Subaqueous calcite deposited on the walls of Devils Hole 2 cave (Nevada, USA) represents a unique archive for geochemical variations within the regional aquifer. Here, we present a 475,000-year record of initial 234U/238U activity ratios in delta notation (δ234U0). Results show a range in values from 1851–1616‰. Variations in δ234U0 coincide with interglacial-glacial cycles over the past 475,000 years. Maximum δ234U0 values correspond to the last five glacial intervals, during which southwest Nevada experienced cool, pluvial conditions. Minimum δ234U0 values correspond to interglacial intervals, during which this region experienced warm, arid conditions. We propose that an elevated water table during glacial periods inundated previously dry bedrock and basin sediments, thereby leaching excess 234U accumulated in these materials. We interpret Devils Hole 2 cave δ234U0 as a proxy for water-rock interactions in this regional aquifer, which is ultimately tied to the surface moisture conditions at recharge zones. The mechanism proposed here serves as a testable hypothesis and possible analogue for future subaqueous speleothem studies in similar hydrogeologic settings. Due to its unprecedented duration, the Devils Hole 2 cave δ234U0 record provides the first paleo-moisture record in southwest Nevada for marine isotope stages 10–12. In addition, high-precision δ234U measurements of modern groundwaters sampled from Devils Hole 2 cave are presented.

The Ash Meadows groundwater flow system (AMGFS) is a large carbonate aquifer located downstream of a potential radioactive waste repository site in southwest (SW) Nevada, USA. Understanding this region’s hydrological variability over longer timescales is required by federal regulations to confidently assess the long-term risk of future radionuclide migration (10 CFR 963). The state of disequilibrium for the naturally occurring uranium isotopes 234U and 238U in groundwaters has been widely used to investigate the modern hydrological parameters of the AMGFS and surrounding aquifers (Paces et al., 2002; Neymark et al., 2005; Bushman et al., 2010; Paces et al., 2013; Paces and Wurster, 2014). Uranium can be added to groundwater systems through (in)congruent dissolution of bedrock and overlying sediments where, in the presence of oxidizing conditions, it readily forms uranyl ions that complex with other ligands (i.e., uranyl carbonate; Langmuir, 1978 and references therein). Disequilibrium occurs when 234U enters waters preferentially due to processes linked to the energetic alpha-decay of parent 238U, such as the ejection of 234U into pore space during alpha-recoil or accelerated diffusion via alpha-recoil tracks (Rosholt et al., 1963; Cherdyntsev, 1971; Kigoshi, 1971; Osmond and Cowart, 1976; Osmond and Ivanovich, 1992; Stirling et al., 2007). The degree of disequilibrium is a function of numerous hydrogeologic parameters, including but not limited to, groundwater source, recharge amount and frequency, water-rock interactions, and flow rate (e.g., Osmond and Cowart, 1976; Andrews and Kay, 1982; Ivanovich et al., 1991; Kronfeld et al., 1994; Toulhoat et al., 1996; Johannesson et al., 1997; Roback et al., 2001; Paces et al., 2002; Neymark et al., 2005; Maher et al., 2006; Bushman et al., 2010; Paces et al., 2013; Paces and Wurster, 2014; Priestley et al., 2018). Thus, studying changes in AMGFS groundwater 234U-238U disequilibrium over long timescales provides valuable insight into the long-term hydrological variability of this region.

Secondary carbonates, such as spring deposits and speleothems, reflect the 234U-238U disequilibrium of the groundwater from which they precipitated. Measured 234U-238U disequilibrium activity ratio [(234U/238U)ACT] is reported here in delta notation, where δ234U is the deviation in permil of (234U/238U)ACT from its secular equilibrium value of 1 [δ234U = ((234U/238U)ACT – 1) × 1000]. Under closed system conditions, measured δ234U (δ234Um) can be corrected for the decay of excess 234U since the time of sample formation (t) in order to determine the initial value of δ234U (δ234U0):
where λ234 represents the decay constant of 234U (2.82206 × 10−6 a−1; Cheng et al., 2013). δ234U0 recorded in secondary carbonates has been used to investigate past changes in regional rainfall amount (Robinson et al., 2004), surface run-off (McGee et al., 2012), and infiltration frequency and/or amount (Ayalon et al., 1999; von Gunten et al., 1996; Bonotto and Andrews, 2000; Hellstrom and McCulloch, 2000; Zhou et al., 2005; Cross et al., 2015; Maher et al., 2014; Oerter et al., 2016).

