Central to the field component of this project was deployment of five automated samplers to collect time series of water samples. The samplers contain a carousel of 24, 500-mL bottles, a pump and distributor arm, and a control module (Figure 2). A solar panel was used to maintain charge on a 12 V battery to run each sampler continuously through the melt season. Sample bottles were modified following published methodology (von Freyberg et al., 2020) to reduce the potential for evaporation to fractionate stable isotopes within the sampler. Samplers were deployed in positions higher than their intake to permit free draining of the intake line, and the sampling inlet was wrapped in 100-µm nylon mesh to prevent coarse material from clogging the pump. All samplers were programmed to collect a 45-mL sample at midnight and noon each day. For 3 days these samples (six samples total) were composited in a single bottle. The carousel of 24 bottles allowed each sampler to run unattended for 72 days, collecting long series of composited samples spanning the melt season.

In the Uinta Mountains (Figure 1B), two samplers were installed at springs discharging from the termini of rock glaciers RG-1 and RG-2 (Figures 2A, B) that were the targets of previous study (Munroe, 2018; Munroe and Handwerger, 2022). RG-1 and RG-2 are tongue-shaped rock glaciers ∼600 m long and ∼100 m wide that are typical of cirque-floor features in these mountains. Both are defined by steep terminal and side slopes ∼20 m tall that exhibit a bilevel stratigraphy of finer material beneath a surface mantle of coarse quartzite boulders. Satellite InSAR (interferometric synthetic aperture radar) analysis indicates that these features move at velocities of ∼10 cm/yr during the summer (Brencher et al., 2021). The springs at the termini of these rock glaciers discharge water with a temperature <1°C consistently through the melt season.

In the La Sal Mountains, three water samplers were deployed (Figures 1, 2). One was positioned at a spring discharging directly from the terminus of the rock glacier in Red Snow Cirque (RSC, Figures 1C, 2C). A second sampler was installed ∼1 km downstream from the RSC sampler along the main stream draining Gold Basin (GB, Figures 1C, 2D). A third sampler was deployed on the Wet Fork of Mill Creek (WF Figures 1D, 2E), which drains a subalpine basin in which rock glaciers are absent. The GB and WF samplers were located at the points where a previous study measured discharge to investigate how the presence of rock glaciers impacts stream hydrographs (Geiger et al., 2014).

To complement the series of composite water samples, automated data loggers were installed at RG-2 in the Uinta Mountains and at RSC in the La Sal Mountains to measure water temperature and specific conductivity every hour for the duration of the water sampler deployment.

To provide information about the properties of precipitation near the water samplers, grab samples of snow were collected on the surfaces of RG-1, RG-2, and GB when the water samplers were deployed. Rain was collected during the deployment period at RG-2, RSC, GB, and WF using samplers designed to eliminate isotope fractionation due to evaporation (Gröning et al., 2012). These rain collectors were deployed within 10 m of the water samplers. Timing and magnitude of precipitation events are constrained by hourly data from automated SNOTEL (snowpack telemetry) sites in each study location. The Chepeta SNOTEL is located <10 km from RG-1 and RG-2 at a similar elevation (Figure 1B), and in the La Sal Mountains, the Gold Basin SNOTEL (GBS) is at a similar elevation <2 km from RSC and GB (Figure 1C).

Deep, long-lasting snowpack renders these locations inaccessible for much of the year, so the water samplers were installed in late June and early July 2022 (Table 1). All five were emptied in early September, and relaunched to run until mid-October, when they were emptied again and deactivated for the winter. The precipitation samplers were emptied in September and again in October. The dataloggers were downloaded in October. All samples for stable isotope analysis were transported from the field in new 15-mL centrifuge tubes, kept cool during shipping, and stored in a refrigerator immediately upon arrival in the laboratory.

In the Department of Earth and Climate Sciences at Middlebury College, samples for isotope analysis were filtered to 0.2 µm and stored in glass vials with Teflon-lined caps. Samples for ICP-MS analysis were vacuum filtered with Whatman Number 1 paper and acidified to pH 2 with trace-element grade HNO₃.

