Large (greater than ∼20 cm) banded HSR pumices with at least two visibly different bands were the targets when collecting pumice samples for analysis. An outcrop ∼21 km E of the junction of Highway 395 and ∼85 km SSW of Burns, OR, hosts an abundance of glassy, relatively fresh banded HSR pumices (Figure 1). This location is close to (<60 km away from) the approximated eruptive center. Few other outcrops of RST are known to contain such large, fresh, and glassy banded pumices, so we exclusively sampled this outcrop.

Fifteen banded pumices were collected based on size, visible variation in color of bands, and width of bands. Pumices with wider bands were preferred for easier separation, though pumices with mm-sized, multicolored bands were common. Banded pumices ranged in their appearance from having multiple shades of light grey to white, multiple shades of medium grey, shades of dark grey and light grey, or several different shades of grey. Two homogeneous white pumices of different textures were also collected for analysis to compare to data from previous studies. The largest unbanded and banded pumices of glassy RST are on the order of 1 m in size. A granitic lithic fragment found within the outcrop of RST in the sample area and obsidian from a pristine densely welded vitrophyre from the area near the Silver Creek location described by Streck and Grunder (1995) was also included for bulk rock analysis.

All 17 pumices were lightly crushed, and the largest bands in the banded pumices were separated based on visible differences in color and possibly texture as compared to surrounding bands. A numbering system was implemented to keep track of different bands of the same sample. A decimal and a number were placed after each original sample ID, where “.1” was considered the lightest appearing band, and each darker band received an increasing number after the decimal. For example, there were three band separates analyzed for sample VS20-122H, so the lightest grey material was given the sample ID of VS20-122H.1, the medium grey material was VS20-122H.2, and the darkest grey material of that sample was labeled VS20-122H.3. Samples were analyzed for ten major and minor element concentrations and 17 trace elements in parts per million (ppm) at Washington State University’s (WSU) Geoanalytical Laboratory. Each sample was powdered at WSU using a low trace element tungsten carbide shatterbox. Powdered forms of each sample are placed in a high temperature oven to vaporize any organic or secondary volatile contents (unrelated to magmatic volatile content) in each sample and to produce a ‘dry’ powder. The percent loss on ignition (LOI) is the weight percent of these volatiles removed during this process. Lower LOI values generally equate to lower secondary water content and thus to data that are more representative of the original composition at the time of emplacement. Major, minor, and some trace element concentrations were measured using WSU’s ThermoARL X-ray fluorescence (XRF) spectrometer using methods described in Johnson et al. (1999). More trace element compositions were determined using WSU’s Inductively coupled plasma-mass spectrometer (ICP-MS).

Streck and Grunder (1997) gathered compositional data from RST HSR pumices using XRF and Instrumental Neutron Activation Analysis (INAA). Five pumices analyzed in Streck and Grunder (1997) were reanalyzed again in this study as a measure of comparison. When plotting results from this study with geochemical data from Streck and Grunder (1997) of the same samples, several elements displayed noticeable offsets. When comparing Nb, Pb, Rb, Sr, and Zr, it is preferred to compare old XRF values to new ICP-MS values, because the INAA method cannot be used for analyzing such elements or lead to much lower precision as compared to elemental concentrations determined by XRF. Percent differences were calculated for all measured elements and oxides where possible. Those with a percent different greater than ∼5% were deemed candidates for recalculation, except when the measured concentrations were <0.05 wt% for oxides or <6 ppm for minor and trace elements (Table 1). Zr, Nb, Pb, and Sr XRF, as well as Ce INAA values from Streck and Grunder (1997) were plotted against new data from this study, and linear regression lines were calculated. The equation from each individual regression line was used to recalculate the old XRF or INAA data for that particular element (Supplementary Appendix A).

