TephraNZ: a major and trace element reference dataset for prominent Quaternary rhyolitic tephras in New Zealand and implications for correlation

Jenni L. Hopkins1, Janine E. Bidmead1, David J. Lowe2, Richard J. Wysoczanski3, Bradley J. Pillans4, Luisa Ashworth1, Andrew B.H. Rees1, Fiona Tuckett1 1School of Geography Environment and Earth Science, Victoria University of Wellington, Wellington, PO Box 600, New Zealand 2School of Science (Earth Sciences), University of Waikato, Hamilton, Private Bag 3105, New Zealand 3240 10 3National Institute of Water and Atmospheric Research, Wellington, Private Bag 14901, New Zealand 4Research School of Earth Science, Australian National University,


Sample preparation
Bulk tephra samples were disaggregated in water for 1-5 min in an ultra-sonic water bath. Clays 205 and ultra-fines (< 5 µm) were rinsed off and samples were then wet sieved using disposable sieve cloths to 125-250 µm shard size or, where necessary, 60-125 µm. Samples were then dried for 12-24 hr at 50º C before mounting in epoxy resin. Seven samples were mounted into individual drill holes (4-mm diameter) in 25 mm epoxy round blocks (a 4:1 ratio of EpoTek 301 resin [A]: hardener [B]). Individual drill holes were then backfilled using the same epoxy mix (see Lowe, 2011, p. 124, for a schematic 210 illustration). Sample blocks were polished using the following sequence: ~3 min in a figure of eight pattern on 800 grit paper with water lubricant to remove the epoxy and break through to the glass shards, ~1 min on 1200 grit paper with water lubricant to remove any large scratches, and ~1 min on 2500 grit paper with water lubricant to begin to reveal the outline of the shards. Blocks were then moved on to the diamond laps with their appropriate lubricant, all at 280 revolutions min -1 rotating the 215 block 90 degrees every 30 s followed by 2 min of ultrasonic bathing at < 24 ºC between each lap stage to remove any loose material on the surface of the blocks: ~ 3 min on 6 µm, ~ 1 min at 3 µm, and ~ 1 min at 1 µm. Blocks were then carbon coated before loading in the electron microprobe system for analysis (EMPA). 220

EPMA method and data reduction
Major element analysis of glass shards was undertaken at Victoria University of Wellington (VUW) by wavelength dispersive X-ray spectroscopy (WDS) on a JEOL JXA8230 Superprobe electron probe microanalyser (EPMA). Broadly the method follows that espoused by Kuehn et al. (2011). Backscatter electron images of each sample were taken and used as block maps to allow the location of 225 EPMA analyses to be replicated for trace element analysis. A defocused circle beam 10 µm in diameter was used at 8 nA to analyse all major elements as oxides (SiO2,TiO2,Al2O3,FeOt,MnO,MgO,CaO,Na2O,and K2O). During standardisation, Na2O was run twice, the second time skipping the peak search to reduce the volatilisation of the element, with the second standardisation value then used. Table 2 shows the EPMA set up and run times. During the analysis, VG-568 was run as a calibration standard, 230 and VG-A99 and ATHO-G were run as secondary standards, with two of each standard (calibration and secondary) analysed between ten sample analyses to monitor machine drift. Initial concentrations were determined using the ZAF correction method, with secondary offline data reduction undertaken to correct for variability in VG-568. Internal correction values were calculated using the GeoREM reference values of VG-568 (Eq. 1) and applied to all the data (Eq. 2). 235 Following this, samples were corrected for deviations from 100 wt.% total, this assumes any variation is https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. 10 NIST-610, BHVO2-G, and ATHO-G. StHS6/80-G was analysed as a secondary standard throughout, and all standards (calibration and secondary) were analysed twice every ten samples. All data were 275 reduced offline using Iolite v.3 TM software (Paton et al., 2011), using 43 Ca as the internal standard value (index channel) and the "Trace_Elements_IS" data reduction scheme (DRS). The data were reduced against ATHO-G as the calibration standard. No post-processing data reduction was necessary for the trace elements data; precision and accuracy were calculated on STHS6/80-G as described above (Eq.

Standardisation method
Multiple calibration standards with different trace element concentrations were analysed to determine which would be most suitable for trace element data reduction. Potential calibration standards included NIST-612, NIST-610, BHVO2-G, and ATHO-G. These were each run twice every ten samples, along with secondary standard STHS6/80-G. Figure 2 shows the STHS6/80-G results of a 285 range of selected, commonly-used trace elements, including Zn (transition metal), Rb (LILE), Zr (HFSE), La (LREE), Yb (HREE), normalised using each of the calibration standards. Overall, the results show that for the lighter masses (e.g. Zn) there is a large variability in the measured STHS6/80-G values across the different standards, but all except BHVO2-G sit within error (2 sd) of the reference value (Fig. 2). For the heavier masses, the variation from the reference value observed within the 290 analysed values decreases, except for NIST-610, which remains highly variable in the middle masses (Rb, Zr, Fig. 2), with variability reducing in the heavier masses (La, Yb, Fig. 2). The data show that the use of ATHO-G as the calibration standard (for data reduction of rhyolites) produces the most accurate and precise data for the secondary standard, for all except the elements with the heaviest masses, and smallest concentrations (e.g. Yb). 295 https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.

