Hydrothermal systems have been common since early in Earth history, form on timescales of minutes to thousands of years, are products of significant heat and mass transfer, and support diverse microbial communities. However, our understanding of how subsurface geological processes, such as reactive transport, phase separation, and shallow groundwater mixing, control hot spring geochemistry and in turn support microbial ecosystems and drive microbial diversification is poorly understood.
Yellowstone's hydrothermal system (hereafter referred to as Yellowstone) hosts the world's largest and most profound example of an active continental hydrothermal system with a surface expression (Hague, 1904;Allen and Day, 1935;White, 1957;Hurwitz and Lowenstern, 2014). A first-order observation of Yellowstone hydrothermal waters is that they have a distinct bimodal distribution of pH (Allen and Day, 1935;White, 1957;Fournier, 1989;Nordstrom et al., 2009;Lowenstern et al., 2012;Hurwitz and Lowenstern, 2014). To explain this bimodal distribution in the pH of Yellowstone hydrothermal waters, decades of research (White, 1957;Fournier, 1989;Nordstrom et al., 2009) have led to a simple and rather elegant model known as "phase separation." At its simplest, the model implies that a deep, hydrothermal reservoir, recharged by meteoric water and infused with crustal and magmatic gases, underlies all of Yellowstone (Rye andTruesdell, 1993, 2007;Kharaka et al., 2002). Heat-induced pressure and density differences provide driving forces that cause the deep hydrothermal fluids to ascend to the surface along high-permeability pathways, such as fractures and faults. Decompression boiling of the fluid ascending from the deep reservoir results in physical separation of a lowdensity vapor phase (made dominantly of steam, >>90%, and trace non-condensable gases) and a denser liquid phase (made dominantly of water, with non-volatile, soluble anions and cations).
The model holds that the two phases diverge and then migrate to the surface along different flow-paths. Fluids derived from the separated liquid tend to be near neutral to moderately basic and maintain their full inheritance of chloride (Cl -), as this ion is unlikely to partition into the vapor phase. These alkaline-chloride hydrothermal fluids also precipitate silica sinter that armors their flow-paths against water-rock interaction. In contrast, the vapor phase input (>>90% steam) is depleted in Cl -and enriched in volatile trace gases such as hydrogen sulfide (H2S) (Lowenstern et al., 2012). Subsequent mixing of the vapor phase with infiltrating, oxidizing, oxygen-rich near-surface groundwater (Fournier, 1989;W. Payton Gardner et al., 2010;Hurwitz and Lowenstern, 2014) converts the trace gas H2S and its derivatives (i.e., native sulfur (S 0 ) and sulfide minerals) to SO4 2-. These distinct water types have also been shown to host taxonomically and functionally dissimilar biological communities. High temperature acid-sulfate springs are dominated by aerobic Archaea (Inskeep et al., 2013;Ward et al., 2017;Colman et al., 2018) and their communities are enriched in functionalities that further facilitate the dissimilation and oxidation of sulfur compounds, thereby enhancing the water's acidity (Mosser et al., 1973;Colman et al., 2018Colman et al., , 2019)). In contrast, high temperature alkaline-chloride communities include both aerobic and anaerobic archaeal and bacterial members (Inskeep et al., 2013;Ward et al., 2017;Colman et al., 2018;Fernandes-Martins et al., 2021) that display a variety of metabolic strategies that are less dependent on dissimilatory metabolism of sulfur compounds and possibly more dependent on arsenic compounds (Fernandes-Martins et al., 2021).
While this first-order geochemical and hydrological paradigm of phase separation and shallow mixing can explain much of the geochemical and biological variation seen in Yellowstone's hydrothermal fluids, our understanding of these processes is limited to a twodimensional perspective. We can measure a pool's surface temperature, chemical composition, and biological characteristics, but our knowledge of the aquifer source lithologies, the subsurface architecture and fluid pathways, and the timescales of the fluid movement through these pathways are unknown. Furthermore, the significance and timescales of reactive transport in controlling the chemistry of phase-separated fluids generated in continental hydrothermal systems and its influence on microbial diversity have yet to be explored. This lack of fundamental information about Yellowstone's hydrothermal processes, in turn, limits interpretations of why such a stark difference exists in the taxonomic and functional diversity of microbial communities that inhabit these spring types.
To better understand how subsurface geological processes, such as reactive transport, phase separation, and shallow groundwater mixing, support microbial ecosystems and influence patterns in microbial diversification, we examine two adjacent, high-temperature pools in Norris Geyser Basin, Yellowstone National Park-Perpetual Spouter (Fig. 1), which is an alkalinechloride spring (pH ~7.5) and an acid-sulfate pool known informally as "Red Bubbler" (pH ~3).
Using this phase-separated hydrothermal system as a natural laboratory, we combine, for the first time ever, near-surface geophysical measurements to examine the architecture, geometry, and porosity of fluid pathways; radiogenic isotopic measurements to establish the different aquifer's lithologies and the significance of gas-water-rock interaction in determining the water's distinctive chemistries; and U-and Th-decay series isotopes to determine the timescales of water-rock interaction. We then apply metagenomic sequencing and informatics analysis of microbial communities inhabiting these springs, in the context of 56 metagenomes from other hot springs, to examine the consequences of variation in these dichotomous hot spring characteristics on the composition and function of microbial communities that inhabit these springs.
This coupling of methods provides a unique understanding of the interconnectedness of physical, chemical, and biological processes within this dynamic system and, more broadly, the bimodal differentiation of hot spring processes and microbial ecosystems observed across Yellowstone and continental hydrothermal systems in general (Brock, 1971). Specifically, we show that the bimodal distribution of phase separated systems is the result of feedbacks between the geochemical and biological (taxonomic and functional) composition of hot springs.
Subsurface processes drive phase separation and control reactive transport on variable timescales as inferred by isotopic measurements, and then this variation drives differences in microbial taxonomy and metabolism, which in turn additively contributes to a hydrothermal system's geochemical bimodality.
The geophysical methods employed for this discussion are 2D-DC resistivity imaging, ground penetrating radar, and surface Nuclear Magnetic Resonance (NMR) sounding. The goal of the geophysical study was to: 1) image any structural controls on existing hydrothermal pathways, 2) identify zones of active water permeability, and 3) potentially image steam or gas zones within the subsurface to understand near surface phase separation or mixing. DC resistivity data were collected in a 2D profile line using an AGI SuperSting R8 system with 112 electrodes at 1 m spacing. These 2D resistivity data were collected with a mixed array composed of dipole-dipole and strong gradient configurations. The two pools were centered within the 2D resistivity profile. Resistivity data were processed using R2 software (Binley, 2015). Inversion parameters include: uniform starting model equal to the average of the apparent resistivity, 140 Ohm m; error model of a = 0.02, b = 0.04; convergence in three iterations to a final error-normalized RMS misfit of 1.00 and depth of investigation (DOI) values were calculated following Oldenburg and Li, 1999. The vertical ground temperature gradient was removed by correcting all resistivity values to a standard of 18°C using the approach of Keller and Frischknecht, 1966 and the interpolated temperature measurements in Y-9 borehole near the Two Pools site (White et al., 1975).
Ground penetrating radar (GPR) data were acquired using a Noggin with 250 MHz shielded antennae (Sensors & Software, Mississauga, Canada), a trace spacing of 0.03 m, a time window of 60 ns and a sampling rate of 2.5 GHz. Data were processed using ReflexW (Sandmeier Geophysical Research, Karlsruhe, DE) using conventional filters and corrections: dewow, linear gain function, correct start-time, frequency bandpass (150 MHz -600 MHz), and Kirchhoff migration (v = 0.065 m ns -1 , determined by minimizing over/under migration artifacts).
Surface NMR data were acquired with a GMR instrument (Vista Clara Inc., Mukilteo, WA). The NMR sounding location was centered on Perpetual Spouter. The primary transmitting/receiving and noise loops were deployed in a "figure 8" pattern to minimize the effect of external electromagnetic noise on the measurements. The "noise compensation loop" was placed NE of the primary loop. Prior to deployment of the primary loops, a proton precession magnetometer was conducted to determine if any large (> 500 nT) magnetic gradients were present at the site. With no large magnetic gradients detected, the NMR system was deployed and acquired with 16 stacks at a local Larmor frequency (2255.6 Hz). NMR data were processed and inverted using MRSMatlab software (Müller-Petke et al., 2016). Inversion parameters included: regularization factor of 6x10 3 determined by L-curve analysis; vertical discretization of 1 m, sensitivity kernel built using resistivity structure determined from 2D-DC Resistivity, final goodness-of-fit Χ 2 = 1.1. Further details on the geophysical methods can be found in Supplementary Methods A.
Sr, Nd, Pb radiogenic isotopes (see supplement in Sims et al., 2013) and 238 U-230 Th were measured by mass spectrometric methods and 226 Ra and 228 Ra were measured by both gamma counting methods and high abundance sensitivity plasma mass spectrometry (Layne and Sims, 2000;Sims et al., 2008aSims et al., , b, 2013;;Scott et al., 2019). 222 Rn and 220 Rn were measured by in situ alpha counting of 218 Po and 216 Po using the Durridge RAD7 (Giammanco et al., 2007;Lane-Smith and Sims, 2013;Giammanco and Sims, 2022) and 223 Ra and 224 Ra were measured by radium delayed coincidence counting (RaDeCC) measurement of Mn-fibers.
Sampling and Chemical purification for Mass Spectrometry Hydrothermal water samples from Perpetual Spouter and Red Bubbler were collected for isotopic analyses during two separate field trips (July 2014 andAugust 2015). Hydrothermal water was pulled into a Luer-lock syringe and filtered through 0.1 μm to remove colloidal particles. Waters were filtered into acid washed LDPE bottles, and each sample was acidified in the field with ~1 ml 6M HCl for every 50 ml of sample water collected.
In the laboratory, filtered hydrothermal water samples were dried in preparation for chemical purifications using column chromatography modified from our laboratories wellestablished procedures (Layne and Sims, 2000;Thirlwall, 2002;Sims et al., 2008a, b;Scott et al., 2019). Dried samples were fluxed with concentrated HNO3 and sat overnight to fully oxidize the sample. Samples were then dried and re-dissolved in nitric acid for separation of U and Th from the bulk sample matrix using a large (10 ml column volume) anion exchange column in nitric acid. The U-Th fraction was dried and re-dissolved in 11M HCl + H2O2 for separation of Th from U using a smaller (1 ml column volume) anion exchange column in HCl. The separated U and Th fractions were dried in preparation for mass spectrometry.
The bulk sample matrix from the first large anion column was dried and re-dissolved in 6M HCl + H2O2 for purification of Fe from the bulk sample matrix. The same large anion column was used to remove Fe from the remaining bulk sample matrix, however the purification uses HCl rather than HNO3. The Fe fraction from this procedure was dried and refluxed in concentrated HNO3 to maintain full Fe oxidation. The Fe procedure was repeated using a smaller (1 ml column volume) anion column in HCl to ensure complete Fe purification. The final Fe fraction was dried in preparation for mass spectrometry.
The remaining bulk sample matrix from the Fe chemistry was dried and redissolved in 1M HBr for purification of Pb from the bulk sample matrix. An anion column (2 ml column volume) in HBr was used, and the purified Pb fraction was dried in preparation for mass spectrometry. The remaining bulk sample matrix was collected and dried for further chemical procedures.
The bulk sample matrix containing the remaining Sr, Nd, and Ra was dissolved in 1M HCl for separation of the three remaining fractions using a cation exchange column (4 ml column volume). Sr, Nd, and Ra were each removed from the cation exchange column in separate aliquots. The Sr and Nd fractions each required one further purification on a SrSpec (0.3 ml column volume) and LnSpec (0.5 ml column volume) column, respectively. Final purification of Ra involves a SrSpec (0.5 ml column volume) column procedure (performed twice) followed by a final anion exchange column (0.5 ml column volume) to remove 228 Th that grew in during the sample processing procedures. The final Sr, Nd, and Ra fractions are all dried in preparation for isotopic and concentration analysis by mass spectrometry.
