Mapping

Glacial features in 19 drainages of the northern Uintas were mapped during the summers of 1998–2000 (Fig. 2). Lateral moraines, end moraines and heads of outwash were identified through map and air-photo interpretation and checked in the field. Glacial limits were correlated between adjacent valleys based on position, elevation, cross-cutting relationships of outwash valley trains and the extent of weathering. Direct absolute ages are not available for these deposits. Mapping was compiled in a GIS, and the dimensions of the glaciers during the Smiths Fork Glaciation were reconstructed from the distribution of ice-marginal features, including terminal and lateral moraines. All glacier reconstructions were field-checked to assess their validity.

Fig. 2. Three-arcsecond digital elevation model of the Uinta Mountains showing drainages of the northern Uintas. Lower elevations are represented by darker shades. The highest elevations in the center of the range are > 4000 m a. s. l.

Calculation of ELAs

Four separate methods were used to calculate the elevations of equilibrium lines for each of the reconstructed glaciers accumulation–area ratio (AAR), toe–headwall altitude ratio, highest elevation of continuous lateral moraines, and cirque-floor elevations.

The AAR method is based on the observation that the majority of modern glaciers have an “accumulation-area” to “total-area” ratio of 0.50–0.65 when they are in equilibrium (Reference Meier and PostMeier and Post, 1962). Geomorphic features delineating the former ice margin are used to determine the area covered by the glacier, and the ELA is then located at the elevation of a line dividing the reconstructed area in the proportions seen on modern glaciers. The AAR technique has been applied to glacier systems all over the world, including New Zealand (Reference PorterPorter, 1975), Hawaii (Reference PorterPorter, 1979), the Cascades of Oregon (Reference ScottScott, 1977) and Baffin Island (Reference HawkinsHawkins, 1985). Reference MeierdingMeierding (1982) determined that a ratio of 0.65 provided the lowest rms error in his analysis of glaciers in the Colorado Front Range. Reference Torsnes, Rye and NesjeTorsnes and others (1993) supported this conclusion with their study of modern glaciers in western Norway. For this study, glacier outlines determined from field mapping were digitized into a GIS from which the area of the glacier was calculated. The elevation of a line that assigned 65% of this glacier area to the accumulation area was determined iteratively.

The AAR method is best applied to glaciers with relatively simple area–altitude distributions. Application of the AAR technique to glaciers with skewed hypsometry, such as ice fields or piedmont glaciers, can yield erroneous results. Previous researchers have attempted to correct for this by using a lower AAR (∼0.50) for glaciers known from reconstructions to have had piedmont glacier morphology (e.g. Reference Brugger, Goldstein, Mickelson and AttigBrugger and Goldstein, 1999). Accordingly, an AAR of 0.50 was used in the calculation of the final weighted ELAs for the five piedmont glaciers in the northern Uintas. As an additional check on the uniformity of glacier hypsometry, the glacier area above a given elevation was determined at 100 m intervals from the terminus to the headwall. The coefficients of determination (r ²) for linear regressions fit to these data were used to evaluate the uniformity of the overall area–altitude distribution of the reconstructed glaciers and their suitability for the AAR method.

The toe–headwall altitude ratio (THAR) method has also been used to determine ELAs. The equilibrium lines on modern glaciers are usually located at an elevation above the terminus that is equivalent to 35–40% of the total elevation difference between the terminus and top of the glacier. The lowest elevation of ice is readily calculated from the elevation of the glacier terminal moraine. The highest elevation that a glacier reached on a cirque headwall is harder to determine, especially in areas of intense postglacial mass wasting. Reference MeierdingMeierding (1982) estimated the upper elevation of glacial ice as the highest point on the cirque headwall with a slope of <60°. He found a THAR of 0.35–0.40 yielded the lowest error and was comparable to the AAR method in its accuracy. In this study the glacier-toe elevation was taken as the lowest elevation along the top of the terminal moraine. The highest extent was taken as the highest contour below the near-vertical part of the upper headwall, and a THAR of 0.40 was applied to locate the ELA.