Paleo-hydrological research in Nevada includes the study of δ234U0 from fossil spring deposits (Quade et al., 1995), soil carbonates (Maher et al., 2014), and dripstone speleothems (Cross et al., 2015). Yet, these deposits are frequently complicated by fragmented preservation, lack of continuous deposition, or difficulties in dating due to detrital contamination. Considering these limitations, we focus on the calcite coatings found in Devils Hole and Devils Hole 2 caves, located ∼200 m apart in the discharge zone of AMGFS. Pioneering work by Ludwig et al. (1992) revealed variations in δ234U0 associated with the last six glacial-interglacial cycles. An investigation into the mechanisms driving these variations has so far not been made. Here, we combine recently published data from Moseley et al. (2016) with new high-precision δ234U0 measurements from Devils Hole 2 cave in order to study variations between 475 and 5 thousand yr B.P. (ka) and propose a possible mechanism driving these changes. In addition, we present high-precision measurements of the modern groundwater δ234U sampled in Devils Hole 2 cave.

Devils Hole (DH) and Devils Hole 2 (DH2) caves are located in the Ash Meadows Oasis in SW Nevada (36°25′N, 116°17′W; 719 m above sea level). The caves are a set of tectonic fissures (Riggs et al., 1994) that were later modified by condensation corrosion (Dublyansky and Spötl, 2015). The entrance to DH2 is 200 m NNE and approximately +25 m in vertical height from DH. DH and DH2 intersect the regional water table at –15 m (DH) and –40 m (DH2) below the surface. Surveying by the authors suggests identical water table elevations in DH and DH2 within 8 cm uncertainty.

Geological and Hydrogeological Setting

DH and DH2 caves are located in the discharge zone of the AMGFS, a large (∼12,000 km2) Paleozoic limestone aquifer dominated by the Cambrian Bonanza King Formation (Winograd and Thordarson, 1975). In the mid-Tertiary, E-W extension generated widespread normal block faulting and N-S oriented fractures in the brittle carbonate sequence (Riggs et al., 1994). Prevailing NW-SE extension over the past ∼10 million years produced subsequent NE-SW oriented fractures, resulting in a highly fractured carbonate rock that provides regional-scale drainage from high-elevation recharge zones to low-elevation discharge zones through an extensive network of subterranean openings (Winograd and Pearson, 1976; Riggs et al., 1994). To the north of our study region, rhyolitic and quartz latitic Tertiary volcanic rocks of the Timber Mountain-Oasis valley caldera complex overlay the carbonate sequence (Frizzell and Shulters, 1990; Fridrich et al., 1994). Quaternary alluvial and lacustrine sediments fill most low-lying basins in the region.

Regional groundwater movement and interbasin exchange occur within the lower carbonate aquifer (Fig. 1), which is alternately confined by young and partly indurated sediments in basins and unconfined beneath ridges (Belcher and Sweetkind, 2010). Groundwater flow is controlled by variations in fracture transmissibility and structural heterogeneity (Winograd and Thordarson, 1975). Additional regional hydrogeological units include upper carbonate aquifers disjointed by basin-fill sediment sequences, lower clastic aquitards, and volcanic tuff aquitards (Winograd and Thordarson, 1975).

The AMGFS is primarily recharged by infiltration of snowmelt and rainfall in the upper elevations of the Spring Mountains (∼500 mm a−1; Winograd and Thordarson, 1975; Thomas et al., 1996; Winograd et al., 1998; Davisson et al., 1999). Groundwater flows northwest toward Frenchman Flat, merging with minor groundwater inputs from the White River flow system sourced from the mountainous region of central Nevada (Fig. 1; Winograd and Thordarson, 1975; Thomas et al., 1996). Additional minor groundwater inputs sourced from Emigrant Valley north of Yucca Flats have also been proposed (Fig. 1; Davisson et al., 1999; Winograd and Thordarson, 1975). From Frenchman Flat, the groundwater follows a high-transmissivity zone southward and ultimately discharges in high volume (38,000 L min−1) along a fault-controlled spring line located in Ash Meadows Oasis and ∼1 km south from DH and DH2 caves (Winograd and Thordarson, 1975). The curved flow path of the AMGFS reflects the presence of an aquitard block at the northwestern end of the Spring Mountains (Fig. 1). Bushman et al. (2010) proposed an alternative groundwater pathway (Fig. 1) sourced from recharge in the Yucca Mountain region that infiltrates the unsaturated zone of volcanic rock before reaching the lower carbonate aquifer, where it flows southward through major north-south oriented faults toward Ash Meadows Oasis. Hydrogeologic conditions have remained static since the Pliocene (Hay et al., 1986).