Principal Component Analysis
To determine the elements that show the most variance with the reference dataset, and therefore the most appropriate (optimum) for compositional separation, we have used principal component 305 analysis (PCA). PCA analysis was run in the coding platform R and RStudio using packages "factoextra", "ggbiplot", "vegan", "cowplot", "rioja", and "ggrepel". Data for Tuhua tephra were removed as these would unnecessarily skew the results due to their distinct geochemistry. Only element values were used (no ratios [e.g. SiO2/K2O] or sums [e.g. Na2O + K2O]). All element values were centred (column mean subtracted from each value) and scaled (value divided by the standard deviation 310 of the column) to allow their variability to be comparable, even when their absolute values are not. PCA was run using the "prcomp" function, and PCA contributions were calculated using "fiz_comp" function. A template of the coding script used can be found in Supplementary Material 1.

Euclidean similarity coefficients
To identify the tephra samples that were most similar, and could therefore pose problems in 315 unique fingerprinting, we ran euclidean similarity coefficients (ESC) analysis. ESC was run on the coding platform R and RStudio using the package "stats". Following the guidelines of Hunt et al. (1995) for ESC analysis, we used non-normalised, mean concentrations of the elements highlighted by the PCA to be the most indicative of variance in the dataset. These values were input as comparison values, and the function "as.matrix.dist" was used to run the "euclidian" statistical method. This method 320 calculates the similarity or samples based on an infinite number of comparison input values. A template of the coding script used can be found in Supplementary Material 1, the output table was manipulated post-production to provide the colour formatting shown in Figure 14.

Results
The averages and their standard deviations for all samples are reported in Table 3; the full 325 reference dataset can be found in Supplementary Material Table 2. All reported values in the text and figures (unless stated otherwise) are recalculated (normalised) to 100% on a volatile-free basis (following Lowe et al., 2017) with the difference between the raw total and 100% being reported as "H2OD" ( Table 3). For best correlation results, we recommend that the full dataset is used in order to see the trends in the geochemical data rather than just the means and standard deviations. 330 https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.
Of the 45 tephra samples, 22 have a 'homogeneous signature', homogeneity being defined here where the standard deviation of the sample is equal to or less than analytical error (2sd of secondary standard, e.g. for FeO = ± 0.23 wt.%, CaO = ± 0.10 wt.%). The majority (~64%) of the samples that have a homogeneous signature are from OVC (e.g. Whakatane, Mamaku, Rotoma) or from calderas 350 older than OVC (~32%), for example: Upper Griffins Road tephra, a correlative of the Whakamaru eruptives, Whakamaru Volcanic Centre (WVC), and Mangapipi tephra, a correlative to deposits of Mangakino Volcanic Centre (MgVC; Fig. 5a). Ten samples show a heterogeneous signature (where standard deviations for both FeO and CaO are greater than analytical errors), with most from a proximal source (~30%), or from tephras deposited in the Whanganui Basin area (40%), and with the remainder 355 being from the Mangaone Subgroup eruptives from the OVC: Hauparu, Maketu, and Ngamotu ( Fig.   5b). Glass shards from four tephra samples show a bimodal signature in major and trace elements, where the populations split into two distinct groups. Tephras showing this phenomenon include Rotorua (OVC), Rerewhakaaitu (OVC), Poihipi (TVC), and Tahuna (TVC). The bimodal signatures of 360 Rerewhakaaitu and Rotorua are well documented , whereas those of Poihipi and Tahuna are newly identified here (Fig. 6). All four of these tephra horizons have their glass-shard bimodal signatures produced predominantly by K2O concentrations, into high (≥ 3.8 wt.%) and low (≤ 3.6 wt.%) populations ( Fig. 6) linked to the crystallisation of biotite minerals. For five of the tephras, we undertook analyses from both proximal and distal samples. These 365 tephras included Whakatane, Rotoma, Waiohau, Rotorua, and Rerewhakaaitu, which are all derived https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. 14 from OVC (Table 1). For Rotoma, Rerewhakaaitu, and Waiohau, the signatures of the proximal and distal deposits are indistinguishable, whereas, for Whakatane and Rotorua the proximal signature is highly variable, and the distal signature is homogeneous but overlapping with part of the extent of the proximal signature (Fig. 7). Similar findings are reported and discussed in more detail for Whakatane 370 tephra in Kobayshi et al. (2005) andHolt et al. (2011); and for Rotorua tephra in Shane et al. (2003a) and Kilgour and Smith (2008). https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.