A third method for estimating ELAs, the lateral moraine (LM) method, involves the identification of lateral moraines in glaciated valleys. For a glacier in equilibrium, ice flows away from the margins in the accumulation area and toward the margins in the ablation area. Sediment is brought to the margins to form moraines, therefore, only in the ablation area. Thus, the elevation of lateral moraines farthest up-glacier should coincide with the ELA. This method may underestimate the ELA where lateral moraines have been affected by postglacial erosion. To reduce errors associated with postglacial erosion of moraines, the ELA was estimated by tracing continuous lateral moraine crests upslope from the terminal moraine on both sides of the valley. This approach allowed projection of lateral moraines across gaps where moraine segments had been removed by mass wasting.

A final method for estimating former ELAs, the cirque-floor (CIR) method, involves measurement of cirque-floor elevations in glaciated drainages. Determining the elevation of cirque floors with a similar aspect yields an estimate of the ELA (Reference FlintFlint, 1971). Such estimates often have large errors because cirque floors reflect the erosive effects of multiple glaciations, all of which may have had different ELAs. Weakly developed lower-elevation cirques are also difficult to identify (Reference MeierdingMeierding, 1982). Well-developed cirques are abundant in the northern Uintas, though, facilitating application of this method and increasing its utility. Therefore, the elevation of the slope break between the headwall and cirque floor in the lowest north-to-east-facing cirque was recorded for each glaciated valley.

Calculation of the weighted ELA

The values derived through each of these four methods were combined to produce a weighted average ELA for each of the Smiths Fork-age glaciers, similar to the method of Reference LockeLocke (1990). Estimates obtained through the AAR method were given a weight of 4, the THAR method 3, the LM method 2 and the CIR method 1. The weights assigned in this study are qualitative, reflecting the estimated accuracy and precision of the methods and the success other researchers have had in their application (e.g. Reference MeierdingMeierding, 1982; Reference LockeLocke, 1990).

AAR values of 0.65 were used in the weighted average for all glaciers except for the South Fork Ashley Creek (No. 1), the East and West Forks Carter Creek (Nos. 3 and 4) and the South and Middle Forks Sheep Creek (Nos. 5 and 6) (see Fig. 1). Reconstructions based on field mapping showed that these five glaciers had a piedmont morphology at their maximum position that skewed their area–altitude distribution. Accordingly, an AAR value of 0.50 was used in the weighted average for these valleys.

Determination of modern temperatures at the ELAs

Mean annual (T _a) and mean summer (T _s; June–August) air temperatures were determined for four SNOTEL sites in the northern Uintas and one remote automated weather station (RAWS) on the Uinta ridge crest. These were combined with average temperatures for Manilla, Utah and Evanston, Wyoming, to produce an elevational transect ranging from 1936 to 3697 m a.s.l. A linear regression for these data was then used to determine the modern T _a and T _s at each of the reconstructed ELAs.

Determination of modern winter accumulation at the ELAs

The water contained within the snowpack at the locations of the ELAs in the northern Uintas was determined based on the methods of Reference PiercePierce (1982) and Reference LeonardLeonard (1984). Snow-accumulation records (Table 1) from 16 SNOTEL sites in the glaciated section of the Uinta Mountains for the years 1986–1999 were utilized (United States National Resource and Conservation Service, http://www.wcc.nrcs.usda.gov/water/wdata.html). The locations of these stations, their mean 1 April water equivalent and their mean annual maximum water equivalent were digitized into the GIS, allowing interpolation of a trend surface representing modern snow accumulation in the Uintas. Linear regressions between SNOTEL elevation and both 1 April and maximum water equivalent were determined, allowing prediction of 1 April and maximum water equivalent at any elevation.

Residuals were then calculated for each SNOTEL site between the measured and predicted values. A second surface was interpolated for these residuals in the GIS, yielding a representation of the deviation from altitude-predicted snowfall. Major departures identify areas in the modern Uintas that are relatively snow-starved, as well as those that receive more snow than would be expected given their elevation.

Precipitation at each of the 19 ELAs was calculated through two steps. First, the linear regression for SNOTEL elevation and mean annual maximum water equivalent was used to estimate the water content of the snowpack based solely on elevation. Then these values were adjusted by querying the surface interpolated from the residuals in the GIS to yield a local correction factor that was applied to the altitude-predicted value. This final estimate represents the modern winter water accumulation (P) at each of the ELAs, taking into account both the elevation and location of the ELA within the modern snowfall distribution.

