Rapid assembly of high-Mg andesites and dacites by magma mixing at a continental arc stratovolcano

Studies of pre-eruptive processes at active volcanoes require precise petrochronological constraints if they are to contribute to hazard assessment during future eruption events. We present petrological and geochemical data and orthopyroxene diffusion time scales for samples from Late Pleistocene high-Mg andesite-dacite lavas (Mg# 53–69) at Ruapehu volcano, New Zealand, as a case study of rapid magma genesis and eruption at a continental arc stratovolcano. Assembly of Ruapehu high-Mg magmas involved the mixing of primitive magmas plus entrained mantle-equilibrated olivines with mid-crustal felsic mush bodies, yielding hybridized magmas with ubiquitous pyroxene reverse-zoning patterns. Orthopyroxene Fe-Mg interdiffusion time scales linked to quantitative crystal orientation data show that most lavas erupted < 10 days after resumption of crystal growth following magma mixing events. The eruption of lavas within days of mixing events implies that pre-eruptive warnings may be correspondingly short.


INTRODUCTION
Arc stratovolcanoes are active foci for research on the generation of intermediate-composition magmas and their frequent, commonly hazardous eruptions. Magma mixing is a process of relevance for both topics because the hybridization of mafic and felsic materials may act as an eruption trigger (e.g., Kent et al., 2010). Mineral zonation patterns that record mixing dynamics can be used to measure pre-eruptive time scales (e.g., Kahl et al., 2011), which are useful parameters for hazard mitigation efforts during modern-day eruption scenarios. Mixing is commonly described for arc volcanic rocks, including some examples of high-Mg andesite and dacite (HMAD). These are defined as having wholerock Mg# values [Mg# = 100 × Mg / (Mg + Fe)] of ≥45, higher than otherwise expected over the SiO 2 range of 54-65 wt% (Kelemen et al., 2003).
While some examples of HMAD volcanism are attributed to a primary mantle origin (e.g., Bry-ant et al., 2011), petrographic studies highlight the role of mafic-felsic magma mixing accompanied by olivine accumulation in disproportionately raising the bulk-rock Mg abundance in other cases (e.g., Streck et al., 2007). Examples of the latter type are valuable for understanding the types and time lines of crustal-level magma hybridization processes. Here we present mineralogical characteristics and orthopyroxene Fe-Mg interdiffusion time scales for a suite of HMAD lavas from Ruapehu volcano, New Zealand, a large andesite-dacite stratovolcano of the Taupo Volcanic Zone with a 200 k.y. history of effusive eruptions ( Fig. 1; Conway et al., 2016).

SAMPLES AND ANALYTICAL METHODS
Samples were collected from HMAD lavas that were erupted from 48 to 37 ka onto the southern and western flanks of Ruapehu volcano (Conway et al., 2016;Fig. 1). They have whole-rock Mg# values of 60-69 (andesites) and 53-57 (dacites), plus elevated Ni (∼30-150 ppm) and Cr (∼100-470 ppm) at equivalent SiO 2 contents when compared with all other lavas from Ruapehu (Conway et al., 2018;Fig. S1 in the Supplemental Material 1 ). Samples used for this study represent ranges in eruption age, composition, geographic distribution, and cooling rate ( Table  S2 in the Supplemental Material). The lavas are non-to moderately vesicular with 15%-35% macrocrystals (>100 µm). Backscattered electron (BSE) images and major-element data for minerals were collected by electron probe microanalysis. Crystal axis orientations for orthopyroxenes used for diffusion modeling were determined by electron backscatter diffraction (EBSD). See the Supplemental Material for details of analytical methods and supplementary data (Table S3).

