Edinburgh Research Explorer Storage, evolution, and mixing in basaltic eruptions from around the Okataina Volcanic Centre, Taupō Volcanic Zone, Aotearoa New Zealand

Abstract


Introduction
Volcanic arcs are characterised by complicated sub-surface architectures that convert basaltic mantle-derived melt into a wide variety of more evolved arc magma compositions (e.g., reviews by Ducea et al., 2015;Grove et al., 2012).Compositional variability can be derived from variations in the primary composition of the mantle melt input, extents of crustal assimilation, type of petrological processes occurring (e.g., crystallisation, degassing, mixing), and the conditions of magma stagnation (pressure, temperature).Static models of the evolution of melt composition in the crust drive compositional variation by changes in temperature (e.g., Annen et al., 2006), whereas dynamic models drive compositional variation by reactive melt percolation (e.g., Jackson et al., 2018); both mechanisms have been used to explain the compositional variability of arc magmas.Reconciling these models requires observations and analysis of volcanic rocks or exhumed crustal sections, which provide snapshots and timeintegrated histories, respectively, of magmatic systems.
Both crustal and erupted materials at arcs are dominated by evolved magma compositions (i.e., andesites to rhyolites) despite the large inputs of basaltic melt required for their formation.Most basalts never reach the surface due to relatively high magma density compared to the surrounding crust.Furthermore, these intrusions cool in the crust and either solidify to gabbroic plutons or generate more evolved magmas that separate and ascend to then erupt or cool to form felsic plutons.Periodic magma mixing (e.g., basalt with rhyolite) may be important in generating intermediate magmas and triggering eruptions (e.g., Laumonier et al., 2014;Sparks et al., 1977).Any basaltic magmas that do reach the surface will have traversed this complicated crustal region, yet unravelling this cryptic differentiation history is not trivial and inevitably requires high resolution, in situ mineral analysis.Here, we utilise microanalytical geochemical methods to collect data on crystals and their melt inclusions to explore the paths taken by basaltic magmas beneath a dominantly rhyolitic caldera.We aim to constrain how and where basaltic magmas are stored within the crust, and what petrological processes affect them.This is important for assessing the current state of magma reservoirs in the crust in the context of geophysical surveys and predicting potential precursory signals before a future eruption at caldera systems.
The Okataina Volcanic Centre (OVC) is one of two currently active caldera systems in the Taupō Volcanic Zone, Aotearoa New Zealand (Taupō Volcanic Centre is the other).From several studies of the rhyolites, the sub-surface architecture below the OVC is known to comprise discrete rhyolitic melt-mush pockets that erupt compositionally distinct magmas within single eruptions (e.g., Cole et al., 2014;Sas et al., 2021;Shane et al., 2008aShane et al., , 2007;;Smith et al., 2004;Storm et al., 2011).Basaltic magmas are key to generating the more evolved magma compositions in the OVC, but little is known about their evolution.Heat and volatiles are assumed to be transferred between basalts and rhyolites to trigger rhyolitic eruptions (e.g., Leonard et al. 2002;Shane et al. 2007Shane et al. , 2008;;Smith et al. 2010), but the initial volatile contents of the basaltic magmas are largely unconstrained.The abundant evidence for basaltic-rhyolitic magma interaction also enables the investigation of how magma mixing is related to basaltic eruption style (e.g., Leonard et al., 2002;Shane et al., 2005).In this study, we combine textural observations with mineral and melt inclusion chemistries to constrain the magmatic compositions (including volatiles), conditions, and processes occurring during crustal storage and ascent of basaltic magmas around the OVC.

Regional setting
The Taupō Volcanic Zone (TVZ) is the most frequently active and productive silicic system on Earth (Wilson et al., 2009).Oblique subduction of the Pacific plate under continental Zealandia leads to the clockwise rotation of the eastern portion of the North Island, resulting in extension in the TVZ, crustal thinning, and basalt underplating (Houghton et al., 1995;Mortimer et al., 2017;Wilson et al., 1995).High rates of basaltic underplating drive the generation of voluminous silicic magma and, together with the relatively thin and faulted crust, enhance magma production and the high frequencyrate of eruptions (e.g., Cole et al., 2014;Price et al., 2005).Extensive crustal contamination occurs, which influences the isotopic composition of both erupted basalts and rhyolites (e.g., Gamble et al., 1993;Graham et al., 1995;Sas et al., 2021;Waight et al., 2017).
Basalts scoria cones and tuffs are volumetrically minor surface featuresat the surface compared to andesites/dacites (which are one order of magnitude greater in volume), and rhyolites (twoto-three orders of magnitude greater in volume) being one and two-to-three orders of magnitude less voluminous than andesites/dacites and rhyolites respectively (Wilson et al., 1995).Basalts of the TVZ are classified as high-alumina and are generated by a combination of rift-induced decompression melting and fluid-induced flux melting (Hiess et al., 2007;Law et al., 2021).Active calderas have high inputs of basalt from the mantle wedge, which is caused by fluidfluxed melting of fertile mantle, i.e. mantle that has not undergone much previous melting (Barker et al., 2020;Zellmer et al., 2020).The mantle source under these calderas is lherzolitic, as the sub-continental lithospheric mantle found further south in the TVZ has been removed by rifting and crustal thinning (Law et al., 2021).,The removal of the sub-continental lithospheric mantle in the northern TVZ causeding the shift to rhyolitic rather than andesitic volcanism (Law et al., 2021).Regions without active calderas have lower inputs of basalt due to either a subdued influence from fluid-fluxing or a more depleted mantle source (Barker et al., 2020;Zellmer et al., 2020).In the latter mechanism, the depleted source is caused by prior melt extraction associated with caldera formation in the region, but these calderas are no longer active (Zellmer et al., 2020).Basaltic eruptions throughout the TVZ are often associated with faults and commonly erupt in association with rhyolitic magmas (Cole, 1970a;Hiess et al., 2007;Nairn and Cole, 1981).Basaltic volcanism exhibits a wide range of eruption style, both within and between individual eruptions and volcanic centres, and shallow conduit processes (including interaction with external, non-magmatic water) are thought to play a major role in determining eruption style (e.g., Carey et al., 2007;Houghton and Hackett, 1984).
The currently active OVC is overwhelmingly rhyolitic, but a diverse range of styles and intensities of basaltic explosive activity is also present within and outside the caldera boundary (Cole et al., 2014;Nairn, 2002) (Error!Reference source not found.Figure 1a).Since ~55 ka there have been at least six basaltic eruptions (and additionally two examples of mafic enclaves and blebs in exclusively rhyolitic eruptions) in this region, ranging from phreatomagmatic to magmatic and Strombolian to Plinian in intensity (Table 1Table 1 and Error!Reference source not found.Figure 1) (Cole et al., 2014;Nairn, 2002).Basaltic Plinian eruptions are rare in the geological record, and Tarawera is the one of the most recent (Cole, 1970a;Nairn, 1979;Rowe et al., 2021;Thomas, 1888;Walker et al., 1984).;Beanland, 1989;Burt et al., 1998;Darragh et al., 2006;Nairn, 2002Nairn, , 1992)); deposit thickness isopleths or extent limit (solid or dashed cruves; Beanland, 1989;Darragh et al., 2006;Nairn, 1992;Pullar and Nairn, 1972) ; and sample locations for this study for the basaltic eruptions (circles).Colour indicates eruption as shown in (b) -eruptions analysed in this study are in colour and other basalts from around the OVC are shown in grey.Inset shows the location of the main map and the Taupō Volcanic Zone (TVZ, shaded area) in the North Island of Aotearoa New Zealand.M = Matahi, where the dotted-grey line is the extent limit; and O = Okareka, where the solid-grey lines are the 1 and 5 cm isopachs.
(b) Qualitative eruption intensity against age (Buck et al., 2003;Hogg et al., 2003;Hopkins et al., 2021;Nairn, 2002;Newnham et al., 2003;Peti et al., 2021) for OVC basaltic magmas -Rerewhakaaitu and Kaharoa do not appear in (a) because they only occur as basaltic enclaves and blebs within a rhyolitic eruption.Notes: *Eruptions analysed in this study.† The Matahi eruption occurred just prior to the Rotoiti Ignimbrite that was the most recent OVC caldera-forming eruption.Volumes (DRE = dense rock equivalent) for Terrace Rd and Rotomakariri are not determined (n.d.), but are likely small due to their limited occurrence (Nairn, 2002).