A kmeans cluster analysis was performed using R Studio on all HSR pumice samples from this study and samples of interest from Streck and Grunder (1997). We used recalculated geochemical data from Streck and Grunder (1997). XRF and INAA analyses conducted by Streck and Grunder (1997) did not measure detectable amounts of some oxides and elements that recent XRF and ICP-MS analysis are able to detect, such as P₂O₅ (XRF), Cr (XRF), and As (INAA). There were also no data reported for elements such as Pr, Gd, Dy, Ho, Er, and Tm, and several samples (i.e., RT219A; RT62A; RT173I; RT173L) did not have measurable Ba, Nb, Y, Pb, Rb, Sr, and Zr INAA data. Because kmeans clustering analysis cannot be conducted when cells in a given spreadsheet are empty (there is no numerical value for that element or oxide), the aforementioned missing elements and oxides were excluded from the analysis, as well as the samples with a significant number of missing values. A principal component analysis (PCA) was then conducted on the same data to determine which elements and oxides have the strongest influence on differentiating the samples, and thus, differentiating each cluster. Elements and oxides with PC1 values >1.5 or < −1.5 were considered the most significant factors in distinguishing samples, and those with PC2 values >2.0 or < -2.0 were also considered. The significance of this subset of elements and oxides was verified by observable differentiation in scatterplots. See Supplementary Appendix C1 for the data spreadsheet used and Supplementary Appendix C2 for the script used in R Studio to conduct both the kmeans clustering and PCA analyses.

Wet liquidus temperatures were calculated for each HSR group using Rhyolite-MELTS software (Gualda et al., 2012). Wet liquidus temperatures were calculated from 1–5 wt% H₂O at 200 MPa and 1–4 wt% H₂O at 150 MPa. Water content ranges were chosen based on the experimentally determined water solubility in rhyolitic melts at ∼850°C (Wallace and Anderson, 2015). Rhyolite-MELTS software was unable to generate wet liquidus temperatures for any Group composition at 150 MPa and 5 wt% H₂O, likely because such water content is not soluble at depths as shallow as 150 MPa.

Storage temperatures were calculated using various equations from Putirka (2008), such as the plagioclase-liquid, alkali feldspar-liquid, clinopyroxene-liquid, and zircon-saturation geothermometers. Appropriate geothermometers for each Group were chosen based on mineral assemblages determined by Streck and Grunder (1997). RST HSRs have euhedral phenocryst phases (Streck and Grunder, 1997), and there is homogeneity within phases (Streck and Grunder, 2008), suggesting these crystals were in equilibrium with the host melt and, thus, are deemed acceptable phases for geothermometers that involved crystal and liquid compositions (e.g., Putirka, 2008). Mineral compositions determined by Streck and Grunder (1997) were used where required, including the clinopyroxene-liquid, plagioclase-liquid and alkali feldspar-liquid geothermometers. Whole rock compositions for all samples within a Group were used where a liquid or glass composition was required, such as the clinopyroxene-liquid, plagioclase-liquid, alkali feldspar-liquid, and zircon-saturation geothermometers. Water content (wt% H₂O) does not influence results for alkali feldspar-liquid temperatures, so 2 wt% H₂O was entered into the calculation spreadsheet but essentially acted as a placeholder. However, water content does make a notable difference in plagioclase-liquid temperatures. It should be noted that HSR Group A is aphyric, but Streck and Grunder (1997) found and analyzed three plagioclase phenocrysts. The single plagioclase-liquid temperature reported for Group A comes from the average of temperatures calculated at 2 wt% and 4 wt% H₂O. For the zircon-saturation geothermometer, the calculated maximum temperature represents the crystallization temperature if the sample does not contain zircon or if it is unknown whether the sample contains zircon. If the sample is known to contain zircon, the calculated minimum temperature represents the crystallization temperature. Streck and Grunder (1997) determined that zircon is present in all HSR Groups except for Group A.

