Introduction
Permafrost, defined as Earth materials that remain below 0°C for more than two consecutive years (French, Reference French2017), is an essential component of cold-region environments and a key indicator of past and present climate. Understanding the former extent and timing of permafrost presence and absence is critical for reconstructing past climates and predicting future feedbacks between permafrost and climate change (Jones et al., Reference Jones, Grosse, Treat, Turetsky, Anthony and Brosius2023). Although instrumental records provide valuable insights into recent permafrost dynamics (Biskaborn et al., Reference Biskaborn, Smith, Noetzli, Matthes, Vieira, Streletskiy and Schoeneich2019), they lack the temporal depth necessary to capture long-term permafrost–climate interactions. Thus, geological proxies are indispensable for extending permafrost records beyond observational timescales. Among these, speleothems have proven to be particularly valuable archives (Vaks et al., Reference Vaks, Gutareva, Breitenbach, Avirmed, Mason, Thomas, Osinzev, Kononov and Henderson2013).
Speleothems, secondary carbonate formations in caves, grow when liquid water percolates through the overlying rock (Fairchild and Baker, Reference Fairchild and Baker2012). However, when permafrost is present, infiltrating water is blocked from reaching cave passages, preventing speleothem growth (Atkinson et al., Reference Atkinson, Harmon, Smart and Waltham1978). Consequently, growth intervals in a speleothem record from a cold environment typically correspond to periods of permafrost absence, while hiatuses indicate permafrost presence (Atkinson et al., Reference Atkinson, Lawson, Smart, Harmon and Hess1986). This relationship has enabled researchers to use speleothems to reconstruct permafrost evolution over orbital timescales. For instance, Vaks et al. (Reference Vaks, Gutareva, Breitenbach, Avirmed, Mason, Thomas, Osinzev, Kononov and Henderson2013) compiled a 500,000 year record of permafrost extent in Siberia, showing that permafrost retreated significantly during interglacial periods, with Marine Isotope Stage (MIS) 11 marking a major thaw. Similarly, speleothems from subarctic Canada demonstrate that past episodes of permafrost thaw closely followed global temperature increases (Batchelor et al., Reference Batchelor, McGee, Shakun, Woodhead, Jost, Arnold, Horne, Kinsley and Freudenburg-Puricelli2024). In a longitudinal transect spanning western Canada, speleothems record permafrost degradation extending into the Arctic during warm interglaciations (Biller-Celander et al., Reference Biller-Celander, Shakun, McGee, Wong, Reyes, Hardt, Tal, Ford and Lauriol2021). Additional studies in Britain and Ireland have revealed that speleothem growth corresponded to interglacial intervals (Atkinson et al., Reference Atkinson, Lawson, Smart, Harmon and Hess1986; Breitenbach et al., Reference Breitenbach, Lawson, Atkinson, Fox, Kwiecien and Henderson2024; Panitz et al., Reference Panitz, Rogerson, Longman, Scroxton, Lawson, Atkinson and Ersek2025), reinforcing the utility of speleothems as indicators of permafrost history.
In contrast to this extensive research on permafrost at high latitudes and in areas impacted by former continental ice sheets, the history of mountain permafrost remains relatively understudied. This situation constitutes an important data gap, because mountain permafrost is particularly sensitive to climate change due to steep temperature gradients and complex topographic influences (Haeberli et al., Reference Haeberli, Noetzli, Arenson, Delaloye, Gärtner-Roer, Gruber, Isaksen, Kneisel, Krautblatter and Phillips2010). Thawing of mountain permafrost has been linked to increased rockfall activity, landslides, and alterations in groundwater flow, which can significantly impact both natural and human systems (Kääb et al., Reference Kääb, Reynolds, Haeberli, Huber, Bugmann and Reasoner2005). Additionally, mountain regions serve as critical water towers, providing freshwater to downstream populations (Viviroli et al., Reference Viviroli, Dürr, Messerli, Meybeck and Weingartner2007), making the study of past mountain permafrost dynamics important for predicting future water availability (Jones et al., Reference Jones, Harrison, Anderson and Whalley2019).
