in the ash abundance for the fine grainsize fractions 0-2 ϕ, suggesting that the physical process of sieving is not creating artificial ash of this size. This data coupled with the qualitative observation that all ash textures are homogeneous (see Online Resources 7), suggests that (1) is highly unlikely and that textures are probably primary. There are no obvious microlite textural features in the ash to indicate whether (2), (3), or (4) is more likely, but the oxidation of olivine seen in both T47 ash and lapilli is more supportive of post-depositional textural modification (iii) and (iv) as described below.
Olivine is one of the minerals that is most effective in recording oxidizing conditions (Haggerty and Baker 1967;Del Moro et al. 2013). Subsolidus transformation of olivine has long been interpreted to reflect alteration under high temperature and oxidizing conditions (Haggerty and Baker 1967;Haggerty 1991). More recent studies have attributed similar subsolidus transformation to either (i) rapid changes in oxygen fugacity during sudden decompression of the magmatic column (Cortés et al. 2006;Knafelc et al. 2019); (ii) convective circulation of magma that is oxidized when exposed to the atmosphere during circulation (Genareau et al. 2010); (iii) exposure of deposits to high temperatures under atmospheric pressures (Del Moro et al. 2013); or (iv) recycling of pyroclasts (D'Oriano et al. 2014). Olivine crystals in both ash and lapilli from the ultra-proximal T47 locality have experienced subsolidus transformations of variable degrees, from minor symplectite intergrowth texture on rims, to the olivine being completely unrecognizable. The olivine transformations suggest that both the ash and lapilli were heated and oxidized to varying degrees. Recent experiments on olivine oxidation suggest that rates of olivine oxidation are temperature-dependent, with nucleation of magnetite and/or hematite occurring in days to weeks at 600 °C and within minutes at 900 °C at atmospheric pressures (Knafelc et al. 2019). Ultra-proximal deposits were likely held above 600 °C for hours during and/or after the eruption ceased, allowing for mineral transformations to occur. Oxidation during rapid ascent in the conduit (i) is unlikely as we would then expect some evidence of olivine subsolidus transformation in T43 pyroclasts. Convective circulation (ii) is unlikely, as previously discussed and recycling (iv) is also unlikely because sample pyroclasts are believed to have chiefly been produced from a sustained (non-pulsatory), highly explosive eruption, and thus precludes significant recycling. Therefore, high temperatures and atmospheric conditions (iii) of the deposits is the most likely scenario.

Estimates of syn-eruptive temperatures for ultra-proximal deposits
Welding is present in multiple packages at T47 (Carey et al. 2007), although samples for this study came from non-welded beds. Welding depends on the clasts' emplacement temperature, the accumulation rate, and the environment of deposition (e.g., solid substrate, hot pre-existing fall material, hot vent, snow, water, etc.). In welded beds adjacent to non-welded sampled beds, temperatures may have been as high as ~ 1000 °C (D'Oriano et al. 2014). Since changes in groundmass textures in basalt may occur as low as ~ 700 °C (D'Oriano et al. 2014), even the tack-welded and non-welded beds probably had sufficiently high temperatures and the oxidizing conditions needed for thermal modification because of the heat released from adjacent welding layers.
The colors of the clasts (both external surfaces and cores) at Tarawera indicate post-depositional thermal oxidation. Selected T47 lapilli and ash are ubiquitously a brightreddish color, whereas T43 pyroclasts are black-gray in color. Reheating experiments show that reddening throughout pyroclasts probably resulted from crystallization of Fe-bearing phases due to oxidizing conditions (D'Oriano et al. 2014). The surface reddening of basalt in ash occurs within ~ 1 h at T > 700 °C in air (D'Oriano et al. 2014) and so take place rapidly during post-depositional heating. The reddish color of clasts is therefore a useful indicator for postdepositional thermal modification.
Groundmass mineral phases can indicate conditions at which the T47 pyroclast textures were modified. The microtextures in T47 ash and lapilli indicate oxidation at a temperature sufficient to promote subhedral-euhedral growth of Fe-Ti oxide microlites and oxidation of olivine (substituted by Fe oxide and forsterite). From experiments, D'Oriano et al. (2014) have found that oxide-rimmed minerals commonly form in reheated clasts by heterogeneous nucleation under highly oxidizing conditions and variable degrees of undercooling at temperatures of 750-1050 °C). Further constraints on clast temperatures come from experimental studies that show hematite exsolution in olivine occurs above 600 °C (in air; P = 1 atm) (Haggerty and Baker 1967;Khisina et al. 1998;Gualtieri et al. 2003). At temperatures above ~ 820 °C, and under oxidizing conditions, olivine reacts to form magnetite species and Mg-rich orthopyroxene (Haggerty and Baker 1967;Champness 1970;Goode 1974;Gualtieri et al. 2003). Based on the cited experimental results, we propose that the subsolidus transformations found in olivine of lapilli and ash in T47 occurred under oxidizing conditions > 600 °C. Such high temperature crystal modification at atmospheric pressures (P = 1 atm) is only possible with short pyroclast transport times, high deposit accumulation rates and perhaps proximity to a hot vent. Using typical basaltic glass transition temperatures of ~ 670-690 °C (Giordano et al. 2005(Giordano et al. , 2008D'Oriano et al. 2013;Deardorff and Cashman 2017), we suggest that the T47 ash and lapilli were held above a minimum temperature of 670 °C to allow some Fe-Ti oxide crystallization of the groundmass. This temperature is consistent with the temperature estimates for subsolidus transformation of olivine from experimental studies (Haggerty and Baker 1967;Khisina et al. 1998;Gualtieri et al. 2003).
Further insight into the temperature and timing conditions for post-depositional modification of basalt products is from cooling versus reheating experiments by Burkhard (2002Burkhard ( , 2005. The experiments suggest that plagioclase will only grow and/or nucleate as a new microlite when T > 930 °C. Overall, T47 clasts have a slightly higher plagioclase content (59-78 area %) compared to T43 clasts (54-59 area %), whereas pyroxene is relatively similar (11-22 area % in T47; 15-20 area % in T43). These results indicate that at least a portion of the ultra-proximal deposits were held > 930 °C for enough time for plagioclase microlites to nucleate and grow in the groundmass.