The technique we have chosen for the short-lived isotopes of Ra ( 224 Ra, 223 Ra and 228 Ra), Mn-fiber Ra extraction, is a reliable and robust method, originally developed for shipboard collection on ocean cruises that utilizes Mn-oxide impregnated acrylic fibers (Mn-fibers) to scavenge Ra from seawater (Moore and Reid, 1973;Moore, 2008). The Mn-fiber extraction technique is particularly well-suited for the measurement of Ra in continental hydrothermal fluids for the following reasons: 1) all four Ra isotopes can be accurately determined from a single Mn-fiber sample; 2) Mn-fibers quantitatively adsorb Ra (Moore and Reid, 1973;Moore, 2008) at relatively fast flow rates, up to 1 liter per minute; 3) radium delayed coincidence counting (RaDeCC) measurement of Mn-fibers is a fast and efficient method for low-activity 223 Ra and 224 Ra, which is essential for achieving the high sample throughput required for such short-lived isotopes; and, 4) Mn-fiber Ra pre-concentration techniques do not inherently require large, expensive high volume pumps and shipboard laboratories, but can be exploited as an easily portable and simple in-situ method compatible with the limitations associated with sampling remote and difficult to access continental hydrothermal features often located in rugged terrain and ecologically fragile environments (e.g. Yellowstone). Following Ra extraction, Mn-fibers are transported from the field to a laboratory (or mobile laboratory for measurement of short-lived Ra) where 224 Ra and 223 Ra activities are measured on a RaDeCC following the methods of Moore (2008). Subsequently the Mn-fibers are ashed for gamma spectrometry measurement of 228 Ra and counted with a high-purity germanium well-type detector using methods detailed in Henderson et al., 2013. Because of the extreme water chemistry of Yellowstone thermal fluids, which ranges in pH from acidic to basic (1.4 -9.8), Eh (-0.231 -0.831 V) and temperature (ambient to 93C, which is boiling at Yellowstone's elevation of ~2,000 m) quantitative extraction of Ra on Mn fiber requires: 1) adequate sample size; 2) efficient Ra extraction and retention on Mn-fibers across a large range of pHs; and, 3) mitigating the potential effects of particulate bound Ra associated with large variations in suspended sediment load in hot spring waters.
All isotopic and solution concentration measurements were completed on the NeptunePlus multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS) at the High Precision Isotope Laboratory at the University of Wyoming. Chemically separated and purified Fe aliquots were dried and then re-dissolved in 1N (5%) HNO3 at least 2 hours prior to analysis. These Fe aliquots were diluted to match standard Fe concentrations, generally ~400 ppb, and Fe isotopes were analyzed using medium-resolution mode and an Apex desolvating nebulizer introduction system. Sample-standard bracketing was used to correct for mass bias.
Each individual sample was analyzed at least three times between individual IRMM-014 standard analyses. Reported errors are two-standard deviation confidence intervals for each hydrothermal fluid sample (Sims et al., 2008a;Dutton et al., 2017).
Purified Sr and Nd were dissolved in 5% HNO3 for isotopic analysis of the 87 Sr/ 86 Sr and 143 Nd/ 144 Nd in static mode using either an Apex or Aridus II desolvating nebulizer. All isotopes of Sr and Nd were measured in individual faraday collectors, and additional collectors were used to monitor and correct for interferences from Rb and Kr for Sr and Ce and Sm for Nd. Internal corrections for mass bias were applied assuming 86 Sr/ 88 Sr = 0.1194 and 146 Nd/ 144 Nd = 0.7219.
The Sr isotopic standard NBS987 and Nd isotopic standards LaJolla and Jndi-1 were measured to assess accuracy, and final isotopic compositions are reported relative to NBS987 87 Sr/ 86 Sr = 0.71024, LaJolla 143 Nd/ 144 Nd = 0.51185, and Jndi-1 143 Nd/ 144 Nd = 0.512107. Purified Pb aliquots were dissolved in 5% HNO3 for isotopic analysis of 208 Pb/ 206 Pb and 207 Pb/ 206 Pb using either an Apex or Aridus II desolvating nebulizer. Thallium was added to each sample to correct for mass fractionation with Pb/Tl ratio target of ~3/1. Lead and thallium isotopes were analyzed in static mode using six Faraday collectors with ratios normalized to 203 Tl/ 205 Tl = 0.418922 to account for instrumental mass bias. A seventh Faraday collector was used to monitor mercury. Lead isotope ratios are reported relative to NBS981 values of Thirlwall (Thirlwall, 2002).
Purified U and Th aliquots were dissolved in 5% HNO3 for isotopic analysis of the 230 Th/ 232 Th and 234 U/ 238 U ratios using an Apex desolvating nebulizer. These isotopic ratios are on the order of 10 -6 and 10 -5 and require the use of combined faraday -ion counter measurements.
Mass bias, ion counter yield, dark noise, and RPQ transmission were corrected for using sample standard bracketing and isotopic standards IRMM-035 for Th and U010 for U. In addition to isotopic ratios, total yields of U and Th were measured relative to ICP CertiSpex concentration standards to obtain the activities of individual nuclides in the hydrothermal fluids. Column chemistry yields for these procedures are no less than 70%; therefore, we assign a conservative error estimate for individual activities of 30%.
Purified Ra aliquots were dissolved in 5% HNO3 for isotopic analysis of the 228 Ra/ 226 Ra ratio using an Apex desolvating nebulizer. The NeptunePlus at the University of Wyoming is equipped with a dual SEM-RPQ collector block, which was used for static measurement of both isotopes of radium (Scott et al., 2019). Mass bias, ion counter yields, and dark noise were accounted for using sample-standard bracketing and an internal 228 Ra/ 226 Ra standard, first calibrated and measured in 2009. Total Ra yields were measured relative to the NIST4967 226 Ra standard. As with U and Th, the column chemistry procedures yield no less than 70% recovery, and we assign a conservative error of 30% to the individual activities of 228 Ra and 226 Ra.
The short-lived Ra isotopes ( 224 Ra and 223 Ra) are measured on a radium delayed coincidence counter (RaDeCC™) which measures 224 Ra and 223 Ra by proxy of the alpha decays from their Rn daughters ( 220 Rn and 219 Rn respectively) via a Lucas cell and attached photomultiplier tube; the delayed coincidence counter discriminates between Rn isotopes (see Moore and Arnold, 1996;Moore, 2008;Moore and Cai, 2013). Each Mn-fiber sample is counted three times: 1) within 0-5 days of extraction for optimal 224 Ra determination; 2) after 8-11 days to reduce uncertainty for 223 Ra associated with "chance coincidence" interference from 224 Ra; and 3) after ~70 days to determine the supported 224 Ra and 223 Ra activities (Moore, 2008).
The longer-lived 228 Ra and 226 Ra are determined by ashing the Mn-fibers in a muffle furnace at 400 ˚C to achieve an appropriate mass and geometry for gamma spectroscopy using 25mm polystyrene vials and a Canberra Industries Inc. high-purity germanium well-type detector. The vials are sealed (with a thin silicon plug and 5mm of epoxy) for 22 days to allow for Rn daughter ingrowth. 228 Ra activity is calculated from 338 and 911 keV (from 212 Pb and 228 Ac respectively) peaks and 226 Ra is determined from 295, 352, and 609 keV peaks (from 214 Pb, 214 Pb, and 214 Bi respectively).
Samples for molecular characterization of planktonic and sediment-associated communities were collected on May 22, 2017, as previously described (Fernandes-Martins et al., 2023). Duplicate sediment samples were collected from each spring using a sterilized spatula and placed in sterile 2.0 mL cryotubes. Duplicate samples of filtered water were collected using a peristaltic pump from a depth of ~20 cm beneath the spring surface. An aluminum in-line filter housing containing a 47 mm diameter, 0.22 µM pore-size Supor filter (Pall, Port Washington, NY) connected to Teflon tubing was used to collect planktonic cells. Prior to placing the sterile filter in the housing, ~2 liters of spring water were pumped through the housing and the tubing to minimize introduction of non-endogenous cells and their genomic DNA. The filter was then placed in the housing using ethanol-sterilized forceps and 6 liters of water was passed through each filter. Filters containing biomass were placed in sterile 50 mL tubes and were immediately frozen on dry ice. Sediments and filters were stored at ~80˚C until further processing. A Fast DNA Spin Kit for Soil (MP Biomedicals, Irvine, CA) was used to extract genomic DNA from each sample following the manufacturer's protocols. DNA extracts were quantified using the Qubit DNA HS Assay kit and a Qubit fluorometer (Molecular Probes, Life Technologies, Madison, WI) and frozen at -20˚C until further processing.
DNA extracts were subjected to shotgun metagenomic sequencing at the Department of Energy Joint Genome Institute (JGI; Walnut Creek, CA) following standard library preparation protocols. Sequencing was conducted on the Illumina Novaseq platform with paired-end 2 x 150 bp sequencing. Reads were quality-filtered, trimmed, and cleaved of Illumina-specific sequencing adapters using BBTools (sourceforge.net/projects/bbmap/) and then read corrected using bfc (v.r181) (Li, 2015) via the standard JGI metagenome processing pipeline. Reads were then assembled into contigs using SPAdes v.3.13.0 (Nurk et al., 2013) with the following specifications: -m 2000 --only-assembler --meta. Metagenome-assembled-genomes (MAGs) were constructed from the assembly, as previously described (Colman et al., 2022). Briefly, the metaWRAP pipeline (v.1.3.2) (Uritskiy et al., 2018) was used to map paired-end reads to the assembled contigs using the BWA aligner and MAGs were generated from assembled contigs (specifying an -l of 2500) based on similarities in sequencing depth coverage profiles and tetranucleotide frequency usage with the MetaBAT v.2 (Kang et al., 2015), MaxBIN v.2 (Wu et al., 2014), andCONCOCT v.1.1.0 (Alneberg et al., 2014) genome binning algorithms. The bin refinement module of metaWRAP was used to select the highest quality MAG dataset for each metagenome from the various binning strategies and by assessing contamination and completeness estimates of MAGs using CheckM v.1.0.5 (Parks et al., 2015). Only medium-high quality draft MAGs (>50% completeness, < 10% contamination) were included in the final analyses. Gene prediction and annotation was conducted using prokka v.1.13 (Seemann, 2014).
Relative abundances were calculated for individual populations with the 'profile' function within CheckM wherein relativized abundances are first calculated from the mapping of reads to individual MAGs and then normalized to estimated MAG sizes. Finally, relative abundances were normalized to all estimated abundances within communities. Whole-metagenome sequence diversity was calculated using a genome bin-independent approach via the Nonpareil (v.3.304) metric and the quality-filtered reads for each metagenome. Briefly, sequence diversity and coverage estimates are derived from Nonpareil sequencing coverage curves and evaluate the level of sequence redundancy within a metagenomic dataset to then approximate the extent of expected diversity that is covered by the actual sequencing effort (Rodriguez-R et al., 2018).
To compare the functional proteins encoded in each MAG, the KEGG Orthology (KO) assignments for each assembled metagenome were collected by comparison against the KEGG automatic annotation server (KAAS) (Moriya et al., 2007). The KO data were used to construct a presence/absence table for each MAG via adaptation of a MATLAB script (https://github.com/dcolman1/). Of the entire KO dataset, only those database annotations associated within the KEGG 'Energy Metabolism' category were used in further analyses to include only data that were most pertinent in discerning the relationship between spring geochemistry and microbial metabolism. A Jaccard dissimilarity matrix was constructed from the normalized dataset using the 'vegdist' function within the vegan package (v.2.4-4) for R and the distance matrix was subjected to principal coordinates analysis ordination in R (Oksanen et al., 2022; https://cran.r-project.org/web/packages/vegan/). In addition, a matrix describing the similarity in KEGG profiles of Red Bubbler and Perpetual Spouter were included within a larger analysis of 54 other chemosynthetic Yellowstone metagenomes that are components of our other ongoing studies. Briefly, KEGG assignments were collected from the Integrated Microbial Genomes (IMG) or KAAS platform for metagenomes and used to build a compositional matrix as described above, but weighted by the instances a KO was identified in a metagenome and only including KO annotations involved in the 'Energy Metabolism' category. The fraction of proteins annotated to the 'Sulfur Metabolism' category among all of the 'Energy Metabolism' annotated proteins was evaluated to assess whether Red Bubbler and Perpetual Spouter populations exhibited trends in "Sulfur Metabolism" consistent with those previously documented for chemosynthetic Yellowstone springs (Payne et al., 2019).
In addition to KEGG assignments for some protein ortholog groups, BLAST-based searches were conducted against the annotations for all MAGs. Specifically, MAGs were queried via BLASTp for specific gene functions, guided by gene contents in closely related genomes to screen for specified metabolic functions using bait sequences specific for the active site subunits for each of the proteins or protein complexes. Positive matches within MAGs were considered as those with an E-value > 1x10 -6 , >30% amino acid homology, and >60% of the length of the BLASTp bait sequence. The assembled metagenome data for the Perpetual Spouter samples are available in IMG under accessions 3300033484 (plankton) and 3300033491 (sediment) and those for the Red Bubbler samples are available in IMG under accessions 3300033476 (plankton) and 3300033830 (sediment).
All data are presented in Tables 123, Figures 1234567and in the Supplemental Tables S1-S5.
Major and trace element chemistry are from (McCleskey et al., 2014;2022b) and presented in Supplemental Materials, Table S1. Supplemental Materials, Table S5 provides isotopic measurements of USGS standard reference materials compared with literature values as a measure of quality assurance.
Phase separation results from decompression boiling and is thus a continuous process that occurs over a multitude of scales. However, for our study we focus on two adjacent phase separated pools Perpetual Spouter and Red Bubbler.