Here, we identify speleothem growth intervals from mountain caves in northern Utah (USA) in order to reconstruct past permafrost conditions in part of the Rocky Mountains. By extending the use of speleothems as paleo-permafrost indicators to mountainous terrains, we provide a novel perspective on permafrost–climate interactions in upland landscapes. Our approach enhances the spatial coverage of paleoclimate reconstructions, providing critical data for refining climate models and improving our understanding of permafrost dynamics in complex upland landscapes. Ultimately, this research contributes to a more comprehensive understanding of how permafrost has responded to past climate changes, which is essential for anticipating its future trajectory in a warming world.
Methods
This study focused on the Uinta Mountains (UM; Fig. 1A), a Laramide-age uplift of Precambrian metasedimentary rocks surrounded by hogbacks of Paleozoic carbonates (Williams, Reference Williams1943; Bradley, Reference Bradley1995). Maximum elevations in the UM are in excess of 4 km above sea level and the range was extensively glaciated during the Pleistocene (Munroe and Laabs, Reference Munroe and Laabs2009), although no glaciers remain today. Local climate records suggest that the modern 0°C isotherm is at an elevation of ∼3300 m; a weather station at 3695 m records a mean annual air temperature of −2.0°C (Munroe, Reference Munroe2006).
(A) Location of Whiterocks and Chepeta Caves in the Uinta Mountains (UM) of northeastern Utah. Other sites mentioned in the text include Bear Lake (BL; Kaufman et al., Reference Kaufman, Bright, Dean, Rosenbaum, Moser, Anderson, Colman, Rosenbaum and Kaufman2009), Great Salt Lake (GS; Oviatt et al., Reference Oviatt, Thompson, Kaufman, Bright and Forester1999; Balch et al., Reference Balch, Cohen, Schnurrenberger, Haskell, Valero Garces, Beck, Cheng and Edwards2005), Leviathan Cave; LC (Lachniet, Reference Lachniet, Oviatt and Shroder2016), Devils Hole (DH; Wendt et al., Reference Wendt, Dublyansky, Moseley, Edwards, Cheng and Spötl2018), Stoneman Lake (SL; Staley et al., Reference Staley, Fawcett, Anderson and Jiménez-Moreno2021), and locations of glacial moraines dated to MIS 6 at Fish Lake (FL; Marchetti et al., Reference Marchetti, Harris, Bailey, Cerling and Bergman2011), and in Wyoming (WY; Dahms et al., Reference Dahms, Egli, Fabel, Harbor, Brandová, de Castro Portes and Christl2018) and Colorado (CO; Laabs et al., Reference Laabs, Licciardi, Leonard, Munroe and Marchetti2020). (B) Entrance of Whiterocks Cave (red arrow) in a cliff of Madison Limestone above the Whiterocks River. Cliff is ∼300 m tall. (C) Photographs of speleothem sample WRC-C (WRC-17-C) in the cave (left) and after being cut and polished (right). Numerous growth hiatuses are visible in the polished sample. (D) Photograph of speleothem WRC-D (WRC-17-D) in the cave (left) and cut and polished (right). Note the growth hiatuses visible as lighter-colored bands.
The solution caves we investigated are developed in Carboniferous limestone on the south slope of the UM (Fig. 1B). Whiterocks Cave has a mapped length of 800 m and a vertical extent of 20 m, with an entrance at 2600 m above sea level. The smaller, deeper Chepeta Cave is developed in the same rock unit at 2600 m ∼4 km to the west. Air temperatures in Whiterocks Cave average 4.5°C a few hundred meters from the large entrance where a seasonal signal with an amplitude of ∼1.5°C is apparent (Supplementary Fig. 1). Farther into the cave, beyond a pair of constrictions, air temperature averages 5.5°C, seasonal cycles span <1°C, and relative humidity is steady at 100% (Supplementary Fig. 1). In contrast, in Chepeta Cave, which has a very narrow entrance, air temperatures are nearly stable, averaging 4.2°C at a point midway through the known vertical extent of the cave system and 4.0°C in the lowest known level (Supplementary Fig. 1). Modern vegetation above both caves is a sparse forest of Douglas fir (Pseudotsuga menziesi) and limber pine (Pinus flexilis) (Fig. 1B). Soils are thin, rocky, and poorly developed. Neither cave was impacted by Pleistocene glaciers (Munroe and Laabs, Reference Munroe and Laabs2009). Mean annual temperature outside the caves is ∼3.5°C with a mean annual precipitation of ∼600 mm (PRISM Climate Group, 2023).
