Field and Laboratory Methods

Sediment cores were extracted using a percussion corer (Reasoner, 1993) operated from a floating platform. Prior to coring, a bathymetric survey was conducted for each lake using a digital depth sounder linked to a GPS receiver to identify the lake depocenter. After the platform was anchored in place, a continuous core (7.5-cm diameter), was taken from the sediment-water interface to the point of refusal. Loose sediment immediately below the sediment-water interface (10–50 cm) was lost during retrieval, and additional poorly consolidated sediment from the upper part of each core was discarded because it was too soft to survive transport undisturbed. The resulting cores ranged from ∼150 to 350 cm long. Cores were cut into ∼50-cm lengths, capped, and transported to the trailhead in specially designed panniers. After shipping, the cores were stored at 5°C until opening.

In the laboratory, cores were split lengthwise, measured, photographed, and described. Samples of approximately 3 cm³ were taken at 1-cm intervals throughout one half of each core for LOI analysis (Dean, 1974). Measurements were made with an automated thermogravimetric analyzer (TGA) that dried, and then heated samples at 550°C until they reached mass constancy. LOI analysis in a TGA eliminates many of the concerns raised by Heiri et al. (2001), including issues of sample position in the furnace (crucibles are constantly cycled around the perimeter of the furnace), length of time each crucible is at high temperature (all crucibles are run through the same steps simultaneously), temperature fluctuations (furnace temperature is tightly regulated and recorded), and differential rates of mass lost (each step continues until all samples reach constant mass).

Chronology

The remaining half of each core was wet-sieved to 500 µm at 1-cm intervals to recover organic fragments suitable for AMS radiocarbon dating. Identifiable fragments included Picea and Abies cones and needles, Salix twigs, and pieces of bark. Lower sections of several cores were dominated by Daphnia ephippia (eggs, hereafter DE) in concentrations sufficient for AMS dating. To date the inorganic basal sediments, pollen was concentrated, using a methodology adapted from Brown et al. (1989) and Mensing (1999), for AMS analysis. The resulting samples ranged from 1 to 20 mg, and consisted of more than 90% Pinus pollen.

AMS results were converted to calendar years with CALIB 5.0 (Stuiver et al., 2005). For developing depth-age relationships, the midpoint of the 2σ age-range with the greatest probability was taken as the calibrated age.

Watershed Variables

Variables describing the morphology and characteristics of the lakes and surrounding watersheds were determined through GIS analysis. Watershed boundaries were digitized from published 1:24,000-scale U.S. Geological Survey (USGS) topographic maps with a 40-foot (12-m) contour interval, supplemented by field inspection. Watershed boundaries were used to clip a 30-m digital elevation model (DEM) of the Uintas acquired from the USGS Seamless Data Distribution System. Slope and aspects grids were derived from the clipped DEM, and values for each watershed were calculated.

Lake outlines were downloaded from the State of Utah Automated Geographic Reference Center and used to calculate lake area, perimeter, and complexity (ratio of lake perimeter to that of a circle with the same area). Maximum depths of each lake were noted during the bathymetric surveys preceding coring. The number of inlet and outlet streams was determined from 1:24,000-scale topographic maps, and the order of the inflowing streams was determined (Strahler, 1952). Hydrologic through-flow was scored for each lake as follows using information gained during field surveys: 1—no inlet or outlet; 2—little or no obvious inflow, weak outflow; 3—modest inflow and outflow; 4—robust inflow and outflow.

The extent of different landcover types in each watershed was determined through analysis of multispectral data captured by the Landsat 5 Thematic Mapper sensor on 2 July 1989 (scene P037R32_5T890702). Bands 2, 3, and 4 were combined in a false-color image, which was subdivided with an unsupervised classification into 16 fields. Information gained through past ground surveys was used to combine these fields into groups representing ice/snow, rock, water, forest, tundra, and wetland. The percent of each watershed occupied by these landcover types was then calculated for each watershed. The area of bedrock outcrops was considered constant over the period of interest. Similarly, given the strong topographic control on snowbank location and size, the area of these features (as a proportion of the total watershed area) was considered relatively constant. The area of forest, tundra, and wetland were not considered further because of the likelihood that they have changed over the past 10,000 years.