Evidence from both DH calcite δ18O and 14C ages from dissolved organic carbon fractions suggests groundwater transit times of <2000 years from the Spring Mountains to DH and DH2 caves (Winograd et al., 2006). Due to the long flow path (>60 km), prolonged residence time, and the retrograde solubility of calcite, the groundwater flowing southwest through DH and DH2 caves is slightly supersaturated with respect to calcite (SI = 0.2; Plummer et al., 2000). Calcite has been continuously depositing upon the submerged walls of both caves in the form of dense mammillary crusts over the past 500 k.y. at a very slow rate of roughly 1 mm ka−1 (Winograd et al., 1992, 2006; Moseley et al., 2016). The groundwater transit time between DH to DH2 cave is ∼5 years based on parameters outlined in Winograd et al. (2006).

δ234U of Modern AMGFS Groundwater

Saturated-zone groundwater δ234U and total U concentrations in the AMGFS region are largely dependent on the rock type of the aquifer (Fig. 1). Waters in Paleozoic carbonate rock aquifers from Oasis valley, Amargosa valley, Spring Mountains, and the easternmost Nevada Test Site have δ234U between 500 and 3000‰. Groundwater δ234U in Quaternary alluvial and Precambrian siliciclastic rock aquifers is between 1900 and 2100‰. Groundwater δ234U from the Miocene volcanic rock aquifers in the Fortymile Wash and Crater Flat areas, in the upland recharge area of Pahute Mesa north of Yucca Mountain, and in down-gradient areas of Amargosa valley are commonly between 3000 and 5000‰. Anomalously high δ234U between 5000 and 7500‰ have been measured in volcanic aquifers in the Yucca Mountain region (Paces et al., 2002). Total U concentrations in groundwater in the AMGFS region are between 0.02 and 10 μg L−1 (Thomas et al., 1991; Paces et al., 2002; Cizdziel et al., 2005).

Modern and Past Climate Regimes

SW Nevada is currently one of the driest regions in North America. High mean annual temperatures (20 °C), high mean evapotranspiration rates (2600 mm a−1) and low mean precipitation rates (15–120 mm a−1) contribute to year-long arid conditions and a lack of groundwater recharge to the AMGFS along valley floors (Laczniak et al., 2001). Vegetation at valley floors largely consists of desert sage and barrel cacti. The higher elevations of the Spring Mountains (above 3000 m above sea level) receive more precipitation (mean 500 mm a−1), have lower mean annual temperatures (6 °C), and support predominately alpine arctic forbs and grasses (Winograd et al., 1998).

The SW United States, including SW Nevada, underwent drastic hydroclimate changes throughout the Quaternary, as best illustrated by the repeated expansion and desiccation of pluvial lakes and wetlands on orbital to millennial timescales (e.g., Oviatt, 1997; Springer et al., 2015). Proxy-constrained model simulations suggest that a southward displacement and intensification of the Pacific storm track increased the amount of wintertime precipitation over the SW United States during glacial intervals (COHMAP Members, 1988; Oster et al., 2015). Increased rainfall contributed to an increase in local moisture availability, defined here as an approximate measure of precipitation minus evaporation (P–E), which prompted increased water table elevations in the AMGFS and surrounding regions (Szabo et al., 1994; Springer et al., 2015; Wendt et al., 2018). During interglacial intervals, a northward recovery of the Pacific storm track contributed to decreased P–E, resulting in low water table elevations (Szabo et al., 1994; Wendt et al., 2018) similar to today.

A 90 cm-long core was drilled from the hanging wall of DH2 cave at +1.8 m above the modern water table. Samples for 230Th dating were hand-drilled along the growth axis of the halved core using 0.3–0.4 mm carbide-tipped drill bits. Powdered sample sizes ranged between 30–50 mg of calcite. 230Th dating was performed at the University of Minnesota, Minneapolis, Minnesota USA. Samples were digested in HNO3 and spiked with a mixed 233U-236U-229Th spike similar to that described in Edwards et al. (1987). Spiked samples were fumed with concentrated HClO4, co-precipitated with Fe, centrifuged, and loaded into anion exchange columns following the methods described by Shen et al. (2002, 2012). Separate U and Th liquid extracts were measured using a ThermoFisher Neptune Plus multicollector–inductively coupled plasma–mass spectrometer (MC-ICP-MS) via a secondary electron multiplier on peak-jumping mode (Shen et al., 2012; Cheng et al., 2013). Chemical blanks were measured with each set of 10–15 samples and were found to be negligible (<100 ag 230Th, <50 ag 234U, <0.5 pg 232Th and 238U). Ages were calculated using the 230Th and 234U half-lives of Cheng et al. (2013). The 234U-238U disequilibrium at the time of deposition (δ234U0) was determined by back-calculating the measured δ234U using its associated 230Th age.