MINERAL CHARACTERISTICS AND ORIGINS Pyroxene
Sharply reverse-zoned orthopyroxene and clinopyroxene are ubiquitous in Ruapehu HMAD lavas (type 1 crystals; Figs. 2A and 2B). Type 1 orthopyroxene and clinopyroxene cores vary over Mg# 67-80, but individual cores are generally unzoned, slightly resorbed, and overgrown by 8-15-µm-thick rims of Mg# 82-89, with outermost higher-FeO reaction rinds. Type 1 pyroxenes occur as single euhedral or broken grains and as crystal clusters that can include ortho-and clinopyroxene ( Fig. 2A). Type 2 pyroxenes are unzoned, with Mg# values equivalent to those of type 1 crystal cores, and commonly occur as clusters (Fig. 2C). Type 3 orthopyroxene crystals have high Mg# cores (like type 1 crystal rims) and display normal zoning toward their rims (Fig. 2D).
Several features imply that type 1 and 2 pyroxene cores were derived from a felsic source. First, average orthopyroxene (opx) core values of Mg# 75 are in equilibrium with relatively evolved melt compositions (Mg# 45), applying a Fe = Mg exchange coefficient, K D [(Fe/Mg) opx /(Fe/Mg) liquid ], of 0.3 (Price et al., 2012), and not with host whole-rock values of Mg# 53-69. Second, type 1 orthopyroxene and clinopyroxene cores are homogeneous, consistent with subsolidus equilibration (Putirka, 2008). Third, orthopyroxene commonly occurs in multi-grain clusters, with rhyodacitic to rhyolitic melt inclusions and interstitial glass ( Conway et al., 2018). This felsic source is inferred to have been largely crystalline (hereafter termed mush). Rim characteristics of type 1 pyroxenes indicate that the felsic mush was dis-turbed, and that outer zones (and type 3 grains) grew from more primitive melts. Broken grains that lack a complete higher-Mg# rim ( Fig. 2B) were likely derived from crystal clusters that were disaggregated during magma hybridization events shortly prior to eruption.

Olivine and Spinel
Type 4 crystals (mostly ∼600 µm, rarely to 4 mm) and glomerocrysts are high-forsterite (high-Fo) olivines that are anhedral, resorbed, and orthopyroxene overgrown, with abundances in samples that correlate with whole-rock Mg# (Table S2). The Fo and NiO contents of their cores indicate equilibrium with peridotitic mantle (∼2500 ppm NiO at 91 mol% Fo [Fo 91 ] ; Fig.  S4). Rims are normally zoned (Fo 76-86 ) across ∼100 µm and slightly to moderately resorbed with adhering <100-µm-wide rinds of type 3 orthopyroxene (Fig. 2E). Euhedral spinels, as large as 100 µm across, are included within most type 4 olivines (Fig. 2E). They are chromian (Cr# 61-75), with Mg# 32-68, 0.2-2.0 wt% TiO 2 , and 10.5-19.0 wt% Al 2 O 3 , indicative of a depleted arc mantle source (Fig. S4). Using overgrowth rates from Coombs and Gardner (2004) and Zellmer et al. (2016), the dimensions of orthopyroxene coronae on type 4 olivines are consistent with their mixing into felsic magma over periods of weeks to months. We infer that the olivines were carried by mantle-derived basalts (∼11 wt% MgO; Barker et al., 2020) that induced the reverse-zoned growth of type 1 pyroxene rims on mixing with felsic mush in the crust. The large, anhedral and sometimes glomerocrystic nature of the olivines suggests that they were not cogenetic with the mafic magma but were instead incorporated from a preexisting cumulate body.
Olivines with Fo 74-78 cores that show resorption and diffuse normal or no zoning and have extensive >100 µm orthopyroxene rinds are classed as type 5 crystals (Fig. 2F). The overgrowth orthopyroxene crystals are mostly type 3, but type 1 crystals are also present (Fig. 2F). Type 5 olivines are interpreted to represent type 4 olivines that experienced prolonged crustal storage resulting in extended orthopyroxene overgrowths and diffusive reequilibration of their cores. The presence of type 1 orthopyroxene as overgrowths indicates that type 5 olivines experienced a final mixing event, consistent with there having been recurring mafic magma recharge pulses into the felsic magma bodies.