Methods
We sampled and analysed material from the Terrace Rd, Rotomakariri, Rotokawau, and Tarawera eruptions as they cover the full range of eruption styles and sizes (phreatomagmatic to magmatic and Strombolian to Plinian) observed around the OVC (Error!Reference source not found.Figure 1b).These eruptions occurred in an active caldera region, but Rotomakariri and Tarawera occurred inside the caldera boundary (along one of the main linear vent zones), whereas Terrace Rd and Rotokawau occurred outside the caldera boundary (Error!Reference source not found.Figure 1a).There are no published melt inclusion data for Terrace Rd, Rotomakariri, and Rotokawau, and only limited published data for Tarawera (Barker et al., 2020;Rowe et al., 2021); melt inclusions have been previously analysed from Okareka and Kaharoa (Barker et al. 2020).
Samples were collected during three fieldwork seasons between 2015 and 2017 (Figures 1a).
Localities for deposits for the Terrace Rd, Rotomakariri, and Rotokawau eruptions were taken from Nairn (2002); exact sample locations and descriptions are given in Supplementary Material (including for Tarawera samples).For Tarawera, samples were collected off the volcano to avoid material that had cooled slowly, which can enhance post-entrapment crystallisation of melt inclusions (e.g., Lloyd et al., 2013).Samples were dried in a lowtemperature oven then sieved into 1 φ size fractions.The clast densities for -3 to -6 φ from Terrace Rd, -4 to -5 φ from Rotomakariri, -3 to -4 φ from Rotokawau, and -4 to -5 φ from Tarawera were measured using the method of Houghton and Wilson (1989).Vesicularity was calculated assuming rock density was equal to that of an anhydrous melt (assumed to approximate the glass density) with the composition of the average whole rock data from the literature (given in Supplementary Material), calculated at room temperature and pressure using DensityX (Iacovino and Till, 2018).Two mean density samples were chosen to make thin sections from (random samples were selected for Rotokawau from a different location due to the small clast size sampled during our fieldwork).
Olivine, pyroxene, and plagioclase mineral separates were analysed using electron probe micro-analysis (EPMA) wavelength dispersive spectrometry (WDS).Unless otherwise stated, all analyses were taken from crystal cores.Melt inclusions from all eruptions were analysed using EPMA-WDS for major, minor, and volatile (S, Cl, and F) elements and for H2O using calibrated volatiles-by-difference (Hughes et al., 2019a).A subset of melt inclusions from Tarawera was analysed for H2O and CO2 using secondary ion mass spectrometry (SIMS) prior to EPMA.To put mineral separate data into context, textural observations on thin sections were made using optical microscopy and scanning electron microscopy (SEM).Some mineral phases (and the groundmass glass for Rotomakariri) in the thin sections were analysed using semi-quantitative (sq) SEM energy dispersive spectroscopy (EDS) (sq-SEM-EDS) and EPMA-WDS to correlate the textures with mineral separates data.
We compiled mineral, melt inclusion, and whole rock data from the literature, particularly from basaltic eruptions not analysed in this study (e.g., Matahi, Okareka, Rerewhakaaitu, and Kaharoa), to expand upon our dataset.Several thermometers (melt, olivine-melt, clinopyroxene-melt, and clinopyroxene-orthopyroxene; Putirka, 2008), oxybarometers (melt Fe 3+ /FeT; Kress and Carmichael, 1991), and barometers (clinopyroxene-melt; Neave and Putirka, 2017, and H2O-CO2 melt concentrations; Ghiorso and Gualda, 2015), as well as modelling using rhyolite-MELTS (Ghiorso and Gualda, 2015;Gualda et al., 2012), were applied to mineral, melt inclusion, and whole rock data from this study and the literature.Data collection and reporting for melt inclusions broadly follows the guidelines of Rose-Koga et al. (2021).Full analytical and calculation details, as well as all data collected and compiled, are provided in Supplementary Material.

Textural and chemical characteristics
Texturally and chemically, Tarawera, Rotokawau, and Terrace Rd scoria are more similar to each other than to scoria from Rotomakariri.evol ved encl ave (from Terrace Rd) cont ai ni ng group t hree pl agi ocl ase and quart z, wi t h some evi dence of react i on wi t h basal t i c mel t at t he margi nsal i gnment of pl agi ocl ase f rom basal t -basal t mi xi ng (bl ue l i ne) and regi on of evol ved mat er i al i s out l i ned i n red.
Terrace Rd, Rotokawau, and Rotomakariri contain abundant macrocrysts, mostly as glomerocrysts (Figure 2Figure 2i-n), whereas Tarawera is almost macrocryst-and glomerocryst-free (see also Law et al., 2021).All eruptions have a similar mineralogy of olivine, plagioclase, and clinopyroxene, with Rotomakariri additionally containing abundant orthopyroxene.Alkali feldspars and quartz were found in all eruptions.Multiple groups of mineral compositions are observed across eruptions, which are outlined in Table 2. Olivine composition varies between eruptions, with a narrow range in forsterite (Fo) content at Terrace Rd (Fo76-79) and Tarawera ( Fo7 9 -85 ) and a wi de r ange at Rot omakar i r i ( Fo7 1-8 2) and Rot okawau ( Fo7 3 -8 2 ) ( F i g u r e 3 Fi gur e 3a-e) .Our dat a suppor t t he f i ndi ngs of Law et al . ( 2021) , wher e ol i vi nes f r om Ter r ace Rd, Tar awer a, and Rot okawau ar e gr oup 1 and t hose f r om Rot omakar i r i ar e gr oup 2 ( Tabl e 2) .Groundmass olivine from Tarawera analysed by Rowe et al. (2021) has lowera similar Fo tohan the macrocryst al s (Fi gur e 3Fi gure 3a).Two groups of cl i nopyroxene are f ound i n al l erupt i ons: group one has hi gh Mg#, whereas group t wo has l ow Mg#Cl i nopyroxene composi t i on i s si mi l ar across al l er upt i ons, i ncl udi ng groundmass clinopyroxene from Tarawera reported by Rowe et al. (2021), with Mg# 69-87 ( Tabl e 2 and F i g ur e 3Fi gur e 3f -k) .Gr oup t wo i ncl udes t he Tar awer a gr oundmass cl i nopyr oxene r epor t ed by Rowe et al . ( 2021) .Or t hopyr oxene i s common onl y i n Rot omakar i r i , occur r i ng as t wo gr oups and i s bi modal i n composi t i on (Figure 3Figure 3h).Group one (hHigh-Mg# orthopyroxene) (83-87) is found as macrocrysts in Rotomakariri (and rarely Rotokawau and Tarawera) and sometimes as inclusions in lower Mg# clinopyroxene grains at Terrace Rd and Rotokawau.Group two (lLow-Mg# orthopyroxene (67-83)), sometimes contains inclusions of apatite, and occurs in all eruptions.Plagioclase displays a wide range of anorthite contentis present in three groups (Table 2 and Figure 3Figure 3k-o).Group oneSome plagioclase grains have cores are very calcic (>An90) cores with coarse sieving and normal zoning to a thin, unsieved, less calcic rim, referred to as high-An plagioclase.Group twoMany plagioclase crystals have lower An (55-90), are mostly unzoned and occur as both macrocrysts and inclusions in clinopyroxene at Terrace Rd, in low-Mg# orthopyroxene at Rotokawau, and in both clinopyroxene and orthopyroxene at Rotomakariri.This plagioclase composition is similar to rims on group onethe highly calcic plagioclase and plagioclase microlites in the Tarawera groundmass (Rowe et al., 2021).For both group one and two plagioclase with >An55, FeO content is high (>0.4wt%) and decreases with increasing An (see Figure S6 in Supplementary Material).There are some Group three plagioclases are with low in An (<55) and FeO (<0.4 wt%), and texturally variable, termed low-An plagioclase.Unlike mineral groupscompositions, which are shared similar across different eruptions, glomerocryst types are unique to individual eruptions.More detailed descriptions of both the minerals groups and glomerocrysts types are provided in the Supplementary Material.
Notes: Blank space indicates textural type not present.

