Devils Hole Calcite Was Precipitated at ±1°C Stable Aquifer Temperatures During the Last Half Million Years

Subaqueous carbonates from the Devils Hole caves (southwestern USA) provide a continuous Holocene to Pleistocene North American paleoclimate record. The accuracy of this record relies on two assumptions: That carbonates precipitated close to isotope equilibrium and that groundwater temperature did not change significantly in the last 570 thousand years. Here, we investigate these assumptions using dual clumped isotope thermometry. This method relies on simultaneous analyses of carbonate ∆47 and ∆48 values and provides information on the existence and extent of kinetic isotope fractionation. Our results confirm the hypothesis that calcite precipitation occurred close to oxygen and clumped isotope equilibrium during the last half million years in Devils Hole. In addition, we provide evidence that aquifer temperatures varied by less than ±1°C during this interval. Thus, the Devils Hole calcite δ18O time series exclusively represents changes in groundwater δ18O values.


Introduction
The oxygen isotope composition of cave carbonates (δ 18 O cc ) is a common tool to reconstruct terrestrial climate (Comas-Bru et al., 2020;Fairchild & Baker, 2012). In particular, speleothems are the primary source of information on Quaternary climatic variability of the Great Basin, southwestern United States (Coplen et al., 1994;Lachniet et al., 2011Lachniet et al., , 20142020;Moseley et al., 2016;Quirk et al., 2020;Shakun et al., 2011;Wendt et al., 2018;Winograd et al., 1988Winograd et al., , 1992Winograd et al., , 2006. Subaqueously formed folia-free mammillary calcites from the Devils Hole groundwater system ( Figure 1) provide a unique, uninterrupted δ 18 O cc record between 4.5 and 570 ka (Winograd et al., 1988(Winograd et al., , 2006. In thermodynamic equilibrium, the δ 18 O cc value is only a function of the carbonate precipitation temperature and the groundwater oxygen isotope composition (δ 18 O gw ). Assuming that the aquifer's temperature did not vary with time and that carbonate growth occurred close to isotope equilibrium, the reconstructed Devils Hole δ 18 O gw time series is thought to reflect changes in meteoric precipitation.
The well-dated Devils Hole record, while showing an excellent agreement with California Current's sea surface temperatures (Herbert et al., 2001;Winograd et al., 2006), displays transitions to interglacial periods ca. 10,000 years before orbital forcing would predict (Ludwig et al., , 1988(Ludwig et al., , 2006. This age discrepancy is not reproduced in other speleothem records from the Great Basin (Lachniet et al., 2011(Lachniet et al., , 20142020). A revised chronology of the Devils Hole δ 18 O time series-based on newly recovered material-suggests that excess 230 Th in the water column, originating from the radioactive decay of 234 U, leads to uranium-series Abstract Subaqueous carbonates from the Devils Hole caves (southwestern USA) provide a continuous Holocene to Pleistocene North American paleoclimate record. The accuracy of this record relies on two assumptions: That carbonates precipitated close to isotope equilibrium and that groundwater temperature did not change significantly in the last 570 thousand years. Here, we investigate these assumptions using dual clumped isotope thermometry. This method relies on simultaneous analyses of carbonate ∆ 47 and ∆ 48 values and provides information on the existence and extent of kinetic isotope fractionation. Our results confirm the hypothesis that calcite precipitation occurred close to oxygen and clumped isotope equilibrium during the last half million years in Devils Hole. In addition, we provide evidence that aquifer temperatures varied by less than ±1°C during this interval. Thus, the Devils Hole calcite δ 18 O time series exclusively represents changes in groundwater δ 18 O values.