Plagioclase and Hornblende
Plagioclase is ubiquitous in the samples, with typical lengths of 300-700 µm, and maximum lengths of 5 mm. Two prominent types show sharply defined reverse (type 6) or normal (type 7) zonation of their rims (outer ∼100 µm; Figs. 2G and 2H). Hornblende (type 8 crystals) occurs only in Te Kohatu member dacite lavas (Conway et al., 2018), reflecting more hydrous source magmas compared with the andesites (Putirka, 2016). They are 200-500-µm-long grains, commonly with adhering plagioclase or pyroxene, and are partially to fully opacitic, with symplectite assemblages of orthopyroxene, plagioclase, and Fe-Ti oxides replacing their outer 20-100 µm rims (Fig. 2I). These crystals yielded an average pressure of 3.5 kbar (∼13 km depth in continental crust) using the Al-in-hornblende barometer of Ridofli et al. (2010). We tested this result against the liquid-amphibole barometer of Putirka (2016), using a melt composition similar to the least-fractionated melt inclusions in Conway et al. (2018), and refined to be in equilibrium with the amphiboles (i.e., 68.5 wt% SiO 2 , 1.3 wt% MgO) using the methods of Scruggs and Putirka (2018). The resulting pressure outputs (average of 3.6 kbar; ∼13 km depth) are identical to those calculated using the Ridolfi et al. (2010) model. The (partial) preservation of hornblende implies rapid magma ascent from these depths (e.g., Auer et al., 2015), thus providing a valuable constraint on the depth of mixing within the crust.

TIME-SCALE MODELING METHODS AND RESULTS
Iron-magnesium (Fe-Mg) interdiffusion timescale modeling was carried out on 54 core-rim boundaries in type 1 orthopyroxene crystals from five andesite samples that met the criteria for accurate modeling (Krimer and Costa, 2017) and are representative of the crystal features in all studied Ruapehu HMAD lavas. Modeling used the calibration of Ganguly and Tazzoli (1994) without an oxygen fugacity correction (after Dohmen et al., 2016) to calculate Fe-Mg interdiffusion coefficient, D Fe-Mg , in orthopyroxene (Cooper et al., 2017). We used a two-pyroxene Fe-Mg exchange temperature of 1100 °C from the geothermometer of Putirka (2008) that was derived from 21 type 1 orthopyroxene-clinopyroxene rim pairs (Fig. S5). Finite-difference methods were used to model profiles with an assumed initial step function (full details in the Supplemental Material). EBSD analysis enabled the assessment of suitable crystal boundaries for modeling; where possible, profiles perpendicular to the c-axis were used (Fig. 3). Modeled time scales are from 0.4 to 27 days, with 50 of 54 crystals recording <10 days (Fig. 3D). Error reflects Monte Carlo simulations of uncertainty, which used ±20 °C (1σ) for the temperature constraint and incorporated uncertainties relating to pixel size, the number of integrated pixel lines across each boundary, and the grayscale values used to define the Mg# values.