With permission from the U.S. Forest Service, top and bottom subsamples were drilled in situ from stalagmites in both caves to determine their suitability for dating and the time period they represent. One sample was also drilled from a layer of flowstone. Speleothems demonstrated to be datable were collected during subsequent fieldwork. Preference was given to removing features that were already broken by seismic activity, previous cave visitors, or other disturbance. In the laboratory, samples were cut lengthwise and polished to reveal their internal structure. Nearly all contained prominent bands of contrasting color inferred to represent growth hiatuses (Fig. 1C and D).
Dating with U-Th focused on bracketing these bands to constrain intervals of active speleothem growth. For each U-Th age, 50–200 mg of calcite powder was drilled from a trench parallel to the growth layers on the stalagmite. The 238U concentrations in the samples averaged ∼330 ng/g with initial δ234U up to ∼3100‰. U-Th dating was performed at the Massachusetts Institute of Technology with methods identical to those described in Biller-Celander et al. (Reference Biller-Celander, Shakun, McGee, Wong, Reyes, Hardt, Tal, Ford and Lauriol2021). These include dissolution of the calcite powders in nitric acid and equilibration with a 229Th-233U-236U spike, precipitation of iron oxyhydroxides to scavenge U and Th from solution, separation of U and Th via ion exchange chromatography, and measurement of U and Th isotope ratios on a Nu Plasma II-ES multi-collector ICP-MS equipped with an Aridus III desolvating nebulizer. Data correction and age calculation were also identical to that described in Biller-Celander et al. (Reference Biller-Celander, Shakun, McGee, Wong, Reyes, Hardt, Tal, Ford and Lauriol2021), including use of 4.4 (± 2.2) × 10−6 as the initial atomic 230Th/232Th ratio.
Results
We generated 108 age determinations distributed across 11 speleothems from Whiterocks Cave and 3 from Chepeta Cave (Fig. 2, Supplementary Figs. 2–14, Supplementary Table 1). The ages span from 631 ± 84 ka to 79 ± 0.3 ka, encompassing ∼550,000 years of paleoenvironmental history. Two-sigma errors on these age determinations are generally less than 2%, >95% of them are less than 10%. The lack of ages younger than 79 ka likely reflects our decision to avoid sampling features that were wet and possibly actively growing. The older limit to the age distribution is the limit of U-Th dating; 12 additional subsamples yielded nonfinite ages beyond this threshold. Although the oldest ages are imprecise, we present them here for completeness (Fig. 2).
Composite dataset of U-Th ages for Whiterocks (WRC) and Chepeta Caves. Ages are presented with 2-sigma uncertainties. Two ages in WRC-4 and WRC-8 were excluded, marked by white X. A kernel density plot (KDE) for the ages is also presented to highlight the strong clustering aligning with interglacial Marine Isotope Stages (MIS). Color shading highlights the clustering of U-Th ages. The two plots at the bottom present June insolation at 60°N (Berger, Reference Berger1978) and a benthic δ18O stack as a proxy for global ice volume (Lisiecki and Raymo, Reference Lisiecki and Raymo2005).
Whiterocks Cave provides the majority of the chronological data, with 89 dates from 11 speleothems (WRC-A, WRC-B, WRC-C, WRC-D, WRC-2, WRC-4, WRC-5, WRC-8, WRC-9, WRC-10, and WRC-11) spanning the entire temporal range of our study. In contrast, Chepeta Cave, which was added later to corroborate the results from Whiterocks, is represented by 19 dates from 3 speleothems (Chepeta-1, Chepeta-2, and Chepeta-3), with ages ranging from 555 to 128 ka (Supplementary Table 1).