These "Two Pools", Perpetual Spouter and Red Bubbler, represent an ideal natural laboratory to investigate phase separation processes: 1) their surficial separation of 14 meters provides optimal scaling resolution for the geophysical measurements; 2) their subsurface geology is well constrained from three nearby boreholes (White et al., 1975); and 3) they have characteristic bimodal chemistries (Fig. 1).
The Two Pools site is in the Norris Geyser Basin (NGB), which is at the intersection of the Norris-Mammoth Corridor structural complex and the eastward extension of the Hebgen Lake fault system (Fig. 1a). Albeit outside the northwest rim of the Yellowstone caldera, NGB is one of Yellowstone's major hydrothermal basins and arguably one of its most active and spectacular. NGB has Yellowstone's highest measured surface and subsurface temperatures (Allen and Day, 1935;White et al., 1975), contains a great diversity of physical features (geysers, mud pots, etc.) and chemical variability, and is one of Yellowstone's most dynamic geyser basins, experiencing considerable uplift and subsidence (Chang et al., 2007(Chang et al., , 2010)). The minimum heat flow across the Norris Geyser Basin is 300-340 W/m 2 with the hydrothermal subbasin hosting Perpetual Spouter and Red Bubbler having an average minimum heat flow of ~310 W/m 2 (Heasler and Jaworowski, 2018). Maximum heat flow measurements for Norris Geyser Basin are variable, dependent on the existence and concentration of hydrothermal features at the surface. NGB also experiences significant periodic hydrothermal disturbances, during which the chemistry, activity, and nature of hydrothermal features in NGB can change drastically over short timescales (weeks to years) (White et al., 1975(White et al., , 1988;;Fournier et al., 2002;Colman et al., 2021). These hydrothermal disturbances are linked to changing interactions between the deep hydrothermal system and the surrounding shallow groundwater system (Fournier et al., 2002), the latter of which is more susceptible to recharge through recent precipitation (Colman et al., 2021). However, the nature and magnitude of the changes in these interactions are not well understood (Fournier et al., 2002).
Three boreholes (i.e., C-II, Y-9, Y-12) have been drilled in NGB (Allen and Day, 1935;White et al., 1975;Jaworowski et al., 2016) to investigate a region in the Yellowstone system dominated by acidic waters (Hole C-II) (Allen and Day, 1935) and high heat flow. Y-9 is the closest borehole to the Two Pools site. Except for Y-9, which starts in a thin surface layer of silica sinter and sinter-cemented glacial gravel and till (down to ~1 meter), Lava Creek Tuff (LCT) is the only lithology present in all three holes (maximum drilling depth of 331.6 meters at Y-12 (White et al., 1975)). Both LCT members, "A" (lower unit) and "B" (upper unit), are present in these boreholes, with the contact between these two units located from 23.8 meters (C-II) to 38.9 meters (Y-9) (White et al., 1975(White et al., , 1988)). C-II is an endmember amongst these holes and as such it is the most representative of a shallow vapor-dominated system and its concomitant hydrothermal alteration (White et al., 1975). Although none of the boreholes penetrated beyond the LCT, the regional geology suggests that a thick sedimentary package of Paleozoic and Mesozoic sandstones, shales and carbonates underlie the LCT in this region (Anon, 1972). For further information on the cores and the downhole measurements, such as change in temperature with depth, which does not follow model geotherms, the reader is referred to (Allen and Day, 1935, White et al., 1975, 1988).
Perpetual Spouter and Red Bubbler are located at the base of Ragged Hills along Tantalus Creek, downstream from the Back Basin of NGB (Fig. 1b). Ragged Hills, which form an elevated south-western margin of NGB, are thermal kames of ice-contact sediments, up to 30 meters thick (Richmond and Waldrop, 1975;White et al., 1988;Fournier et al., 2002;Jaworowski et al., 2006). These thermal kames formed at the end of the Pinedale glaciation (22-13 ka) when melting ice deposited sand and gravel within an active hydrothermal area, leaving behind thick clastic deposits that were indurated and cemented by opal and retain a fairly high porosity on a variety of length scales (Richmond and Waldrop, 1975;White et al., 1975White et al., , 1988;;Fournier et al., 2002;Jaworowski et al., 2006Jaworowski et al., , 2016)). The surface of the Two Pools site is hard and rocky with each spring emerging from distinct pools. Both are precipitating mineral deposits: Perpetual Spouter is depositing silica sinter, Red Bubbler is depositing alternating layers of hematite and silica -much like banded iron formations.
Perpetual Spouter, named by Hague andothers in 1904 (Hague, 1904), is an alkalinechloride pool with a temperature of ~90°C, a pH of ~7.5, low SO4 2-(~37 mg/L) and high Cl - (~790 mg/L) (Supplemental Materials, Table S1). The conduit of Perpetual Spouter is lined with silica sinter precipitate to an unknown depth. This silica sinter lining greatly limits shallow water-rock interaction between Perpetual Spouter's waters and the underlying LCT.
Na/K geothermometry estimates the Perpetual Spouter reservoir at about 210 °C using the Truesdell equation and at about 250 °C using either the Fournier or the Giggenbach equation (Tuesdell and Fournier, 1976). Because silica is being precipitated during upflow, the preferred temperature is 250 °C (White et al., 1992). Despite periodic thermal disturbances in other systems in NGB, Perpetual Spouter's geochemical composition has remained remarkably constant over both long and short timescales (Fournier et al., 2002). Measurements of tritium show that Perpetual Spouter's waters (measured from 1967-1976) are best modeled as a mixture of 7% tritiated water with a short residence time (10 years), and thus presumed to be shallow recharge, and 93% tritium-free waters which are hypothesized to be from a deep reservoir (Pearson and Truesdell, 1978;Gardner et al., 2011). Perpetual Spouter's flow rate appeared to be consistent over the time of measurement.
Red Bubbler is an acidic, ephemeral pool with a temperature of ~88°C, a pH of ~3, high SO4 2-(~207 mg/L) and moderate Cl -(~340 mg/L) (Supplemental Materials, Table S1) (McCleskey et al., 2014;2022b). This relatively high Cl -concentration for an acid-sulfate system requires some remixing with boiled Cl --enriched waters, with the most likely candidate being water from the Perpetual Spouter system. While Red Bubbler's chemistry has remained fairly consistent, its water level exhibits considerable variability from overflowing to nonexistent, to filling and refilling with regular periodicity, over a variety of timescales (hours to seasonally). In addition to Red Bubbler, there are other nearby acid-sulfate springs emanating within Tantalus Creek, and nearby high-temperature fumaroles upgradient on the flank of Ragged Hills indicating that Red Bubbler, albeit the dominant feature, is just part of a complex and highly dynamic vapor-dominated system that is being influenced by shallow groundwater recharge.
Not surprisingly, the Two Pool's distinct chemistries result in contrasting concentrations of solutes. The dissolved ion concentrations of elements such as Pb, Fe, Nd, Sr, U, Th, and Ra in the acidic Red Bubbler waters are much higher than in waters from Perpetual Spouter, often by an order of magnitude or more, whereas elements such as Ca, Na, Cl, F, Li, B, Br, Cs, and As are all higher in Perpetual Spouter.
GPR imaging (Fig. 2a) revealed convex up, subparallel, moderately continuous reflections which attenuate at <2 m depth that we interpret to be silica-cemented, clastic kame deposits associated with the Ragged Hills, localized Fe deposits associated with Red Bubbler, and silica sinter associated with Perpetual Spouter. The 2D-DC resistivity results show shallow prominent, but discontinuous, regions of high resistivity down to ~4 m that can be interpreted as regions of either: 1) lower bulk porosity than the surrounding material or 2) increased gas content within a laterally consistent bulk porosity. Knowledge of the surface geology and nearsurface geology from boreholes (White et al., 1975) leads us to interpret these high resistivity zones as areas with lower bulk porosity, likely regions of silica sinter and sinter-cemented clastic glacial kame deposits. Directly below Perpetual Spouter (Fig. 2b) is a narrow, vertically oriented, low resistivity zone (~80 Ohm m) that is interpreted to represent the primary conduit feeding the spring from a reservoir >10 m below the surface. Likewise, immediately below Red Bubbler, in the top 2 m, is a low resistivity zone (~30 Ohm m) that extends southwest towards Ragged Hills, (Fig. 2b) parallel to the ground surface. The surface NMR data (Fig. 2c-f) show a decrease in water content with depth that, assuming saturation, would indicate a decrease in porosity and/or a decrease in the extent of weathering.
The isotopic measurements are tabulated in Tables 1-3 and shown in Figures 345. Radiogenic Nd, Sr, Pb and Fe isotopic compositions and U-and Th-decay series abundances of Red Bubbler and Perpetual Spouter and other relevant acidic and neutral chloride hot springs are tabulated in Tables 1 and 2. Measurements of nearby and regional volcanic and sedimentary rocks from potential aquifers for these waters are tabulated in Table 3; these data come from measurements made at the UW High Precision Laboratory and published data from other studies (Leeman et al., 1977;Doe et al., 1982;Hildreth et al., 1991;Hildreth et al., 1984;Sturchio et al., 1993).
The isotopic compositions of Perpetual Spouter and Red Bubbler are starkly different, particularly with regards to 208 Pb/ 206 Pb (Fig. 3a; Table 1). Red Bubbler's acid-sulfate waters exhibit a 208 Pb/ 206 Pb signal that is essentially identical to that of the Lava Creek Tuff, whereas Perpetual Spouter's alkaline-chloride waters present a substantially lower 208 Pb/ 206 Pb that is similar in value, but not identical, to that of the Paleozoic and Mesozoic sedimentary rocks buried beneath the Lava Creek Tuff in the NGB.
In contrast to 208 Pb/ 206 Pb, the 87 Sr/ 86 Sr of volcanic and sedimentary rocks show considerable overlap and are by themselves less sensitive for distinguishing the source characteristics of Yellowstone's hydrothermal waters. Because of low analyte concentrations for the amount of water collected (150 ml), the 143 Nd/ 144 Nd isotope measurements for Perpetual Spouter are unreliable and thus not reported. Nonetheless, and importantly, 87 Sr/ 86 Sr and 143 Nd/ 144 Nd of Red Bubbler's acidic waters are identical to those of Lava Creek Tuff (Tables 1&3). Finally, we note that consistency in Pb and Sr isotopic compositions measured in Perpetual Spouter and Red Bubbler over the two-year time period suggests these two pools were compositionally in a steady-state.
To confirm these isotopic differences, we extended our Sr and Pb radiogenic isotopic study to several other springs within Yellowstone National Park (Fig. 3). The extended suite includes: nearby acid-sulfate waters also hosted in LCT; the alkaline-chloride feature Beryl Spring, which is believed to represent the quintessential deep hydrothermal water beneath Yellowstone (Nordstrom et al., 2004), and is also hosted in the LCT; and, finally, waters from Mammoth Terraces and Boiling River, whose unique calcium-bicarbonate-sulfate compositions require them to have interacted extensively with the deeper Mesozoic and Paleozoic rocks (Kharaka et al., 1991;Fouke et al., 2000). The acid-sulfate samples have high 208 Pb/ 206 Pb, whereas the alkaline-chloride waters of Beryl Spring and calcium-bicarbonate-sulfate waters from Mammoth Terraces and Boiling River have low 208 Pb/ 206 Pb. We note that 208 Pb/ 206 Pb measurements from the drill core of Y-10 at Mammoth Terraces (Leeman et al., 1977) yield identical values as our measurements of calcium-bicarbonate-sulfate waters from Mammoth Terraces and Boiling River.
We also measured stable Fe isotopes in Red Bubbler and Perpetual Spouter waters and in both the LCT and the Madison Limestone (Tables 3). Our measurements for Perpetual Spouter were effectively below detection limit. For Red Bubbler 56 Fe/ 54 Fe and 57 Fe/ 54 Fe are identical to LCT, within uncertainty, indicating that chemical weathering in this acidic system is mobilizing Fe without isotopic fractionation.
We use measurements of 228 Ra and 226 Ra to determine the timescales of water-rock interaction, and measurements of both long-lived ( 238 U, 232 Th and 230 Th), and short-lived ( 224 Ra, 223 Ra and 222 Rn) isotopes to constrain model parameters and to validate assumptions implicit to the application of the ( 228 Ra/ 226 Ra) chronometer (Fig. 4 & 5; Table 2 & 3). All nuclide abundances and ratios are given in activities, denoted by parentheses (e.g., ( 228 Ra/ 226 Ra)), to provide comparable units within a decay chain.
Activity ratios of ( 228 Ra/ 226 Ra) in Perpetual Spouter and Red Bubbler waters are drastically different from each other. Perpetual Spouter's alkaline-chloride waters have a much lower ( 228 Ra/ 226 Ra) of 5.9 ± 3.1 (n = 3); Red Bubbler's acid-sulfate waters have a very high ( 228 Ra/ 226 Ra) of 45.0 ± 6.4 (n = 3). With the exception of 222 Rn (which we suspect, as discussed in detail below, is degassing over the whole Two Pools region), the activities of the individual nuclide abundances are consistently higher in Red Bubbler than they are in Perpetual Spouter.