390
Total iron expressed as FeO.

395
https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. Figure 8 shows a chondrite-normalised spider plot of all the trace element data for the glass shards analysed. The majority of the data plots along a common pattern of variable concentrations of 400 HFSE, LILEs and LREEs, but they show more consistent concentrations of HREEs. Of note are peaks in Ba, Nd, Pr, and a negative Sm anomaly. Sr shows the largest variability in concentrations from < 1 to ≤100 ppm, likely caused by a variability in feldspar crystallisation (Pearce et al., 2004). Several different patterns are observable within this full data suite pertaining to individual samples. The obviously different signature is that for glass from Tuhua tephra which shows a low concentration of Ba 405 (< 10 ppm) and Sr (<1 ppm) in comparison with values for the rest of the samples, and with high concentrations of all other elements, especially the REEs (Fig. 8). Analyses of glass shards from the Maketu tephra can also be identified by their very high Ba values (> 1000 ppm), mid-range Nb values (between those of Tuhua and the general trend), and much higher concentrations of all other elements (Fig. 8). We also note Er and Lu peaks, which pertain to glasses from the Te Rere tephra, that sit at the 410 higher concentration levels of the general trend ( Fig. 8; Table 3), and samples from Ngamotu, Rotoehu/Rotoiti and Earthquake Flat that sit at the lower overall trace element concentration levels of the general trend, but with high Ba values (Fig. 8). For the tephras where both proximal and distal samples of glass have been analysed for trace elements, the HFSEs (including Zr, Hf, Th, and Ti) and LILEs (including Rb, Sr, and Cs) can be used to maintain the heterogeneity between the proximal and 415 distal samples, whereas the HREE and the LREE tend to have a lower variability (Fig. 7).

445
In many cases, the major element concentrations in glass are sufficient to allow different tephras to be distinguished, a result consistent with the findings from much previous work both in New Zealand and elsewhere (e.g. Lowe et al., 2017). PCA results for the glass-shard major elements ( Fig. 9) show that PC1 and PC2 explain 82.7% of the variance within the data. When scaled, concentrations of TiO2, Al2O3, CaO, and SiO2 for PC1 ( Fig. 9), and MnO, K2O, FeOt, and SiO2 for PC2 ( Fig. 9) are shown to 450 have the highest contribution to the variance and are therefore most appropriate for distinguishing between tephra deposits for the reference dataset as a whole (Fig. 9). These major elements, especially CaO, FeO, and K2O, have long been recognised as being useful to distinguish many New Zealand late Quaternary tephras form one another (e.g. Lowe, 1988;Shane, 2000;Alloway et al., 2013), the presence of TiO2, Al2O3, and MnO are somewhat unusual. In a number of cases (discussed below) however, 455 major element concentrations are shown to overlap for certain tephra horizons, and thus trace elements and trace element ratios are investigated to provide additional variables to use as discriminants. PCA was also applied to scaled trace elements with the results suggesting that PC1 and PC2 could explain 66.0 % of the variability in the trace element data with Tb, Ho, Dy, Y, Sm, Nd, Tm, Yb, Hf, and Gd the ten highest contributors to PC1, and Rb, Th, Sr, U, Cs, Ta, Pb, La, Ce, and Pr highlighted as the ten 460 highest contributors the variance of PC2 (Fig. 10). Therefore, statistically these trace element concentrations and ratios of these trace elements have the potential to be the most useful in distinguishing the individual tephra horizons when using their glass-shard compositions alone.

Source-specific major and trace elements
The central TVZ contains nine recognised calderas, each with different eruption histories, but all having produced large magnitude/volume tephra-producing rhyolitic eruptives. Some of the calderas are 495 attributed to single caldera collapse events (Rotorua, Reporoa, and Ohakuri), others to composite collapse events that overlap spatially but not temporally (Mangakino and Kapenga), but the majority to multiple collapse events over an extended period of time (Maroa, Okataina, Taupō, and Whakamaru) ( Fig. 1; Wilson et al., 1995aWilson et al., , 2009Barker et al., 2021). Although the calderas are mostly discrete in space, evidence from multiple eruptions has shown their plumbing systems may be linked tectonically 500 (e.g. Wilson et al., 2009;Allan et al., 2012). Hence, the ability to trace a tephra deposit to a caldera source through glass-shard geochemistry alone could be challenging. The results of the PCA analysis suggest that tephra sourced from the TVC can be distinguished from those of a proposed Mangakino source (MgVC) (Fig. 9). Using SiO2/K2O vs. Na2O+K2O ratios, the glass shards of the TVC tephras generally have higher SiO2/K2O and lower Na2O+K2O ratios in 505 comparison to those of the equivalent oxides for MgVC-sourced tephra (Fig. 11a). This information is important, but because of the age differences for the calderas (TVC ~ 0.32 Ma to present, and MgVC ~1.6 Ma to 1.53 Ma and ~1.2 Ma to 0.95 Ma2 1995), the use of this distinction is likely more important for discussions on mantle source dynamics rather than for geochemical correlation of tephra deposits. Previous studies have suggested that the geochemical characteristics of glass shards from TVC 510 and OVC tephra deposits can be distinguished from after the eruption of the Kawakawa/Oruanui (KOT) to the present day using fO2 of Fe-Ti oxides and minerals (Shane, 1998), pumice and lava compositions (Sutton et al., 2000), and glass chemistry (Froggatt and Lowe, 1990). Our results also show there is a bimodality in the TVC glass-shard data as a whole and that the post-KOT tephra deposits from the TVC and OVC are quite different whereas the pre-KOT tephra from OVC and TVC are similar (Fig. 11c&d).