Calculation of the differences between modern and LGM paleoclimate

The values for modern T_s and P for the locations of each of the ELAs were compared with the range of T_s and P found at modern glacier equilibrium lines (Reference LeonardLeonard, 1984; Reference Brugger, Goldstein, Mickelson and AttigBrugger and Goldstein, 1999). Because two variables are involved, this analysis cannot yield a unique solution. However, the range of changes in T_s and P required to shift these parameters from their modern values to those known to exist at modern glacier ELAs can be determined.

ELA estimates

Estimates of Smiths Fork-age ELAs obtained by the four methods described above are shown in Table 2. Given contour intervals of 20 ft (6 m) on the lower-elevation quadrangles, 40 ft (12 m) on the higher-elevation maps, and 50 ft (15 m) on the tightest network interpolated in the GIS, these estimates are considered accurate to within ±7 m. The CIR method yields the highest mean ELA (3300 m a.s.l.), while the THAR method yields the lowest mean estimate (3050 m a.s.l.).Values for the AAR and LM methods are intermediate (3090 and 3120 m a.s.l., respectively). This result is different than that obtained by Reference LockeLocke (1990) who reported that the LM method underestimated the others by about 230 m, suggesting that LGM lateral moraines in the northern Uintas may be better preserved than those in western Montana. Standard deviations are similar for the AAR, THAR and LM methods, suggesting that these techniques were more successful in estimating the former ELA trend surface.

Statistics allowing comparison of the results obtained by the four methods for the Smiths Fork Glaciation are given in Table 3 (after Reference LockeLocke, 1990).The means are statistically indistinguishable (at α = 0.001) between the AAR and THAR methods (t-statistic 24, P = 0.005), AAR and LM methods (t-statistic −1.56, P = 0.137) and THAR and LM methods (t-statistic −67, P = 0.002). In contrast, because the CIR method yielded higher estimates of ELAs, differences between the means derived from this method and the other three are highly significant (P < 0.001).

Weighted ELAs

Figure 3 illustrates the elevation of the weighted ELAs along the north slope of the range. The mean weighted ELA is 3100 m a.s.l. with a range of 2860–3230 m a.s.l. and a standard deviation of 97 m. ELAs are lowest at the western end of the north slope (∼2900 m a.s.l.). In contrast, ELAs for the three glaciers in the center of the north slope are 80–130 m above the mean value. The ELA gradient over the western 60 km of the study area is 6 m km⁻¹ (370 m elevation change). The Smiths Fork-age ELAs also descend in elevation (∼ 3 m km⁻¹) at the eastern end of the range, where most of the glaciers had equilibrium lines 10–40 m above the mean (∼100 m lower than those in the center of the range).

Fig. 3. Smiths Fork-age weighted ELAs across the north slope. The Uinta ridge crest is shown for reference.

Lapse rates and modern summer temperatures at the ELAs

Based on the data from three SNOTEL sites, the RAWS and the records from Manilla, Utah and Evanston, linear regressions relating station elevation to T_s and T _a were determined:

where T is in °C, and z is the station elevation in meters. The summer adiabatic lapse rate for the northern Uintas is therefore 6.7°C km⁻¹, while the mean annual rate is 5.3°C km⁻¹. Based on Equation (1), T _s at the 19 Smiths Fork-age ELAs ranges from 8.7° to 11.2°C. From Equation (2) the modern 0°C summer isotherm is projected to lie above an elevation of 4500 m a.s.l., while modern mean annual temperatures at all the ELAs range from −0.8° to 1.2°C.

Modern winter snow accumulation at the ELAs

Predicted water equivalents (in mm) are related to SNOTEL site elevation (z) as follows:

Annual Maximum = 0.4326z − 859

The average date of the maximum water equivalent at the 16 SNOTEL sites ranges from to 26 March to 7 May. Figure 4 illustrates that the higher-elevation stations record their maximum after 1 April (gradient of 40 days km⁻¹, r ² = 0.78, P < 0.001). Accordingly, the annual maximum value was used as a proxy for winter snow accumulation instead of the 1 April measurement, to avoid the bias inherent in using the 1 April value for stations of different elevations (e.g. Reference LockeLocke, 1989).