These differences vary amongst different nuclides according to their solubilities and half-lives and thereby provide information about assumptions (e.g., the solubility of Th) and critical model parameters, such as recoil rates.
As possible rock source lithologies for our reactive transport modeling, we also provide a compilation (Table 3) of values for [U], [Th], and ( 232 Th/ 238 U) for unaltered LCT (Doe et al., 1982;Hildreth et al., 1984Hildreth et al., , 1991) ) and altered LCT (as represented in Y-12 drill core from NGB (Sturchio et al., 1993)), as well as several different possible local sedimentary rock lithologies.
Because of a lack of knowledge of the actual sedimentary rock(s) involved, we calculated an average Paleozoic-Mesozoic sedimentary rock composite for modeling purposes using the average [U], [Th] and ( 232 Th/ 238 U) values from Sturchio et al., 1993.
A total of 4 -5 x 10 8 sequence reads were generated from Perpetual Spouter and Red Bubbler water and sediment microbial communities. Estimation of genomic diversity coverage based on read sequence redundancy revealed exceptionally high genomic coverage (99.6%, 99.9%, 99.9% and 99.9% for the four metagenomes, respectively), suggesting that additional sequencing effort would not recover substantial additional genomic (and thus, taxonomic) diversity, unlike what is typically observed for metagenomic analyses of other microbial environments (Rodriguez-R et al., 2018). Assembly of the reads and genome-based binning from the metagenomes generated 48 genome bins (i.e., metagenome-assembled-genomes; MAGS), including 15 and 20 from Perpetual Spouter water and sediment communities, respectively, in addition to four and eight from Red Bubbler water and sediment communities, respectively (Supplemental Materials, Table S3). Of the entire metagenomic read datasets, 61.7%, 76.7%, 87.3%, and 68.8% of reads were mapped to the MAGs, respectively, suggesting a high level of fidelity between the MAG dataset to the overall metagenomic read dataset. Consequently, the MAG-based representation of the communities is reflective of the expected diversity in the systems given the high level of sequencing effort for relatively non-diverse communities, nearly 100% recovery of estimated genomic diversity, and a high degree of the read datasets mapped to the reconstructed MAGs. Ordination of a matrix describing the dissimilarity in protein homologs involved in energy metabolism that are encoded in these 48 MAGs reveals two highly resolved groups that largely correspond to each spring (Fig. 6a). The planktonic and sediment MAGs from Red Bubbler form a tight group, while the planktonic and sediment communities from Perpetual Spouter are more functionally distinct from each other, with many Perpetual Spouter sediment MAGs clustering with those from Red Bubbler sediments and water (Fig. 6a). Consistent with previous observations that an enrichment of genes encoding proteins involved in dissimilatory sulfur metabolism is a distinguishing feature of communities inhabiting acidic Yellowstone springs relative to those in circumneutral/alkaline springs (Colman et al., 2019;Payne et al., 2019), the Red Bubbler communities were substantially enriched in proteins involved in dissimilatory sulfur metabolism whereas the Perpetual Spouter communities were less enriched (Fig. 7). Communities from both springs clustered among other springs with similar pH, when considering 54 other metagenomes from 46 springs spanning much of the geochemical diversity of Yellowstone (Fig. 7; Supplemental Materials, Table S2). Thus, the communities of the two pools were representative of the variation in community composition that is observed across the broader Yellowstone geothermal system.
To further assess putative taxonomic and functional differences among communities, MAGs were subjected to metabolic reconstructions from which their putative functions were predicted (Figs. 6b & 6c; Supplemental Materials Table S4). Despite similar functional profiles of the Red Bubbler planktonic and sediment communities (Fig. 6a), they have highly different taxonomic compositions. Both communities are dominated by Archaea primarily within the Sulfolobales class, albeit different Sulfolobales taxa dominated the water and sediment communities (Fig. 6b). Specifically, the planktonic community was almost completely dominated (>99% of the community) by an uncultured Sulfolobales 'Acd1' population that represented < 5% of the sediment community. The sediment community was taxonomically more evenly distributed and comprised MAGs closely related to characterized aerobic Sulfolobaceae-like Sulfurisphaera (i.e., the QEFN01 group), Saccharolobus, and Acidianus, in addition to the anaerobes Acidilobus, Vulcanisaeta, and Thermoproteus (Supplemental Materials, Table S3). Despite taxonomic differences, MAGs comprising both the planktonic and sediment communities in Red Bubbler encoded similar suites of proteins that are involved in the dissimilatory reduction or oxidation of various sulfur compounds that include hydrogen sulfide (H2S), thiosulfate (S2O3 2-), and native sulfur (S 0 ) (Fig. 6b; Supplemental Materials, Table S4).
Sediment MAGs, including that for the most dominant sediment Sulfolobales population, also encode homologs of sulfocyanin enzymes that allow for aerobic oxidation of ferrous iron (Fig. 6b; Supplemental Materials, Table S4). The capacity for either arsenite (As(III)) oxidation or arsenate (As(V)) reduction, in addition to the reduction of oxidized nitrogen compounds (e.g., nitrate, NO3 -; nitrite, NO2 -, or nitrous oxide, N2O) were not observed in any of the Red Bubbler MAGs (Fig. 6b; Supplemental Materials Table S4).
In contrast to Red Bubbler populations, the dominant populations represented by MAGs in the Perpetual Spouter planktonic and sediment communities are taxonomically similar and comprise both archaeal and bacterial populations (Fig. 6c; Supplemental Materials, Table S3).
The dominant MAGs in both plankton and sediment communities (~42 and 72% of communities, respectively) are closely related to an uncultivated taxon affiliated with the Caldarchaeales group (formerly 'Aigarchaeota'). Metabolic reconstruction suggests these Archaea to be aerobic and facultatively autotrophic. The Perpetual Spouter Caldarchaeales MAGs encode several homologs of enzymes needed for the 3-hydroxypropionate/4-hydroxybutyrate pathway of CO2 fixation and homologs of enzymes putatively enabling the reduction of As(V), in addition to the oxidation of H2S and carbon monoxide (Fig. 6c, Supplemental Materials, Table S4). The putative ability of populations to reduce As(V) was also indicated among many other Perpetual Spouter MAGs that were primarily recovered from sediment communities. Moreover, populations with the potential ability to oxidize As(III) were also present in the water and sediment communities of Perpetual Spouter, namely an abundant Thermocrinis MAG. The potential ability to reduce oxidized nitrogen compounds (e.g., NO3 -, NO2 -, or N2O) was also prevalent among MAGs in both the sediment and planktonic communities of Perpetual Spouter. The capacity for Fe(II) oxidation was implicated in some Perpetual Spouter populations, but only those in very low abundance (< ~1%) of the total community.
In this contribution, we examine two adjacent, high-temperature pools in Norris Geyser Basin, Yellowstone National Park-Perpetual Spouter (Fig 1), which is an alkaline-chloride spring (pH ~7.5) and an acid-sulfate pool known informally as Red Bubbler (pH ~3). We combine, for the first time ever, near-surface geophysical measurements to examine the architecture, geometry, and porosity of fluid pathways; radiogenic isotopic measurements to establish the different aquifer's lithologies and the significance of gas-water-rock interaction in determining the water's distinctive chemistries; and U-and Th-decay series isotopes to determine the timescales of water-rock interaction. We then apply metagenomic sequencing and informatics analysis of microbial communities inhabiting these springs to examine the consequences of variation in these dichotomous hot spring characteristics on the composition and function of microbial communities that inhabit these springs. This coupling of methods provides a novel understanding of the interconnectedness of physical, chemical, and biological processes within this dynamic system and more broadly, the bimodal differentiation of hot spring processes and microbial ecosystems observed across Yellowstone and elsewhere.
Our geophysical imaging provides unique information on subsurface architecture critical to understanding fluid movement in Yellowstone hydrothermal systems (Pasquet et al., 2016;Bouligand et al., 2019;Smeltz et al., 2022;Ciraula et al., 2023b, a) and more specifically in this Two Pools system. Given the decrease in porosity (Fig. 2c) and relaxation time (i.e., decrease in fracture aperture/pore size, Fig. 2d) with depth observed by surface NMR, and the general decrease in resistivity with depth observed with 2D-DC resistivity, we conclude that the deeper water is very solute-rich, exists in small pores or fresh fractures, and that there is some mixing with meteoric water towards the surface in larger pores or weathered fractures. This is evidence of an overall decrease of LCT alteration with increasing depth and is consistent with visual and petrological investigations of borehole Y-9 (White et al., 1975).
The surface NMR data (Fig. 2c, d) allow us to make general interpretations about the specific surface area of the formation as a function of depth. While we cannot quantitatively interpret flow from surface NMR measurements, co-interpretation of the porosity and relaxation time (a proxy for surface area to volume ratio) is known to be a qualitative indicator of permeability as a formation flow property (e.g., Walsh, 2008). From the ground surface to ~8m-deep, the substrate has high water content/porosity (~0.15 m 3 /m 3 ) and relatively long T2* (~0.32 s), indicative of a porous matrix substrate. From ~8 m to ~18 m below the surface, there is a decrease in water content/porosity (~0.1 m 3 /m 3 ) and a corresponding decrease in T2* (~0.06 s) that would be consistent with a granular matrix material with smaller pores; i.e., less weathering than the material between 0-8 m, assuming identical parent material. Finally, below 18 m, we observe a further decrease in porosity (~0.05 m 3 /m 3 ) and a corresponding increase in T2* (>0.1 s) that are consistent with fractured material that has low water content, but relatively high bulk porosity due to open fracture apertures that are larger than the pores of granular material. This has implications for reactive transport because the transition from highly weathered porous material to a substrate that is more closely related to fractured rock is likely to also correspond to a change from lower permeability and higher surface area (slower flow/transport, higher geochemical reactivity) to higher permeability and lower surface area (faster flow/transport, lower geochemical reactivity). While the 1D surface NMR dataset only allows us to consider this in a general sense due to the lack of spatial information, the 2D Electrical Resistivity Tomography (ERT) image (Fig. 2b) adds the extra dimension that allows the interpretation that heterogeneous development of porous, permeable substrate in the subsurface leads to areas of low permeability/low weathering (i.e., >10 3 ohm m, Fig. 2b) between zones of higher permeability/higher weathering (i.e., <10 3 ohm m, Fig. 2b).
While near surface geophysical imaging does not explicitly reveal direct evidence for phase separation occurring above the depth of investigation (Fig. 2), our results provide critical information necessary for understanding the subsurface architecture beneath the Two Pools. Our near surface geophysical imaging shows: 1) distinct pathways for different fluids beneath each of the Two Pools, and 2) shallow subsurface flow paths coming down from the Ragged Hills. Importantly, this shallow water recharge helps to establish the distinct chemistry of these Two Pools. When the shallow oxidized water mixes with the dispersed vapor it too helps to oxidize and acidify RB's water. And as we discuss in detail below, it is this oxidative process that facilitates reactive transport and ultimately leads to the assembly of taxonomically and functionally divergent thermophilic microbial communities.
Albeit close together (~14 meters), and likely connected at some level below this study's geophysical depth of investigation, Perpetual Spouter and Red Bubbler have starkly different 208 Pb/ 206 Pb. These data imply that hydrothermal waters below the Norris Geyser Basin originate from a deep aquifer hosted in sedimentary rocks and that after phase separation, reactive transport processes completely shift the geochemical and isotopic compositions of Red Bubbler's waters.
Red Bubbler's 208 Pb/ 206 Pb, which is akin to the LCT, exemplifies the significant influence of shallow reactive transport. The low resistivity zone identified by the geophysical imaging immediately beneath Red Bubbler and extending southwest towards Ragged Hills is interpreted as weathered LCT that acts as a lateral conduit for meteoric water moving from the Ragged Hills to Red Bubbler. We suggest that transport through this weathered unit enables mixing of oxidized meteoric water with steam and trace H2S vapor, and its derivatives (S 0 and sulfide minerals), ascending from below Red Bubbler. The fluid is then oxidized in the near-surface to generate sulfuric acid (H2S + 2O2 => SO4 2-+ 2H + ), a process that is mediated, at least in part, by lithoautotrophic microorganisms (Mosser et al., 1973;Nordstrom et al., 2005Nordstrom et al., , 2009;;Colman et al., 2018) and is consistent with the observed ephemeral behavior of Red Bubbler and the enrichment of populations with the ability to aerobically catalyze H2S/S 0 oxidation (Fig. 6b). The acidity, in turn, promotes weathering of host rock, including leaching of acid-soluble elements (e.g., Pb, U, Fe, etc.) that then leads to high metal solute concentrations. In this case, the acidity of Red Bubbler's waters have leached enough Pb out of the LCT to completely overprint the 208 Pb/ 206 Pb of the Red Bubbler waters (Fig. 3b, Tables 1 and 3) that would otherwise be expected to be similar to Perpetual Spouter based on Cl -measurements. In other words, both springs are sourced by fluids from the deep hydrothermal reservoir, but the acidity that develops in Red Bubbler, driven by oxidants available in infiltrating surface-derived fluid, leads to leaching of elements that overprints the original isotopic signatures from the deep hydrothermal reservoir.