Melt inclusions
The Crystallisation, diffusion, and bubble-formation can alter major and volatile element chemistry of melt inclusions after entrapment (e.g., Barth et al., 2019;Barth and Plank, 2021;Bucholz et al., 2013;Danyushevsky et al., 2000Danyushevsky et al., , 1988;;Dungan and Rhodes, 1978;Gaetani et al., 2012;Gaetani andWatson, 2002, 2000;Hartley et al., 2014Hartley et al., , 2015;;Lowenstern, 2003Lowenstern, , 1995;;Moore et al., 2015;Nielsen et al., 1998;Rasmussen et al., 2020;Roedder, 1979;Saper and Stolper, 2020;Schiano, 2003;Sobolev and Shimizu, 1993;Wallace et al., 2015).On the basis ofBased on mineral-melt exchange equilibria, most melt inclusions were not in equilibrium with their host crystal indicating post-entrapment crystallisation (PEC) has occurred, except three clinopyroxene-and six plagioclase-hosted melt inclusions(Figure S3a).This indicates some post-entrapment crystallisation has occurred.The effects of PEC were corrected for using the method of (Collins et al., (2022) and only data with <|15|% PEC-correction are used (see Supplementary Material for details).The effect of post-entrapment crystallisation on major, minor, and volatile element trends was evaluated by adding back the composition of the host mineral until equilibrium between the calculated melt composition and host mineral composition was achieved (further details provided in Supplementary Material).The mean/maximum post-entrapment crystallisation correction (excluding plagioclase-hosted melt inclusions) are 6/23 % for Tarawera, 5/14 % for Terrace Rd, 3/9 % for Rotomakariri, and 4/28 % for Rotokawau.Although Fe-Mg diffusion may also be important, it is not possible to evaluate its effect on the clinopyroxene-hosted melt inclusions currently; for this reason, we focus on trends in oxides other than MgO and FeO.
Even when assuming the maximum degree of post-entrapment crystallisation (without Mg-Fe diffusion) for each clinopyroxene-, orthopyroxene-, and olivine-hosted melt inclusion, trends in major and volatile element chemistry do not change from those observed using the raw data (details in Supplementary Material).For instance, the positive correlation between SiO2 and Al2O3 and K2O, and negative correlation between SiO2 and CaO are robust.Moreover, Rotomakariri melt inclusions remain much more evolved than the other melt inclusions.Hence, these trends reflect pre-entrapment processes for major elements.For this reason, uncorrected (i.e., raw) melt inclusion compositions are used throughout, and we focus on SiO2, Al2O3, CaO, and K2O (Figure 4).The effect of 10 % crystallisation on plagioclase-hosted melt inclusions was modelled to see its effect on trends in melt inclusion chemistry.This showed that the low Al2O3 concentrations are likely due to post-entrapment effects, but the difference in CaO compared to clinopyroxene-hosted melt inclusions is likely a pre-entrapment feature (Figure 4).for volatiles contained in co-existing vapour bubbles (i.e., composition and size of vapour bubbles were not measured) to reconstruct bulk melt inclusion compositions.CO2 is greatly affected by bubble formation, whilst H2O, S, and Cl are less affected due to lower partitioning into the vapour phase and/or potential kinetic effects (e.g., Hartley et al., 2014;Maclennan, 2017;Moore et al., 2015;Rasmussen et al., 2020;Wallace et al., 2015).Rather than add additional uncertainty related to reconstructing the original melt composition, we assume CO2 concentrations represent minimum estimates of the CO2 content of the melt, and do not try and fit degassing trends to our data.Bulk (i.e., melt + bubble) H2O content can additionally be altered by diffusion into or out of the melt inclusion (e.g., Barth et al., 2019;Barth and Plank, 2021;Bucholz et al., 2013;Gaetani et al., 2012;Hartley et al., 2015Hartley et al., , 2014)). .The possibility of de/rehydration is considered for each eruption.
Basaltic to basaltic-andesite melt inclusions are similar in in group one and group two clinopyroxenes and olivines from Terrace Rd, Rotokawau, and Tarawera, but were not found  2020) and some of these melt inclusions overlap with the whole rock data (Error!Reference source not found.Figure 4).There is a wide-range of Al2O3 at a given SiO2 (15-19 wt% Al2O3 at 50 wt% SiO2) with no correlation (Error!Reference source not found.Figure 4a).CaO decreases with increasing SiO2, with some melt inclusions having higher CaO at a given SiO2, which overlaps with the whole rock data (Error!Reference source not found.Figure 4b).K2O increases with SiO2 for both melt inclusion and whole rock data (Error!Reference source not found.Figure 4c).Basaltic to basaltic-andesite melt inclusions from Tarawera, melt inclusions have a wide range in H2O contents (0-5.5 wt%), whereas H2O concentrations at Terrace Rd (2.0-4.4 wt%) and Rotokawau (0.8-3.1 wt%) have a more limited range (Error!Reference source not found.Figure 5a and e).(2.2-4.8 wt% H2O) The lack of correlation between H2O and K2O suggests crystallisation accompanied by concentration or degassing has been overprinted by diffusive water-loss and hence H2O contents are a minimum (Error!Reference source not found.Figure 5e).Chlorine concentrations are lowest for Terrace Rd (~1000-1100 ppm), t hen and Tarawera ( ~850-1700 ppm), and have si mi l ar chl or i ne concent r at i ons ( 1110-1880 and 630-1870 ppm Cl respect i vel y) t hat ar e l ower t han Rot okawau has t he hi ghest ( ~12501300-2730 2300 ppm) Cl, which positively correlates with K2O (Error!Reference source not found.Figure 5b).Total sSulfphur (ST) has a similarly wide range in all three eruptions (~350-3980 3300ppm ST; Error!Reference source not found.Figure 5c) and fluorine concentrations are also similar (~30290-10100 ppm F; Error!Reference source not found.Figure 5d), neither of which correlate with K2O.CO2 (74-831433-756 ppm) was measured for a subset of Tarawera melt inclusions only (Error!Reference source not found.Figure 5f).For S > 1000 ppm, S and Cl positively correlate, whereas for S < 1000 ppm they negatively correlate (Error!Reference source not found.Figure 5g   in clinopyroxene (this study; Rowe et al., 2021), olivine (Barker et al., 2020), plagioclase, orthopyroxene and quartz (Rowe et al., 2021); olivine-hosted melt inclusions for Okareka and Kaharoa (Barker et al., 2020); glass analyses from mafic blebs from Rerewhakaaitu (Shane et al., 2007); and melt inclusions and groundmass glass from OVC rhyolites (Johnson et al., 2011).