Individual speleothems reveal complex patterns of deposition, but sequential sampling along growth axes reveals consistent age progressions. A few features exhibit growth over relatively narrow time windows, for instance WRC-B, which was active for 10,000 years between 127.5 ka and 117.8 ka. However, top and bottom ages demonstrate that most speleothems record intermittent growth over hundreds of thousands of years. For example, WRC-C spans ∼395,000 years (631 to 236 ka), WRC-8 encompasses 419,000 years (541 to 122 ka), and Chepeta-1 extends 397,000 years (525 to 128 ka). These extraordinarily long histories are punctuated by substantial growth pauses responsible for the banding visible when the speleothems were cut and polished.
Analysis of the full set of 108 U-Th ages using kernel density estimation (Weij et al., Reference Weij, Woodhead, Hellstrom and Sniderman2020) with a bandwidth of 10,000 years reveals a pronounced clustering (Fig. 2). The strongest peak is at 328 ka, followed by 123 ka, 238 ka, and 392 ka, with a fifth broad peak centered on 514 ka.
Discussion
Correlation with the global marine isotope record
Our dataset of U-Th ages for speleothems from Whiterocks and Chepeta Caves exhibits clear temporal patterns in growth, with distinct growth intervals and hiatuses in both caves that correlate strikingly with glacial–interglacial cycles recorded in the global marine isotope record (Lisiecki and Raymo, Reference Lisiecki and Raymo2005; Railsback et al., Reference Railsback, Gibbard, Head, Voarintsoa and Toucanne2015). The youngest cluster of ages (ca. 123 ka) aligns strongly with the peak of the last interglaciation, MIS 5e (Fig. 2). The second-youngest cluster (ca. 238 ka) aligns with MIS 7e at 240 ka. The cluster centered at 328 ka, represented by the largest number of speleothems from both caves, corresponds with the peak of MIS 9e. The slightly older cluster at 392 ka aligns with MIS 11c, one of the longest and warmest interglaciations of the past 500,000 years, which featured a broad peak around 400 ka. Finally, despite the wide error bars on the oldest U-Th ages, the broad cluster of ages centered on 514 ka aligns with MIS 13, and ages from near the base of speleothem WRC-C correspond with MIS 15 (Fig. 2).
Equally notable is the alignment of significant growth hiatuses with major glaciations in the benthic isotope record. Of the 108 U-Th ages, only 11 (10%) fall outside interglacial periods or fail to overlap with interglaciations at 2-sigma (Fig. 2). Five of these (from WRC-C, WRC-5, and Chepeta-1) date to ∼280–300 ka, aligning with the prominent MIS 9a interstadial (Railsback et al., Reference Railsback, Gibbard, Head, Voarintsoa and Toucanne2015). Two ages from WRC-D and WRC-5 fall within the MIS 10 glaciation, and two from WRC-5 and Chepeta-1 are during MIS 6, but all are from times when summer insolation was relatively high (Fig. 2). The MIS 6 age from Chepeta-1 is also from the base of a short stalagmite (<4 cm) growing off the side of the main feature, suggesting that this temporary resumption of growth was accompanied by a change in the position of the drip on the cave ceiling, which may reflect opening of a different pathway for infiltrating water during this window of permafrost thaw. In the remaining two cases, samples drilled near the tops of stalagmites yielded anomalously young ages, which we exclude from further analysis. One from the top of WRC-8 dated to late in MIS 8 (260 ka), but when re-drilled and analyzed again, yielded an age (328 ka) matching the peak of MIS 9 (Fig. 2). The age of this rerun sample also agreed within error with an age 0.3 mm below, so we consider the original 260 ka result to be an outlier. Similarly, a sample from WRC-4 returned an age late in MIS 10 (349 ka); however, when drilled 0.3 mm lower and rerun, that age shifted to MIS 11 (394 ka). In both cases, we suspect that the anomalously young ages reflect inclusion of younger calcite precipitated on the stalagmite top or in cracks and pore spaces just beneath the surface. The samples yielding the anomalous ages also have substantially higher 232Th concentrations, suggesting inclusion of material from the stalagmite top or from cracks.