To measure the δ234U of modern groundwater within DH2 cave, two water samples were collected at approximately –0.2 and –1.3 m below the water table. 4 L of water were sampled using a sterile polyethylene hand pump and separated into 2 L collection bottles at each depth. Both pump and collection bottles were soaked in dilute HCl and tested for possible background contamination prior to collection. Samples were transported to the Death Valley Park Aquatic Ecology laboratory in Pahrump, Nevada and acidified using 1 mL of HNO3 per 1 L of water approximately two hours after collection. 2 L from each depth were passed through a 0.2 μm pore-sized filter. Water sample uranium measurements were performed at the University of Minnesota. 2 g subsamples, corresponding to 6 ng 238U, were spiked with a 233U-236U tracer (Cheng et al., 2013). Uranium was collected following chemical methods described by Chen et al. (1986) and isotope ratios were measured on a ThermoFisher Neptune Plus MC-ICP-MS via a secondary electron multiplier on peak-jumping mode. Total procedural blanks including filtering were found to contribute ∼1 pg 238U and <0.3 fg 234U, and were well within instrumental uncertainties.

A total of 100 δ234U0 values were calculated along the DH2 core. Results from the first 51 δ234U0 values were previously published in Moseley et al. (2016) (Table DR21). The remaining 49 δ234U0, U concentrations, and 230Th ages are presented here for the first time (Table DR1; see footnote 1). 230Th ages range from 4.89 ± 0.45 ka to 476 ± 11 ka. A single calculated δ234U0 value was identified as a statistical outlier (Q test) and omitted from the data set. DH2 δ234U0 during this time ranged from 1851–1616‰. Uranium concentrations of DH2 calcite averaged 566 ± 270 (1σ) ng g−1 over the past 475 k.y. Growth rates were calculated using a linear interpolation. Negative rates due to occasional age reversals were omitted. Results show that DH2 δ234U0 does not correlate with U concentration (R2 = 0.02) or growth rate (R2 = 0.02) over the past 475 k.y. (Fig. 2). The range and average value of the DH2 δ234U0 data agree within mean uncertainty (0.85%) with the previously published δ234U0 data from DH cave (Fig. 2), which spans between 630 ± 109 ka and 5.2 ± 0.2 ka (Ludwig et al., 1992). Five maxima in DH2 δ234U0 are identified at ca. 475 ± 11, 374 ± 6, 278 ± 2, 185.1 ± 0.7, and 43.2 ± 0.2 ka corresponding to glacial marine isotope stages (MIS) 12, 10, 8, 6, and 2, respectively. DH2 δ234U0 minima occurred at 410 ± 6, 327 ± 3, 239.2 ± 1.5, and 128.8 ± 0.4 and 4.90 ± 0.05 ka corresponding to interglacial MIS 9, 7, 5, and the Holocene, respectively (Fig. 3).

The δ234U of modern groundwater collected in DH2 cave is 1762 ± 2‰ and U concentrations were measured as 3.085 ± 0.005 μg L−1 (Table 1). DH2 modern groundwater δ234U values falls within the observed range of calcite δ234U0 variations. All past and modern measurements presented in this study are within the range of modern groundwater δ234U collected throughout the AMGFS region (Fig. 1).

Link to Regional Hydroclimate

The DH2 δ234U0 time-series reveals a close link with local moisture availability over the past 475 k.y. Periods of elevated δ234U0 values (defined here as δ234U values greater than modern values) coincide with periods of increased P–E in SW Nevada, as indicated by DH2 water-table elevations greater than +5.5 m relative to the modern water table (r.m.w.t.) between 391 ± 7 and 342 ± 5, between 320 ± 3 and 250 ± 2, and between 218 ± 1 and 157.8 ± 0.7 ka (Fig. 3; Wendt et al., 2018). During the last glaciation, elevated δ234U0 values (>1760‰) between 83.1 and 13.0 (±0.3) ka coincide with increased P–E in SW Nevada, as indicated by marsh and spring deposits in the Las Vegas valley (Quade, 1986; Quade and Pratt, 1989; Quade et al., 1995; Springer et al., 2015), speleothem growth in Pinnacle Cave (Lachniet et al., 2011), spring deposits in Indian Springs valley (Quade and Pratt, 1989), paleo-ecological reconstructions from southern Nevada pack-rat middens (Thompson et al., 1999), and water table elevations above +5 m r.m.w.t. in DH2 and DH caves (Szabo et al., 1994; Wendt et al., 2018). During interglacial intervals, decreased δ234U0 values (<1760‰) coincide with periods of decreased P–E, as indicated by water table low-stands (below +1.8 r.m.w.t.) at 410 ± 6, 327 ± 3, 239.2 ± 1.5, and 128.8 ± 0.4 ka. We argue that DH2 δ234U0 reflects hydrological variability in SW Nevada over glacial-interglacial timescales.