Plain Language Summary
The oxygen isotope composition of cave carbonates records changes in Earth's climate. However, the reliability of such records depends on how stable the carbonate precipitation environment was. Here, we use a novel method called dual clumped isotope thermometry that can provide simultaneous information on a carbonate's growth temperature and whether any additional fractionation processes affected its oxygen and clumped isotope signatures. Specifically, we investigated the Devils Hole caves, which provide a reference oxygen isotope time series for North America. We find that groundwater temperature did not change significantly in the last half-million years. Variations in the oxygen isotope composition of the deposited carbonates solely reflect variations in the oxygen isotope composition of the groundwater.
ages that are too old (Moseley et al., 2016). Alternatively, variations in the temperature of the subaqueous environment and the kinetics of carbonate mineralization could result in a δ 18 O cc record that does not accurately represent changes in δ 18 O gw and, thus, could contribute to the apparent age discrepancy.
Multiple lines of indirect evidence suggest both a stable subaqueous temperature and close to equilibrium calcite precipitation conditions for most of the Devils Hole record (Coplen, 2007). The Devils Hole caves formed tectonically at the Ash Meadows groundwater aquifer's fault-controlled discharge area   (Figure 1). The partially submerged caverns provided an isolated carbonate growth environment. When the caves opened to the surface, mammillary calcite formation ceased (Plummer et al., 2000). Based on the large size of the Ash Meadows aquifer (>12,000 km 2 ), the groundwater temperature has been postulated to remain constant on longer timescales. Kinetic isotope fractionation in most speleothems is related to the degassing of CO 2 (aq) from the groundwater (Affek et al., 2008;Affek & Zaarur, 2014;Daëron et al., 2011;El-Shenawy et al., 2020;Fairchild & Baker, 2012;Guo & Zhou, 2019;Hansen et al., 2013;Hendy, 1971;Kluge & Affek, 2012;Kluge et al., 2014). The mammillary calcites in the Devils Hole caves exhibit growth rates below 1.5 mm kyr −1 (Winograd et al., 2006), orders of magnitude slower than growth rates of subaerial speleothems. Due to these slow growth rates, the DIC-H 2 O-CaCO 3 system (dissolved inorganic carbon (DIC)) may have had sufficient time to attain isotope equilibrium (Dreybrodt & Scholz, 2011; BAJNAI ET AL.   (Thomas et al., 1996) over the region's generalized geology (Denny & Drewes, 1965;Horton et al., 2017). Winograd et al., 1992). Previous clumped isotope measurements (∆ 47 ) from Devils Hole Cave 2 hinted at a stable (30.6 ± 2.6°C) groundwater temperature between 65 and 180 ka (Kluge et al., 2014), corroborating that δ 18 O cc values predominantly record variations in δ 18 O gw values. However, like their oxygen isotope composition, the clumped isotope composition of carbonates can also be affected by kinetic fractionation (Bajnai et al., 2020;Daëron et al., 2011Daëron et al., , 2019Guo & Zhou, 2019). The ∆ 47 and δ 18 O cc measurements alone do not allow one to resolve potential kinetic effects in carbonate formation temperatures, especially in terrestrial environments where the δ 18 O values of the parent fluid are highly variable. Without independent evidence for insignificant contributions of reaction kinetics to the clumped and oxygen isotope record, the inferred temperature stability remains uncertain.
Recent developments in mass spectrometry enable simultaneous, high-precision analysis of the mass-48 CO 2 isotopologue (mostly 12 C 18 O 18 O; ∆ 48 ) and the mass-47 CO 2 isotopologue (mostly 13 C 18 O 16 O; ∆ 47 ), derived from carbonate minerals (Fiebig et al., 2019). In ∆ 47 versus ∆ 48 space, thermodynamic equilibrium at any given temperature is represented by a single point. Deviations from thermodynamic isotope equilibrium related to temporal variations in reaction kinetics proceed along trajectories specific to the underlying reaction mechanisms, for example, CO 2 (aq) degassing (Guo, 2020;Guo & Zhou, 2019). Simultaneous ∆ 48 and ∆ 47 measurements on a single carbonate phase, termed dual clumped isotope thermometry, enable identifying the nature and extent of kinetics involved in carbonate mineralization (Bajnai et al., 2020).
In this study, we analyzed the ∆ 47 and ∆ 48 values of 10 mammillary calcites from the Devils Hole caves spanning the last half million years. We investigated the potential relevance of kinetically induced variations on δ 18 O cc values and whether variable reaction kinetics and crystallization temperatures could affect the accuracy of the time series.