l t i c w i t h e i t h e r t h e s a m e ( s i n g l e g r a i n ) o r l o w e r A l 2 O 3 . T h e y a r e v o l a t i l e -p o o r i n c o m p a r i s o n t o c l i n o p y r o x e n e -h o s t e d m e l t i n c l u s i o n s , a n d K 2 O n e g a t i v e l y c o r r e l a t e s w i t h S T a n d C l ( s i n g l e g r a i n f r o m t h i s s t u d y ) . T h e b a s a l t i c -a n d e s i t e i n c l u s i o n s h o s t e d i n o r t h o p y r o x e n
Rotomakariri melt inclusions hosted in group two clinopyroxene and group one orthopyroxene and groundmass glass are mostly andesitic (two are dacitic), with low CaO and Al2O3 and high K2O (Error!Reference source not found.Figure 4).H2O and ST are lower than most of the basaltic to basalticandesite melt inclusions, although similar to the low-sulfur (ST <1000 ppm) set of clinopyroxene-hosted melt inclusions (Error!Reference source not found.Figure 5a and c).Chlorine is high and similar to Rotokawau (Error!Reference source not found.Figure 5b); fluorine is much higher than any of the basalts (Error!Reference source not found.Figure 5d).At Tarawera, a few clinopyroxene-hosted melt inclusions are also andesite-dacite, but have higher Al2O3 , similar CaO, and lower K2O than Rotomakariri melt inclusions (Error!Reference source not f ound.Fi g ur e 4) .A s i ng l e andes i t e mel t i ncl us i on hos t ed i n an Na -r i ch pl ag i ocl ase f r om T ar aw er a has si m i l ar A l 2 O3 t o t he o t h e r T ar aw er a a n d es i t e -da c i t e m e l t i n c l u s i o ns , a l t h o ug h i t s K 2 O r es em b l es R o t oma k a r i r i a n d es i t e m e l t i n c l us i o ns .D a c i t e-r R hy o l i t e i c m e l t i n c l u s i o ns a n d g r o u ndm as s g l as s a r e as s o c i a t ed w i t h group two orthopyroxene and quartz from Rotokawau and Tarawera (Error!Reference source not found.Figure 4) and, have low H2O, variable Cl, and have very low ST (0-70 ppm), and low F (Error! Reference source not found.Figure 5).
5 Pre-and syn-eruptive storage, evolution, and mixing of multiple magmas The bulk magma (i.e., whole rock) composition erupted in basaltic eruptions (and found as basaltic enclaves in rhyolitic eruptions) from around the ŌVC is similar (Figure 4).However, the different compositional groups of clinopyroxene, orthopyroxene, and plagioclase, in combination with different melt inclusion compositions, indicate that multiple components are found across these basaltic eruptions (Figure 3 and Figure 4).Therefore, it is useful to group these components when discussing magmatic evolution during crustal storage.Based on the mineral and melt inclusion compositions there are five different components.There are two basalt to basaltic-andesite components found in all eruptions (Basalt-1 and Basalt-2); two andesite components that are much less common and more specific to the Rotomakariri (Rm) and Tarawera (Tw) eruptions (Andesite-Rm and Andesite-Tw); and a minor amount of rhyolite component found in all eruptions (Table 3).These components are repeatedly sampled as the groups of mineral types and melt inclusion compositions are common to many different eruptions.Similar melt inclusions (albeit more primitive) and mineral chemistries are found in other basalts from around the ŌVC (e.g., Kaharoa, Okareka, Matahi, and Matahina) and even in Taupō Volcanic Centre (TVC) basaltic material (e.g., Oraunui), showing that these are common features within the TVZ (Allan et al., 2017;Barker et al., 2020;Deering et al., 2011;Rooyakkers et al., 2018;Wilson et al., 2006).Each component may reflect differences in source (e.g., initial magma composition due to degree of slab influence), storage conditions (e.g., pressure, temperature, oxygen fugacity), processes (e.g., varying degrees of cooling-or decompression-induced crystallisation or crustal assimilation), physical state (mush-like in the crust, or solidified as an intrusion or after eruption at the surface) or combinations thereof.We use oxy-thermo-barometry (Figure 6), rhyolite-MELTS modelling (Figure 7), and comparison to experiments to explore the crystallisation conditions of Basalt-1, Basalt-2, and Andesite-Rm melts (calculation details are in Supplementary Material).As the same components occur in eruptions separated spatially and temporally, these sets of conditions must be common around the ŌVC even though magmas themselves were not sourced from the same spatio-temporal reservoir.The textures observed in each eruption reflect different processes during ascent, such as magma mixing and microlite crystallisation.