Statistics

Mean, median, and modal values were calculated for each LOI record starting at 10,000 cal yr BP; earlier parts of the records (i.e., pre-10 ka BP) were ignored to avoid comparing basal sediments from one lake with organic-rich sediment from another. Standard deviations and the coefficient of variation (as a percent) were also computed. Given the complexity and variability of records from the different lakes, these descriptive statistics were considered most useful for describing the LOI records. The correlation between the watershed variables calculated in the GIS and the LOI statistics was assessed with Spearman's Rank Correlation, and the robustness of the significant correlations was evaluated by iterative recalculation of the correlations with one lake excluded each time. The 12 lakes were then divided by visual inspection into 3 groups with common LOI patterns, and the differences between these groups were evaluated with the Kruskal-Wallis test.

Sedimentology

The locations of the 12 lakes are given in Figure 1, and information about each lake is provided in Table 1. The lakes range from 3.7 to 12.8 m in depth, from 4.0 to 14.6 ha in area, and from 2957 to 3474 m in elevation. Cores 04-08 and 04-09 are from lakes located above treeline. All cores penetrated more than 2 m below the sediment-water interface, and recovery ranged from 62 to 98% (Table 1).

Figure 1.

Shaded relief map of the Uinta Mountains showing locations of the lakes cored in this study. A—Marshall Lake (core 04-01), B—Hoover Lake (04-02), C—Pyramid Lake (04-03), D—Swasey Lake (04-06), E—Spider Lake (04-07), F—Little Superior Lake (04-08), G—North Star Lake (04-09), H—Elbow Lake (04-04), I—Reader Lake (05-01), J—Taylor Lake (05-10), K—Upper Lily Pad (06-01), and L—Lower Lily Pad (06-02). Inset shows the location of the Uinta Mountains in northeastern Utah.

Table 1

Locations and properties of lakes cored in the southern Uinta Mountains.

Almost all of the cores contain a similar bipartite stratigraphy of gyttja overlying a lower section with reduced organic content. The gyttja ranges from black (10YR 2/1) to very dark grayish brown (10YR 3/2) to dark reddish gray (2.5YR 3/2) in color, and is usually massive near the top of the recovered core. Some sections contain faint millimeter-scale laminations reflecting oscillations between slightly lighter and darker sediment. In some lakes, particularly 04-03, the gyttja is vesicular.

Organic content was determined for 2850 samples. LOI values range from less than 2% in basal silty sand (04-02) to 40% in gyttja (04-04) (Table 2). Mean organic contents per lake decrease with elevation, with values of 10% or less in cores from the highest lakes (04-09 and 05-10).

Table 2

Loss-on-Ignition (LOI) values, and lake and watershed variables.

Chronology

A total of 71 AMS radiocarbon determinations were made (Table 3). The majority of these dates are in stratigraphic order and yield sedimentation rates of 0.5 to 0.1 mm yr⁻¹ (50 to 10 cm·10⁻³ yr⁻¹). In core 04-01 bulk sediment and a conifer needle from the same stratigraphic depth (47 cm) returned similar ages, suggesting that there is no hard-water effect in this lake, or that if there is one, the effect is limited to shifts of a century or two. A similar situation is expected in the other lakes given the lack of carbonate bedrock in their watersheds, which supports the validity of dating DE.

Table 3

Radiocarbon dates for the southern Uinta Mountains lake cores.

The dates returned on pollen concentrates yielded more equivocal results. For instance, in core 05-01, pollen returned an age statistically indistinguishable from a conifer needle at the same depth (218–220 cm). In other cases the pollen dates appear too old, for instance in core 04-06, where the pollen dates (ca. 18 ka BP) is impossible to reconcile with the cosmogenic ¹⁰Be surface-exposure age of 18 ka BP determined for the terminal moraine 20 km downvalley (Munroe et al., 2006). In contrast, some pollen dates appear too young relative to other ¹⁴C analyses on the same cores (e.g., 04-02, 04-04, 05-01). Because it is not clear how to consistently interpret the pollen dates, and because the majority of them are from near-basal sediments that are not the focus of this report, they were ignored and depth-age models were based on linear sedimentation rates between dated non-pollen samples, with extrapolation to the core top and bottom as necessary. In core 04-01, the age of a pronounced LOI minimum was shifted slightly to match the dated minimum (4.4 ka BP) in core 04-03 given the similarity of these two records, a synchronous LOI minimum in core 04-02, and the close proximity of all three lakes.