Study Area and Materials
We investigated mammillary calcites from the Devils Hole (36.425355°N, 116.291447°W) and the Devils Hole Cave 2 (36.427122°N, 116.291166°W) subaqueous caves, situated in the Amargosa Desert, Nevada, USA ( Figure 1). Both caves formed tectonically in Cambrian carbonates and dolomites Winograd & Thordarson, 1975). The aquifer recharges from two sources: Over 60% of the groundwater originates from the nearby Spring Mountains, ca. 80 km to the northeast, and up to 40% of the groundwater comes from the Pahranagat Valley, ca. 140 km to the north (Thomas et al., 1996). Based on 14 C ages, the groundwater travel time is estimated to be 2,200 years and 5,900 years, respectively (Thomas et al., 1996). The primary source of recharged water in the Spring Mountains is snowmelt (Winograd et al., 1998). The δ 18 O values of modern winter meteoric precipitation in southern Nevada range from −12.5‰ to −14.5‰ (Ingraham et al., 1991).
Material for this study was cut from the three cores studied by Coplen (2007); Winograd et al. (1988Winograd et al. ( ), (1992Winograd et al. ( ), (2006. These cores were explicitly chosen as they pertain to the original Devils Hole time series. Core DH-11 is from Devils Hole, and cores DHC2-8 and DHC2-3 are from Devils Hole Cave 2. There are no measurable differences between the two caves, neither in modern groundwater temperature nor in contemporaneous δ 18 O cc values (Winograd et al., 2006). The three cores were retrieved from 60 to 20 m below the present water level. The water level in the caves dropped by ca. 9 m since the last glacial period, which has been linked to changes in regional moisture availability (Szabo et al., 1994;Wendt et al., 2018). However, in the cores studied here, there are no folia and there is no evidence for dissolution, recrystallization, or any depositional hiatus (Kolesar & Riggs, 2004), indicating that they were continuously submerged since the middle Pleistocene. For clarification, we note here that Kluge et al. (2014) and Moseley et al. (2016) investigated cores containing interlayered folia and mammillary calcite from Devils Hole Cave 2, which were collected at or above the modern water table; they are different from those studied here, which were formed more than 20 m below the modern water table. The calcite slabs, taken from the cores, were first cleaned in an ultrasonic bath using de-ionized water, dried in a vacuum oven at 30°C, and finally homogenized using an agate mortar and pestle.

Dual Clumped Isotope Thermometry
Clumped isotope analyses were performed on a Thermo Scientific 253 Plus gas source isotope ratio mass spectrometer connected directly to a fully automated carbonate acid digestion system (90°C) and CO 2 purification line, detailed in Fiebig et al. (2019). The samples were measured along with carbonate standards (ETH-1, ETH-2, and ETH-3) in two sessions between September 2019 and June 2020. To complement the measurements in this study, we included five replicate analyses of sample DHC2-8 reported in Bajnai et al. (2020), thus, all samples were measured in 9-23 replicates. The clumped isotope values were calculated using the IUPAC isotope parameters (Brand et al., 2010;Daëron et al., 2016) and are reported on the carbon dioxide equilibrium scale at 90°C (indicated by subscript "CDES90"). Data correction of the carbonate samples for all measurement periods consisted of three steps, that is, correction for non-linearity, correction for scale compression, and correction for variation in the acid reaction environment, similarly as described in Bajnai et al. (2020). An extended description of the data processing steps is provided in the supporting information. All data of the analyses of both samples and reference materials are provided in Datasets S1-S2.