The timing and duration of elevated DH2 δ234U0 is coincident with periods of elevated lake levels (e.g., Ku et al., 1998; Bacon et al., 2006; Benson et al., 2013; McGee et al., 2012; Oviatt, 1997) and increased vadose-zone infiltration (Maher et al., 2014; Cross et al., 2015) recorded in a diverse array of archives across the wider SW United States. Curiously, however, the interpretation of δ234U0 recorded in DH2 speleothems is opposite to the interpretation of δ234U0 recorded in subaerial speleothems in this region (e.g., Cross et al., 2015; Denniston et al., 2007; Shakun et al., 2011; Lachniet et al., 2011, 2014), and thus merits discussion. The mechanisms controlling the δ234U in subaerially formed speleothems (i.e., dripstones) are attributed to the frequency and amount of surface recharge (Cross et al., 2015 and references therein). Because 234U is produced in the unsaturated zone at a constant rate, 234U accumulates in soil and bedrock during periods of infrequent and/or low recharge amount (i.e., low P–E), such that infiltrating waters during these dry periods contain high δ234U values. During periods of frequent and/or high recharge 234U is diluted, resulting in lower δ234U values. Overall, a negative correlation between δ234U0 values and P–E is observed in subaerial speleothems from the SW United States (Cross et al., 2015; Denniston et al., 2007; Shakun et al., 2011; Lachniet et al., 2011, 2014). Conversely, δ234U0 recorded in DH and DH2 calcite (subaqueously deposited) demonstrates a positive correlation with regional P–E. Our results clearly demonstrate that previously described interpretation of speleothem δ234U0 does not universally apply to all depositional settings. In order to understand the link between DH and DH2 δ234U0 and regional hydroclimate changes, we propose and discuss the following possible controlling mechanisms.

Mechanisms Controlling δ234U0

Possible mechanisms controlling groundwater δ234U in DH and DH2 caves include: (1) changes in the chemical or physical properties of groundwater, (2) changes in groundwater flow rate or path length due to tectonic activity, (3) changes in groundwater source, and (4) exposure to previously unsaturated bedrock and/or sediments due to fluctuations of the local water-table. The relatively constant U concentration and growth rate recorded in DH2 calcite argue against past changes in the chemical or physical properties of groundwater (e.g., pH, redox state, temperature, or partial pressure of CO2). The synchronicity between DH2 δ234U0 and hydroclimatic variations in SW Nevada argue against mechanisms associated with tectonic activity, as these changes would result in randomized changes in δ234U0 over time. The remainder of this discussion will investigate mechanisms (3) and (4).

Changes in Groundwater Source

The modern groundwater in DH2 cave (δ234U = 1762 ± 2‰) is largely sourced from Spring and Sheep Mountain recharge (Fig. 1; Winograd and Thordarson, 1975), whose groundwater δ234U signatures range from 500 to 3000‰ (Thomas et al., 1991; Paces et al., 2002; Cizdziel et al., 2005), typical of carbonate-rock aquifers in this region. Additional groundwater pathways to the AMGFS from northern regions have been proposed (see Study Site section), including inputs originating from volcanic terrains, which show anomalously high groundwater δ234U (up to 7500‰) values (Paces et al., 2002).

During glacial periods, increased recharge in lower-altitude regions (relative to Spring Mountains) may have affected the hydraulic head and mixing proportion of different groundwater flow paths contributing to AMGFS. For example, increased recharge in the northern volcanic terrains may have increased this region’s groundwater input to the AMGFS, thereby elevating groundwater δ234U0. Under this mechanism alone a direct correlation would be expected between DH2 δ234U0 and P–E in volcanic terrains north of DH and DH2 caves (e.g. Yucca Mountain). Instead, DH2 δ234U0 reach maximum values mid-glacial cycle (ca. 43.2, 185.1, 278, and 374 ka) followed by a gradual decrease toward interglacial values despite continued DH2 water table high stands (Wendt et al., 2018) and high mean annual precipitation in the Yucca Mountain region (Thompson et al., 1999) until the termination of the associated glacial cycle (Fig. 4). For example, during the last glaciation, DH2 δ234U0 commenced decreasing toward interglacial values at 43.2 ± 0.2 ka, despite maximum wet conditions of the Yucca Mountain region from 27 to 11.5 ka (Thompson et al., 1999) and continuous water table high-stands (+5.5 m r.m.w.t.) until 19.85 ± 0.04 ka (Wendt et al., 2018).