515
Most glass shards erupted after the KOT event from the TVC have low SiO2 (≤ 77 wt.%) and less variable K2O (~3 wt.%), and higher values for all other major elements in comparison to those of the glass shards erupted from the OVC (Fig. 11c&d). In comparison, tephra erupted from the TVC and OVC prior to, and including the KOT, do show a large amount of overlap in their glass geochemical signatures. For OVC, there is a high density of samples that have their SiO2 concentrations at ~78 wt.%; 520 https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. however, there is a high variability in SiO2 overall, with Maketu, Hauparu, and Ngamotu of the Mangaone subgroup plotting with SiO2 concentrations ≤~76 wt.%, and the remaining Mangaone subgroup samples (Unit L, Awakeri, and Mangaone) clustering at SiO2 = 76-77.5 wt.% (FeO =~1.2 wt.%, K2O =~2.8 wt.%, Al2O3 =~13 wt.%, CaO =~1.2 wt.%), a finding consistent with those of Smith et al. (2005) who divided the Mangaone subgroup into 'old' and 'young' eruptives on the basis of low 525 and high SiO2, respectively, and also unlike the other OVC sourced samples that plot around SiO2 =~77.5-79 wt.%, (FeO = ~0.8-0.9 wt.%, K2O = 2.75-4.5 wt.%, Al2O3 =~12-13 wt.%, CaO =~0.5-1.0 wt.%; Fig. 11c&d). Analyses from the Rotoehu/Rotoiti tephra deposits plot independently from those of other OVC eruptives for this time period. However, they overlap with those of some TVC-tephraderived glass compositions (Poihipi, Tahuna, Okaia, and KOT). The Rotoehu/Rotoiti tephra deposits 530 have a markedly homogeneous geochemical signature, and are also much older than TVC eruptions (Table 1). Hence, coupled with the thickness of the deposits, it is likely that a tephra linked to the Rotoehu/Rotoiti eruption would be obvious to distinguish through stratigraphy and age combined with the geochemistry. The TephraNZ dataset presented here also includes analyses of glass of samples from tephras 535 erupted from the Kapenga Volcanic Centre (KVC; Earthquake Flat eruption), Rotorua Volcanic Centre (RoVC), and Whakamaru Volcanic Centre (WVC). In addition, some older tephra deposits have been recorded in the Whanganui Basin and elsewhere. These are well-known beds but their caldera sources are not yet defined (Alloway et al., 1993;Pillans et al., 1994Pillans et al., , 2005Shane et al., 1996;Rees et al., 2018;2019). Figure 12 shows a comparison plot for the data from KVC, RoVC, and WVC with those regions 540 populated by glass data from samples from the OVC, TVC and MgVC sources. Overall, the samples plot with a lower SiO2/K2O ratio (≤ ~25), similar to that of the MgVC-sourced tephra, which seems to be indicative of older sources in comparison to the OVC and TVC. The samples potentially linked to RoVC (Bussell, 1986;Bussell and Pillans, 1997) show different geochemical compositions. For example, Kakariki-tephra-derived glass has slightly higher SiO2 ≥ 78 wt.%) in comparison to that of the 545 Ararata Gully tephra (SiO2 ≤ 77wt %), suggesting that they are likely derived from different eruptions, but potentially the same source (Mamaku Ignimbrite reportedly has variable geochemical phases; Milner et al., 2003). Glass from the KVC sample (Earthquake Flat tephra) has a very homogeneous signature in the major elements, but a more variable signature in the trace elements, both of which overlap with OVC-and TVC-source signatures. There is a very large spread for the data from the 550 unknown samples, precluding the ability to specify their source based simply on major and trace elements alone. Nevertheless, their glass compositional signatures are clearly more similar to those of the older MgVC sourced tephra, in comparison to those of the younger TVC and OVC deposits, as would be expected based on their known age range ( Table 1).