Fig. 4. Average date of the maximum annual snowpack as a function of SNOTEL site elevation.

Figure 5 shows the surface interpolated from the annual maximum values. The predicted values for the ELAs determined from Equation (4) range from 380 to 540 mm. But Figure 6 illustrates that individual SNOTEL sites receive from −32% to +52% of the predicted values and that residuals at each of the ELAs range from −25% to +21%. Correction of the predicted values at the ELAs using these residuals yields estimates of P from 357 to 586 mm.

Fig. 5. Distribution of mean annual maximum SWE (in mm) across the Uintas for 1986–99. Lighter shading represents greater mean annual maximum SWE. The surface was interpolated from the 16 SNOTEL sites (triangles). The Smiths Fork-age glaciers are outlined, and their weighted ELAs are highlighted. The greatest modern winter accumulation is in the centers of the north and south slopes and at the far western end of the range.

Fig. 6. Surface representing percentage residuals from the altitude-predicted (Equation (4)) mean annual maximum SWE across the Uintas for 1986–99. Symbols are the same as in Figure 5. The greatest negative deviation is in the northeast and just west of center on the south slope. The greatest positive deviation is at the extreme western end of the range.

Modern snowfall distribution

Based on the SNOTEL data, the modern snowfall distribution varies considerably across the Uintas, with some areas receiving 30% more or less snow than would be expected based on their elevation (Fig. 6). These substantial local variations in snow accumulation are likely related to prevailing winds and moisture-transport patterns.

The spatial variability of precipitation in the Uintas is further demonstrated by the coefficient of determination (r ²) for the regression relating elevation and mean 1 April snowpack (Equation (3)), 0.55, which is below the range found for 13 mountain ranges in Colorado by Reference LeonardLeonard (1989) (r ² = 0.58–0.99), and at the lower end of the range for 24 mountain ranges in the Great Basin by Reference Zielinski and McCoyZielinski and McCoy (1987) (r ² = 0.52–0.98). The regression between elevation and annual maximum snowpack (Equation (4)) is slightly stronger (r ² = 0.60), suggesting that late-spring storms affect the entire range, balancing differences produced earlier in the winter by local storms.

The Uintas are also generally snow-starved today compared with other mountain ranges of the western United States. Mean 1 April water equivalent values in the Uintas are considerably lower than in most of the Great Basin ranges studied by Zielinksi and McCoy (1987). In addition, the gradient (from Equation (4)) of annual water equivalent values in the Uintas, 0.43 m km⁻¹, is slightly lower than the 0.5 m km⁻¹ considered typical of isolated topographic barriers, and significantly lower than the 1.2 m km⁻¹ characteristic of the northern Rocky Mountain ranges (Reference Porter, Pierce, Hamilton and WrightPorter and others, 1983). The lower gradient in the Uintas may be partially due to the numerous upwind mountain ranges in the Great Basin, which may diminish the amount of moisture reaching the Uintas. The modern SWE gradient (1.37 m km⁻¹) on the western slope of the Wasatch Range, immediately west of the Uintas, is the steepest in the area studied by Reference Zielinski and McCoyZielinski and McCoy (1987), indicating the efficacy of this range at intercepting westerly-derived moisture.

Additional support for this interpretation comes from the surface interpolated from the snowpack residuals (Fig. 6), which reveals that the eastern half of the northern Uintas receives less snowfall than the western end of the range. Reference MitchellMitchell (1976) delineated a major climatic boundary, termed the “winter boundary”, which is present from November through March across northern Utah and the north slope of the Uintas. Because atmospheric circulation along the winter boundary is dominantly west to east, orographic precipitation probably depletes the moisture content of air masses as they pass along the Uintas, causing less snowfall on the eastern summits relative to the western ones (Reference HoughtonHoughton, 1979).