Watershed Variables, Loi Patterns, and Statistical Analysis

The variables calculated for each watershed are presented in Table 2. Statistical evaluations of correlations between watershed variables and LOI values for the past 10,000 years are given in Table 4. Lake elevation, hydrologic through-flow, watershed elevation, percent ice/snow and bare rock, and the area of upstream lakes in the watershed are strongly correlated with average LOI (mean, median, mode) over the past 10,000 years. Correlations are not significant between LOI and lake dimensions (depth, area, perimeter, complexity), between LOI and watershed/lake ratios (area, depth), or between LOI and watershed slope.

Table 4

Strength of correlations between LOI over the past 10,000 years and watershed variables.

Visual inspection supports division of the 12 LOI records into 3 groups with distinct patterns: Steady, Trending, and Rising (Fig. 2). Steady lakes (cores 04-02, 04-07, 04-09, and 05-10) feature fairly stable LOI values after the initial latest Pleistocene rise. Trending lakes (cores 04-01, 0-04, 04-06, 04-08, 06-01) feature greater LOI variability organized into multi-centennial to millennial-scale increasing and decreasing trends. Rising lakes (cores 04-03, 05-01, 06-02) feature monotonically increasing LOI, interrupted by spikes to higher or lower LOI values. Significant differences exist in average LOI, through-flow, extent of upstream lakes, maximum watershed elevation, and ice/snow cover between the three groups of lakes (Table 5).

Figure 2.

Loss-on-ignition records for the 12 lakes in this study. All records are plotted at the same scales (LOI from 0 to 40%, age from 0 to 14,000 cal yr BP). Records are labeled by their patterns (Steady, Trending, Rising), and mean values for the past 10,000 years are shown by dashed lines. All lake outlines are shown with a 200-m scale bar.

Table 5

Significance of differences in mean values between LOI patterns.

Loi Values

The organic contents determined for these cores are equivalent to those reported by other studies from similar settings. For instance, Zielinski (1989) documented LOI values of 10 to 25% in Miller Lake (3234 m a.s.l.), and Fall et al. (1995) reported Holocene LOI values between 10 and 20% in Rapid Lake (3134 m a.s.l.), both in the Wind River Range (Wyoming), ∼250 km north of the Uintas.

Correlations between Watershed Variables and Loi Records

Lake elevation, through-flow, watershed elevation, extent of ice/snow and bare rock, and the area of upstream lakes are strongly correlated with average LOI over the past 10,000 years (Table 4). The statistical strength of these relationships decreases slightly when individual lakes are removed from the data set, but most retain P-values < 0.10. This situation suggests that although some lakes, particularly those at higher elevations, exert a somewhat disproportionate influence on the statistical significance of the correlations, the overall relationships are robust.

All of the variables that are notably correlated with average LOI are interrelated through hydrology. Watersheds at higher elevation hold more late-lying ice and snow, contain more exposed rock, and receive greater amounts of precipitation. As a result, more water runs off these landscapes and through the higher lakes. The positive correlation (Spearman's rho = 0.528, P = 0.078) between elevation and through-flow, which was assessed on the basis of observed inlets and outlets, indicates the extent to which the geomorphology of the higher lake basins has evolved to convey greater volumes of water.

Greater through-flow contributes to lower LOI values because stream water temperatures are low in these watersheds due to their high elevation, long duration of seasonal ice cover, and lingering summer snowmelt. The flux of this cold water into the lakes suppresses aquatic productivity and encourages oligotrophic conditions. Furthermore, given the low nutrient status of soils derived from quartzite-derived glacial deposits in the Uintas (Bockheim et al., 2000), greater water volumes may lead to overall dilution of the already minimal dissolved nutrient load, reducing aquatic productivity and corresponding sediment LOI. In contrast, lakes not fed by lingering snowmelt can be more readily warmed by the sun, and evaporation in closed basins may concentrate nutrients. Both processes would lead to enhanced productivity and higher LOI values.