ORIGINS OF RUAPEHU HMAD MAGMAS
The mineralogical features (Fig. 2) and geochemical trends (Fig. S1) defined by Ruapehu HMAD samples yield a model of mixing between a mafic end member (basaltic melt with olivine xenocrysts) and a felsic end member (rhyodacitic mush bodies). Our model for Ruapehu HMAD eruptives (Fig. 4) starts with mantle-derived magmas, which incorporated olivine xenocrysts that crystallized below or at the base of the crust (>25 km depth; Salmon et al., 2011). These olivines were carried into felsic mush bodies, resulting in resorption, normal zoning, and pyroxene overgrowth over weeks to months (e.g., Coombs and Gardner, 2004). The crystal mushes likely represent the residues of magmas erupted since ca. 160 ka, which is the lower limit of ages for lavas from this part of the edifice (Gamble et al., 2003), and therefore have inferred mid-crustal (∼20-9 km) loci in the central axis of the volcanic plumbing system. Seismic tomography suggests a transition to feeder dikes above this (Rowlands et al., 2005), and hornblende barometry constraints of ∼13 km depth are consistent with mixing in the mid-crust (Fig. 4). We infer that mafic magma intruded and mobilized the mid-crustal storage zone: crystal mushes were disaggregated and the rims of type 1 pyroxene formed and then quenched by eruption within days to weeks. These rapidly instigated eruptions resulted in substantial construction of the edifice (Fig. 1A). Similar petrological and geochemical features suggest a common mode of magma genesis for HMAD eruptives from White Island (Mandon, 2017), Pukeonake (Beier et al., 2017) and Hauhungatahi (Cameron et al., 2010) (Fig. 1A). Moreover, magma-mixing origins proposed for HMAD in global examples are based on similar features of xenocrystic high-Fo olivine and sharply reverse-zoned orthopyroxene and clinopyroxene (e.g., Ohba et al., 2010;Streck and Leeman, 2018), suggesting that a hybrid origin for HMAD may be widespread.

UTILITY OF DIFFUSION MODELING FOR HAZARD PREPAREDNESS
Petrochronologic studies of olivine routinely incorporate crystal-axis data due to olivine's well-defined anisotropic diffusivity tensor (e.g., Kahl et al., 2011). While diffusion anisotropy is not fully constrained for orthopyroxene (Krimer and Costa, 2017), we show here that its effects can be limited using quantitative crystal-orientation data. By analyzing thin sections, crystalaxis orientation measurements were performed rapidly by EBSD and integrated with BSE imagery and analytical data, while petrologic context was retained. Our results indicate eruptions occurred 0.4-27 days (with 50 of 54 crystals recording eruption <10 days) after the initiation of overgrowths following magma mixing events   (Fig. 3). For optimally oriented crystals, the average time scale is 4.1 days (n = 5) for holocrystalline sample CC089 and 1.3 days (n = 3) for glassy sample CC174. This contrast holds for the full data set, with average time scales of 5.9 days for 33 crystals from holocrystalline samples and 1.2 days for 21 crystals from glassy samples. Our data imply that post-emplacement cooling can skew short-term modeled diffusion time scales for crystals collected from the interiors of slowly cooled lava flows. Crystals from glassy samples are more likely to record accurate magma mixing time scales unaffected by continued diffusional relaxation during lava emplacement and prolonged cooling. Our results   indicate that if magma ascent yields detectable geophysical signals, the eruption of lava with volumes large enough to impact local infrastructure could occur with only days of warning at Ruapehu. Although the mixing-to-eruption period occurred over days to weeks in each case, 40 Ar/ 39 Ar ages for lavas indicate that this process occurred repeatedly, with repose periods of centuries or more between successive eruptions at that time (Conway et al., 2016). Recognizing similarly recurrent patterns in future eruptive scenarios would provide greater confidence for long-term hazard forecasting.

CONCLUSIONS
We conducted a petrological and geochemical investigation of high-Mg andesitedacite (HMAD) lava flows that were erupted at Ruapehu volcano, as a case study of rapid magma genesis and eruption at a continental arc stratovolcano. The HMAD magmas were generated by mixing between mantle-derived, high-Fo olivine-bearing mafic magmas and mid-crustal felsic mush bodies. Disruption of the mush by mafic recharge is recorded by disaggregated orthopyroxene crystals, which display sharp reverse zoning of their rims. Magma mixing-to-eruption time scales, based on Fe-Mg interdiffusion modeling constrained by crystalorientation data, range from 0.4 to 27 days, with 50 of 54 crystals recording <10 days. Lava flows are the dominant feature of Ruapehu volcano; however, no effusive eruptions of lava onto its flank have occurred in recorded history. The rapid mobilization and eruption of mid-crustal magmas reported here provides valuable context for eruptive scenarios at active arc volcanoes and indicates that the onset of future effusive eruptions at Ruapehu may occur with very little warning.