Discussion
The groundwater temperature in the caves has been remarkably invariant in the last 80 years. For March 1985, the most precise measurement published up to date reports 33.7(±0.2)°C in the main chamber of Devils Hole at depths between 5 and 37.5 m below water level (Plummer et al., 2000). This value is in good agreement with temperatures reported for November 1954, ranging from 33.9°C to 34.0°C (at 2-29 m below the modern water table), and for April 1961, ranging from 34.0°C to 34.3°C (at 1.3-27 m below the water table) (Hoffman, 1988). Earlier studies reported similar temperatures, that is, 32.8°C-33.9°C for the 1930-1947 period (Miller, 1948), and 33.5°C for December 1966 (Dudley & Larson, 1974). Furthermore, the covariation of fluid inclusion δ 2 H and δ 18 O cc values hint at constant temperatures through the penultimate full glacial and peak interglacial   Landwehr et al. (2011)). The uncertainties represent 95% confidence intervals (Fernandez et al., 2017).  (Coplen, 2007;Winograd et al., 1992). During the deposition of the most recent sample DHC2-8 (4.5-16.9 ka), the δ 18 O cc values remained constant (within ±0.1‰), and the calcite growth rate did not exceed 0.2 mm kyr −1 (Coplen, 2007;Winograd et al., 2006). The groundwater's temperature when DHC2-8 precipitated is thought to be identical to its modern temperature (33.7°C) (Coplen, 2007). Overall, the clumped and oxygen isotope compositions of DHC2-8 are considered to best represent equilibrium values at 33.7°C among Earth-surface carbonates (Coplen, 2007;Daëron et al., 2019;Tripati et al., 2015;Wostbrock et al., 2020).
However, it needs to be considered that the exact position of equilibrium in ∆ 47 (CDES90) versus ∆ 48 (CDES90) space is not yet known with confidence, as the discrepancy between the theoretical and the experimental equilibrium ∆ 47 (CDES90) values at 33.7°C indicate. Within their 95% confidence interval, the measured ∆ 48 (CDES90) and ∆ 47 (CDES90) values of most samples from the Devils Hole caves are indistinguishable from the corresponding values of sample DHC2-8 (Figure 2; although sample DH-11 109.4 shows a 0.002‰ difference in its ∆ 48 (CDES90) value from the value of DHC2-8, this minor difference is likely a result of statistical uncertainty, that is, arising from the 5% probability that both samples have indistinguishable ∆ 48 (CDES90) , especially because the ∆ 47 (CDES90) values of these two samples are indistinguishable from each other; Figure 2). However, unlike DHC2-8, all other samples plot slightly above the theoretical ∆ 47 (CDES90) versus ∆ 48 (CDES90) equilibrium line (Figure 2). If kinetics prevailed, these would be associated with CO 2 (aq) degassing from the groundwater rather than CO 2 absorption (Kluge et al., 2014). Theoretical modeling and experimental results indicate that CO 2 (aq) degassing simultaneously shifts ∆ 47 values toward lower and ∆ 48 values toward higher values with increasing growth rates (Bajnai et al., 2020;Guo, 2020;Guo & Zhou, 2019). Therefore, if involved, CO 2 (aq) degassing kinetics should cause departures below the equilibrium line, contrary to what is observed ( Figure 2). This conundrum implies that the theoretical calibration overestimates the equilibrium ∆ 48 (CDES90) values. Unless the thermodynamic equilibrium ∆ 47 (CDES90) and ∆ 48 (CDES90) values at 33.7°C are accurately determined, it is impossible to provide definitive evidence that DHC2-8 and all other samples attained full internal equilibrium at 33.7°C.