Mineral separates of quartz, feldspar, titanomagnetite, and clinopyroxene from each Group were targets for three oxygen isotope analysis using University of Oregon’s Laser fluorination line with 35W Newwave CO₂ IR laser run in O₂ mode. Oxygen isotope values were gathered to calculate storage temperatures. Samples chosen for this analysis are from Streck and Grunder (1997), including MS-91-34E (Group E), MS-91-165A (Group C), MS-91-173B (Group D), and MS-91-173C (Group E). No phenocrysts were analyzed from Group A because it is aphyric. Titanomagnetite was separated and analyzed for Groups B-E, clinopyroxene and feldspar were analyzed for Groups C-E, and quartz was analyzed for Groups B and C. Temperatures were calculated using the equation and appropriate coefficient (A) from Chiba et al. (1989) for quartz-titanomagnetite (qtz-tmt), quartz-clinopyroxene (qtz-cpx), quartz-albite (qtz-alb), albite-clinopyroxene (alb-cpx), albite-titanomagnetite (alb-tmt), and clinopyroxene-titanomagnetite (cpx-tmt). It should be noted that coefficients from Chiba et al. (1989) were experimentally determined under higher pressure (15–16 kbar) conditions than what we estimate for the RST HSR magmas, so these are considered in addition to several other geothermometers when determining magma storage temperatures.

Storage pressures were calculated using SiO₂ wt%, K₂O wt%, and Na₂O wt%, representing quartz, albite, and orthoclase components. CIPW normative values for the average composition were calculated using the CIPW normative value calculator from Bolte et al. (2015). These values were plotted on a haplogranitic Qtz-Ab-Or ternary diagram (Holtz et al., 1992) to determine the storage pressures and equivalent storage depth for each Group.

At first glance, plotting the original XRF and ICP-MS data from Streck and Grunder (1997) with our new XRF and ICP-MS data showed significant offset between several elements, most notably Zr and Nb, where older XRF concentrations for each element were consistently higher than new ICP-MS concentrations (Figure 3). Results from calculating the percent difference (% diff.) between XRF and INAA data from Streck and Grunder (1997) and our new XRF and ICP-MS data determined that Zr, Nb, Ce, Pb, Sr, and Zr were elements that needed to be recalculated (Table 1). Ba was not recalculated, because the large difference of 84% is only observed within the Group with the lowest concentration of Ba, Group A. The percent difference of Ba in Groups B–E ranges from only 2.9–0.2%. See Supplementary Appendix A for full recalculation methods and all recalculated values. These recalculated values are used from this point forward in the study unless otherwise noted.

Based on the XRF and ICP-MS results, all 17 samples analyzed, including the two homogeneous pumices, are high-silica rhyolites (>75 wt% SiO₂), except two band separates (VS20-120A.2 and VS20-122G.2) that had between 70 and 73 wt% SiO₂. These may be a mixture of a dacite magma or mafic component in the system and, thus, are excluded from analyses to focus on HSR petrogenesis.

In accordance with Streck and Grunder (1997), the majority of the HSRs appeared to be broadly segregated into groups with compositional gaps, which is readily apparent with respect to FeO*, Ba, Nb, and Zr compositions. Major element oxides other than FeO* do not prove to be useful in segregating Groups, and the trace element variability among all HSRs analyzed is astounding. See Supplementary Appendix B for full XRF and ICP-MS results.

Based on scatterplots of numerous elements and oxides, combined with re-examining the hand sample from which the material was separated, it was clear that some separates were accidental physical mixtures of two or more Groups during the preparation process (Figure 4). It is beneficial to exclude these samples from the dataset entered into the statistical analyses, because they will blur the geochemical distinction parameters that would otherwise be clear in generating clusters.

Plotting geochemical data from VS20-120E.1 suggested it may be a mixture of Groups A and B. Two band separates were analyzed from sample VS20-120E. Upon closer examination of the hand sample from which the bands were separated, the separate labeled as the lighter material (VS20-120E.1) seems to include material of possibly two different magmas; one appears slightly lighter and more stretched than the other.

Band separate VS20-120E.2, the darkest band analyzed from sample VS20-120E.2, appeared to be a potential mixture of D and E with some parameters like Ba and Eu/Eu*, but it plotted with strong similarity to Group D with respect to elements like Nb and Zr. The dark grey material targeted for this band separate appears homogeneous, and calculated mixing proportions range dramatically. Thus, we conclude that this band separate is not likely to be a mixture of multiple magmas, and it is likely a Group D HSR magma.

Scatterplots of geochemical data from VS20-122C.1 suggested it could be a mixture of Groups A and B, but with a number of outlying concentrations with various elements and oxides. Two band separates were analyzed from sample VS20-122C. The pumice from which the material was separated has thinner, more convoluted banding relative to other pumices analyzed, making physical isolation of one color or texture of material difficult.