Homogeneous, heterogeneous, and bimodal samples
Fingerprinting of glass shards for correlation relies on the ability to distinguish between different deposits and therefore a homogeneous signature that is distinct from all other samples is the ideal 'fingerprint'. However, sometimes there is more complexity in the geochemical data and heterogeneity can develop in a tephra deposit through a number of mechanisms (Lowe, 2011): 565 (1) Variability in the magma body itself (e.g. Nairn, 1992;Nairn et al., 2004;Smith et al., 2004;Kobayashi et al., 2005;Shane et al., 2008;Charlier and Wilson, 2010;Klemetti et al., 2011;Cole et al., 2014); (2) Proximal vs. distal complexity, linked to (1) (e.g. Manning, 1996;Shane et al., 2003a;Holt et al., 2011);570 (3) Post-or syn-depositional reworking (e.g. Schneider et al., 2001) For example, the heterogeneous signature identified for the Kaharoa tephra agrees with previous findings for this eruption. Nairn et al. (2004) and Sahetapy-Engel et al. (2014) reported that tephra compositional variability within the Kaharoa deposits shows sequential tapping of a stratified magma 575 body coupled with syn-eruptive changes in dispersal patterns. In general, this is likely one of the reasons why some of the proximal tephra deposits analysed in this study have a more variable geochemical signature in comparison to those of their distal counterparts (Fig. 7). Although the proximal deposits record the detail in the eruption progression, the distal deposits tend to record the very largest phase of the eruption (e.g. Walker, 1980) but differences can be expected to occur 580 according to the azimuths of wind direction during an eruption and the number and degree of interconnectedness of magma bodies involved in the eruption (e.g. Walker, 1981;Kilgour and Smith, 2008;Sahetapy-Engel et al., 2014;Storm et al. 2014;Rubin et al., 2016). The tephrochronological principle is much more likely to utilise distal unknown deposits, and therefore we suggest that using the distal signature (or signatures) maybe more appropriate for 585 correlation. In general, distal tephras are more chemically homogeneousbut with some notable and well-documented exceptionsand this attribute therefore allows them to be traced over large areas (Manning, 1996). Alternatively, the identification of heterogeneity or bimodality in distal tephras, once recognised, can be an additional useful characteristic for fingerprinting (e.g. Shane et al., 2003aShane et al., , 2008Lowe et al., 2017). These statements, however, rely on the tephra being identified as a primary deposit, 590 and not reworked. Reworking is commonly seen in paleofluvial deposits, for example those in the Whanganui Basin, and in other environments prone to mixing such as in surficial soils. This reworking can mix tephra from multiple eruptions, and can cause highly variable glass chemistry within a single deposit (e.g. Shane et al., 2005Shane et al., , 2006. Fluvial reworking can be commonly identified by sedimentary structures within the deposit, for example, ripples or cross bedding indicative of fluvial transport and 595 deposition (e.g. Shane, 1994;Schneider et al., 2001), over thickening of deposits (e.g. Vucetich and https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. , 1969;Lowe, 2011), or through shard morphology, for example anomalously large shards or rounding of shards (e.g. Leaphy, 1997). Heterogeneous signatures (where the standard deviation of the analyses is greater than the analytical error) in major element compositions were identified for ten of the tephra deposits: Kaharoa, 600

Pullar
Taupō Y5 Proximal (P), Whakatane P, Hauparu, Maketu, Ngamotu, Fordell, Onepuhi, Birdgrove, and Ototoka. Our data show that for some samples, specific trace elements and trace element ratios have lower geochemical variability (Fig. 13a). The elements that work best to separate out the individual units within a deposit with a heterogeneous signature reflect the minerals that have formed during fractional crystallisation of the melt. Because of this, different elements or element ratios work for 605 different tephras. For example, for Kaharoa, Sr acts to effectively reduce the variability of the signature, whereas for Taupō Sr can be used to maintain the heterogeneity in the sample (Fig. 13a). Bimodality was identified for four of the tephra horizons analysed: Rotorua (OVC), Rerewhakaaitu (OVC), Poihipi (TVC), and Tahuna (TVC). For all four of these, K2O concentration causes the bimodality, and therefore trace elements with similar chemical properties reinforce the 610 bimodality (for example, LILEs Rb, Sr, and Cs; HFSEs Zr, or REE Eu), whereas most other trace elements do not show this bimodal signature (Fig. 13b).