Additional evidence against this mechanism includes the observed decoupling between DH2 δ18O and DH2 δ234U0 during glacial periods. Recharge from the Yucca Mountain region that reached the lower carbonate aquifer would have likely mixed with southward-flowing groundwaters sourced from Pahute Mesa, which are depleted in δ18O relative to discharging AMGFS waters by 1–2‰ (Davisson et al., 1999). An increase in groundwater contribution from this region is therefore expected to have enriched δ234U signatures while simultaneously depleting the δ18O of discharging waters (and of calcite precipitating from this water). Instead, DH2 δ234U0 decoupled from DH2 δ18O ∼40 k.y. prior to the enrichment of δ18O associated with Termination II and ∼30 k.y. prior to Termination I (Fig. 4).

The DH2 δ234U time series cannot account for past changes to the input of southeast-flowing groundwaters from the White River Flow System or Yucca and Frenchman Flats, because these regions contain groundwaters with similar U concentrations and δ234U signatures to Spring and Sheep Mountain ranges (Thomas et al., 1991; Paces et al., 2002; Cizdziel et al., 2005) (see supplementary information for further discussion). Overall, the timing of DH2 δ234U0 decline during glacial periods does not support changes in groundwater inputs from volcanic terrains as a primary mechanism for recorded DH2 δ234U0 variations, and suggests that the contribution of excess 234U is sourced locally.

A commonly used hydrologic tracer not included in this study is the strontium isotopic composition. The mean 87Sr/86Sr value for the larger springs at Ash Meadows (accounting for 83% of discharge) is 0.7125 ± 0.0002 (2σ) (Stuckless et al., 1991). Previous studies on the Sr isotope budget of regional groundwater indicate indistinguishable 87Sr/86Sr values between the AMGFS carbonate aquifer (0.7084–0.7191) and in the volcanic aquifers in the vicinity of Yucca Mountain (0.7093–0.7119) (Bushman et al., 2010; Stuckless et al., 1991). Assigning a most probable source of water based on potential variations in Sr isotopes recorded in DH calcite would therefore require a more in-depth investigation.

Exposure of Previously Unsaturated Bedrock and/or Basin Sediments

An alternative mechanism involves the inundation of previously unsaturated bedrock and/or basin-fill sediments during periods of high P–E. Wendt et al. (2018) document >9.5 m rises of the DH2 water table during cool and wet glacial cycles over the past 350 k.y. Considering this, the AMGFS lower carbonate aquifer can be divided into two zones: a lower zone that remains saturated throughout both glacial (high water table) and interglacial (low water table) intervals, and an upper zone that is saturated exclusively during glacial intervals. During interglacial intervals, we propose that the unsaturated upper zone undergoes a build-up of 234U along mineral grain and fracture surfaces (Fig. 5). Due to the absence of passing groundwaters, mobile 234U atoms remain in damaged lattice sites or are ejected directly into the neighboring pore space where they remain associated to mineral surfaces. During glacial intervals, a high water table re-exposes the upper zone to oxidizing groundwaters, such that the accumulated mobile 234U is transported downstream. The resulting increase in groundwater 234U concentrations is ultimately registered in DH2 calcite as increased δ234U0 without significant changes to 238U concentrations.

During periods of high P–E, it is additionally possible that: (1) Quaternary-Tertiary lacustrine sediments, alluvial sediments, and volcanic tuffs that overlay the primary carbonate rock aquifer would become partially saturated (exact volume unknown) due to an increased water table elevation and (2) local recharge to the AMGFS may occur in Quaternary sediments and volcanic tuff filling basin floors. Under this scenario the availability of mobile 234U would further increase, as the fraction of mobile 234U that is susceptible to preferential leaching increases in proportion to decreasing grain size. The previously unsaturated upper bedrock and overlying Quaternary sediments will henceforth be referred to as the “upper zone,” and are assumed constant in physical composition over this study interval.