Sample VS20-120A.1 consistently plots between Groups B and C on element and oxide scatterplots, but usually slightly more similarly to Group B. The hand sample of this pumice has relatively thin and convoluted banding, so isolation of a single composition was difficult in the preparation process.

To objectively determine how these HSRs would be grouped based on composition, a PCA was first conducted on all samples (excluding the aforementioned outliers). This resulted in the samples being placed into eight clusters (Figure 5; Supplementary Appendix C). In accordance with findings from Streck and Grunder (1997), the PCA determined that the most influential factors (PC1 > 1.5) in distinguishing the HSRs were Ba, Eu/Eu*, Eu, FeO*, Hf, Zr, with Cs, Nb, Rb, Ta, Th, Y, and Yb also being identified as significant factors (PC1 < -1.5). Scatterplots verify the significance of these parameters in clustering the samples.

A kmeans cluster analysis without presetting the desired number of clusters resulted in HSRs being placed into eight clusters (Figure 5; Supplementary Appendix C). As the PCA indicated, the primary components that influenced cluster formation were Ba, Eu/Eu*, FeO*, Zr, and Nb. All 11 samples categorized into cluster 6 were those identified as Group A by Streck and Grunder (1997) and samples from this study with the lowest Ba (10–51 ppm), Eu/Eu* (0.18–0.22), FeO* (0.74–0.89 wt%), and Zr (161–182 ppm), and the highest Nb (31.6–36.2 ppm). This also includes homogeneous white pumice VS20-122J, obsidian sample RNL 1903. All samples within cluster 6 are considered Group A HSRs. Granite lithic fragment MS-20-05 was not included in the kmeans cluster analysis, but it possesses compositional similarity to Group A and will be categorized as such.

All 10 samples categorized into cluster 8 were those identified as Group B in Streck and Grunder (1997) and samples from this study with 115–225 ppm Ba, 0.24–0.29 Eu/Eu*, 1.14–1.38 wt% FeO*, 242–308 ppm Zr, and 27.2–29.0 ppm Nb. All samples within cluster 8 are considered Group B HSRs.

All 13 samples categorized into cluster 4 were those identified as Group C in Streck and Grunder (1997) and samples from this study with 367–730 ppm Ba, 0.30–0.42 Eu/Eu*, 1.37–1.80 wt% FeO*, 338–375 ppm Zr, and 24.1–26.3 ppm Nb. This also includes homogeneous light grey pumice VS20-120H. All samples within cluster 4 are considered Group C HSRs.

Two of the three “Group D” samples from Streck and Grunder (1997) that could be included in the kmeans cluster analysis were included in cluster 7, as well as six other samples from this study. Compositions for samples within cluster 7 are 1,201–1,274 ppm Ba, 0.44–0.48 Eu/Eu*, 1.57–1.85 wt% FeO*, 389–401 ppm Zr, and 21.9–23.0 ppm Nb. All samples within cluster 7 are considered Group D HSRs. Two samples from this study were categorized into cluster 5, including VS20-122H.3 with 1,370 ppm Ba, 0.49 Eu/Eu*, 1.95 wt% FeO*, 401 ppm Zr, and 23.6 ppm Nb, and VS20-122A.2 with 1,298 ppm Ba, 0.47 Eu/Eu*, 1.83 wt% FeO*, 397 ppm Zr, and 22.8 ppm Nb. VS20-120E.2 is the only sample within cluster 1, with 1,541 ppm Ba, 0.53 Eu/Eu*, 1.94 wt% FeO*, 410 ppm Zr, and 22.8 ppm Nb. Sample RT4A, originally identified as a ‘Group D’ sample in Streck and Grunder (1997), is the only sample categorized within cluster 3. RT4A has 1116 ppm Ba, 0.47 Eu/Eu*, 1.95 wt% FeO*, 400 ppm Zr, and 21.5 ppm Nb.