The isotopic compositions 87 Sr/ 86 Sr and 143 Nd/ 144 Nd of Red Bubbler's acidic waters are also akin to the values measured in LCT, however, it is Red Bubbler's high 208 Pb/ 206 Pb that uniquely provides conclusive evidence of shallow water-rock interaction between Red Bubbler's waters and LCT.
In stark contrast, Perpetual Spouter's alkaline-chloride waters have substantially lower 208 Pb/ 206 Pb (Fig. 3), are much less acidic, and its conduit is lined with silica sinter (>1 meter and likely deeper), armoring against shallow water-rock interactions. The low 208 Pb/ 206 Pb values in Perpetual Spouter's waters are most similar to the stratigraphically deeper Paleozoic and Mesozoic sedimentary rocks that underlie the LCT in NGB (Fig. 3b). That Perpetual Spouter's waters are not identical to the 208 Pb/ 206 Pb of the sedimentary rock lithologies suggests either shallow mixing with water that has undergone water-rock interaction with LCT, or the influence of dissolved, locally derived dust (n.b. all samples are filtered to 0.1 microns when sampling).
The very low Pb concentrations in waters from Perpetual Spouter (lower than Red Bubbler by more than an order of magnitude) make either option a possibility. However, we note that tritium measurements in waters from Perpetual Spouter point to small amounts (~7%) of shallow admixing with young, tritiated water (Pearson and Truesdell, 1978;Gardner et al., 2011).
Therefore, the low 208 Pb/ 206 Pb for Perpetual Spouter strongly suggests that the alkaline-chloride waters acquired a strong component of their Pb isotopic signal from water-rock interaction with deep sedimentary rock lithologies.
These conclusions are substantiated by our supplemental measurements of 87 Sr/ 86 Sr and 208 Pb/ 206 Pb in nearby, hydrothermal springs. The acid-sulfate springs, which are hosted in LCT, again have 208 Pb/ 206 Pb similar to LCT, consistent with significant shallow water-rock interaction.
Additionally, the alkaline-chloride Beryl Spring, considered to be an archetypical representation of the deep hydrothermal water, and the calcium-bicarbonate-sulfate waters in Mammoth Terraces and Boiling River, which have interacted extensively with limestones and anhydrites of the Mississippian Madison group (Kharaka et al., 1991;Fouke et al., 2000), have low 208 Pb/ 206 Pb, similar to the measured 208 Pb/ 206 Pb values for Paleozoic and Mesozoic limestones and sediments. 208 Pb/ 206 Pb is an important and diagnostic time-integrated radiogenic tracer of a rock's 232 Th/ 238 U ratio (Sims and Hart, 2006). Of all the chemical measurements made in Yellowstone's hydrothermal waters, it is the contrasting 208 Pb/ 206 Pb signatures preserved by these two adjacent hydrothermal features which uniquely require and identify two distinct lithological influences.
The elevated 208 Pb/ 206 Pb presented by Red Bubbler results from increased interaction with the LCT, which has a high 232 Th/ 238 U, while the lower 208 Pb/ 206 Pb seen in Perpetual Spouter's waters indicates a stronger influence of the more deeply buried Paleozoic and Mesozoic sedimentary rock units which have a much lower 232 Th/ 238 U characteristic of carbonates. We therefore posit that the fluids emanating from Perpetual Spouter are more indicative of the original hydrothermal reservoir hosted in the deep Paleozoic and Mesozoic sedimentary rocks. Whereas the fluids emanating from Red Bubbler, likely began with an isotopic signature like that of Perpetual Spouter, but then post phase separation shallow water-rock interaction with the volcanic LCT, which has a higher 232 Th/ 238 U, imparted a higher 208 Pb/ 206 Pb on Red Bubbler's waters. Amazingly, there have been surprisingly few measurements of radiogenic isotopes in Yellowstone's hydrothermal waters. As such, our isotopic results, particularly 208 Pb/ 206 Pb, demonstrate the importance of source lithology and reactive transport for establishing the geohydrobiological feedbacks that ultimately control the bimodal patterns in the Two Pools geochemistry and microbial diversity. Specifically, our 208 Pb/ 206 Pb measurements uniquely identify the "deep" hydrothermal aquifer's lithology as sedimentary and not volcanic in character, contrary to what is typically assumed about the lithology of the deep aquifer, and they quantitively demonstrate the extent to which acid-sulfate waters are chemically weathering the near-surface Lava Creek volcanic tuff and inheriting its chemical and isotopic signature. Furthermore, as shown below in our modeling of water rock-interaction timescales, this 208 Pb/ 206 Pb fingerprinting is both self-consistent and critical for our reactive transport modeling.
While reaction-path models can predict the chemical reactions during water-rock interaction, knowledge of the timescales of water-rock reaction and fluid transport are poorly known. Constraining these timescales is critical to understanding the dynamics of these systems.
The U-and Th-decay series systematics provide useful Quaternary chronometers (Sims et al., 2021) and can help to determine the duration of water-rock reactions and fluid transport on timescales of tens to hunreds of thousands of years (Kadko and Moore, 1988;Kadko et al., 2007).
Within the U-and Th-decay series (Fig. 4) is a quartet of Ra isotopes with large differences in their half-lives: 226 Ra (t1/2 = 1600 years), 228 Ra (t1/2 = 5.75 years), 224 Ra (t1/2 = 3.63 days) and 223 Ra (t1/2 = 11.43 days). These vastly different half-lives, coupled with strong partitioning of Ra into fluid during water-rock reactions, provide a novel chronometer to quantify the timescales of gas-water-rock interaction in Yellowstone's convective hydrothermal system. Models utilizing Ra isotopes to establish timescales of water-rock interaction have been successfully applied to terrestrial groundwater systems (Krishnaswami et al., 1982), deep ocean hydrothermal systems (Kadko and Moore, 1988), Icelandic hydrothermal waters (Kadko et al., 2007) and Yellowstone's hydrothermal waters (Clark and Turekian, 1990;Sturchio et al., 1993).
Remarkably, and of historical significance, in 1906, only ten years after Ra was first discovered and isolated by the Curries at the turn of the twentieth century, scientists for the USGS measured Ra in Yellowstone's thermal waters to determine the "hydrography of the Yellowstone hydrothermal system" (Schlundt and Moore, 1909;Schlundt and Breckenridge, 1938). In this study, we use measurements of 228 Ra and 226 Ra to determine the timescales of water-rock interaction, and measurements of both long-lived 238 U, 232 Th and 230 Th and short-lived 224 Ra, 223 Ra and 222 Rn to constrain model parameters and to test assumptions implicit to the application of the ( 228 Ra/ 226 Ra) chronometer.
Conceptually, Yellowstone's convective hydrologic cycle can be broken into four distinct stages.
Stage 1: Meteoric water enters the ground, presumably as recharge from the nearby Madison Plateau and the Absaroka and Gallatin Mountain Ranges (Rye andTruesdell, 1993, 2007;Kharaka et al., 2002;W Payton Gardner et al., 2010;W. Payton Gardner et al., 2010). Shallow groundwater recharge can serve as a diluent, or even as an oxidant, to deep thermal waters; however, meteoric water contains minimal concentrations of both Th and Ra (generally << 1 ppb). Thus, the ( 228 Ra/ 226 Ra) chronometer does not address, nor is it influenced by, the timescales of Stage 1.
Stage 2: The downward percolating groundwater infiltrates Yellowstone's magmaticallyheated rock; these waters chemically react with this hot rock and are infused with magmatic gases rising from below. It is here in the 'deep homogenous hydrothermal reservoir' that hightemperature and high-pressure conditions facilitate fluid-gas-rock interactions. 208 Pb/ 206 Pb isotopes indicate that outside the caldera this deep reservoir is made up of Paleozoic and Mesozoic sediments and carbonates. Radium isotopes are transferred from aquifer rocks (made up of either primary or secondary minerals or both) into solution, through both direct rock dissolution and alpha recoil. Stage 2 is when the clock starts for the (foot_0 Ra/ 226 Ra) chronometer.
Stage 3: Upon heating in Stage 2, the chemically-infused near-supercritical water rises relatively quickly to the surface on account of temperature and pressure gradients. Clark and Turekian, 1990 assumed, albeit without direct evidence, that the water rises instantaneously to the surface without time for decay, thus preserving the Ra isotope signal acquired during the time of deep high-temperature water-rock interaction. However, given the short half-life of 228 Ra of 5.77 years, the possibility for some decay of 228 Ra exists and needs to be considered and/or ruled out by other constraints.
Stage 4: The rising water undergoes decompression boiling and separates into an alkaline-chloride fluid phase and a steam-dominated vapor phase. The phase-separated fluids then subsequently mix to various degrees with oxidizing shallow meteoric water. In the case of the Two Pools -these phase-separated waters emerge as the alkaline-chloride Perpetual Spouter and the acid-sulfate Red Bubbler. Importantly, these different compositions cause different degrees of water-rock interaction. Acidic waters corrosively weather the shallow LCT, resetting the chemical and isotopic composition of the water and the Ra chronometer. Alkaline-chloride waters precipitate silica sinter and armor themselves against water-rock interaction, thus preserving a deeper signal.
The differences inferred for fluid sources between Perpetual Spouter and Red Bubbler suggest that timescales of water-rock interaction differ considerably between these two hot springs. Essentially, the isotope data indicate that the water in Perpetual Spouter and Red Bubbler represent two different stages and hydrological regimes in the Yellowstone hydrothermal system (Fig. 3). Perpetual Spouter's alkaline-chloride water represents the deep hydrothermal reservoir that is minimally perturbed by water-rock interaction en route to the surface (Nordstrom et al., 2009). Thus, for Perpetual Spouter's alkaline-chloride boiled waters, we interpret our Ra measurements as providing information on the timescales of water-rock interaction in the "deep homogeneous hydrothermal reservoir" (i.e., Stage 2 with perhaps some chronometer provides information on the timescales of recent, shallow water-rock interaction occurring in Stage 4 after phase separation.
Radium nuclides can enter Yellowstone's hydrothermal waters by any combination of the following four processes: (1) dissolution of aquifer solids, (2) in situ radioactive decay of the dissolved parent nuclides, (3) direct recoil across the solid-liquid boundary as a result of production by radioactive decay in the solid, and (4) desorption from solid surfaces. Radium nuclides will leave the system by either one of two processes: 1) decay; or 2) adsorption onto a surface. An important assumption implicit within ( 228 Ra/ 226 Ra) chronometer calculations is the assumption that U-and Th-decay series isotopes are in equilibrium in the unaltered aquifer source rock. In our application here, this assumption is initially quite reasonable given that the half-lives of the relevant progeny isotopes (Fig. 4) are very short compared to the ages of both the sedimentary rocks we are positing to be the host of the deep aquifer (the underlying deep sedimentary rocks are late Paleozoic and Mesozoic) and the LCT (640 ka) that the acid-sulfate waters are interacting with. That said, progressive dissolution will disrupt this disequilibrium state in the rock, strip out U preferentially, and ultimately favor the enrichment of shorter-lived nuclides in the fluids, including 228 Ra.
The non-steady state solution describing the activity of a Ra daughter in solution is indicated by 𝐴 𝐷 𝐿 , and can be calculated from the expression:
Eq. 1
Where activity is defined as n, representing the isotope's decay constant, , multiplied by the number of atoms, n, of that isotope, and has units of decays per unit time for a given quantity.
Ignoring the effects of adsorption/desorption, the total activity for a Ra isotope in solution can be cast as:
Where the production terms (P) for Ra isotopes (atoms • time -1 • kg of water -1 ) going into Yellowstone's hydrothermal fluid are the algebraic sum of: P1) in situ decay of parental Th isotopes in solution; P2) input of Ra from rock alteration and dissolution by corrosive, hightemperature and low pH hydrothermal fluids; and P3) recoil of Ra isotopes into the fluid from decay of their Th parents in the host rock. Radium loss is mainly through decay and, to a very limited extent, adsorption depending on the composition of the phase-separated fluids. Our formalism, at some peril, ignores the effects of adsorption/desorption. Ignoring the effects of adsorption/desorption is reasonable for Red Bubbler's acidic waters, but potentially problematic for alkaline-chloride waters like Perpetual Spouter, particularly if the process is isotope selective.