Polybaric crystallisation and mixing
Basalt-1 encompasses mMost of the mineral and melt inclusion analyses in this study, namely type one clinopyroxene and their melt inclusions, An50-90 type two plagioclase, and their melt inclusions; and the groundmass material (except for Rotomakariri) belong to a basaltic magma (Figure 3Figures 3 and Error!Reference source not found.Figure 4).Olivine compositions show Basalt-1these melts are not mantle-derived, but have already undergone crystallisation deeper in the system (Law et al., 2021) .Bot h gr oup one and t wo ol i vi nes as def i ned by Law et al . ( 2021) coul d have been der i ved f r om Basal t -1, wher e gr oup t wo ol i vi nes cr yst al l i sed deeper i n t he syst em and ar e sour ced f r om cumul at es .Al t er nat i vel y, gr oup t w o ol i vi nes may der i ve f r om a separ at e basal t i c mel t .crystallisation is responsible for the compositional range of whole rock and melt inclusion data.The narrower spread in compositions and temperatures for whole rock and minerals compared to melt inclusions is consistent with the basaltic bulk composition of the system.Conversely, melt inclusion compositions reflect local changes in temperature and associated crystallisation, recording the melt present in the primary mush system near the solidus.The mushy nature of storage is also evidenced by abundant glomerocrysts in Terrace Rd, Rotomakariri, and Rotokawau (Figure 2Figure 2i-n).A wide range of pressures (~7 kbar to surface) is derived from melt-clinopyroxene barometry, with most estimates <3 kbar (Figure 6a).The highest H2O-CO2 measurements require some melts to be derived from at least 3 kbar (Figure 6ab).These estimates overlap with pressure-temperature estimates for Tar awer a f r om Rowe et al . ( 2021) ., and i mpl y pol ybar i c st or age of basal t i c magmas , especi al l y gi ven t he l ar ge model er r or s associ at ed wi t h t hi s bar omet er ( ±0. 14 GPa; Put i r ka, 2008) .The i ncr ease i nhi gh Al 2 O3 wi t h i ncr easi ng Si O2 r equi r es pl agi ocl ase cr yst al l i sat i on t o have been suppressed, suggesting differentiation at higher pressures (e.g., Blatter et al., 2013;Marxer et al . , 2021;Münt ener and Ul mer, 2018a;Nandedkar et al . , 2014).The l i mi t ed l i t erat ure whol e rock Fe 3+ / FeT dat a i mpl y rel at i vel y oxi di sed condi t i ons (~ΔNNO+1).Rhyol i t e-MELTS modelling suggests the compositions of melt inclusion and whole rock data are not formed by fractional crystallisation of a single parental melt composition at a unique pressure and oxygen fugacity (Figure 4Figure 4d-f).However, the range in melt inclusion compositions is bracketed can be derived by fractional equilibrium crystallisation during cooling (to 101050 °C) from a single melt composition deeper than 0.3 GPa at relatively oxidised (ΔNNO=0 to +2) and H2O-saturated conditions at depths between 1 and 7 kbar (Figure 4Figure 4d-f).
The oxygen fugacity range from Rhyolite-MELTS agrees with Tthe limited literature whole rock Fe 3+ /FeT data, which impliesy relatively oxidised conditions (~ΔNNO+1).The handful of andesite-dacite Tarawera melt inclusions Andesite-Tw is chemically distinct from Andesite-Rm and melt inclusions record a lower temperature of ~9850-1000 °C ( pr essur es coul d not be est i mat ed f r om t he avai l abl e dat a, Fi gur e 6ba) and.As evi dence f or Andesi t e-Tw i s onl y f ound i n a f ew mel t i ncl usi ons at Tar awer a, i t i s not consi der ed vol umet r i cal l y i mpor t ant ar ound t he ŌVC ( Fi gur e 4) .Rhyol i t e-MELTS model l i ng did not recreate this composition (Figure 4Figure 4d-f) from the same initial magma composition used for Basalt-1 and Andesite-Rm.crystallisation of a single composition of basaltic magma (Figure 6a).Given the large model uncertainties (e.g., ±1.4 kbar for cpx-m; Putirka, 2008), individual magma reservoirs cannot be distinguished.Additionally, the detailed mineral compositions (e.g., olivine; Figure 3Figure 3a-e) and glomerocryst textures are distinct (Figure 2Figure 2i-n) indicating evolution in discrete, isolated pods prior to eruption, consistent with the temporal and spatial spread of eruptions.The range in major element composition at a single SiO2 (especially obvious in Al2O3; Figure 4Figure 4d) is then a result of mixing between these basaltic magmas.Interaction between basaltic magmas is also evidenced by the tFirstly, there is evidence for the mixing of multiple basaltic magmas.Textures in the scoria, which are indicativee of mixing between different batches of basaltBasalt-1 and Basalt-2 that have subtly different crystallisation conditions (i.e., come from different places in the magmatic system) or decompression histories (e.g., T-H2O conditions).At Tarawera, there are multiple instances of Basalt-1,this includinges the carrier melt (as represented by the macrocryst-poor whole rock composition), and the low/high-ST melt inclusions (Error!Reference source not found.Figure 5g).A similar picture applies to Rotokawau, where mingled groundmass textures suggest multiple basaltic carrier melts from Basalt-1 (Figure 2Figure 2c).For Terrace Rd, the small glomerocrysts could be phenocrystic or antecrystic, whereas the large glomerocrysts, as well as the large orthopyroxene crystals (Figure 2e, i, and m), are antecrystic and there is evidence for multiple melts in the groundmass (Figure 2Figure 2e, i, and m).Hence, these basalts evolved from a single, primitive, oxidised magma in distinct, isolated, mushy pods due primarily to cooling-induced crystallisation, with some addi t i onal degassi ng, and mi xed dur i ng ascent .I t i s not cl ear whet her al l t he ant ecr yst i c mat er i al came f r om t he same pl ace or event and how much mel t was t r anspor t ed wi t h t he mi xi ng event , al t hough t her e i s evi dence f or mul t i pl e mel t s i n t he gr oundmass.Despite broadly similar melt chemistry between eruptions, the detailed mineral compositions (e.g., olivine and clinopyroxene, Figure 3) and glomerocryst textures are distinct (Figure 2i-n) indicating evolution in discrete, isolated pods prior to eruption, consistent with the temporal and spatial spread of eruptions.
Amphibole was not observed in the eruptions studied here; it has been described only in the groundmass of basaltic and gabbroic enclaves from the Kaharoa eruption, where amphibole crystallisation is thought to have been triggered by a late-stage increase in H2O, possibly due to interaction with rhyolite (Leonard et al., 2002).Most of the melt inclusions at Tarawera, Rotokawau, and Terrace Rd record temperatures that are too high (>1050 °C) for amphibole stability despite their relatively high H2O concentrations (Figure 7Figure 7) (Foden and Green, 1992).This suggests that basalt-rhyolite mixing prior to the Kaharoa eruption moved the magma into the amphibole stability field by cooling, rather than by increasing its H2O content.For Tarawera and Rotokawau, melt inclusions can be divided into two sub-groups based on S concentrations above and below ~1000 ppm (Error!Reference source not found.Figure 5g), .requires two separate regimes of crystallisation as previously observed by Rowe et al. (2021).We suggest that these regimes correspond to isobaric cooling and decompression-induced degassing (Error!Reference source not found.Figure 5g).Concentrations of S and Cl increase in the melt during crystallisation for melt inclusions with >1000 ppm S (Error!Reference source not found.Figure 5g),.This behaviour indicating indicates these elements behaved incompatibly (i.e., were not part i t i oned i nt o coexi st i ng sol i ds or exsol ved fl ui ds), (Fi gure 5).The magma may ei t her have been vol at i l e-undersat urat ed, such t hat t here was no f l ui d phase f or Cl or S t o part i t i on i nt o, such t hat or f l ui d-mel t part i t i on coeff i ci ent s for S and Cl at these conditions were very low (e.g., Gennaro et al., 2021;O'Neill, 2020;Tat t i t ch et al . , 2021;Thomas and Wood, 2021) .I f t he magma was i ni t i al l y f l ui d-under sat ur at ed, t hi s woul d cont r ast w i t h most ar c r egi ons wher e hi gh magmat i c CO2 concent r at i ons r esul t i n f l ui d-sat ur at i on deep i n t he cr us t ( e. g ., Wal l ace, 2005) .Rot ok awau and Tar aw er a mel t i ncl us i ons w i t h <1000 ppm ST show t he same t rend fori ncreasi ng chl ori ne Cl but t he opposi t e t rend forwi t h decreasi ng sul fur S (i .e. , decreasi ng ST wi t h cryst al l i sat i on, Error!Reference source not found.5.1.3Similar storage conditions and volatiles prior to eruptions of varying style Despite the similar pre-eruptive compositions and conditions of the basaltic magma, there is a wide variation in eruption style, and therefore no systematic relationship between storage conditions and eruption style (Error!Reference source not found.Figure 1b and 6).Bamber et al. (2019) suggested moderate storage temperatures (<1100 °C) are important for generating basaltic Plinian eruptions, which occur at Tarawera, but are also found for the smaller intensity eruptions (Figure 6).Volatile concentrations (H2O, Cl, and S) and trends are also similar between basaltic eruptions around the OVC (Error!Reference source not found.Figure 5).High H2O concentrations suggest H2O exsolution was important during ascent, which may drive basaltic Plinian eruptions (Bamber et al., 2019;Pérez et al., 2020).However, high H2O concentrations are found across the range of eruption styles and are therefore not unique to Tarawera (Error!Reference source not found.Figure 5a and f).Both Rotokawau and Tarawera have a population of melt inclusions that display degassing trends, and this population may have been missed at Terrace Rd where fewer melt inclusions were analysed (Error!Reference source not found.Figure 5g).The unique degassing path for Plinian eruptions compared to other explosive eruptions proposed by Moretti et al. (2018) led to lower Cl but higher S in less explosive eruptions compared to more explosive eruptions due to the differences in dehydration and sulphide-saturation that occur during crystallisation.However, the observed differences in S and Cl concentration around the OVC do not relate to eruption style: Rotokawau has higher S and Cl but eruption intensity intermediate between Terrace Rd and Tarawera, and there is no evidence for sulphide-saturation (Error!Reference source not found.Figure 5g).High CO2 concentrations are thought to be important for generating (sub-)Plinian basaltic eruptions (e.g., Allison et al., 2021;Sable et al., 2009), which could be important around the OVC.Unfortunately, our CO2 data for Tarawera are likely compromised by bubble formation and we do not have sufficient data to compare against smaller eruptions (Error!Reference source not found.Figure 5f).
External influences within the crust could also influence eruption style of basaltic magmas around the OVC.Basaltic eruptions around the OVC are tectonically controlled, as evidenced by the linear nature of their eruptive vents underlain by dikes (Nairn and Cole, 1981).Hence, these eruptions may be triggered by earthquakes, especially given the high melt H2O contents and mushy-nature of storage, which could also influence eruption style (e.g., Hamling and Kilgour, 2020;Seropian et al., 2021).Additionally, the presence and physical state (e.g., viscosity) of large silicic bodies in the crust could affect basaltic eruption style by impeding (or not) the ascent of basaltic magmas to the surface.The tectonics in addition to the complex nature of the crust around the OVC may therefore be important for generating the wide variety of basaltic eruption style observed in the region.