Slight variances in the compositions of VS20-120E.2, RT4A, VS20-122H.3, and VS20-122A.2 provide an explanation for why the statistical analysis essentially categorized these samples as separate and distinct groups and how this can be reconciled. RT4A was originally categorized as a Group D HSR in Streck and Grunder (1997). The PCA identified Ba as one of the top five most influential components for distinguishing rhyolites. Both VS20-120E.2 and RT4A plot with striking similarity to Group D samples in almost all compositional parameters, except for Ba, where there are slight variations (∼85–270 ppm difference). The broad similarity but variance in one critical element may be the cause for RT4A and VS20-120E.2 being isolated from all other Groups. Cluster 5 consisted of only two samples: VS20-122H.3 and VS20-122A.2. With respect to the most distinguishing parameters (e.g., Zr, Eu/Eu*, Nb), these samples plot similarly to Group D. These samples only vary slightly from Group D composition with respect to a few elements, such as Ta, TiO₂, FeO*, and Ba. As with samples VS20-120E.2 and RT4A, these few and slight variations may have been sufficient to result in a categorization as a separate cluster in the analysis. We conclude that the four samples distributed among kmeans clusters 1, 3, and 5 should be combined with those in cluster 7, and all should be considered Group D HSRs.

All 12 samples categorized into cluster 2 were those classified as Group E in Streck and Grunder (1997) and samples from this study with the highest Ba (1835–1999 ppm), Eu/Eu* (0.56–0.68), FeO* (1.83–2.18 wt%), and Zr (420–451 ppm), and the lowest Nb (21.1–22.7 ppm). All samples within cluster 2 are considered Group E HSRs.

In total, there are 12 samples classified as Group A, 11 as Group B, 14 as Group C, 14 as Group D, and 12 as Group E (Figure 5). Samples VS20-120E.1 and VS20-122C.1 are classified as mixtures of Group A and B magmas, and sample VS20-120A.1 is considered a mixture of B and C magmas with some possible contamination of the low-silica VS20-120A.2 magma. The two homogeneous pumices included in this study possess compositions that strongly represent a single HSR Group type: VS20-120H, a light grey, minimally stretched pumice is composed of Group C magma, and VS20-122J, a white, super stretched pumice, is composed of Group A magma.

Individual banded pumices contain at least two and as many as four HSR magma types (Figure 6; Supplementary Appendix D). For example, VS20-122B contains bands of Group C and D magmas, VS20-120C contains Group A, C, and E magmas, and VS20-122E contained Group B, C, D, and E magmas. As previously mentioned, material was separated from a pumice and considered a unique band relative to the other bands in an individual sample if it possessed a discernable difference in color and/or texture. Each analyzed separate did, in fact, represent a distinctly different composition to other bands within that sample, except for the accidental physical mixtures.

Average compositions (Table 2) were used in mass balance equations to produce mixing lines based on mixing proportions in intervals of 10% from 90–10% (Figure 7) as one possible test to determine whether mixing of endmember Groups can produce intermediate Groups (i.e., mixing of Group C and Group A to produce Group B). These calculations failed to produce intermediate compositions, most notably with concentrations of La, Sm, Ba, FeO*, Eu/Eu*, TiO₂, and Al₂O₃. Band separates previously identified as outliers or potential mixtures, including VS20-120E.1, VS20-122C.1, and VS20-120A.1, notably fell on or near some of these mixing lines. However, because standard mixing lines are calculated based only on the concentrations of two elements at a time, they alone are not enough to conclusively state whether these outlier analyses are or are not simply mixtures of multiple Groups.

In order to determine whether these outliers are, in fact, some mixture of two Groups, a simple mixing equation was arranged to calculate the potential mixing proportion of each element proportion ( % ) of   X   in   mixture = C m i x – C Y C X – C Y ⋅ 100 where C_mix is the element concentration of the potential mixture, C_X is element concentration of one endmember composition in the mixture, and C_Y is the element concentration of the other endmember composition in the mixture. This equation is arranged such that the resulting value is the percent of C_X in the mixture. The calculated average composition for each Group was used to address each sample of interest. If a sample is a mixture of two endmember components, the calculated proportions of each element should be similar (within ∼10%).