Very low Th/Ra in both Red Bubbler and Perpetual Spouter equate to negligible Ra ingrowth from the decay of aqueous parental Th, even in Red Bubbler's acidic waters where Th is slightly soluble (Tables 1 and 3). Thus, the ingrowth production term, P1, can be omitted, and equation 2 reduces to:
P2, the input of the Ra isotopes (𝐴 𝐷 𝐿 ) as a result of high-temperature water-rock chemical reactions/dissolution is given by:
where 𝐴 𝑃 𝑅 = activity of parent nuclide in host rock (dpm • g -1 ) and W = water/rock ratio (g • g -1 )
which is defined as the number of grams of fluid required to dissolve one gram of rock. The 10 -3 factor is to account for the rock's activity being given in (dpm • g -1 ), whereas P2 is in units of (atoms • time -1 • kg of water -1 ). Because is in the denominator, nuclide input from rock dissolution will be most important for the longer-lived nuclides, thereby producing low ( 228 Ra/ 226 Ra) relative to the equilibrium rock ratio (Fig. 4 inset). Finally, when W approaches infinity P2 becomes less significant.
P3, the recoil rate (R) of a nuclide, is given by the expression:
Eq. 5
Recoil is most important for short-lived nuclides and in isolation this process will produce high ( 228 Ra/ 226 Ra) (Fig. 4 inset). Of all the production terms, recoil input is one of the most uncertain and difficult to quantify. Most studies use the 222 Rn activity to approximate the recoil supply rate of all alpha decay products (Krishnaswami et al., 1982;Kadko and Moore, 1988;Clark and Turekian, 1990;Kadko and Butterfield, 1998) because 222 Rn is chemically inert and thus not affected by adsorption or other secondary reactions. Furthermore, 222 Rn has a very short half-life (t1/2 = 3.85 d) and will quickly reach a steady state in solution (P𝜆 = 𝜆𝐴). However, application of 222 Rn as a recoil proxy for Ra isotopes requires the use of a recoil efficiency coefficient of 222 Rn relative to Ra isotopes (expressed here as ). Measurements of 224 Ra (t1/2 = 3.64 days) can provide another estimate of the Ra recoil from the aquifer rock into the reacting fluid. 224 Ra has a half-life similar to that of 222 Rn and is produced from the alpha decay of 228 Th (t1/2 = 1.9 years), which is in turn produced by 228 Ra decay. Since recoil favors isotopes with short half-lives and assuming, reasonably, that very little 228 Th is in solution, the 224 Ra in the circulating fluid is being produced primarily by recoil from rock and is thus another good proxy of recoil. Summing these two production terms and expressing equation 3 in terms of ( 228 Ra/ 226 Ra) activity in the fluid provides the final expression for the Ra chronometer used in this study: Eq. 6
When solving equation 6 as a function of time, and under typical conditions where both recoil and dissolution are contributing to the fluid's composition, the functional form of this equation starts with ( 228 Ra/ 226 Ra) rising rapidly over time to reach an apogee, then quickly decreases over time (Fig. 4 inset). The maximum ( 228 Ra/ 226 Ra) and the functional form of its decrease over time are controlled by: 1) the relative half-lives of 228 Ra and 226 Ra; 2) ( 232 Th)/( 238 U) in the aquifer source rock; 3) the 228 Ra and 226 Ra recoil rates, R; and, 4) the extent of rock dissolution, which is parametrized as the chemical water/rock ratio, W (g • g -1 ).
Red Bubbler and Perpetual Spouter waters have strikingly different chemistries and isotopic signatures. A critical feature of this chemical distinction is the vastly different ( 228 Ra/ 226 Ra) between the two pools. Perpetual Spouter has an average ( 228 Ra/ 226 Ra) of 5.97 ± 3.09, and Red Bubbler has an average ( 228 Ra/ 226 Ra) of 45 ± 6 (Table 2). We argue based upon radiogenic isotopes and geophysical observations that post-phase separation reactive transport processes impart this bimodal geochemical and isotopic difference, with Perpetual Spouter and Red Bubbler providing information on two separate regimes beneath the NGB hydrothermal system. Perpetual Spouter's alkaline-chloride waters are dominated by fluids representing a deep hydrothermal reservoir where water-rock interaction was occurring with Paleozoic and Mesozoic sedimentary rocks, whereas Red Bubbler's acidic waters are dominated by fluids suggestive of shallow water-rock interaction with the LCT. Thus, in the following we consider Perpetual Spouter and Red Bubbler separately when applying the Ra chronometer.
In the following Ra chronometer modeling, we explore reasonable values for the determinant variables: 1) 232 Th/ 238 U of the aquifer source rock, 2) recoil rate (R), and 3) water/rock ratio (W). Justifications for our choice of these variables for each of the two systems, as well as the sensitivity of our results to uncertainties in these variables, is detailed in Supplementary Material B.
Geophysical imaging, combined with geochemical and isotopic data, indicate that Red Bubbler waters (pH ~ 3) represent the mixing of deep hydrothermal waters, vapor phase gases, and shallow oxidizing groundwater flowing in from the adjacent Ragged Hills. Movement of the resulting acidic fluids along their reaction pathway through the LCT significantly weathers the LCT. Red Bubbler's high metal solute concentrations (including U, Ra, and Ba) and its radiogenic Sr and Pb isotopes similar to LCT are evidence of high solute input from shallow water-rock reaction. In Red Bubbler's waters, the Ra chronometer provides information on the timescales of shallow water-rock interaction occurring after phase separation.
Because the Pb, Nd, Sr and Fe isotopic compositions of Red Bubbler's waters are indistinguishable from those of LCT, and because this system is acidic, we model two endmember scenarios: water-rock interaction with unaltered LCT; and, water-rock interaction with altered LCT, as measured in borehole Y-9 (Sturchio et al., 1993). Because of progressive alteration during water-rock interaction, it is likely that the ( 238 U/ 232 Th) of the aquifer source rock is intermediate to these two endmembers. We also note that in acidic systems, the rock and mineral surfaces are likely to be protonated and thus repellant toward Ra cations; as such, it is reasonable to ignore the effects of adsorption/desorption when calculating the mean timescales of water-rock interaction from Red Bubbler's ( 228 Ra/ 226 Ra). Finally, in our modeling, we assume that the non-reactive transport time from the deep reservoir (i.e., once chemical reaction between fluid and rock has ceased) is negligible relative to the half-life of 228 Ra. This assumption is reasonable given the near-surface water-rock interactions that are resetting the ( 228 Ra/ 226 Ra) chronometer.
Red Bubbler's ( 228 Ra/ 226 Ra) ranges from 37.9 -50.3 (Fig. 5, Table 2), with an average of 45 ± 6. Assuming that U-and Th-decay series progeny are in equilibrium in the LCT source rock, the ( 228 Ra/ 226 Ra) production ratio of LCT (1.7 ± 0.15), is inferred from our measurements of ( 232 Th/ 238 U). This much lower theoretical rock ( 228 Ra/ 226 Ra) requires that water-rock interaction processes, namely through recoil, preferentially enrich 228 Ra in the fluid relative to 226 Ra. Thus, Red Bubbler's very high ( 228 Ra/ 226 Ra) requires recent and short-lived water-rock interaction with significant 228 Ra input from recoil. This temporal constraint holds for both unaltered and altered LCT.
In these calculations, recoil rate, R, and water/rock ratios, W, are inversely coupled in producing a given ( 228 Ra/ 226 Ra). This inverse relationship exists because W is in the denominator of the rock dissolution production term (Eq. 4), so when W values are high, the production term, P2, becomes increasingly less relevant. There are unique combinations of W and R values capable of producing a given ( 228 Ra/ 226 Ra). In Red Bubbler's case we use the maximum ( 228 Ra/ 226 Ra) of 50.3 for our modeling. The higher ( 232 Th/ 238 U) of altered LCT requires lower W and R values to produce the high ( 228 Ra/ 226 Ra) of Red Bubbler's waters.
For all scenarios and all samples, calculated mean water-rock residence times for Red Bubbler are less than 100 years (Fig. 5a). The different aquifer rock lithologies examined have vastly different ( 232 Th/ 238 U) (Table 3), which has a significant effect on required recoil rates (Supplementary Material B), water/rock ratios, the form of the model curves, and calculated water-rock residence times (Fig. 5a). Average unaltered LCT has a ( 232 Th/ 238 U) of 1.67 ± 0.15 and calculated water-rock interaction times range from 15 years to 44 years. Altered LCT (based on borehole Y-12) has a much higher ( 232 Th/ 238 U) of 4.8 because of the much greater solubility of hexavalent U; this higher ( 232 Th/ 238 U) shallows the form of the model curve resulting in slightly longer calculated mean residence times ranging from ~16 years up to 90 years.
Intermediate or progressive alteration (and U removal) between unaltered LCT to altered LCT (assuming Y-12 as an endmember) will result in model curves and model ages intermediary between these two endmembers.
Decisively, across a wide range of parameter space, and regardless of whether the aquifer source rock is unaltered or altered LCT (both are consistent with Red Bubbler's Pb isotopes), Red Bubbler's exceptionally high ( 228 Ra/ 226 Ra) requires recent and short water-rock interaction times (<100 years) (Fig. 5a). In the context of Red Bubbler's hydrological regime, this timescale represents the average length of time, post phase separation, that Red Bubbler's waters were sufficiently acidified to start chemically weathering the LCT. We note that this timescale is an average as it integrates across the range of porosities and timescales experienced by the collective emergent waters. However, we also note that the extent of weathering must be significant enough that the isotopic compositions are buffered, as the measured values were constant over the four-year period of our measurements.
Alkaline-chloride waters, such as those emanating from Perpetual Spouter (pH ~ 7.5), precipitate silica sinter along their conduits, which armors against shallow water-rock interaction. Beyond the effects of volatile loss during boiling, alkaline-chloride waters are minimally perturbed en route to the surface and are thus posited to represent the deep, homogeneous, hydrothermal reservoir (Nordstrom et al., 2009). Thus, for alkaline-chloride waters, the Ra chronometer provides perspective on the timescales of deep water-rock interaction before phase separation. Regardless, and importantly, the 208 Pb/ 206 Pb of Perpetual Spouter's waters reflect a geochemical signal that suggests mixing with components from a deep (below LCT) sedimentary rock-hosted aquifer. These sedimentary rocks have low U and Th concentrations and very low ( 232 Th/ 238 U), in contrast to the local volcanic rocks, in particular the LCT, that have higher U and Th concentrations and much higher ( 232 Th/ 238 U). Therefore, using the U and Th abundances of the sedimentary rocks makes it possible to model the timescales of Perpetual Spouter's waterrock interaction by using the Ra chronometer. This is in stark contrast to the earlier modeling of Clark and Turekian (1990), who, in the absence of radiogenic isotopic data, assumed that the cations, including Ra, in all Yellowstone hydrothermal fluids were derived from average volcanic rock. There are several possible local sedimentary rock units, or even a combination of them, that can serve as the deep aquifer represented by Perpetual Spouter's waters. Their [U],
[Th] and ( 232 Th/ 238 U) are tabulated in Table 3. Because of our lack of knowledge of the specific sedimentary rocks that are hosting deep aquifer reservoir(s), and several of them could be involved, we calculated an average Paleozoic-Mesozoic sedimentary rock composite for modeling purposes using the average [U], [Th] and ( 232 Th/ 238 U) values from Sturchio et al., 1993. Perpetual Spouter's ( 228 Ra/ 226 Ra) ranges from 2.5 -8.41, with an average of 5.97 ± 3.09 (Fig. 5b, Table 2). This range encompasses the ( 228 Ra/ 226 Ra) values measured by Clark and Turekian, 1990 in three other NGB alkaline-chloride hot spring waters, all reported as pH = 7 (Hydrophane = 2.50 (± 0.49); Medusa = 3.89 (± 0.29); and Opalescent Springs = 4.43 (± 0.70); average of 3.61 (± 1.00)).
For alkaline-chloride systems, Clark and Turekian, 1990 assumed that, in the deep system, 228 Ra and 226 Ra are in equilibrium between the solution and the adsorbed phase and that these waters then ascend rapidly to the surface unperturbed by subsequent processes. While it is likely that the different Ra daughters reached equilibrium in the deep sedimentary rock aquifer, changing conditions during fluid ascent, including phase separation and mixing with shallow groundwater, disrupt this equilibrium steady-state. Thus, it is likely that Ra is being deposited/adsorbed (Sturchio et al., 1993) in the near-surface environment and that such effects need to be considered in calculated water-rock interaction timescales. Fortunately, our measurements of ( 224 Ra/ 222 Rn) in Perpetual Spouter's waters provide a quantitative measure of Ra adsorption/precipitation. Both 224 Ra and 222 Rn have short and similar half-lives; therefore, both nuclides will be recoiled at a rate similar to their production ratio, as inferred from the ( 232 Th/ 238 U) of the source aquifer. Perpetual Spouter's ( 224 Ra/ 222 Rn) is about an order of magnitude lower than the ( 232 Th/ 238 U) of regional sedimentary or volcanic rock lithologies, including composites (Table 3). This apparent depletion of ( 224 Ra/ 222 Rn) relative to its inferred production ratio suggests Ra loss, perhaps in part due to the incorporation of Ra into silica sinter during its deposition. The effect that adsorption and precipitation processes will have on calculated mean water-rock interaction ages depends entirely on whether these shallow processes are altering the ( 228 Ra/ 226 Ra) established in the deep reservoir during stage 2. If 228 Ra and 226 Ra are being precipitated and adsorbed in their relative isotopic abundances, the ( 228 Ra/ 226 Ra) will remain unchanged and calculated residence times will remain unperturbed.