ELAs and implications for Smiths Fork-age paleo-climate

Several studies have presented data on the range of mean summer temperature (T _s) and winter precipitation (P) values at modern glacier ELAs (Reference LoeweLoewe, 1971; Reference LeonardLeonard, 1989; Reference Ohmura, Kasser and FunkOhmura and others, 1992). Figure 7 shows an envelope that encloses most of the modern data points. The boundaries of this envelope are defined by the following equations from Reference Kotlyakov and KrenkeKotlyakov and Krenke (1982) and Reference LeonardLeonard (1989):

Fig. 7. Climate-space diagram for Smiths Fork-age ELAs. Modern T _s and P at the ELAs plot in the lower right. If modern T _s were unchanged, P would need to be increased 1000% to move the modern climate conditions into the range seen at modern ELAs. At the other extreme, a decrease in T _s of 10°C would be required if modern P values remained constant. Increases of 200 % or 300 % in P would require depressions of T _s 8° and 7°C, respectively.

The values for modern T _s and P at the Smiths Fork-age ELAs in the northern Uintas (listed inTable 3) plot in the lower right of this diagram. If modern T _s were unchanged, P at the ELAs would need to be increased 1000% to move current climate conditions into the envelope of modern ELA conditions. At the other extreme, a decrease in T _s of 10°C would be required if modern P values remained constant. Increases of 200% or 300% in P would require depressions of T _s of 8° and 7°C, respectively. This analysis assumes that the distribution of snowfall during the LGM was similar to the modern distribution and that the change in P was uniform at all 19 ELAs.

Reference SeltzerSeltzer (1994) completed a more sophisticated analysis linking the change in ELA (ΔELA) with changes in winter accumulation, heat input from radiation, and temperature. If it is assumed that the majority of ablation occurs during the summer, that the heat input from radiation is independent of elevation (Reference SeltzerSeltzer, 1994) and that the overall radiation input at the LGM was similar to the modern value (Reference BergerBerger, 1978), then Seltzer’s equation 19 simplifies to:

where ΔELA is in meters, ΔP is in mm and ΔT _S is in °C.

Based on Equation (1), the modern T _s range between the 19 Smiths Fork-age ELAs is 2.5°C (Table 4). If the lapse rate for mean summer temperature was the same at the LGM, the western ELAs must have received ∼2000 mm more precipitation each winter than those in the center of the range to compensate for the warmer T _s value at lower elevations. Referencing the P increase for each of the ELAs relative to the highest ELA (Middle Fork Beaver Creek, MFBC_ELA) reveals that the lower ELAs received 5–2100 mm more P than MFBC_ELA. Numerical simulations of LGM climate in the vicinity of the western Uintas have indicated that precipitation was likely increased by a factor of 2 during the winter (Reference Timofeyeva, Clement, Craig, Wilson and BuffaloeTimofeyeva and others, 1999). Figure 7 shows that if P were doubled at MFBC_ELA, T _s there would need to be depressed from 8°C to 5.5°C to reach the range of conditions at modern ELAs. Assuming a LGM P that was double the modern value at MFBC_ELA, the estimated P values for the 18 lower ELAs range from 943 to 3038 mm (Table 5). These values are greater than those interpreted by assuming a uniform change in P for all 19 data points (Fig. 7) because they consider the independent effects of T _s and P on the elevation of the former equilibrium lines. If these data are plotted in climate-space, with a uniform temperature decrease of 8°C, all but one of the data points fall outside the envelope enclosing conditions at modern glacier ELAs (Fig. 8). Reducing the temperature decrease to 5.5°C allows all 19 data points to fit, suggesting that the actual LGM temperature depression was somewhere between 8° and 5.5°C.

Fig. 8. Climate-space diagram showing possible T _s and P changes for the Smiths Fork Glaciation. Values for P were computed from Equation (7) assuming 200% P at the highest ELA and the modern summer lapse rate. Only one of the data points falls within the envelope when plotted with a uniform 8°C temperature depression. However, a T _s depression of 5.5°C allows all the points to fall within the envelope, suggesting that the actual LGM temperature depression (taking into account non-uniform changes in P) may have been closer to 5.5°C.