Indistinguishable tephras
Euclidean similarity coefficient (ESC) analysis was used on all glass-shard reference data for tephras from Rotoiti/Rotoehu to Kaharoa in addition to the PCA and geochemical investigation to 615 determine those samples that have indistinguishable element concentrations at similar ages (Fig. 14). Poihipi and Tahuna (SC = 15) and Mamaku and Rotoma-D (SC=3) come up with significantly low similarity coefficients, hence suggesting that these samples will be indistinguishable in both major and trace elements. When trace element ratios are run through the SC analysis, Waimihia and Unit K (SC=6), Waimihia and Poronui (SC=9), Unit K and Poronui (SC=5), and Karapiti and Waimihia (SC=7) show significantly low similarity coefficients. In addition, when simple geochemical assessment 625 is applied, similarities are observed between Taupō and Waimihia, and Waiohau, Rotorua, and Rerewhakaaitu (Table 5).
These results suggest that for Poihipi and Tahuna, and Mamaku and Rotoma tephras, trace element ratios in glasses could enable them to be distinguished. Figure 15a shows that for Poihipi and Tahuna the best separation (although some overlap remains) is seen in the ratios La/Yb vs. Ba/Y; in 630 addition, Tahuna also shows a bimodality in Ba/Th ratio which is not seen for Poihipi. For Rotoma and https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License.
Mamaku, the tephras can be separated (although some overlap remains) using Ba/Th vs. Rb/Sr and Rb/Zr vs. Rb/Sr (Fig. 15b). Rotoma and Rotoehu/Rotoiti are very similar in their glass-shard major elements, but can be distinguished using specific, but a wide range, of trace elements (Fig. 15c). They are also very different in age, hence should not be too difficult to distinguish on the basis of stratigraphy 635 or dating. Waimihia and Unit K (Taupō Subgroup) tephras are very difficult to distinguish, and their relative similarity in age (3382 ± 50 and 5088 ± 73 cal. yr BP, respectively; Lowe et al., 2013) and mineralogy could see them misidentified if dates were unavailable or imprecise. Geochemical investigation beyond the PCA and SC analyses of glass shows that Lu, Sc, Mn, and Co can be used to 640 geochemically distinguish these two tephras ( Fig. 15d), indicative of fractional crystallisation of differing amounts of clinopyroxene, plagioclase, and amphibole during the eruptive events. Although not identified by the SC analysis directly, Poronui (11,195-51 cal yr BP) and Karapiti (11,501 ± 104 cal. yr BP) tephras also have comparable age, geochemistry, and mineralogy; thus using major, trace, and trace element ratios these two tephras remain indistinguishable. Glass shards from the three 645 Holocene tephras, Waimihia, Poronui, and Karapiti, also have very similar trace element and traceelement ratios but, as for Waimihia and Unit K, they can be distinguished with Lu, Sc, Mn, and Co, where Waimihia has higher Sc, Lu, and Mn, but lower Co in comparison to those of the Poronui and Karapiti tephras. They can also be distinguished simply with a biplot of FeO vs. CaO, or Na2O+K2O or SiO2/K2O, or SiO2, where the Waimihia samples in general have lower FeO, Na2O+K2O, SiO2/K2O, and 650 higher CaO, and SiO2 in comparison to the equivalent values for Poronui and Karapiti samples ( Fig.   15e, Table 5).
Geochemical investigation and PCA analysis also highlights the similarity of the Waiohau, Rotorua, and Rerewhakaaitu tephras. There is added complexity with these samples as we have both proximal and distal deposits to compare, where, as discussed previously, the proximal samples will 655 likely be more heterogeneous. Glass analyses of the Waiohau tephra show it can be distinguished from those for the Rotorua and Rerewhakaaitu tephras using a range of trace elements and trace-element ratios. In addition, the Rotorua and Rerewhakaaitu tephras are observed to be bimodal for some elements. The Waiohau also has different mineralogy from that of Rotorua and Rerewhakaaitu tephras (Froggatt and Lowe, 1990;Lowe et al., 2008). Conversely, the Rotorua and Rerewhakaaitu tephras are 660 usually indistinguishable in geochemistry and mineralogy, and therefore accurate dating and stratigraphic super-positioning would have to be relied upon to distinguish them (Fig. 15f, Table 5).

Figure 13. Biplots to show examples of how trace elements in glass enable manipulation of heterogeneous and bimodal geochemical data. Panel (a) shows analyses of glass from Kaharoa and Taupō tephras, both of which show a heterogeneous signature with most
665 major elements (presented on a normalised basis). Sr has a low variability for Taupō, but does not for Kaharoa tephra, conversely, Ba has a low variability for Kaharoa, but does not for Taupō. Panel (b) shows the bimodal signature created for Tahuna tephra using K2O composition; this is also seen for Cs, but is not for Ba.

for alternative elements). Plots show examples of the elements that enable these tephras to be separated (a) Poihipi and Tahuna (from TVC); (b) Mamaku and Rotoma (from OVC); (c) Rotoma and Rotoiti/Rotoehu (from OVC); (d) Waimihia and Unit K (from TVC); (e) Waimihia, Poronui and Karapiti -note that Poronui and Karapiti are indistinguishable using glass-chemistry;
(f) Waiohau, Rerewhakaiitu and Rotorua -note that Rerewhakaiitu and Rotorua are indistinguishable using glass chemistry. All major element data presented on a normalised basis, and total iron is expressed as FeO.
KOT, Okaia, and Unit L (Mangaone Subgroup) show indistinguishable major elements in their constituent glass shards, and very similar trace elements. The TephraNZ samples have been compared to existing published data and are complementary in major elements (e.g. Sandiford et al., 2002;Smith et al., 2002;Lowe et al., 2008;Allan et al., 2008;Molloy, 2008;20). This is 685 the first time trace element glass data have been published for Unit L and Okaia tephras. Our results show that Unit L glass shows bimodality in Rb/Zr, Ba/Th, Ce/Th and Y/Th and in this way it can be distinguished from the KOT and Okaia tephras ( Table 5).