Mixing and entrainment during ascent influenced by eruption style
There is variable extents of mixing of multiple basaltic magmas prior to eruption (and also entrainment of xenocrystic high-An plagioclase described in Section 5.2).The implication is that a carrier magma interacted with multiple different basaltic magma bodies as it ascended through the crust, picking up crystals en route.This is also seen in differences in oxygen isotope compositions between crystals and groundmass in these eruptions (Law et al., in review).The timescales of these interactions were likely very short (e.g., to preserve multiple groundmass textures and produce the sharp rims of lower-An plagioclase around high-An plagioclase, Figure 2Figure 2c and e-h), and probably occurred during pre-eruptive magma ascent.The extent of mixing is correlated with eruption style and crystals were entrained during ascent.Lower intensity eruptions (Terrace Road, Rotokawau) contain a high proportion of macro-crystals, whereas

Field Code Changed
Tarawera has a negligible crystal cargo (0.5 vol%, Sable et al., 2009).The carrier melt entrained more crystals as it passed through the mushes in smaller eruptions than in the Plinian eruption.This difference likely reflects the faster ascent rate of Plinian magmas, rather than the cause per se of varying eruption style.
These eruptions (including Rotomakariri) additionally show entrainment of rhyolitic material (e.g., low-Mg# orthopyroxene, low-An plagioclase, quartz, alkali feldspars, and rhyolitic melt inclusions), which has a similar composition to OVC rhyolitic eruptions (Figure 3Figures 3 and Error!Reference source not found.4).The rhyolitic material appears to have been incorporated at a late-stage of magma ascent (e.g., sharp boundaries between basaltic and rhyolitic material, Figure 2Figure 2p), probably when the basaltic magma "punched" through previously erupted, cold residual rhyolite domes (i.e., solid material in the crust or erupted at the surface, then buried).This diversity of magma types and mixing dynamics sampled both in individual eruptions and across eruptions from around the OVC reflects the interplay between basaltic magma ascent rates and the distribution, composition, and rheological state of magma bodies both vertically and horizontally.As mixing timescales appear to be short for the basalts that reach the surface, precursory signals to basaltic explosive eruptions could be limited, as suggested by the observations of the Tarawera 1886 C.E. eruption (Keam, 1988).In summary, Basalt-1 is volatile-rich and evolved from a single, primitive, oxidised magma in distinct, isolated, mushy pods due primarily to cooling-induced crystallisation, with some additional degassing during ascent.High anorthite plagioclase (>An90) can be indicative of hydrous conditions (e.g., Panjasawatwong et al., 1995;Takagi et al., 2005).The high magmatic water contents would occur as melting is driven by fluid-fluxing of a fertile mantle in active calderas (e.g., Barker et al., 2020;Zellmer et al., 2020).Plagioclase-liquid hygrometry using Waters and Lange (2015) suggests 5-7 wt% H2O in the melt, yet the analysed melt inclusions are almost anhydrous (Figure 5).This suggests hydrogen loss from the melt inclusions via diffusion, either during storage in a low-H2O melt or degassing during ascent (e.g., Hamada and Fujii, 2007).The higher temperatures compared to Basalt-1 (up to ~1250 °C) recorded by the melt inclusions would then reflect their low H2O content due to dehydration (Figure 6a).Alternatively, the high anorthite content could be due to high Ca/Na in the melt and not reflect high water contents in the melt (e.g., Panjasawatwong et al., 1995).In this case, the low H2O and high temperatures could be characteristics of the primary melt.Their occurrence as iInclusions of high-An plagioclase in clinopyroxene indicates plagioclase crystallisation before clinopyroxene, which occurs at lower H2O.This may reflect the decompression melting source that is thought to dominate in intracaldera regions (Barker et al., 2020;Zellmer et al., 2020).Hence, decompression melting could also be a minor component of active calderas.However, further investigation is needed to distinguish between these different potential sources, in particular measuring melt inclusion compositions from these high-An plagioclase grains for volatiles and trace elements.