As a check on the correlation of speleothem growth with interglacial conditions, an independent sample t-test was used to evaluate whether marine δ18O values were significantly lower at the times corresponding to the U-Th ages relative to average values over the past 632,000 years. Bootstrapping with 10,000 iterations was employed to accommodate the uncertainty associated with the speleothem ages, and a P value of 0.05 was used as the significance threshold. The mean value of δ18O corresponding to the speleothem ages is 3.9‰, compared with the long-term (632,000 year) average of 4.1‰ (Lisiecki and Raymo, Reference Lisiecki and Raymo2005). This difference is significant (P <0.001) confirming that the speleothems in our study grew preferentially when global ice volume was below the average for the past ∼630 ka. A marine δ18O value of 4‰ appears to represent the threshold below which speleothem growth in these caves is possible, which is consistent with the mean marine δ18O value of 3.8‰ for interglaciations (MIS 1, 5, 7, 9, 11, 13, and 15) versus 4.5‰ for glaciations (MIS 2, 6, 8, 10, 12, and 14). An identical analysis reveals that mean values of June insolation values at 60°N (Berger, Reference Berger1978), when lagged 4500 years to account for the timescale of amplification by greenhouse gases (Ruddiman, Reference Ruddiman2006), are significantly higher (P < 0.001) at times corresponding to speleothem ages relative to the long-term average (503 vs. 493 W/m2).
Mechanism
Our data indicate that conditions suitable for speleothem growth in Whiterocks and Chepeta Caves were uncommon during the past 600,000 years; for most of this time, speleothems were inactive. Moreover, speleothem growth occurred overwhelmingly during interglaciations, meaning that the mechanism responsible for cycling speleothem growth on and off reflects a local response to global glacial–interglacial fluctuations.
Several possible explanations can be considered for why glacial–interglacial cycles could be recorded by speleothem growth intervals in these caves. Reduction in cave ventilation driven by physical reconfiguration of connections to the surface could cause pCO2 in the cave atmosphere to rise to the point where carbonate precipitation is inhibited and speleothem growth would cease (James et al., Reference James, Banner and Hardt2015). However, the same pattern of growth intervals is seen in both Whiterocks and Chepeta Caves, despite their different geometries, and it seems unlikely that ventilation was synchronously and repeatedly restricted in both caves, which were not impacted by alpine glaciation. Similarly, variation in the ability of the groundwater to dissolve carbonate bedrock could be driven by changes in the abundance of vegetation and soil above the caves (Dreybrodt, Reference Dreybrodt1999), with reduced vegetation and soil cover (and therefore reduced acidity of infiltrating soil water) during colder intervals leading to a cessation of speleothem growth. Records of past vegetation (or soil) in the vicinity of these caves over long timescales are not available. However, even though modern vegetation above these caves is sparse and soils are thin (Fig. 1B), stalactites broken by visitors over the past few decades are growing new soda straws, indicating that bedrock dissolution and speleothem growth are possible at these sites even with minimal vegetation and soil. Decreased precipitation during glacial intervals could conceivably have reduced infiltration to the extent that water failed to reach the caves. However, proxy evidence and modeling suggest that effective moisture (e.g., the balance between precipitation and evaporation) in this region during at least the last glacial maximum was somewhat greater than modern (Lora and Ibarra, Reference Lora and Ibarra2019; Oster and Ibarra, Reference Oster and Ibarra2018), and it is reasonable to assume similar conditions during older glaciations. Thus, a mechanism other than aridity is needed to explain cessation of speleothem growth.