However, if 228 Ra equilibrates faster with the mobile pool than does 226 Ra (Sturchio et al., 1993), the fluid's ( 228 Ra/ 226 Ra) will increase, and the calculated mean water-rock interaction times will appear to be shorter. Consequently, for Perpetual Spouter, the calculated water-rock interaction times should be considered minimum values.
For our calculations of mean water-rock interaction timescales for Perpetual Spouter, we examine: three lithologies -average sediment, unaltered LCT, and altered LCT (not shown); two water/rock ratios, W = 250 and W = 125; and, two recoil rates, R = 500 dpm/l and 1,000 dpm/l (See Supplementary Material B for discussion of parameter choices and sensitivity analyses).
Calculated mean water-rock residence times for Perpetual Spouter are greater than 100 years for all scenarios and all samples (Fig. 5b). The various aquifer rock lithologies examined have vastly different ( 232 Th/ 238 U) (Table 3), which has the most significant effect on determining the fluid ( 228 Ra/ 226 Ra) and resulting calculated water-rock residence times (Supplementary Material B).
Changing R and W across the range chosen has a smaller effect on calculated residence times (Fig. 5b). Average Mesozoic sediment has the lowest ( 232 Th/ 238 U) of 0.695, and the shortest calculated mean residence times for Perpetual Spouter waters, ranging from ~125 years to ~530 years. Average unaltered LCT has a considerably higher ( 232 Th/ 238 U) of 1.67 (± 0.15), requiring longer residence times of ~330 years to roughly 1,500 years. Altered LCT (Y-12) has an even higher ( 232 Th/ 238 U) of 4.8 (Sturchio et al., 1993) and requires even longer calculated mean residence times of ~1,500 years to ~1,800 years for Perpetual Spouter. Note that in this later case the age is indeterminant for sample Perpetual Spouter #140917, as the fluid ( 228 Ra/ 226 Ra) is below the calculated steady-state ( 228 Ra/ 226 Ra) productivity ratio as inferred from the altered LCT ( 232 Th/ 238 U) of 4.8. Because the 208 Pb/ 206 Pb isotope ratios suggest that Perpetual Spouter waters are best modeled as mixtures of waters that have come from the deeper underlying Paleozoic to Mesozoic sedimentary rocks (Fig. 5b) with waters that have interacted with LCT (or other Yellowstone volcanic units), Perpetual Spouter's water-rock interaction timescales are also likely intermediate between these two endmembers.
These calculations ignore the effects of adsorption/desorption or coprecipitation on ( 228 Ra/ 226 Ra). If, during Ra precipitation or adsorption, 228 Ra equilibrates faster with the mobile pool than does 226 Ra (Sturchio et al., 1993), the fluid's ( 228 Ra/ 226 Ra) will increase, and the calculated mean water-rock interaction times will appear to be shorter and would be considered minimum values/ages. However, if the time of ascent in Phase 3 is long relative to the half-life of 228 Ra, then the fluid's ( 228 Ra/ 226 Ra) will decrease, and the calculated mean water-rock interaction times would be considered maximum values/ages. Nonetheless, Perpetual Spouter's low ( 228 Ra/ 226 Ra) indicates a mean water-rock interaction time of hundreds to thousands of years (Fig. 5b), consistent with Perpetual Spouter waters coming from Yellowstone's deep hydrothermal system where the waters were interacting with the underlying Paleozoic and Mesozoic sediments (Fig. 3b). These time constraints for Perpetual Spouter are consistent with, but much more definitive than, tritium constraints that only define Perpetual Spouter waters to be dominantly (>93%) tritium-free (Pearson and Truesdell, 1978;Gardner et al., 2011) and thus predominantly pre-1945 (i.e., pre-bomb) (Cauquoin et al., 2016). Ultimately, it is not surprising that Yellowstone's deep hydrothermal water has a convective timescale of several hundreds of years. However, given the high outflux of alkaline-chloride water, and the high recharge into the Yellowstone hydrothermal system from the Madison Plateau and the Absaroka and Gallatin Mountain Ranges, it seems unlikely that these waters' reaction times are thousands to tens of thousands of years old (Rye andTruesdell, 1993, 2007;Kharaka et al., 2002;W Payton Gardner et al., 2010;W. Payton Gardner et al., 2010).
In summary, this study is the first-time anyone has determined the timescales of waterrock interaction, focusing on phase-separated waters. The observed high 228 Ra/ 226 Ra in Red Bubbler's acid-sulfate waters requires recent and short, shallow water-rock interaction timescales (10s of years) coming from a reservoir with high ( 232 Th/ 238 U), whereas Perpetual Spouter's neutral chloride waters, which are coming from the deep hydrothermal reservoir with low ( 232 Th/ 238 U), have much older ages and longer water-rock interaction time scales (100's-1,000's of years). Establishing the timescales of water-rock interaction is an important and hard-todetermine parameter in natural systems, yet it is a critical component of the Domköhler number for reactive transport. Demonstrating the viability of the 228 Ra/ 226 Ra chronometer used here is important for understanding phase separation in Yellowstone's hydrothermal system and has applicability across a range of groundwater and hydrothermal studies.
Phase separation and its influence on reactive transport and the chemistry of springs has a profound effect on the taxonomic and functional composition of their resident microbial communities, as revealed here for Red Bubbler and Perpetual Spouter (Fig. 6), which are representative of broader Yellowstone-wide variation in microbial communities (Fig. 7). The planktonic and sediment communities from Red Bubbler were comprised entirely of Archaea, consistent with previous studies indicating that Archaea dominate the most acidic and highest temperature springs in Yellowstone (Inskeep et al., 2013;Colman et al., 2018) and in other globally distributed continental geothermal systems for instance in New Zealand, Iceland, China, and Japan (Hou et al., 2013;Ward et al., 2017;Colman et al., 2023). In contrast, dominant planktonic and sediment populations in Perpetual Spouter comprised both Archaea and Bacteria. This set of observations is attributed to the combination of high temperature and acidity that together impose chronic stress on microbial cells, a condition that has been suggested to select for Archaea over Bacteria during the assembly of those communities (Valentine, 2007;Colman et al., 2018). The combination of high temperature and acidity is also likely responsible for the lower taxonomic diversity associated with Red Bubbler, as has been suggested for other acidic hot springs in Yellowstone (Inskeep et al., 2013;Colman et al., 2018;Fernandes-Martins et al., 2023) and other globally distributed acidic springs (Power et al., 2018;Colman et al., 2023).
Previous studies suggest that the acidification of hot springs in Yellowstone is driven by a series of geobiological feedbacks that involve both abiotic and biotic components (Mosser et al., 1973;Nordstrom et al., 2005Nordstrom et al., , 2009;;Colman et al., 2018). H2S, which is enriched in vapor phase gases that source acidic springs, can be oxidized abiotically at high temperature by oxygen that is made available by mixing with recently infiltrated meteoric water (Brock et al., 1972;White et al., 1988;Colman et al., 2018). Abiotic oxidation of H2S generates S2O3 2-(Eq. 5), which is unstable in waters with pH <6.0 (Xu and Schoonen, 1995) and rapidly disproportionates to form sulfite (HSO3 -) and S 0 (Eq. 6) (Nordstrom et al., 2005), the latter of which accumulates due to its low solubility and slow reactivity with water at temperatures of <100°C (Nordstrom et al., 2005).
2HS -+ 2O2 → S2O3 2-+ H2O
Eq. 5 H + + S2O3 2-→ S 0 + HSO3 - Eq. 6 HSO3 -+ 1/2O2 → SO4 2-+ H + Eq. 7 S 0 + 3/2O2 + H2O → SO4 2-+ 2H + Eq. 8 HSO3 -is also unstable at high temperature in the presence of oxygenated waters with a pH < 4.0 (Colman et al., 2020) and oxidizes abiotically to form SO4 2-via Eq. 7 (Nordstrom et al., 2005).
Importantly, these abiotic reactions (Eqs. 5-7) do not generate net acidity, since a mol of H + is consumed in Eq. 6 and a mol of H + is generated in Eq. 7. Rather, it is the O2-dependent oxidation of S 0 (Eq. 8) that is responsible for the generation of acidity and microbial activity mediates this process at temperatures of <100°C (Mosser et al., 1973). The functional capabilities of dominant populations in Red Bubbler, which importantly allow for O2-dependent oxidation of S 0 (Fig. 6b), is consistent with this model for hot spring acidification. As spring acidification progresses, the stabilities of intermediate H2S oxidation products (S2O3 2-and HSO3 -) continue to decrease, thereby increasing the amount of S 0 available for additional acid generation. These geobiological feedbacks are responsible, at least in part, for the unique chemistry and biology associated with Red Bubbler and that differentiate them from Perpetual Spouter, and more broadly, the bimodal differentiation of hot spring microbial ecosystems observed across Yellowstone and elsewhere (Fig. 7) (Inskeep et al., 2013;Hou et al., 2013;Power et al., 2018;Moreras-Marti et al., 2021;Colman et al., 2023). The enrichment for aerobic S 0 oxidation metabolic potential in metagenomes from a variety of acidic high temperature hot springs in Yellowstone (Colman et al., 2019) indicates that this geobiological feedback is responsible for the bimodal distribution of spring pH across Yellowstone, and one that is established upon the foundation of phase separation and reactive transport processes described above.
A comparison of the taxonomic and functional composition of planktonic and sediment communities in Red Bubbler when compared to those of Perpetual Spouter points to another striking consequence of phase separation and reactive transport on the assembly of microbial communities in hot springs. Whereas communities from Red Bubbler exhibited similar functional compositions, but distinct taxonomic compositions, those from Perpetual Spouter were far more differentiated at a functional level (Fig. 6c). Previous work has shown that plankton and sediment communities in hot springs are often differentiated at a taxonomic level, with the presence of solid phase minerals in sediments and greater access to infused atmospheric oxygen in waters suggested to be the predominant reasons for this differentiation (Colman et al., 2016;Fernandes-Martins et al., 2021). From a deterministic perspective on community assembly, there are only two logical mechanisms that could lead to this observed differentiation between planktonic and sediment communities in hot springs: 1) adaptations that allow organisms to better compete for nutrients with differential availability in the water column versus sediments of springs, as described above; or, 2) sourcing of planktonic cells from the subsurface at a rate that exceeds the growth rate of cells adapted to conditions in the hot spring waters. We believe that both mechanisms are at play, with mechanism-one being more important in differentiating communities inhabiting circumneutral to alkaline springs and mechanism-two being more important in differentiating communities inhabiting vapor phase-sourced springs. As mentioned above, Perpetual Spouter is likely sourced by water from the deep hydrothermal reservoir, with transport times estimated to be on the order of hundreds of years. The long residence time of these waters in the deep reservoir allow for extensive interaction with minerals in the bedrock that hosts this aquifer, which drives the system to anoxia. Those fluids are often also enriched in As(III), leached from rhyolite bedrock that hosts the hydrothermal aquifer (Stauffer and Thompson, 1984;McCleskey et al., 2022a). Furthermore, the conduits feeding deep hydrothermal water to circumneutral to alkaline springs like Perpetual Spouter are armored by precipitated silica (Vitale et al., 2008;Gibson and Hinman, 2013), a feature that would be expected to prevent extensive infusion of oxidized meteoric fluids that have recently infiltrated the surface. In effect, this would limit the chemical reactions that can support microbial metabolism to those that are less dependent on O2 from atmospheric influx due to its low solubility of O2 at high temperatures (Amend and Shock, 2001). The dominant organism within the Perpetual Spouter community was affiliated with the currently uncultivated Ca. Calditenuis aerorheumensis (Caldarchaeales), an organism that is inferred based on metagenomic data to be facultatively aerobic and facultatively autotrophic. Metabolic inference suggests that these organisms may be able to fuel autotrophic energy metabolism via oxidation of carbon monoxide or sulfide coupled potentially to As(V) or O2 reduction. The apparent ability of the dominant Perpetual Spouter population in waters and sediments to switch between aerobic (O2 reduction)
and anaerobic (e.g., As(V) reduction) respiratory strategies may be a consequence of the low and possibly variable levels of O2 in high temperature waters of circumneutral to alkaline springs.
Importantly, the amount of total arsenic as arsenite (As(III)) in circumneutral to alkaline springs tends to be >50%, although it can be as high as 100% (Stauffer and Thompson, 1984;McCleskey et al., 2005McCleskey et al., , 2014McCleskey et al., , 2022a,b),b). The total amount of arsenic in Perpetual Spouter was nearly a factor of 3 greater than that of Red Bubbler, and ~50% of this was as As(V) (Supplemental Materials Table S1). It is possible that conversion of As(III) to As(V) is catalyzed via co-inhabiting Thermocrinis populations whose MAGs encode As(III) oxidizing capability.