Interpretation of vesicle textures
Vesicle textures are commonly used to reconstruct magma ascent history (e.g., Cashman and Mangan 1994;Mangan and Cashman 1996;Carey et al. 2009Carey et al. , 2012Shea et al. 2010). Slight differences in density distributions between 1886 ultra-proximal and medial sections are likely to reflect: (1) T47 is dominated by clasts derived from a vent erupting at high intensity, but may contain some clasts derived from adjacent vents (Sable et al. 2006b), whereas at T43 clasts were erupted in powerful fashion and all traveled to the full height of the Plinian column, but could be from up to four vent sources (e.g., Sable et al. 2006b), and (2) some T47 clasts have larger vesicles in their core than T43 clasts, which we infer reflects higher degrees of post-fragmentation bubble expansion at T47 after deposition. Vesicle size distributions (Fig. S2) are similar between T47 and T43 lapilli, but vesicle morphologies are slightly different (e.g., Fig. 5), driven by the protrusion of microlites within the vesicles in T47 ejecta that we propose formed during post-depositional crystallization. We suspect that in the ultra-proximal environment at Tarawera, post-depositional crystallization modified and overprinted primary microtextures.

Conclusions
Post-depositional thermal modification has modified the textures of pyroclasts in ultra-proximal deposits from the 1886 Tarawera eruption. The microtextures in ultraproximal ash and lapilli indicate that they were oxidized at temperatures sufficient to promote subhedral-euhedral growth of Fe-Ti microlites and subsolidus transformation of olivine, probably > 600 °C for at least tens of minutes after the eruption ceased. The higher content of plagioclase microlites in ultraproximal pyroclasts suggests longer cooling times > 930 °C compared to pyroclasts deposited in the medial environment. Pyroclasts that were ejected higher into the volcanic plume and deposited at a significant distance from the vent did not undergo significant Fe-Ti oxide crystallization and lack subsolidus transformation of olivine. High accumulation rates of hot and coarse tephra in the ultra-proximal environment, in addition to convective heat from the vent, led to the required high post-depositional emplacement temperatures, and reduced deposit cooling rates (Fig. 10). Welding beds above and below the studied beds could have provided additional heat for late post-depositional crystallization.
To our knowledge, there have been no previous systematic studies demonstrating such textural changes from ultra-proximal to more distal settings. We suggest that post-depositional modification of groundmass textures of both ash and lapilli by retained and vent-derived heat, or by heat loss from adjacent welding beds, in the near-vent environments may be common for explosive basaltic eruption deposits globally. This is evident by the bright-reddish color of ultra-proximal deposits, e.g., 12-13 January 2011 Mt. Etna (Andronico et al. 2014);September 2008Stromboli (D'Oriano et al. 2011; and Krafla welded layer (Calderone et al. 1990). Decompression textures of ultra-proximal basaltic eruption products are thus likely to be overprinted. We strongly recommend examining juvenile pyroclasts in medial or distal environments when available instead.
Acknowledgements Sandrin Feig is thanked for his support with SEM and EPMA analyses, which were carried out at the Central Science Laboratory at the University of Tasmania. We thank Wendy Cockshell for locating our requested samples from the rock store at the University of Hawaii. Thin sections were kindly prepared by Scott Wagner. Comments from two anonymous reviewers greatly improved the quality of this manuscript. Thanks to Corrado Cimarelli for handling the manuscript.

Author contributions Not applicable.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions. HM's research was funded by a University of Tasmania Graduate Research Scholarship. The research of RJC and BFH was supported by subcontracts with the Geological and Nuclear Sciences (GNS), New Zealand. JDLW's Tarawera work was supported with a GNS Science contract.
Code availability Not applicable.

Declarations
Ethics approval and consent to participate Not applicable. Plinian plume transporting material to medial and distal locations T47 T43 Fig. 10 Sketch (not to scale) illustrating thermal modification in the ultra-proximal area of Tarawera, versus the absence of such processes at medial sites. In the ultra-proximal environment, a combination of rapid accumulation and burial of coarse pyroclasts, convective ventderived heat (yellow arrows), and the movement of hot gases through the deposits allowed heat retention of the rocks followed by slow cooling rates. The pyroclasts that rose well into the plume cooled rapidly and were deposited fresh in an environment far enough away from the vent and with low enough accumulation rates so that no thermal modification processes occurred. Black dashed line through ultra-proximal deposits represents current crater walls following latestage phreatomagmatic activity which excavated earlier deposits. Ox: Fe-oxide, Plag: Plagioclase, Ol: Olivine, Fo: Foresterite. Scale bar: 5 µm Consent for publication All authors consent to publication.

Conflict of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.