690
This foundation dataset, derived in a formalised way, is unique in New Zealand and provides researchers with new avenues of research. It is our hope that the foundation dataset can be improved and expanded with analyses of other known deposits, and that a subsidiary catalogue of accurately correlated geochemical samples can be added to bolster the dataset. As noted earlier, it is beyond the scope of this paper to dive too deeply into the detail of the data but we feel that it will provide the basis 695 for countless projects in the future. Below we highlight some of the current gaps which we think would benefit from further research.

Further statistical analysis
We have applied simple ordination and statistical analyses to this dataset; however, we believe 700 that further rigorous statistical analysis could be applied. Firstly, the analyses we present in this publication have been applied to mean values for each of the tephra samples (e.g. data from Table 3); there is no reason why these simple tests could not be applied to the full dataset, using all the individual values analysed for each sample. Secondly, for simplicity we chose to split up the assessment of major and trace elements, these could be run concurrently. Third, we chose very basic tests (PCA and ESC) to 705 fit with our requirements, however there is likely some more appropriate statistical test that could be applied to get the most out of this exceptional dataset. For example, (extended) Canonical Variates Analysis (CVA); applying CVA to PCA results could determine optimal discrimination between multivariate data for single tephra deposits. This discrimination will increase the ability to identify an unknown tephra based on its similarity to known signatures plotted in multivariate space (e.g. 710 discriminant function analysis; Tyron et al., 2009;Lowe et al., 2017;Bolton et al., 2020).

Whanganui Basin correlatives
A number of the tephras reported in this research were sampled from the Whanganui Basin, an uplifted Plio-Pleistocene basin margin sequence that preserves as many as 45-superposed cyclothems 715 deposited since ~3 Ma (Naish and Kamp, 1997;Naish et al., 1996Naish et al., , 2005Carter and Naish, 1998;Carter et al., 1999;Pillans, 2017;Grant et al., 2018Grant et al., , 2019Tapia et al., 2019). The tephra deposits within the basin contribute to the robust chronological framework that has been constructed for this region (Seward, 1976;Beu and Edwards, 1984;Alloway et al., 1993;Naish and Kamp, 1995;Shane et al., 1996;Saul et al., 1999;Pillans et al., 1994Pillans et al., , 2005Naish et al., 1996Naish et al., , 2005Rees et al., 2018Rees et al., , 2019. 720 These tephras also record a critical time in New Zealand's volcanological historythe transfer between activity from the Coromandel Volcanic Zone to the Taupō Volcanic Zone (Briggs et al., 2005). Deposits from this period are generally poorly exposed at source, and thus distal tephras could provide an insight into the eruptive history, geochemical evolution, and potentially even caldera evolution during this period (Houghton et al., 1995). Most of the tephras reported in this research are well known and well 725 dated, which is why they were included in the study. However, most do not have a known source caldera or source eruptives, or have only been variably correlated to other deposits in New Zealand (e.g. Lowe et al., 2001;Pearce et al., 2008). There are also numbers of tephra deposits in the Whanganui Basin that have yet to be studied, and thus a research project that is tephra focused, rather than using it as an accessory to a different line of enquiry, is timely. 730

IODP and ODP correlatives
At present there is a wealth of information that has yet to be fully investigated in the tephra deposits in ODP Leg 181 Sites 1122, 1123, 1124, 1125(Carter et al., 2003Allan et al., 2008) and IODP Expedition 372 and 375 sites U1517 and U1520 (Pecher et al., 2018;735 Saffer et al., 2018). Pioneering work includes that undertaken by Watkins and Huang (1977) and Nelson et al. (1985) and findings from more 'local' marine coring expeditions include those reported by Shane et al. (2006). The new reference material built by this project will allow more definitive identification and correlation of tephras within these cores, specifically post-2 Ma. However, the reports currently published on these deposits suggest that there are many more tephra deposits to be found in these 740 marine and offshore sites than we have in the TephraNZ dataset (Carter et al., 2003;Holt et al., 2010Holt et al., , 2011 (2005) reported an additional six tephra deposits that are correlated between the cores, but not to onshore equivalents, leaving potentially ~81 tephra horizons within the ODP cores that are uncorrelated. This 755 information could provide a detailed investigation into the timing and evolution of the TVZ eruptions that is unobtainable from onshore deposits.