The apparent abilities of many other populations in Perpetual Spouter to utilize arsenic compounds as reductants (e.g., As(III); Thermocrinis) or oxidants (e.g., As(V); Ca. Calditenuis aerorheumensis) may also reflect the adaptations of microbial populations to survive in hot springs sourced by the deep hydrothermal aquifer. The lack of such an ability in the Red Bubbler populations may conversely reflect differences in fluid sourcing and reactive transport between the Two Pools, which exerts primary influence on their geochemistry and thus the availability of more thermodynamically preferable substrates (e.g., iron and sulfur compounds) (Shock et al., 2010) to fuel microbial metabolism. Cultivation of the Caldarchaeales populations and additional investigation of arsenic metabolism in this and other lineages would allow confirmation of their ability to carry out metabolisms predicted based on analyses of their MAGs.
In contrast to Perpetual Spouter, the fluids that source Red Bubbler are channeled through a diffuse transport network, as shown by near surface geophysical measurements reported herein.
This diffuse flow allows for mixing of reduced fluids (gases) with recently infiltrated oxidized meteoric waters, which would be expected to support microbial activity if other conditions (e.g., temperature constraints) are met (Shock and Holland, 2007;Shock et al., 2010). Despite the ephemeral nature of Red Bubbler, as described above, it was flowing during the time that samples for molecular analyses were collected. As such, the observed taxonomic disparity, coupled with the limited residence time associated with Red Bubbler waters during the time of sample collection, point to the dominant MAG in the planktonic phase (Sulfolobales 'Acd1') as possibly being sourced from the subsurface of this spring, rather than being specifically adapted to localized conditions in the water column. Consistently, a recent analysis of subsurface waters in another spring within the NGB, Cinder Pool, demonstrated that the Sulfolobales Acd1 group increased in abundance with waters at depth (up to 21 m) relative to surface waters (Colman et al., 2022). In support of this argument, it is unlikely that stark chemical differences (e.g., acidification of meteoric water by nearly four orders of magnitude) can develop between the sediment and water column on such a short time scales of fluid residency in Red Bubbler.
Nonetheless, and as stated above, metabolic reconstruction of the dominant archaeal populations in both the planktonic and sediment communities in Red Bubbler indicate that they are supported largely by oxidation of reduced sulfur compounds suggesting that they share the same metabolic niche. We envision that abiotic (or potentially biotic) oxidation of sulfide in the near subsurface of Red Bubbler drives the deposition of S 0 along the diffuse transport network that sources this spring. The Sulfolobales 'Acd1' population can oxidize this S 0 in the subsurface to generate acidic waters that emanate from the source of this spring; Sulfolobales 'Acd1' sequences recovered in the plankton phase of Red Bubbler are thus attributed to sloughed subsurface cells.
We suggest that this is due to physiological adaptations (e.g., temperature) that delineate the distribution of taxonomically distinct and functionally equivalent archaeal taxa to different ecological compartments (subsurface vs. surface). Additional physiological studies, including those with cultured Sulfolobales 'Acd1' cells that dominant waters versus Sulfolobales 'QEFN01-2' cells that dominates sediments, will help to assess this intriguing hypothesis.
Finally, the acidification of near surface ground waters as they are transported through the diffuse transport network that sources Red Bubbler is responsible for enrichment of those waters with iron (6.7 to 7.7 mg L -1 ), the majority (>75%) of which is in a reduced state (Supplemental Materials Table S1) that, based on Fe isotopic data presented here, is leached from local LCT. The outflow channel is lined with iron oxyhydroxides (Fig. 1B) suggesting rapid oxidation and precipitation of iron oxyhydroxides iron. In waters with pH of < 4.0, such as those in Red Bubbler, Fe(II) oxidation with oxygen is slow and requires a microbial catalyst (Edwards et al., 1999). The presence of Archaea in the sediments of Red Bubbler that encode homologs of enzymes that allow for Fe(II) oxidation (e.g., sulfocyanins) to support their energy metabolism is thus attributable to acid leaching of this reduced metal by fluids in the near surface during their transport toward the spring source. Indeed, the dominant Sulfolobales population in Red Bubbler sediments appears to at least partially support primary productivity through aerobic Fe(II) oxidation (Fig. 6b). Leaching of Fe and its availability to support microbial activity in Red
Bubbler is enabled by the process of phase separation that allows for the concentration of H2S in vapor phase gas that is then oxidized through a series of geobiological feedbacks that drive fluid acidification. The absence of homologs involved in Fe(II) oxidation in abundant Perpetual
Spouter populations is thus attributed to the lack of Fe(II) in those waters (0.06 mg L -1 ;
Supplemental Materials Table S1) due to the insolubility of iron at circumneutral pH, its instability in the presence of oxygen at circumneutral pH, and the limited water-rock interactions that these fluids experience during their transport to the surface, as outlined in previous sections.
While these results describe the specific communities of Red Bubbler and Perpetual Spouter, they are reflective of other microbiological observations across Yellowstone hot springs (Fig. 7) (Meyer-Dombard et al., 2005;Boyd et al., 2013;Inskeep et al., 2013;Colman et al., 2018Colman et al., , 2019) ) which have demonstrated stark differences in the taxonomic and functional compositions of communities in hot springs with varying pH (i.e., due to phase separation). The communities of high-temperature acid-sulfate springs are comprised almost entirely of aerobic sulfur-metabolizing Archaea (Inskeep et al., 2013;Colman et al., 2018;Power et al., 2018) that are supported by functionalities that may be responsible for the acidification of those springs (Colman et al., 2018(Colman et al., , 2019)). In particular, microorganisms are the catalysts that drive the oxidation of S 0 , which results from incomplete oxidation of H2S and which is otherwise stable in the absence of biology at temperatures of <100°C (Nordstrom et al., 2005). In contrast, hightemperature alkaline-chloride springs host communities comprising both archaeal and bacterial members (Meyer-Dombard et al., 2011;Inskeep et al., 2013;Fernandes-Martins et al., 2021) that are largely supported by aerobic and anaerobic metabolisms that can involve a variety of different electron donors including arsenate (Fernandes-Martins et al., 2021) that is leached from rhyolite and is enriched in this water type (Stauffer and Thompson, 1984;McCleskey et al., 2022a). Per the phase separation model and for reasons outlined above, the lack of fluid mixing on diffuse flow paths sourcing springs like Perpetual Spouter, combined with the limited availability of O2 and/or H2S in those fluids, does not promote enrichment of sulfur-oxidizing populations and the acidification of spring waters.
Alexander von Humboldt profoundly stated in 1863, that "Everything is Interconnected" (von Humboldt, 1863). Our combined, multi-disciplinary measurements, from the two adjacent, phase-separated pools, Red Bubbler and Perpetual Spouter, provide, for the first time, a fourdimensional (spatial and temporal) understanding of the interconnectedness between hydrological, geochemical, and biological processes occurring before, during and after phase separation (Fig. 8). We uniquely show that Perpetual Spouter's alkaline-chloride waters preserve a signal of deep, protracted (hundreds of years) water-rock interaction with underlying Paleozoic and Mesozoic sedimentary rocks; whereas, in Red Bubbler's acid-sulfate waters, this deep, sedimentary geochemical signal is overprinted by shallow, ongoing, reactive transport with the LCT. Further, near-surface geophysical measurements indicate shallow meteoric recharge is integral in establishing the geobiological feedbacks that oxidize and acidify Red Bubbler's acid-sulfate waters and promote shallow, ongoing water-rock interaction and chemical weathering;
whereas, Perpetual Spouter's alkaline-chloride waters host both aerobic and anaerobic communities that largely reflect the influence of the deeper, anaerobic hydrothermal waters.
Our geophysical and isotopic data highlight causal links between subsurface geological processes and the assembly and diversification of thermophilic microbial communities. As precipitated silica that inhibits extensive infusion of oxidized meteoric fluids, thereby fostering communities capable of anaerobic metabolism. However, the 3 H data (Pearson and Truesdell, 1978;Gardner et al., 2011) also require roughly 7% mixing of shallow groundwaters.
Furthermore, Perpetual Spouter's slightly elevated 208 Pb/ 206 Pb (Fig. 3b), relative to the 208 Pb/ 206 Pb in the purported deep sedimentary reservoir, requires a mixing component that could be either water that has interacted with the LCT, or dissolution of dust that influences the water's composition. If dust is the primary contributor to the 208 Pb/ 206 Pb signal then this result has implications for nutrient supplies including organic carbon and phosphorous (Aarons et al., 2017;Aciego et al., 2017), possibly from microbial necromass. This might help to explain the prevalence of putative heterotrophs in Perpetual Spouter. In any case, the 3 H and 208 Pb/ 206 Pb data demonstrate that Perpetual Spouter is a mixture of mostly deep anoxic waters and some shallow oxidized water having nucleogenic 3 H and some component (dust or water) with higher 208 Pb/ 206 Pb than inferred for the deep hydrothermal sedimentary reservoir. As a result, the table is set for both anaerobic and aerobic heterotrophic populations in circumneutral to alkalinechloride springs.
Our results here represent a seminal example of interconnectedness within a continental hydrothermal system. When examined as an interconnected geohydrobiological system, our data illuminate the geochemical consequences of reactive transport processes and shallow meteoric recharge on the diversification and maintenance of thermophilic microbial communities that inhabit surface and near-surface ecological niches along the fluid flow-paths of this phaseseparated hydrothermal system. To this end, our collective results provide unique and direct evidence of the causative connections between subsurface geological processes and biological diversification. And, as such, our integration of different, yet complementary, geological, geophysical, geochemical, and molecular microbial data informs a unique and holistic interpretation of the development and evolution of phase-separated hydrothermal systems in a way not possible by any single method of study (Fig. 8).
While our results are specific to Yellowstone they are of global significance. Phase separation is a universal process in hydrothermal systems that have acted as critical evolutionary hot spots driving the diversification of early life on Earth. Yellowstone hosts the world's largest and most quintessential surface expression of a continental hydrothermal system. Thus, by extrapolation, our study's intellectual advances apply to most hydrothermal systems on Earth and likely throughout the solar system. As such, this research reaches far beyond the interest of any single discipline and speaks to a broader understanding of Earth system science.
Finally, we speculate that these same processes are likely to have taken place in continental hydrothermal systems throughout Earth history, albeit the role of O2 in hot spring acidification is likely more recent (Andersen et al., 2015).
0.01 0.1 1 10 100 1000 0 100 200 300 400 500 600 700 ( 228 Ra/ 226 Ra) Time (Years) Recoil Only Dissolution Only Unaltered LCT tal Figure 2 Supplemental Figure C2
Ra Chronometer Parent Nuclides Recoil Proxies Neutrons Protons 234 U 245.4 ka α 235 U 704 Ma α 238 U 4.47 Ga α 234 Pa 1.18 m β - 234 Th 24.1 d β - 232 Th 14 Ga α 231 Pa 3.276 ka α 230 Th 75.38 ka α 231 Th 25.52 h β - 228 Ra 5.75 a α 228 Ac 6.13 m β - 228 Th 1.91 a α 227 Th 18.68 d α 227 Ac 21.772 a α/β - 226 Ra 1.6 ka α 223 Ra 11.43 d α 224 Ra 3.66 d α 223 Fr 22.0 m β - 222 Rn 3.825 d α 220 Rn 55.6 s α 219 Rn 3.96 s α 215 Po 1.8E-3 s α 218 Po 3.05 m α 216 Po 0.15 s α 235 U Decay Chain 238 U Decay Chain 232 Th Decay Chain 0 10 20 30 40 50 60 0 50 100 150 200 ( 228 Ra/ 226 Ra) Mean Residence Time (Years) Y-12, Alt. LCT, W =25; R = 0.48k Av. LCT, W = 25, R = 10k A) Red Bubbler 140917 150814 140724 Figure 5 B) Perpetual Spouter Perpetual Spouter 0 5 10 15 20 25 30 35 0 100 200 3 ( 228 Ra/ 226 Ra) Mean Residence Ti 150814 140724 140917 100 150 200 Residence Time (Years) Y-12, Alt. LCT, W =25; R = 0.48k Av. LCT, W = 25, R = 10k 150814 140724 B) Perpetual Spouter l Spouter 0 5 10 15 20 25 30 35 0 100 200 300 400 500 ( 228 Ra/ 226 Ra) Mean Residence Time (Years) AMS W = 250, R = 0.5k AMS W = 125, R = 0.5k AMS W = 250, R = 1k LCT W =250, R = 0.5k LCT W = 125, R = 0.5k LCT W = 250, R = 1k 150814 140724 140917 Figure 5
Ra decay during transport in Stage 3). In Red Bubbler's acid-sulfate waters, water-rock interaction during reactive transport through the LCT resets the composition of the fluids, including its Ra concentrations and isotopes. Thus, with Red Bubbler's waters, the Ra
Ra (fg/g)------238 U (dpm/g)