If temperature variations have occurred in the Devils Hole caves, one could expect them between peak glacial and interglacial periods. For this reason, we investigated samples DH-11 44.5 and DH-11 189.9, representing the maximum and the minimum of the Devils Hole δ 18 O cc time series, that is, periods of interglacial and glacial maxima, respectively (Figure 3a). Despite the presumably maximum surface temperature spread, the ∆ 47 (CDES90) and the ∆ 48 (CDES90) values of these two samples remain indistinguishable ( Figure 2, Table 1). The lack of systematic variation in the measured ∆ 48 (CDES90) and ∆ 47 (CDES90) values, together with the observation that all samples are indistinguishable from each other in ∆ 47 (CDES90) versus ∆ 48 (CDES90) space, strongly implies that reaction kinetics related to CO 2 (aq) degassing-if having been effective at all-were invariant during the formation of the half-million-year-long Devils Hole record. An alternative scenario would require that temperature and kinetic effects compensated each other. Considering the effects of temperature and CO 2 (aq) degassing kinetics on ∆ 47 (CDES90) , ∆ 48 (CDES90) , and δ 18 O cc , one can demonstrate that such a scenario is unreasonable. For example, a 2°C decrease in temperature, that is, from 34°C to 32°C, would correspond with shifts of +0.005‰ in ∆ 47 (CDES90) and +0.002‰ in ∆ 48 (CDES90) based on the theoretical calibration (Bajnai et al., 2020;Hill et al., 2014), and with a shift of +0.38‰ in δ 18 O cc based on Coplen (2007 (Coplen, 2007). The timing of Termination II, that is, the penultimate glacial/interglacial transition, occurs at 142 ka in Devils Hole δ 18 O cc record (black arrow) (Winograd et al., 2006) and at 130 ka in the LR04 benthic δ 18 O stack (red arrow) (Lisiecki & Raymo, 2005).
Because all data points are indistinguishable, one can determine the 95% confidence interval of the mean ∆ 47 (CDES90) values of the samples: ±0.003‰. This ∆ 47 (CDES90) uncertainty translates into a maximum temperature variability of ±1°C around 33.7°C, independent if the experimental (Petersen et al., 2019) or the theoretical calibration (Bajnai et al., 2020;Hill et al., 2014) is used (Figure 3b). During inorganic carbonate precipitation, it is reasonable to expect that the clumped and the oxygen isotope compositions of the carbonates are affected by identical kinetic processes. In the case of Devils Hole, the constant carbonate precipitation environment suggested by the clumped isotope data implies that the δ 18 O cc values of the samples are similarly not affected by kinetic variations. The analytical uncertainty of the δ 18 O cc measurements of the Devils Hole record is ±0.07‰ (Winograd et al., 2006), and the temperature sensitivity of the oxygen isotope fractionation between calcite and water is 0.19‰°C −1 around 30°C, independent of the calibration of choice (Coplen, 2007). Consequently, the observed constant temperatures attest to an extremely stable subaqueous environment and that δ 18 O cc variations exceeding 0.20‰ (as represented by the root of the square sum of the analytical reproducibility and temperature sensitivity) reflect changes in δ 18 O gw , not temperature ( Figure 3c).
This enables one to examine if the maximum possible variability in the groundwater temperature, that is, ±1°C, could contribute to a debated feature of the Devils Hole record, namely that it shows ice-age terminations preceding other archives by ca. 10 ka (Moseley et al., 2016;Winograd et al., 1988Winograd et al., , 1992Winograd et al., , 2006. Specifically, Termination II occurs at 142 ka in the Devils Hole record (Winograd et al., , 2006 and at 130 ka in the LR04 composite benthic δ 18 O record (Lisiecki & Raymo, 2005), which is primarily a proxy of global (deep) seawater temperature and ice volume. The reconstructed Devils Hole δ 18 O gw value at 142 ka is −13.55 (±0.20)‰ and at 130 ka is −12.55 (±0.20)‰ (Figure 3c). The observed change in δ 18 O gw between these two dates, that is, 1.00‰, significantly exceeds the total uncertainty of 0.40‰ associated with δ 18 O gw determinations. Therefore, variations in kinetic fractionation or groundwater temperature cannot be the reason for the particular timing of the ice-age termination shown by the Devils Hole record.
In conclusion, we have demonstrated using dual clumped isotope thermometry that neither variations in temperatures nor variations in reaction kinetics influenced the Devils Hole calcite δ 18 O time series to any significant degree. Combined ∆ 47 and ∆ 48 measurements enable the characterization of the carbonate precipitation environments for speleothems and other climate archives and, therefore, strengthen the interpretation of the recorded climatic information. Slowly precipitating subaqueous calcites may accurately record changes in groundwater isotope compositions even during glacial-interglacial cycles.

Data Availability Statement
All study data are deposited at https://zenodo.org/record/4671734