Considering the strong synchrony of growth intervals in both caves, our interpretation is that permafrost inhibited water infiltration for much of the past ∼600,000 years. Only under peak interglacial conditions is this permafrost eliminated, allowing water to reach the cave passages and speleothem growth to resume. Numerical modeling indicates that >500 km2 of the modern Uinta landscape, including locations within 12 km of Whiterocks and Chepeta Caves, has a >50% chance of permafrost occurrence (Obu et al., Reference Obu, Westermann, Bartsch, Berdnikov, Christiansen, Dashtseren, Delaloye, Elberling, Etzelmüller and Kholodov2019). Smaller patterned ground features, including non-sorted circles within this area of likely permafrost occurrence, appear active today (Fig. 3), and extensive permafrost was certainly present at higher elevations in the UM in the past, as demonstrated by relict large-scale patterned ground and soil evidence for pervasive cryoturbation (Munroe, Reference Munroe2006, Reference Munroe2007). Therefore, permafrost is present in the UM today and clearly covered a much wider area in the past.
Evidence for past and present permafrost at high elevations in the Uinta Mountains (UM). (A) An aerial view of polygonal sorted patterned ground on a broad ridgeline above modern tree line. Polygon diameters are generally 5–8 m. (B) A vertical aerial view of sorted stone polygons. Width of polygons is ∼5 m. (C) Ground-level view of modern unvegetated active non-sorted circles (highlighted by dashed oval). Knife for scale is ∼20 cm long. (D) Soil pit (50 cm deep) with irregular horizon boundaries testifying to strong cryoturbation that moved dark surface soil downward and brought light-colored, organic-poor subsoil toward the surface.
In gridded climate data for a ∼2000 km2 area centered on the caves, spanning an elevation range from 1550 m to 3860 m, the modern lapse rate for mean annual air temperature (MAAT) is 5.0°C/km (PRISM Climate Group, 2023). These data indicate that the modern 0°C isotherm is at an elevation of ∼3300 m, whereas the MAAT on the ridgeline above the caves (∼2800 m) is ∼2.5°C (Fig. 4A). A depression of MAAT of >2.5°C would, therefore, lower the 0°C isotherm to intersect the ground surface above the caves (Fig. 4B), whereas it is currently located at a higher elevation (Fig. 4C). Attempts to derive a direct link between MAAT and the occurrence of mountain permafrost have been confounded by the wide array of factors influencing ground temperature, including topographic aspect, snow depth, and vegetation cover (Guodong and Dramis, Reference Guodong and Dramis1992). Nonetheless, the position of the 0°C isotherm is often considered a first-order approximation of the permafrost limit in mountain landscapes. Support for this assertion comes from modeling, where the average elevation of locations in the UM with a 50% chance of modern permafrost is 3465 m, corresponding to a MAAT of approximately −0.7°C (Obu et al., Reference Obu, Westermann, Bartsch, Berdnikov, Christiansen, Dashtseren, Delaloye, Elberling, Etzelmüller and Kholodov2019). Additional modeling suggests that the ridgelines above Whiterocks and Chepeta Caves have a modern mean annual ground temperature (MAGT) of ∼5°C (Ran et al., Reference Ran, Li, Cheng, Che, Aalto, Karjalainen and Hjort2022), which is consistent with the air temperatures measured within the caves (Supplementary Fig. 1). From the available evidence, it is reasonable to conclude that a MAAT depression of ∼3°C to 5°C would be sufficient to form permafrost above Whiterocks and Chepeta Caves (Fig. 4B). For comparison, a coupled 2D mass balance and ice-flow numerical modeling approach concluded that air temperatures in the UM during the last glacial maximum (MIS 2) were depressed 4–6°C (Laabs et al., Reference Laabs, Plummer and Mickelson2006). Similarly, a paleoclimate estimate based on glacial equilibrium line reconstructions suggested a temperature depression of 5.5°C (Munroe and Mickelson, Reference Munroe and Mickelson2002). Collectively, modern observations of small-scale active patterned ground and the distribution of extensive relict patterned ground, combined with modeling of permafrost likelihood, modern MAGT, and glacial paleoclimate conditions, strongly support the interpretation that permafrost formed above Whiterocks and Chepeta Caves during glacial intervals, causing speleothem growth to cease.