In order to evaluate the compatibility of the proposed mechanism to the DH2 δ234U0 record, we present a simple conceptual model that estimates the nonlinear response of groundwater δ234U variations to water-table fluctuations during a glacial-interglacial cycle (Fig. 6). Following the end of an interglacial interval, rising water table elevations during a glacial inception would increase the exposure of the upper zone to groundwater and thus increase groundwater δ234U0. Water table elevations first reach maximum height (H) as δ234U0 attain peak values (Umax). After reaching maximum height, water table elevations remain at or near maximum height throughout the remainder of the glacial interval (tH). Following Umax, δ234U0 values gradually decrease throughout the interval tH in response to a depletion of mobile 234U in the upper zone. In total, the conceptual model shown in Figure 6 resembles the triangle-shaped peaks observed in the DH2 δ234U0 time series during glacial intervals.

Three lines of evidence from the DH2 δ234U time series support this mechanism. (1) According to the proposed model, the timing of DH2 δ234U0 maxima (i.e., Umax) is expected to coincide with periods in which DH2 water table elevations first reached maximum height (H) within a glacial cycle. We define DH2 local Umax values as 1851 ± 32‰ (MIS 10), 1835 ± 13‰ (MIS 8), between 1825 ± 6‰ and 1828 ± 5‰ (MIS 6), and 1822 ± 3‰ (MIS 4–2). Indeed, the timing of DH2 local Umax associated with MIS 8 at 278 ± 3 ka agrees within age uncertainties with the timing of water table maxima (+8 m r.m.w.t.) at 273 ± 2 ka. The timing of DH2 local Umax associated with MIS 6 between 197.2 and 185.1 ± 2 ka coincides with the timing of water table maxima (+9.5 m r.m.w.t.) at 190.0 ± 0.7 ka. Finally, the timing of DH2 local Umax associated with MIS 4.2 at 43.2 ± 0.2 ka coincides with the timing of water table maxima (+9 m r.m.w.t.) at 44.0 ± 1.0 ka (Fig. 4).

(2) According to the proposed model, higher water table maxima (i.e., H) associated with a single glacial cycle would increase the volume of upper-zone saturation and thereby capture more 234U, which would result in higher groundwater δ234U. The maximum height of DH2 water table in a glacial cycle is similar (>+8 m r.m.w.t.) between MIS 8, 6, and 4–2. An exception is the short pluvial interval associated with MIS 7d (ca. 230–220 ka), during which the DH2 water table reached above +3.1 m but no higher than +6.5 m r.m.w.t. (Wendt et al., 2018). Corresponding DH2 δ234U0 values did not exceed 1748‰ during this time, suggesting that lower water table maxima at MIS 7d did not saturate a sufficient volume of the upper zone to shift δ234U toward higher values.

(3) According to the proposed model, prolonged periods of low water table levels would extend the accumulation interval of 234U within the upper zone such that Umax values are expected to scale proportionally to the duration of the previous interglacial (tI). Determining the duration of the last four interglacials using the DH2 water table record is complicated by the low data resolution during MIS 9e and lack of data during MIS 11. Instead, we use global sea level reconstructions from Spratt and Lisiecki (2016) which show a close agreement with the timing of DH2 water table fluctuations on glacial-interglacial timescales (Wendt et al., 2018). For the purposes of this study, we set the interglacial sea level boundary to ≥–5 m relative to modern day. Using this, the duration of the last four interglacials are approximated to 16 ka (MIS 11c), 9 ka (MIS 9e), 8 ka (MIS 7a–c), and 5 ka (MIS 5e). By comparing each interglacial duration to the following glacial Umax value (selecting 1828 ± 5‰ as absolute local Umax for MIS 6), a positive correlation (R2 = 0.98) is observed (Fig. 7). This result supports the hypothesis that longer interglacial intervals provide a longer period of 234U accumulation in the upper zone, which in turn results in higher δ234U0 maxima during the following glacial interval.

Overall, we argue that the timing and amplitude of δ234U0 variations point to the inundation of the upper zone as the primary driving mechanism of DH2 δ234U0. Future groundwater hydrology and geochemical modeling is required to test this hypothesis. Budget calculations are largely limited by the unknown surface area of bedrock fractures, through which 234U is contributed to the groundwater system. Additional major unknowns must also be addressed, such as the differential contribution of uranium from bedrock and overlying sediments, potential past variations in hydraulic head and flow rate, changes in preferential flow routes over time, and the extent and effect of calcite deposition along fracture surfaces in the aquifer.