Stable isotope measurements were made with a Los Gatos 45-EP Triple Liquid Water Isotope Analyzer. Accuracy of the instrument is 0.4‰ for δD and 0.1‰ for δ¹⁸O. Samples were run against a bracketing set of 4 standards and calibrated with a cubic spline function relative to Standard Mean Ocean Water (SMOW). Each sample was analyzed 10 times, and the first 2 injections were discarded to minimize cross-over. Additional standards were run as unknowns as an internal check on the results. Results were compared with the Global Meteoric Water Line (GMWL) (Craig, 1961). Values of d-excess were calculated as d-excess = δD-(8 × δ¹⁸O) (Dansgaard, 1964).

Hydrochemical characterizations were made with a Thermo iCap ICP-MS. Samples were run against standards derived from NIST Standard Reference Material 1643f “Trace Elements in Water”. An in-house standard was used to determine the abundance of Si and Ti, which are not present in 1643f. Standards were run after every 10 unknowns and a linear correction was applied to compensate for instrument drift. Interpretation focused on elements that consistently exhibited concentrations >1 ppb.

Isotope data were calibrated using Los Gatos Research LWIA Post Analysis Software v.4.5.0.6. Principal component analysis in SPSS v.28.0 using a varimax rotation was employed to reduce the hydrochemical data. Datalogger datasets were filtered to average daily values in Hoboware Pro v.3.7 and exported as.csv files.

The time series of water samples collected from RG-1, RG-2, and RSC reveal that these systems evolve between different primary water sources as the melt season progresses. In early summer, water discharging from the rock glaciers has depleted isotope values consistent with a source dominated by melting snow (Table 2). This interpretation is logical given observations of lingering snow on the surface of the rock glaciers when the water samplers were deployed in late June/early July (Table 1), and similar situations have been reported for other rock glaciers (Krainer and Mostler, 2002). As the summer progresses this visible snow disappears, and a water source with less negative values of δ¹⁸O becomes more significant (Figure 3). Transient excursions of isotopes to higher values corresponding to major precipitation events are clear (Figures 4, 5), and are logical given the amount of rain and the less depleted signature of rain from the precipitation collectors (Figure 3). However, the spikes in δ¹⁸O corresponding to precipitation dissipate in <2 weeks, indicating relatively rapid drainage of water through the system (Krainer et al., 2007), and are superimposed on an overall trend of rising isotope values as the influence of snowmelt wanes (Krainer and Mostler, 2002). Electric conductivity, along with abundances of Ca and Mg, rises through the summer, which could indicate that water in late summer spends more time in contact with fresh weatherable minerals within the rock glacier, as has been noted in prior studies (Caine, 2010). Given the geochemical stability of the quartzite, these weatherable minerals may be sourced from the trace amounts of argillite present in the rock glacier catchment or from the deposition of eolian dust (Litaor, 2022), which is known to impact the Uintas (Munroe, 2014; Munroe, 2022). Ratios of Ca/Mg also shift during the melt season, dropping slightly in the Uinta rock glaciers and rising in the La Sals. This pattern requires that the sources of water are changing, not just their proportions. Additionally, the maintenance of cold (<1°C) temperatures in the rock glacier discharge indicates that these systems are thermally buffered throughout the summer, likely due to the presence of internal ice within the rock glacier (Berger et al., 2004; Brighenti et al., 2021a). High values of d-excess in the rock glacier water are consistent with the melting of ice that has undergone multiple freeze-thaw cycles (Steig et al., 1998; Williams et al., 2006). Collectively, these observations support the interpretation that water discharging from rock glacier springs in early summer is dominated by snowmelt, that internal ice becomes important as the melt season progresses, and that summer precipitation events drain rapidly through the system causing short-term changes in water characteristics that fade over the course of a few days to at most a few weeks.