High-An plagioclase xenocyrsts
Group one plagioclase composition is not only found in basaltic material from around the ŌVC since ~55 ka, but also in basaltic material from the ~26.5 ka TVC Oruanui eruption (Allan et al., 2017;Rooyakkers et al., 2018;Wilson et al., 2006) and the ~330 ka ŌVC post-caldera deposits following the Matahina eruption (Deering et al., 2011).The ubiquity of group one plagioclase in spatially and temporally separated ŌVC (and TVC) basalts requires common crystallisation conditions.In summary, Basalt-2 is primitive and could either be hydrous and derived from fluid-flux mantle melting or dry and derived from decompression mantle melting.Further investigation is needed to unravel these processes.

5.455.3
Evolved magmas: Andesite-Rm, Andesite-Tw, and RhyoliteRotomakariri Rotomakariri consists of mostly Andesite-Rmandesite melt inclusions and matrix glass, containing group two clinopyroxene, and group onehigh-Mg# orthopyroxene, but also olivine and both medium-and high-An plagioclase , their melt inclusions, and the groundmass material (Fi gure 3Figures 3 and Error!Reference source not found.Figure 4).The high-An plagioclase are xenocrystic, as described in Section 5.2.Clinopyroxene-melt t her mobar omet r y suggest s cr yst al l i sat i on at 1050 ± 18 °C and 5 ± 3 kbar and The occur r ence of gr oup t wo cl i nopyr oxene i n ot her er upt i ons suggest s Andesi t e-Rm, al t hough uncommon i n t he ŌVC, i s not uni que t o Rot omakar i r i ( Fi gur e 3f -j ) .Tt wo-pyr oxene thermobarometry suggestss higher pressures (~0.6 ± 0.4 GPakbar, with large model uncertainties of ±0.3.2GPakbar) and overlapping temperatures (~1000950-1100 °C) (Figure 6b)., This is unusually hot for an andesite.Rhyolite-MELTS modelling suggests Andesite-Rmthe andesitic melt inclusions and matrix glass can form from a similar initial magma composition as Basalt-1the basalt characteristic of Tarawera, Terrace Rd, and Rotokawau.However, fractional equilibrium crystallisation is to a lower T (~950 °C), shallower (0.1 MPakbar, which contrasts markedly with the two-pyroxene barometry but within error of the clinopyroxene-melt barometry), and under more reducing conditions (ΔNNO-1 to 0) (Figure 7Figure 4Figure 4d-f).
Rotomakariri melt inclusion H2O contents are very low, but this could indicate diffusive loss of H2O, which is supported by many Rotomakariri melt inclusions being crystallised (these were not analysed; Figure 5Figure 5a).The low Cl and ST concentrations indicates partitioning into a coexisting fluid (Figure 5Figure 5b, c, and g).This is expected at low pressures and the hot, dry melt

Field Code Changed
Formatted: Font: 12 pt conditions observed; especially for sulphur S in more evolved melt compositions (e.g., Clemente et al., 2004;Gennaro et al., 2021;O'Neill, 2020;Tattitch et al., 2021;Thomas and Wood, 2021) (Figure 5).Overall, this suggests Rotomakariri was a basaltic magma like Tarawera, Terrace Rd, and Rotokawa, which had entrained xenocrystic high-An plagioclase.It then stalled at shallow levels, resulting in low volatile contents and further cooling, crystallising to an andesite melt composition prior to eruption.Similar storage conditions and volatiles prior to eruptions of varying style 6 The magmatic architecture around the OVC The mineral and melt inclusion compositions measured in Tarawera, Terrace Rd, Rotokawau and Rotomakarir are common to many different OVC eruptions.For instance, similar melt inclusions (albeit more primitive) and mineral chemistries are found in other basalts from around the OVC (e.g., Kaharoa, Okareka, Matahi, and Matahina) and even in Taupō Volcanic Centre (TVC) basaltic material (e.g., Oraunui), showing that these are common features within the TVZ (Allan et al., 2017;Barker et al., 2020;Deering et al., 2011;Rooyakkers et al., 2018;Wilson et al., 2006).As the same basaltic magmas are occurring in eruptions separated spatially and temporally, these sets of conditions must be common around the OVC even though magmas themselves were not sourced from the same spatio-temporal reservoirs.
Combining the evidence from barometry and mixing textures suggests a crust full of individual magma reservoirs around the OVC, that are variously sampled during eruption.Despite large model uncertainties, pressures derived from clinopyroxene-melt and H2O-CO2 barometery and rhyolite-MELTS modelling lie within the TVZ crust assuming a crustal density of 2700 kgm - 3 and a Moho at 25-30 km or 7-8 kbar pressure (Bannister et al., 2004).This suggests that basaltic magmas, in addition to rhyolitic magmas, are stored and evolve polybarically within the crust.This agrees with current geochemical and geophysical constraints from previous Tarawera clinopyroxene barometry (1-3 kbar, with some >7 kbar, reported in Sable et al., 2009) and the presence of partial melt bodies at similar depths around the OVC, such as at 6-16 km using receiver functions (Bannister et al., 2004), 10-20 km (as shallow as 8 km beneath Waimungu) using electrical resistivity inversions (Heise et al., 2016(Heise et al., , 2010)), and 8-10 km from earthquake swarms attributed to a basaltic dike intrusion (Benson et al., 2021).Additionally, conceptual models based on petrological modelling invoke mafic sheets residing at 11-15 km, with some isolated pods found at 8-6 km depths (Cole et al., 2014;Deering et al., 2010).Large uncertainties in clinopyroxene-melt barometry mean individual magma reservoirs cannot be identified using this method.However, the mineral textures and compositions suggest evolution in isolated reservoirs, where each batch has its own distinct composition reflecting their individual histories.
These observations suggest that a thick, crustal mush -containing a wealth of magma types in individual, isolated pockets of magmas -is mostly trapping the ascending basalts in the crust that fuel magmatism around the OVC.This model likely applies more generally to active calderas in the TVZ and is similar to other arc settings, such as the Andean Puna plateau, resulting in the dominance of compositionally-evolved volcanism (e.g., Delph et al., 2017;Kay et al., 2010).However, the extensional regime of the TVZ is clearly important in allowing some of these basalts to reach the surface and erupt explosively.
The few basalts that do make it to the surface have passed through the complicated crustal mush and carry the signature of these interactions in their crystal cargo.This highlights the use of basaltic mineral and melt inclusion chemistry as windows into the sub-surface in silicic magmatic regions, extending its application from using olivine-hosted melt inclusions to understand mantle melting dynamics (e.g., Barker et al., 2020) to analysing clinopyroxene-hosted melt inclusions to gain insight into crustal processes.Combining data from multiple eruptions separated spatially and temporally has highlighted that similar processes are important around the OVC for potentially the last ~30 ka.

Author Contributions
ECH, JDB, HMM, and GK conceived the project idea.ECH and SL collected and processed the data.All authors contributed to data interpretation.ECH led manuscript production with further contribution from all authors.