Inferred mechanism for climatic cycling of speleothem growth intervals in Uinta Mountain caves. (A) Modern relationship between elevation and mean annual air temperature (MAAT) in the region surrounding the caves (PRISM Climate Group, 2023). The lapse rate is 5°C/km, and the modern 0°C isotherm is at an elevation of ∼3300 m. The ridgeline above the caves at ∼2800 m likely has a modern MAAT of ∼2.5°C. (B) A depression of MAAT greater than ∼3°C would drive formation of permafrost in the epikarst above the caves, cutting off water infiltration and stopping speleothem growth. (C) Warming that raises the 0°C isotherm and degrades permafrost above the caves would permit renewed water infiltration and renewed carbonate precipitation.
Regional comparison
Long climate records overlapping with the time period represented by our data from Whiterocks and Chepeta Caves are rare; however, there are some important points of reference (Supplementary Fig. 15). For example, Bear Lake, located approximately 200 km to the northwest of our study area, was apparently low and disconnected from the Uinta-draining Bear River during the interglacial conditions of MIS 7 and MIS 5 (Kaufman et al., Reference Kaufman, Bright, Dean, Rosenbaum, Moser, Anderson, Colman, Rosenbaum and Kaufman2009). In contrast, the lake level was high during the MIS 6 glaciation (Kaufman et al., Reference Kaufman, Bright, Dean, Rosenbaum, Moser, Anderson, Colman, Rosenbaum and Kaufman2009). Although the record from Bear Lake is difficult to interpret due to its active tectonic setting and a rudimentary age model, low stands in lake record appear to correspond with times of speleothem growth in our study, consistent with warmer conditions. Similarly, the sedimentary record from Great Salt Lake in northwest Utah indicates high water during MIS 6 as well as during MIS 12 (Oviatt et al., Reference Oviatt, Thompson, Kaufman, Bright and Forester1999; Balch et al., Reference Balch, Cohen, Schnurrenberger, Haskell, Valero Garces, Beck, Cheng and Edwards2005), corresponding with times when speleothem growth in Whiterocks and Chepeta Caves stopped. Numerical modeling suggests that episodes of high lake level in the Great Salt Lake basin, reflecting greater effective moisture, were driven primarily by depressed summer temperatures that reduced evaporation (Belanger et al., Reference Belanger, Amidon, Laabs, Munroe and Quirk2022). Regional records of alpine glaciation during MIS 6 are fragmentary, but cosmogenic surface-exposure dating of moraines in southern Utah (Marchetti et al., Reference Marchetti, Harris, Bailey, Cerling and Bergman2011) and in neighboring Wyoming and Colorado (Dahms et al., Reference Dahms, Egli, Fabel, Harbor, Brandová, de Castro Portes and Christl2018; Laabs et al., Reference Laabs, Licciardi, Leonard, Munroe and Marchetti2020) indicates an advance culminating in MIS 6, when speleothems in our study were not growing.
Our interpretations from Whiterocks and Chepeta Caves are also consistent with long cave-based paleoclimate records from the region (Supplementary Fig. 15). For instance, to the west in Nevada, the composite Leviathan chronology documents warm conditions during MIS 5e, when speleothems were active in the caves we studied (Lachniet et al., Reference Lachniet, Denniston, Asmerom and Polyak2014; Lachniet, Reference Lachniet, Oviatt and Shroder2016). Farther to the southwest, the water table in Devils Hole, Nevada, was higher during glacial intervals, reflecting a cooler and wetter climate when speleothems in our study were dormant (Wendt et al., Reference Wendt, Dublyansky, Moseley, Edwards, Cheng and Spötl2018; Jackson et al., Reference Jackson, Steidle, Wendt, Dublyansky, Edwards and Spötl2023; Steidle et al., Reference Steidle, Wendt, Dublyansky, Edwards, Li, McClure, Moseley and Spötl2024). Conversely, the water table at Devils Hole dropped synchronously with the onset of interglacial conditions, consistent with a warming climate that facilitated reactivation of speleothems in Whiterocks and Chepeta Caves associated with the loss of permafrost.