Importance of pluvial Lake Bonneville

The reconstructed mean ELA of 3100 m a.s.l. during the Smiths Fork Glaciation compares favorably with the value of 3000 m proposed previously for the Uintas by Reference Porter, Pierce, Hamilton and WrightPorter and others (1983, fig. 4-2) and with the range 2620–3310 m proposed by Reference SchlenkerSchlenker (1995, table 15). Moreover, Reference FlintFlint (1971, fig. 18-4) indicates that the lowest elevation of cirque floors in the Uintas is about 3050 m a.s.l. However, the non-uniform east–west variation in ELA (Fig. 3) was not previously recognized.

Summits at the western end of the Uintas are the first to intercept moisture advected by prevailing westerly winds along the winter boundary defined by Reference MitchellMitchell (1976). Yet the modern P increase at the western end of the range compared with the range center is considerably less than the 800–2000 mm required to explain the 370 m decrease in ELA between these two locations (from Equation (7)). Moreover, adjusted values of P do decrease somewhat across the study area, but the values at the highest ELAs (459, 479, 521 mm) are similar to the values at the lowest ELAs (483 and 456 mm).

Therefore, a change in prevailing wind direction or proximity to a moisture source is required to explain the more dramatic increase in winter precipitation in the western Uintas during the Smiths Fork Glaciation. Because orographic shielding by upwind mountains reduces downwind precipitation (Reference HoughtonHoughton, 1979), ELAs should rise in a downwind direction if the local lapse rate remains constant. The eastward rise in ELA along the western half of the north slope documented in Figure 3 suggests that prevailing winds did carry moisture from west to east during the LGM, as they do today. Support for this interpretation comes from Reference MoranMoran (1974) and Reference Zielinski and McCoyZielinski and McCoy (1987) who concluded that circulation patterns across the northern Great Basin in the vicinity of the winter boundary during the late Pleistocene were similar to those today. In addition, climate simulations using a global circulation model suggest that the North American jet stream and storm tracks were displaced southward by the Laurentide ice sheet (COHMAP members, 1988), reinforcing a dominantly westerly airflow through this region.

If prevailing winds were carrying moisture to the Uintas from the west during the LGM, then the presence of pluvial Lake Bonneville ∼100 km upwind of the western Uintas would have greatly increased moisture availability (Fig. 9). Lake Bonneville reached its maximum surface elevation of 1550 m a.s.l. around 15 kyr BP, covering an area of 51 700 km² (Reference O’ConnorO’Connor, 1993). Although the lake reached its maximum extent a few thousand years after the LGM, it was present during the Smiths Fork maximum and was certainly larger than the modern Great Salt Lake (Reference Oviatt, Currey and SackOviatt and others, 1992; Reference OviattOviatt, 1997). Numerical simulations of the LGM paleo-climate in the Bonneville basin suggest that the mean monthly maximum temperatures in the basin were below 0°C only in January and December (Reference Craig, Timofeyeva and ClementCraig and others, 1997; Reference Timofeyeva, Clement, Craig, Wilson and BuffaloeTimofeyeva and others, 1999). This result indicates that open water was likely present in the early and late winter. This moisture source, combined with the abrupt topography of the Wasatch Front and western Uintas, probably produced tremendous lake-effect snow. This extreme loading of winter precipitation would have forced lower ELAs at the western end of the Uintas despite the warmer T _s values at the lower elevations.

Fig. 9. Modern atmospheric circulation in the western United States. Solid black line is the winter boundary (Reference MitchellMitchell, 1976).

LGM ELA depression

Figure 8 demonstrates that a twofold increase in P at MFBC_ELA during the LGM would have been accompanied by a 5.5°C temperature depression and non-uniform precipitation changes at ELAs across the northern Uintas (Reference Craig, Timofeyeva and ClementCraig and others, 1997; Reference Timofeyeva, Clement, Craig, Wilson and BuffaloeTimofeyeva and others, 1999). The magnitude of LGM ELA depression (AELA) can be determined from these data by substituting values of P (estimated from the elevation difference between each equilibrium line and MFBC_ELA) and a T _s decrease of 5.5°C into Equation (7). Estimates of ΔELA for the LGM obtained this way range from 650 to 1250 m, with a mean of 900 m, suggesting that previous estimates (∼1475 m from Reference SchlenkerSchlenker, 1995; ∼1000 m from Reference Richmond, Wright and FreyRichmond, 1965), which were based solely on inferred summer temperature changes, are too great.