It should be noted that mixing proportion calculations, though a common method for determining the likelihood and proportions of mixture between two materials, do not provide meaningful results when the two materials of interest possess great similarity. Especially for trace elements of low concentrations (<10 ppm), small variations will result in a wide range of mixing proportions. For example, Group A has an average of 2.25 ppm Tb and 4.55 ppm U, and Group B has an average of 2.62 ppm Tb and 3.36 ppm U. When calculating the mixing of A and B to make VS20-122C.1 with 2.29 ppm Tb and 3.63 ppm U, the resulting proportions are low for Tb (∼10%) and are high for U (∼79%). With this noted, we report all results, and it is apparent that La, Ce, Nd, Sm, Eu, Hf, and Zr are most consistently within a ± 10% range of mixing proportions calculated for each given sample.

Calculated mixing proportions for VS20-120E.1 vary widely overall from ∼10 to 85%. However, of the 21 trace elements analyzed, 14 (La, Ce, Nd, Sm, Eu, Tb, Th, Hf, Ta, U, Rb, Cs, Sc, and Zr) fall within the range of an ∼27–45% proportion of Group A mixed with Group B.

Mixing proportions calculated for VS20-122C.1 also range widely from ∼10 to 80%. Of the 21 trace elements analyzed, seven (La, Ce, Nd, Sm, Eu, Hf, and Zr) fall within the range of an ∼27–41% proportion of Group A mixed with Group B. Five other trace elements (Lu, Ba, Th, Rb, and Cs) fall within the range of ∼62–71% proportion of A mixed with B.

Calculated mixing proportions for VS20-120A.1 overall range between ∼18 and 67%, with 9 of 21 trace elements calculated to have an ∼52–67% proportion of Group C mixed with Group B, including La, Ce, Nd, Sm, Eu, Tb, Ba, Ta, and U. Calculating mixing proportions to test if VS20-120A.1, instead, could be a mixture of A and D also displayed large overall variance (∼18–80%) but slightly better results, where 11 of the 21 trace elements (La, Ce, Nd, Th, Nb, Hf, Ta, U, Rb, Cs, and Zr) ranged between 64 and 80% proportion of Group A mixing with Group D. As mentioned previously, compositional analysis of VS20-120A.2 revealed that it is not a HSR, with only ∼73 wt% SiO₂. Inability to generate a reasonable mixing proportion of any magma types to generate VS20-120A.1 may be because it is a product of mixing with a non-HSR magma, possibly such as a dacite magma.

Wet liquidus temperatures were calculated using the average composition of rhyolites within each Group. Temperatures determined using Rhyolite-MELTS (Gualda et al., 2012) when conditions varied from 150 to 200 MPa and 1–5 wt% H₂O show mostly a decrease in calculated range overall from Group E to Group A, where Group E temperatures range from ∼850 to 970°C, from ∼833 to 970°C for Group D, ∼812–975°C for Group C, ∼790–980°C for Group B, and ∼768–990°C for Group A (Figure 8). Varying water content has a greater impact on calculated wet liquidus temperatures than varying pressure. The lowest water contents produce the highest temperatures for each Group. At 1 wt% H₂O at both 200 and 150 MPa, the overall trend shows a decrease in temperature from Group A to Group E. At 2 wt% H₂O at both 200 and 150 MPa, there is no discernable trend in wet liquidus temperatures from Group A to Group E. With 3–5 wt% at pressures where applicable, the trend shows a general increase in temperature from Group A to Group E.

Calculations from the Putirka (2008), δ¹⁸O, and Jorgenson et al. (2021) geothermometers yielded average storage temperatures for all Groups from ∼600 to 1,300°C, with temperatures increasing from Group A to Group E (Figure 8; Table 3). Pressure variations had little impact on the resulting temperature when using the plagioclase-liquid geothermometer for Group A, so the calculations were performed at 200 MPa. Considering all calculated temperatures, temperatures range from 787 to 842°C for Group A, 792–896°C for Group B, 811–1,135°C for Group C, 600–981°C for Group D, and 800–1,300°C for Group E. All δ¹⁸O results are considered accurate except for the titanomagnetite value for Group E (Table 4). δ¹⁸O of titanomagnetite for Groups B, C, and D were 2.79‰, 2.83‰, and 2.71‰, respectively, and was 4.17‰ for Group E. The relatively high δ¹⁸O value for Group E is likely due to an impure titanomagnetite separate.