During the melt season of 2022 the RG-1 and RG-2 water samplers were deployed in the same locations as for a previous study in 2021 (Munroe and Handwerger, 2022). Inter-comparison of these datasets, therefore, allows the interannual stability of the hydrochemical trends in rock glacier water to be considered (Figures 8, 9). Overall, δ¹⁸O and d-excess followed similar patterns in 2021 and 2022 (Figure 9). Values of both are low at the beginning of the melt season and exhibit a steady rise upon which transient increases related to precipitation are superimposed, similar to findings reported by previous studies (Krainer and Mostler, 2002; Krainer et al., 2007). Values of δ¹⁸O are similar to snow in late June and early July, and the tendency of values at RG-1 to be slightly higher than RG-2 is present in both years. Values of d-excess start close to 10 and rise to ≥16‰ in both seasons. Short term decreases in d-excess align with major precipitation events, particularly in the 2022 record from RG-2 (Figure 9).

In a key study to consider multiple years of water chemistry from the same rock glacier springs (Brighenti et al., 2021a), found that patterns of changing EC and hydrochemistry occurred with a similar timing during each melt season, with the magnitude of EC elevation and trace element enrichment corresponding with the date when snowmelt ended and with peak summer temperatures. In this regard it is notable that at the Chepeta SNOTEL site near RG-1 and RG-2, peak snow water equivalent (SWE) was nearly identical in the spring of 2021 (270 mm) and the spring of 2022 (287 mm), therefore although the exact date when snowmelt ended each year is unknown, it likely occurred at a similar time. The overall 2022 water year was wetter than 2021; by 1 July 2022, a total of 544 mm of water had fallen at Chepeta during the preceding 9 months, in contrast to 401 mm at that point in 2021. However, the generally consistent patterns in stable isotopes through these two summers in the Uintas (Figure 9) indicate that the seasonal evolution of rock glacier meltwater is controlled more by when snowmelt ends and less by absolute amounts of precipitation.

The two contrasting rock types in the locations where the water samplers for this project were deployed permit the influence of lithology on rock glacier water chemistry to be assessed (Figure 8). In the quartzite terrane of the Uintas, EC is consistently lower than in the igneous terrane of the La Sals (Figures 4, 5). Even though EC values rise through the summer at both RG-2 and RSC, peak values at RG-2 (15 μS/cm) are still below the lowest values at RSC (∼25 μS/cm). Values of Ca and Mg are also notably different between the two sites (Figure 7). In the Uintas, Ca is consistently <5,000 ppb, whereas at the RSC site in the La Sals, Ca starts around 5000 and rises to 15,000 ppb by October. Values of Ca/Mg are also considerably different, averaging 3 in the Uintas in contrast to >10 at RSC. The clearest illustration of the role of lithology is seen in the biplot of principal components determined for the hydrochemical data (Figure 6). The Uinta sites (RG-1 and RG-2) plot in an overlapping field that is completely separate from the La Sal rock glacier (RSC) and the sampler slightly downstream in Gold Basin (GB). The complete lack of overlap between the samples from the two sites emphasizes the significant role played by lithology in controlling the hydrochemistry of rock glacier water.

Previous work has come to similar conclusions regarding the importance of lithology in controlling the chemical characteristics of rock glacier meltwater. For example, the tendency of rock glaciers to alter the properties of water flowing through them has been established by field studies (Giardino et al., 1992). High concentrations of metals have been noted in the water draining from rock glaciers (Thies et al., 2017; Liaudat et al., 2020), and rock glaciers have been reported as sources of acid rock drainage (Ilyashuk et al., 2014; Ilyashuk et al., 2018). The hydrochemistry of rock glacier water has also been confirmed to be significantly different than water draining from glaciers (Fegel et al., 2016), particularly in terms of higher abundances of Ca and Mg (Williams et al., 2006), and direct comparison of rock glacier springs located above different rock types has revealed strong contrasts in the abundance of major ions and trace elements (Brighenti et al., 2021a). The results of this study corroborate these findings by illustrating a clear distinction between the hydrochemistry of rock glacier water in the quartzite Uintas and the igneous terrane of the La Sals.