Interpreting DH2 δ234U

Variations in DH2 δ234U0 over the past 475 k.y. provide insight into hydrological changes in SW Nevada on glacial-interglacial timescales. We interpret DH2 δ234U0 as a proxy for the interaction of groundwaters with previously unsaturated bedrock and sediments. The amount and nature of water-rock interactions in the AMGFS are a function of water table elevations, which in turn reflect recharge amount and surface P–E conditions at recharge zones. Intervals of anomalously low and high δ234U0 values may therefore shed light on past periods of extreme P–E conditions. The lowest recorded DH2 δ234U0 values (1615 ± 3‰) between 128.8 and 121.4 (±0.4) ka coincide with DH2 water table low stands that reached below modern elevations (Wendt et al., 2018) and the highest DH/DH2 δ18O values recorded over the past 500 k.y. (Winograd et al., 1992, 2006; Moseley et al., 2016), suggesting exceptionally low P–E conditions in the AMGFS recharge regions associated with MIS 5e. The highest recorded DH2 δ234U0 values (1850 ± 30‰) coincide with the lowest DH δ18O values over the past 500 k.y. (Winograd et al., 1992, 2006), suggesting exceptionally high P–E associated with MIS 10. Curiously, modern groundwater δ234U and calcite δ234U0 during the Holocene have relatively high interglacial values (Fig. 3) despite low water table elevations and enriched (∼–15.3‰) δ18O (Moseley et al., 2016; Wendt et al., 2018). This may be due to the partial opening of the cave (via roof collapse) at ca. 4.5 ka (Winograd et al., 2006), which potentially introduced surface waters that contained higher δ234U due to interactions with upper zone sediments.

More importantly, the DH2 δ234U0 record provides the first record of hydroclimate conditions in SW Nevada from MIS 10–12. Maximum δ234U0 values at 370 ± 7 ka indicate that water table high-stands first reached maxima associated with MIS 10 at this time (Fig. 3). Low DH2 δ234U0 (1640 ± 30‰) centered at 410 ± 6 ka indicate low P–E in SW Nevada associated with MIS 11c. Considering the larger data uncertainties, the transition from MIS 11 to MIS 10 (+0.17‰) is similar in amplitude to the transition from MIS 5e to the last glaciation (+0.20‰), indicating a dramatic shift in the hydroclimate of SW Nevada between MIS 11 and MIS 10 comparable to the last interglacial-glacial transition.

Groundwater δ234U measured in DH2 cave (1762 ± 2‰) falls within the observed range of modern groundwater in the AMGFS region and past δ234U0 variations recorded in DH2 calcite. DH2 δ234U0 variations are synchronous with past changes in the regional hydroclimate, such that elevated δ234U0 values (>1760‰) between 392 and 343 (±7), between 320 and 250 (±3), between 217.7 and 157.8 (±1.3), and between 83.1 and 13.0 (±0.3) ka coincide with periods of high local water table elevations. Likewise, decreased δ234U0 (<1760‰) correlate to periods of low local water table elevations at 410 ± 6, 327 ± 3, 239.2 ± 1.5, and 128.8 ± 0.4 ka (Fig. 4). We propose that elevated water table levels increased the exposure of upper zone bedrock and sediments to groundwater during glacial intervals, resulting in the leaching of mobile 234U that accumulated during the previous interglacial interval. Evidence for this mechanism includes similar timing in maximum DH2 δ234U0 and the DH2 water table elevations during glacial intervals, as well as a correlation between maximum glacial DH2 δ234U0 values and the duration of the previous interglacial interval. We interpret DH2 δ234U0 as a proxy for the interaction of groundwaters with previously unsaturated bedrock and sediments, which is ultimately tied to the surface P–E conditions at recharge zones. Due to its unprecedented length, the DH2 δ234U0 record provides the first hydroclimate record in SW Nevada to encompass MIS 10–12. The mechanism proposed serves as a testable hypothesis that may be applicable to future subaqueous speleothem studies in similar climatic and hydrogeologic settings.

This work was supported by the Austrian Science Fund project number FP263050 (to C.S.) and by the National Science Foundation project number 1602940 (to R.L.E.). This research was conducted under research permit numbers DEVA-2010-SCI-0004 and DEVA-2015-SCI-0006 issued by Death Valley National Park. We thank K. Wilson for assistance in the field and P. Zhang for assistance in the laboratory. Thanks also to M. Ottman for his insight and an anonymous journal reviewer for constructive criticism.

1GSA Data Repository item 2019299, uranium-series data, is available at or by request to
Science Editor: Bradley S. Singer
Gold Open Access: This paper is published under the terms of the CC-BY license.