Loi Patterns and Watershed Variables

The repetition of Steady, Trending, and Rising LOI patterns in lakes from different sectors of the range, combined with adjacent lakes having different LOI curve types, suggests that aspects of the surrounding watershed strongly influence organic sedimentation, possibly to an extent sufficient to overprint shifts in the lacustrine environment driven by regional climatic changes. Table 5 presents a comparison of LOI statistics and watershed variables from these three groups. Once again the commonality between variables exhibiting significant differences is related to hydrology. Overall, lakes with Steady LOI patterns have lower average LOI values. These lakes are found in watersheds that attain greater elevations, and their inlet/outlet configuration attests to greater volumes of through-flowing water. The watersheds around these lakes also contain significantly more ice/snow cover, and a larger area of upstream lakes; both of which represent temporary storage in the hydrologic cycle. Steady LOI records, therefore, are associated with lakes that are hydrologically active, accommodating large volumes of cold through-flowing water that encourage oligotrophic conditions. Given the constancy of their LOI values over the period of interest, lakes with Steady LOI records have apparently had higher levels of hydrologic activity throughout the Holocene.

Lakes with Rising LOI patterns tend to be found at the lowest elevations within generally smaller watersheds (Table 5). They are also surrounded by significantly less late-lying snow, are not connected to upstream lakes, and have only minimal development of inlets and outlets. Together these characteristics encourage warmer water and greater productivity within these lakes, leading to enhanced accumulation of organic sediment. Furthermore, two of the three lakes in this group are considerably shallower than the rest (∼4 m, Table 2). When the length of the cores retrieved from these lakes is compared with their modern depths, it is apparent that theses lakes have lost approximately half their depth since the early Holocene. Thus the increasing LOI values characteristic of these lakes likely reflect their progressive shallowing and eutrophication over time.

Lakes with Trending LOI patterns have watershed and lake properties that are intermediate between the other groups (Table 5). In particular, their average LOI values overlap with those common to the Rising group, but they are found at generally higher elevations and feature at least some evidence for through-flowing water. On the other hand, they have fewer inflows and a considerably smaller area of upstream lakes and late-lying snow than Steady LOI lakes, so they are not dominated by their open hydrology. In the absence of overriding external (i.e., geohydrologic) or internal (shallowing, eutrophication) factors, the shifting LOI trends in these lakes may reflect paleoclimate variability at multi-centennial to millennial scales, especially for trends that are similar in lakes from different areas. For instance, four of the five Trending lakes feature LOI values that peak in the early to middle Holocene. Because these lakes are located all across the southern Uintas (Fig. 1), the shifting trends in their records may reflect a limnologic response to regional climate forcing, perhaps the early Holocene insolation maximum (Berger, 1978; Bartlein et al., 1998). Evidence from other proxies suggests that the distribution of high-altitude vegetation in the Uintas shifted in response to warmer temperatures in the early Holocene. Specifically, the only conifer needle found in the highest lake (core 04-09), which is located ∼100 m above modern treeline, dates to 7250 cal yr BP, and palynological evidence for a higher-than-modern treeline in the northern Uintas during the early Holocene was reported by Munroe (2003).

Implications for Paleolimnology

It is useful to consider which types of mountain lakes would be most valuable from a paleolimnology perspective. The answer, of course, depends on the objectives of a particular study, yet the exploration of the Uinta lakes data set presented here supports a few generalizations. First, lakes in active hydrologic settings will likely have LOI records that are insensitive to minor-to-modest climate changes because the large amounts of through-flowing water buffer them against changes in water level, temperature, and salinity that may affect productivity. Second, records from lakes with minimal through-flow in low-elevation basins may be dominated by continuous eutrophication driven by progressive shallowing. Major climatic shifts may still be recorded in the sediment of these lakes; for instance, the two departures to lower LOI values in core 04-03 at ca. 8000 and ca. 4000 BP (Fig. 2) represent notable disruptions in the overall increasing LOI over the past 10,000 years that are paralleled by similar fluctuations in neighboring lakes with different LOI patterns (cores 04-01 and 04-02). However, the possible role of changing internal factors (i.e., depth, volume, nutrient status) would need to be considered for the development of accurate paleoclimate records from lakes in these settings. Finally, lakes with moderate values in a few key categories (i.e., elevation, through-flow, upstream lake area, ice/snow cover) may offer the best potential for paleolimnological investigations in high mountain environments. LOI records from these lakes are neither Steady nor Rising, suggesting that they are not dominated by through-flow or eutrophication. Instead, variations in LOI trending over multi-centennial to millennial scales imply that lake sediments in these settings are recording environmental responses to paleoclimate variability.