The three water samplers in the La Sals can be used collectively to investigate the extent to which stream water in catchments with rock glaciers differs from water in watersheds lacking these features (Figure 8). The RSC sampler collects water discharging directly from a rock glacier terminus, and the GB sampler collects water a short distance downstream along the main stream in this basin (Figure 2). In contrast, the WF sampler collects water along the main stream in a watershed where rock glaciers are absent. This difference was exploited by previous work that focused on the role of rock glaciers in controlling the character of stream hydrographs (Geiger et al., 2014). Figure 5 reveals the similarities between the RSC and the GB isotope time series, and Figure 6 emphasizes the affinity of the hydrochemistry at these two sites. Clearly the water in the stream at the GB sampler strongly reflects the character of the rock-glacier derived water entering the stream system upvalley. At WF, in contrast, both the isotope and hydrochemical data are notably different from at GB. Values of δ¹⁸O at WF are low and stable (Figure 5), resembling snow throughout the melt season despite the lack of snow on the ground surface in this watershed after late June. In Figure 6 the hydrochemistry of the water at WF is also demonstrated to be significantly different than the overlapping water at RSC and GB. In Figure 7, it is clear that values of Ca and Mg are dramatically higher at WF than at other sites, and that values of Ca/Mg are lower than at RSC and GB. These differences are summarized in Figure 10 where values of δ¹⁸O are plotted against PC-1 for the three samplers in the La Sals, emphasizing the similarity of water at RSC and GB, and the divergence of WF.

Long-term studies have noted rising values of Ca and Mg in streams fed by rock glaciers and attributed this pattern to melting of buried ice (Caine, 2010). Meltwater derived from this ice may participate in weathering of Ca-bearing minerals in the bedrock (Williams et al., 2006), or be involved in weathering of Ca-bearing eolian dust (Litaor, 2022). Overall degradation of permafrost due to rising ground temperatures has been invoked as a major mechanism behind changing water chemistry in alpine watersheds (Colombo et al., 2018; Liaudat et al., 2020). The results reported here from the La Sal Mountains support this prior work, emphasizing the degree to which the presence of rock glaciers influences the characteristics of water downstream.

The approach employed by this study was successful in generating time series of water samples from springs and streams in the study areas. One limitation in the experimental design, however, is the fact that snowmelt was already underway when the samplers were deployed. This situation is an unavoidable product of the inaccessibility of these locations in the winter, and the fact that the samplers cannot operate in subfreezing temperatures. Nonetheless, future studies employing this methodology should endeavor to deploy samplers as early as possible in the summer to ensure that as much of the snowmelt period is sampled as possible.

The remote location of the samplers also required compositing of samples to stretch the 24-bottle carousel out over as long an interval as possible. The resulting set of 3-day composite samples is sufficient for evaluating long-term trends over the course of the entire melt season, but lacks the resolution necessary to identify shorter-term changes in response to precipitation events (Krainer and Mostler, 2002), or perhaps to accelerated melt driven by particularly warm days (Brighenti et al., 2021a). Samplers capable of collecting a greater number of samples and/or more frequent visits to empty the carousel of 24 bottles would allow a finer-scale sampling interval for identifying more rapid shifts in water chemistry.

A third limitation is that discharge measurements were not collected, therefore it is not possible to identify discharge trends over the course of the melt season or transient changes in association with precipitation events. Water discharge from rock glaciers is notoriously difficult to quantify (Krainer and Mostler, 2002; Krainer et al., 2012) because of the tendency for water to drain below the ground surface through coarse blocks and talus. In the present study, installation of weirs was not approved, so even minimum discharge estimates are, unfortunately, unavailable. Nonetheless, future work could perhaps be focused on alternate locations where temporary weirs might be permitted.