Mineral compositions
The TephraNZ reference dataset is only populated by glass major and trace element analyses at 760 present. This is because glass geochemistry is one of the most frequently used and accessible tools for tephra correlation. Aerodynamic sorting of tephra componentry through transportation adds to the favourability of glass shards as the dominant tool because glass shards tend to be the only phase that is found at both proximal and distal sites. However, previous New Zealand-based studies have specified how mineral assemblages and their geochemical compositions can be used to distinguish certain tephras 765 and their source (e.g. Nairn and Kohn, 1973;Lowe, 1988;Froggatt and Lowe, 1990;Froggatt and Rogers, 1990;Shane, 1998;Shane et al., 2003b;Allan et al., 2008;Lowe et al., 2008;Lowe, 2011). For example, the mineral cummingtonite, where predominant, is a known identifier for tephras from the Haroharo complex of the OVC (Whakatane, Rotoma, Rotoehu/Rotoiti (Table 4); Ewart, 1968;Lowe, 1988;Froggatt and Lowe, 1990). At present, ferromagnesian mineralogical assemblages (following 770 Froggatt and Lowe, 1990;Smith et al., 2005;Lowe et al., 2008) for all the TephraNZ samples younger than and including Rotoehu/Rotoiti have been published (see Table 4). Extending this tabulation to include the older samples would add another useful criterion to the correlation toolbox . Additionally, the fractional crystallisation of plagioclase, biotite, amphibole, zircon, hydrous mineral phases, or Fe-Ti oxides has been shown to be the key impactor on the trace element chemistry 775 (Shane, 1998;Allan, 2008;Turner et al., 2009Turner et al., , 2011. Thus the prevalence of these minerals is also an important potential fingerprinting tool. The information on the mineralogy of the tephras is not only useful for fingerprinting but also can be used in determining the characteristics of the magma source components, and potentially provide estimates for the temperature, pressure, and oxidation states of the magmatic system before eruption (e.g. Lowe, 1988Lowe, , 2011Shane, 1998 (Ward, 1967;Pain, 1975;Vucetich et al., 1978;Iso et al., 1982;Froggatt, 795 1983;Manning, 1996;Lowe et al., 2001;Newnham et al., 2004;Allan et al., 2008;Briggs et al., 2006;Lowe, 2019;B. Laeuchli pers. comms. 2020). However, at present the authors are not aware of a detailed, up-to-date study into the primary compositions of these tephra deposits.  Shane et al., 1994;Houghton et al., 1995;Black et al., 1996;Tanaka et al., 1996;Milner et al., 2003), Kukumoa Subgroup (~0.22-0.05 Ma; Manning, 1996), and Tikotiko Ash (~0.125 ka; Lowe, 2019). A number of these studies are outdated, and with improved methodologies (major and trace element analysis, potentially of melt inclusions where preserved, dating techniques, 805 and other measures to help construct time frames such as via phytolith studies to determine glacial vs interglacial periods) it could be timely to further investigate this period of (apparent) deficit.

Conclusions
Major and trace element geochemical compositions of glass shards for a large suite of prominent, widespread New Zealand rhyolitic tephras have been analysed systematically and published 810 for the first time as "TephraNZ". TephraNZ is a foundation dataset for collating geochemical data about New Zealand tephras. The foundation reference dataset is made up of known deposits that have their ages quantified through independent methods, or are from the type sites where tephras were first defined, or well-documented reference sections. Detailed methodology is reported to allow subsequent research to acquire comparable data to those in this database. Principal component analysis indicates 815 https://doi.org/10.5194/gchron-2020-34 Preprint. Discussion started: 9 November 2020 c Author(s) 2020. CC BY 4.0 License. 38 that for the TephraNZ database, as a whole, major elements CaO, TiO2, K2O, and FeOt are responsible for the variability in PC1 and PC2 space. For trace elements, Tb, Ho, Dy, Y, Sm, Nd, Tm, Yb, Hf, Gd and Rb, Th, Sr, U, Cs, Ta, Pb, La, Ce, Pr are responsible for ~ 69% of the variability, and trace element ratios, Ba/Zr, Ba/Hf, Ba/Eu, Zr, Nb, Rb/Zr, Ba/Y, Zr/Th, Zr/ Yb, Nb/Y, Zr/Y, and Ba/Th, Ce/Th Y/Th, Ba/La, Ba/Ce, Zr/Th and Sr/Nb are responsible for ~73 % of the variance. Euclidean similarity 820 coefficients can also be used to distinguish between some geochemically similar glass analyses. However, further detailed geochemical investigation is required to distinguish others. Geochemically indistinguishable tephras (on the basis of both major and trace element glass-shard compositions) are identified as Taupō and Waimihia; Poronui and Karapiti; Rotorua and Rerewhakaaitu; and KOT and Okaia. Only Poronui and Karapiti are noted as entirely indistinguishable, with other methods of 825 characterisation listed as alternative options, including mineralogy, age, and stratigraphic relationships.

Author contribution
JLH and RJW designed the project. DJL and BJP contributed samples from previous field campaigns, and DJL provided guidance on new and existing field locations for sample collection. JLH and JEB undertook the field work, lab work, analysis and data reduction. ABHR advised on statistical analysis 830 and R-coding. LA supervised and helped JEB develop LA-ICP-MS analysis and data reduction. FT supervised and helped JEB develop sample mounting and polishing procedures. JLH wrote the manuscript with contributions from all co-authors.

Competing interest
The authors declare that they have no conflict of interest.