Figure 2
Figure 2 Annotated back-scattered electron (BSE) scanning electron microscope (SEM) images of scoria textures .Each column is a separate eruption: Terrace Rd (a, e, i, m, and additionally p at the bottom of the far-right column),

Figure 3
Figure 3 Histograms of mineral chemistry showing fraction of crystal core analyses in each compositional bin (number of analyses n indicated for each panel).Each column represents a different mineral phase, labelled along the top: (a-e) olivine -forsterite content, (f-j) pyroxene -Mg# (filled histograms are for cpx, except unfilled bars in (h) that are for opx, with different opx populations labelled in grey), and (k-o) plagioclase -anorthite content (different plg populations are labelled in grey in (l) but apply to all eruptions).Each row represents an individual eruption, which are labelled down the left-hand side and shown using colour: (a, f, k) Tarawera (orange), (b, g, l)

Figure 4
Figure 4 Major element systematics for melt composition data for (a) Al2O3, (b) CaO, and (c) K2O vs. SiO2.Symbol shapes distinguish between melt inclusion (PEC-corrected, hosted in olivine = square, clinopyroxene = circle, plagioclase = diamond, orthopyroxene = up-triangle, or quartz = down-triangle) and groundmass glass i n R o t o m a k a r i r i ( E r r o r !R e f e r e n c e s o u r c e n o t f o u n d .F i g u r e 4 ) , a l t h o u g h o l i v i n e -h o s t e d m e l t i n c l u s i o n s a t T e r r a c e R d h a v e h i g h e r C a O .T h e r e i s n o t r e n d i n m e l t c o m p o s i t i o n w i t h c l i n o p y r o x e n e M g # , a l t h o u g h t h e t w o R o t o k a w a u m e l t i n c l u s i o n s h o s t e d i n M g # 7 6 c l i n o p y r o x e n e h a v e t h e m o s t e v o l v e d m e l t c h e m i s t r y .A t R o t o k a w a u a n d T h e m o s t b a s a l t i c o f t h e s e m e l t inclusions overlap with the higher SiO2 range of the olivine-hosted melt inclusions from Barker et al. ( ), whereas S and H2O do not correlate (Error!Reference s o u r c e n o t f o u n d .F i g u r e 5 h ) .B r o a d l y , t h e r e i s a p o s i t i v e c o r r e l a t i o n o f K 2 O w i t h H 2 O , S T a n d C l ( F i g u r e 5 ) .A t R o t o k a w a u a n d T a r a w e r a , t h e r e i s a s e c o n d p o p u l a t i o n o f m e l t i n c l u s i o n s w i t h S T < 1 0 0 0 p p m w h e r e H 2 O a n d S T ( b u t n o t C l ) n e g a t i v e l y c o r r e l a t e w i t h K 2 O .T a r a w e r a m e l t i n c l u s i o n s h o s t e d i n g r o u p o n e p l a g i o c l a s e a r e b a s a e h a v e s i m i l a r C a O , b u t d i f f e r e n t A l 2 O 3 , t o c l i n o p y r o x e n e -h o s t e d m e l t i n c l u s i o n s ( R o w e e t a l ., 2 0 2 1 ) .

Figure 5
Figure 5 Volatiles systematics for melt composition data for: (a) H2O, (b) Cl, (c) S, (d) F, and (e) H2O (zoomed in scale compared to a) vs. K2O; (f) CO2 vs. H2O; and (g) Cl and (h) H2O vs. S. Symbol shapes distinguish between melt inclusion (PEC-corrected, hosted in olivine = square, clinopyroxene, = circle plagioclase = diamond, orthopyroxene = up-triangle, or quartz = down-triangle) and groundmass glass (cross).Different eruptions are indicated by symbol colour (Tarawera -Tw = orange, Rotokawau -Rk = green, Rotomakariri -Rm = blue, Terrace Rd -Tr = purple, other OVC basalts -o-Ba = grey, and OVC rhyolites -Rhy = black).Uncertainties for our data are indicated in each panel as two standard deviations of precision based on repeat analys es of VG-2 over all analytical sessions for EPMA data (note the K2O error bar is small) or two standard deviations of the minimum precision based on repeat analyses of standards over all analytical sessions for SIMS data in (f).Trends for different processes are shown with grey arrows and labelled in (e) and (f).Data sources: melt inclusion and groundmass glass for Terrace Road, Rotomakariri, and Rotokawau (this study); Tarawera melt inclusions hosted

Figure 6
Figure 6 Pressure and temperature estimates for each eruption (shown by colour, other OVC basalts include Kaharoa -K, Okareka -O, and Matahi -M).(a) Pressure vs. temperature from: cpx-m using whole rock data (coloured outlines are assuming 0 wt% H2O and filled regions are 5 wt%) and melt inclusion data (circles), where the standard errors of estimate (SEE) is shown in the bottom-right corner by the thick black lines with a circle in

Figure 7
Figure7Temperature vs. H2O.Curves are maximum temperature amphibole is stable at for a given bulk H2O content of the system from(Foden and Green, (1992) for different pressures (written on each line).Symbols are melt inclusion data (see Figure4Figure4 for interpretation of the symbol shape and colour -only data from this study are plotted, which are PEC-corrected), where temperature is derived from the melt composition and measured H2O concentration is plotted), which is a minimum for the system.Uncertainties are indicated for T (±1SEE) and H2O (±2 sd of the precision based on repeat analyses of secondary standard VG2) in the top corner.
Fi gure 5g).Decreasing pressure during ascent drives crystallisation and degassing, forming a fluid that s eques t er s S, but not Cl ( e. g ., L es ne et al ., 2011) .T he m e l t i n i t i a l l y c o n t a i n ed 17 00 p pm ST a n d ~700 p pm C l ( ~1 200 p pm C l f o r R o t ok aw au ) , b u t t h e m ax i m um c o n ce n t r a t i o n s a r e ~3000 p pm ST a n d 2000 p pm C l ( 2800 p pm Cl f o r R o t ok aw au ) .I n s um m ar y , t h e b a s a l t i c m ag m a i s v ol a t i l erich, This would suggesting these melts are derived from the high-degrees of fluid flux melting associated with caldera regions, as inferred by Barker et al. (2020) and Zellmer et al. (2020).

Figure 8
Figure 8 Ca/Na ratios in plagioclase (and approximate An): (a) from all analysed grains; and (b) calculated for equilibrium with melt inclusions from this study.Solid vertical linesCircles are calculated Ca/Naplg in equilibrium with average whole rock data for each eruption (circles), whereas dashed vertical linesstars are in equilibrium with primary melt

Table 1
Basaltic eruptions and magmas from around the Okataina Volcanic Centre (OVC) since the last caldera- vast majority ofMost glass analyses in this study come from melt inclusions hosted in Rowe et al .( 2021)minor olivine-, plagioclase-, and orthopyroxene-hosted melt inclusion and groundmass glass analyses (Error!Reference source not found.Figures4 and Error!Reference source not found.5;additionaldataareshown in FigureS7Supplementary Material).From our study, most glass analyses are from Tarawera, followed by Rotokawau then Rotomakariri, with a handful from Terrace Rd.The Tarawera dataset is supplemented with melt inclusion data f r om Bar ker et al .(2020)andRoweetal .(2021).Mel t i ncl usi ons show consi der abl e r ange i n composi t i on.Ter r ace Rd, Rot okawau, and Tar awer a mel t i ncl usi ons ar e pr edomi nant l y basal t i c t o basal t i c-andesi t e i n composi t i on, wher eas Rot omakar i r i gl asses ar e ent i r el y andesi t i c ( Fi gur e 4) .