Finally, the uniquely long sedimentary archive from Stoneman Lake, Arizona (Supplementary Fig. 15) records the presence of a freshwater lake during MIS 2, 6, 10, and 12, implying cooler conditions with greater effective moisture (Staley et al., Reference Staley, Fawcett, Anderson and Jiménez-Moreno2021). In contrast, dry conditions prevailed during MIS 5, 7, 9, 11, and 13 when speleothems were active in the caves we studied (Staley et al., Reference Staley, Fawcett, Anderson and Jiménez-Moreno2021).
Together, this array of regional records corroborates the speleothem evidence for permafrost-controlled infiltration changes in the UM. Speleothem dormancy in Whiterocks and Chepeta Caves aligns with cooler, glacial conditions across the interior western United States, whereas intervals of speleothem growth consistently correspond with evidence for warmer temperatures during interglaciations. This coherence across diverse proxies supports the interpretation that the timing of speleothem growth reflects large-scale climatic controls on mountain permafrost, not local anomalies unique to these two caves, and underscores the broader regional significance of the patterns observed in our dataset.
Implications
Our results indicate that for most of the last 600,000 years, permafrost was widespread in the UM at elevations 500 m or more below its modern, restricted extent. The pattern of U-Th ages supporting this interpretation, consistent across multiple speleothems from two separate caves, demonstrates that permafrost in this region was a persistent barrier to infiltration during cooler periods, despite greater effective moisture at those times. Conversely, the clear temporal correspondence between speleothem growth and warm intervals in the marine isotope and insolation records (Fig. 2) indicates that mountain permafrost responds sensitively to orbital-scale climate variability and has undergone dramatic fluctuations in extent over glacial–interglacial cycles.
Our findings are relevant to understanding the stability of mountain permafrost under ongoing and future warming. Whereas most paleo-permafrost studies have focused on high-latitude regions (Vaks et al., Reference Vaks, Gutareva, Breitenbach, Avirmed, Mason, Thomas, Osinzev, Kononov and Henderson2013) or areas near former continental ice sheets (Batchelor et al., Reference Batchelor, Orland, Marcott, Slaughter, Edwards, Zhang, Li and Cheng2019; Biller-Celander et al., Reference Biller-Celander, Shakun, McGee, Wong, Reyes, Hardt, Tal, Ford and Lauriol2021; Breitenbach et al., Reference Breitenbach, Lawson, Atkinson, Fox, Kwiecien and Henderson2024; Panitz et al., Reference Panitz, Rogerson, Longman, Scroxton, Lawson, Atkinson and Ersek2025), our work shows that speleothems can also serve as effective archives of past permafrost dynamics in alpine environments. As warming accelerates, the potential for widespread thaw of mountain permafrost poses significant risks to slope stability, groundwater systems, and seasonal hydrology in high-elevation landscapes (Knight and Harrison, Reference Knight and Harrison2023). By providing a long-term record of thermal and hydrological shifts in the subsurface, our study highlights the climatic thresholds that must be crossed to eliminate permafrost and reinitiate infiltration, a process already underway in many mountain regions today (Mollaret et al., Reference Mollaret, Hilbich, Pellet, Flores-Orozco, Delaloye and Hauck2019). This approach can be extended to other carbonate terranes worldwide, offering a novel perspective on cryosphere history in mountainous settings, and opening another frontier for studying permafrost–climate feedbacks in complex topography (Haeberli, Reference Haeberli2013).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/qua.2026.10076
Acknowledgments
The authors thank G. Carling, E. Norris, and C. Plunkett for field assistance, and S. Burns for assistance in the laboratory. JM was supported by an award from Fulbright Austria in 2018 and a Guest Professorship from the University of Innsbruck in 2024. Comments from L. Anderson, M. Rogerson, and an anonymous reviewer were very helpful in improving the article.
Conflict of interest statement
The authors have no competing interests to declare.