Plotting CIPW normative values for average Group compositions on the Qtz-Ab-Or ternary diagram resulted in a wide range of pressures (Figure 9), with Groups A and C at ∼200 MPa (∼7 km depth), Groups B and D at ∼500 MPa (∼17 km depth), and Group E at ∼600 MPa (∼21 km depth). The obsidian sample categorized as Group A (sample RNL 1903) and the granitic lithic fragment (sample MS-20-05) with a similar composition to Group A were also plotted on the Qtz-Ab-Or ternary and yielded pressures of ∼175–200 MPa (∼5–7 km depth). XRF analysis of the obsidian sample had a significantly low percent loss-on-ignition (LOI ∼ 0.6%), so the nearly 1:1 Na/K ratio is considered as a representation of the true Na/K ratio for Group A. Therefore, the pressure estimates for the granite and obsidian samples are interpreted as more accurate for Group A. All high-silica rhyolite pumices have higher LOI (∼2-4%) and variable Na/K ratios and thus are likely affected by alkali exchange (cf. Streck and Grunder, 1997). Hence, this skewing of the Na/K ratio is assumed to be the case for all RST HSRs, so Na/K ratios were adjusted to 1:1 for each Group average and plotted on the same Qtz-Ab-Or ternary. After these adjustments, Group B and C are also at ∼200 MPa (∼7 km depth), but Group D remained at ∼500 MPa (∼17 km depth), and Group E remained at ∼600 MPa (∼21 km depth).

The clinopyroxene “cpx no liquid” geobarometer from Jorgenson et al. (2021) was used to calculate storage pressures for Groups B–E, because these Groups contained clinopyroxene (Figure 9). Clinopyroxene compositions come from INAA of 100–300 grains by Streck and Grunder (1997). Group B clinopyroxenes had a calculated pressure of 175 MPa (∼6.13 km depth), 217 MPa (∼7.52 km depth) for Group C, and 200 MPa (∼7.00 km depth) for Groups D and E.

Detailed compositional analysis (XRF and ICP-MS) of RST HSRs and subsequent statistical analysis in this study confirms the compositional Groups and compositional gaps first proposed by Streck and Grunder (1997). There is little evidence for mixing among these Groups observed in the geochemical data, which agrees with the conclusion of Streck and Grunder (1995) that these banded pumices represent distinct magmas that erupted together during the same eruption.

We consider all calculated storage temperatures between 700°C and 1,000°C to be valid, and thus, only temperatures within this range are included when calculating the average storage temperatures for each Group (Table 3). The significantly high clinopyroxene-titanomagnetite and albite-titanomagnetite temperatures of about 1,166°C and 1,300°C, respectfully, for Group E reflect the erroneously high δ¹⁸O titanomagnetite value for Group E. Temperatures from thermometers that involve albite vary widely and produce unrealistically low temperatures like 600°C and unrealistically high temperatures like 1,135°C for Group C. These erroneous temperatures are likely related to these thermometers from Chiba et al. (1989) being calibrated for high-pressure conditions. Additionally, Chiba et al. (1989) does not provide thermometers for Na-sanidine or anorthoclase feldspars. Albite thermometers were used because it was the closest composition to feldspars within the RST HSRs. However, this could also explain the wide range of calculated temperatures. Other calculated temperatures for Group D and E indicate similar storage temperatures, so it is assumed that the quartz-titanomagnetite and albite-titanomagnetite temperatures for Group E would likely be similar to Group D. Thus, when calculating the average storage temperature for Group E assumes the same albite-clinopyroxene and albite-titanomagnetite temperatures of Group D. The average calculated storage temperature for Group A is 805 ± 38°C, Group B is 826 ± 42°C, Group C is 877 ± 60°C, Group D is 880 ± 63°C, and Group E is 895 ± 73°C (Table 3).

All calculated storage temperatures using phase-liquid equations from Putirka (2008) suggest a broad increase of ∼805°C to ∼895°C from Group A to Group E. This agrees with estimated temperatures from Streck and Grunder (1995; 1997) ranging from ∼800 to 880°C. The Jorgenson et al. (2021) clinopyroxene no-liquid geothermometer results for Groups B-E are comparatively low, which may be due to the general increase in Fe at lower temperatures and the significance of Fe concentration in the pressure calculation.

Assuming this overall temperature increase from Group A to Group E is accurate, then wet liquidus temperatures would parallel this trend at greater water contents (≥3 wt% H₂O), implying either that all RST HSR magmas contained a minimum of 3 wt% H₂O or that water content must decrease from Group A to Group E. Combining wet liquidus temperatures with storage temperatures, Group E magma is the hottest, and magmas get progressively slightly cooler to Group A magmas.

Various geobarometers (Holtz et al., 1992; Jorgenson et al., 2021) show a slight increase in pressures from Group A to Group E, where Group A was stored at ∼6.1 km depth and Group E was stored at a ∼7 km depth. With calculated temperatures varying among different geobarometers, it can be generalized that these magmas were generated and stored at approximately similar depths of ∼6.5 km.

Melt densities for each HSR magma were calculated to further understand how these magmas would reside together in the crust (Figure 10). Melt densities are expected to increase with increasing Fe content, so a decrease in density is expected from Group E with the highest average FeO* content (∼2 wt%) to Group A with the lowest average FeO* content (∼0.8 wt%). Keeping temperature, pressure, and water content consistent among all Groups (800°C; 200 MPa; 3 wt% H₂O), calculated melt densities for adjusted and pre-adjusted Na/K ratios differed only slightly (∼0.005–0.008 g/cm³). Restoring the Na/K ratio to 1:1 (to make up for a slight Na-K exchange during glass hydration, cf. Streck and Grunder, 1997) resulted in a slight decrease in densities that ranged from ∼2.269 to 2.246 g/cm³ from Groups E–A. Continuing to use the compositions with the restored Na/K ratio, densities calculated at 800°C, 3 wt% H₂O, and decreasing 245–125 MPa from Groups E-A displayed an overall decrease (∼2.284–2.251 g/cm³) that highlights the decrease of density at lower pressures, particularly with Groups A and B. Densities calculated at 200 MPa, 3 wt% H₂O, and average storage temperatures for each Group (from Group E at 895°C and Group A at 805°C) also showed little variation compared to conditions where temperatures among Groups were consistently at 800°C. However, as expected, increasing temperature decreased the density of these magmas, so the calculated range of densities between less evolved (Group E) and more evolved (Group A) magmas becomes slightly muted (∼2.271–2.259 g/cm³). Keeping temperature (800°C) and pressure (200 MPa) consistent among Groups but increasing the water content from 2–4 wt% H₂O from Group E to Group A resulted in a drastic difference in calculated melt densities, where densities decreased from Group E (∼2.318 g/cm³) to Group A (∼2.227 g/cm³). These results show that water content is the most significant variable influencing melt densities of these HSR Groups. However, since increasing water content and temperature can have opposite effects on density, densities were also calculated where all three parameters were adjusted using our new pressure, temperature, and water content estimates. When using average storage temperatures for each Group, increasing pressure 125–245 MPa from Groups A-E, and decreasing water contents 2-4 wt% form Groups A-E, the overall trend was similar to when only varying water contents among Groups, but densities for each Group were slightly lower, ranging from ∼2.217 g/cm³ for Group A and ∼2.310 g/cm³ for Group E. These calculations underscore the significance of water content on the densities of RST HSRs.

If water content does, in fact, increase 2-4 wt% from Group E to Group A and we assume storage pressures of 200 MPa for all Groups, then calculated wet liquidus temperatures also show an overall increase of ∼115°C: ∼905°C for Group E, ∼870°C for Group D, ∼835°C for Group C, ∼821°C for Group B, and ∼798°C for Group A (Figure 11). These temperatures are approximately within the range of calculated storage temperatures for each Group (Figure 8).

Combining the plethora of previously acquired data and data from this study provides a wealth of information on magma storage conditions that can be useful in deciphering the magma storage configuration of the RST. However, it is apparent that the RST is derived from a complex system. From here, we synthesize all data to propose the most likely magma storage configuration of the RST.