THE TERRACES OF JACK CREEK, MADISON COUNTY, MONTANA

James P. Bearzi¹ and William W. Locke; Department of Earth Sciences, Montana State University, Bozeman, MT 59717

(¹Presently with NM State Department of Environment, 1190 St. Francis Drive, Santa Fe, NM 87503)

ABSTRACT

Jack Creek, a tributary to the Madison River of southwestern Montana abutting the classic Cedar Creek alluvial fan, is the site of an exceptionally well-developed terrace sequence. The morphology of the terraces suggests multiple working hypotheses of terrace origin and age which are in part testable through data on landform evolution (soil development and slope evolution). Possible explanations for terrace formation include tectonic activity, base level control, and climatic modification of load and discharge within the stream basin. Each of these should result in distinctively different terrace morphologies. Similarly, climatic fluctuations across Pleistocene and Holocene time should result in a distinctive sequence of ages of terraces.

The long profiles of ten nested terraces of Jack Creek show significant variability in gradient, suggestive of fluctuations in the ratio of load to discharge. However, soil and scarp development criteria suggest only two distinct episodes of terrace formation. The most logical interpretation of the landforms and supporting data is that the deposition of the highest basin fill, in which the terraces are formed, is primarily the result of climate-induced aggradation. Erosion and cutting of the terraces of Jack Creek was a response to lowering of local base level (the Madison River), which in turn was dominantly a response to glacial / interglacial transitions, including the Pleistocene-Holocene climate change, and secondarily a response to river migration.

INTRODUCTION

The Ennis, MT 15' topographic quadrangle (USGS, 1949) is well known to most geologists and physical geographers because of the exceptionally well-formed (and well-labeled!) Cedar Creek alluvial fan, which dominates the southern half of the quadrangle. What is less often recognized is the terrace sequence of comparable dimensions which surrounds Jack Creek, to the north of Cedar Creek on the same quadrangle (Figure 1). Just as the Cedar Creek fan offers educational value in the clarity of its expression, so do the Jack Creek terraces. And, just as the Cedar Creek fan hides subtleties of great interest, so do the Jack Creek terraces. The educational value of the Jack Creek sequence is clear at many levels:

• at the introductory level (physical geology/geography), they must be recognized;
• at the advanced undergraduate level (geomorphology), they can be measured;
• and at the graduate level (Quaternary stratigraphy), they can be explained.

FIGURE 1. Jack Creek study area.

The physiographic settings of the fan and terraces are similar (Figure 1). Both lie at the north end of the upper Madison valley, southwestern Montana. The apex of the fan lies at about 1800 m (6000 ft) and that of the terraces at about 1700 m (5600 ft), and they terminate at the Madison River (local base level) at about 1500 m (4900 ft). The apices are defined by the Madison fault system, which has been active in Quaternary time (Schneider, 1985), and has generated about 7 km of relief between basement rocks interpreted beneath the basin floor (Egbert, 1960) and that exposed in the adjacent Madison Range. The valley floor receives less than 40 cm (16 in.) of precipitation annually, but the streams head at elevations exceeding 3200 m (10,500 ft), which receive >100 cm (40 in.) of precipitation annually. The bedrock of both Cedar and Jack Creek basins is dominantly Cretaceous shales (baked to hornfels in the vicinity of the Tertiary intrusions which support Lone, Cedar, and Fan Mountains, the high points of the drainages). Both drainages supported Quaternary glaciation (Alden 1953; Grabb, 1977), more extensively in the Cedar Creek drainage, and both are affected by mass movement of the shales, more extensively in the Jack Creek drainage.

RECOGNITION

In a typical Physical Geology/Geography course, recognition of landforms is the major goal. The terraces can be recognized by the distinctive "sawtooth" pattern of contour lines, alternating from perpendicular to the creek on the floodplain and terrace treads to parallel to the creek across terrace risers (Figure 2). The best way for introductory students to recognize the terraces is to draw a profile utilizing both master and intermediate contours which cuts some of the terraces. The resolution of the 1949 USGS Ennis 15' quadrangle will not allow the identification of all of the terraces, but 7.5' quadrangles (Cherry Lake, Ennis, Ennis Lake, and Fan Mountain), now available in provisional form, are a distinct improvement. Including the modern channel, from 7 to 11 terraces (depending on base map scale and imagination!) rise from creek level to approximately 65 m (200 ft) above it. Questions which logically might arise from such an exercise include:

• "Why do we show different numbers of terraces?" (subsequent erosion or deposition);
• "Why is the same terrace at different elevations?" (stream origin, former gradient);
• "Why is the same terrace at different heights above Jack Creek?" (stream origin, difference between former and present gradient); and
• "Why (on a linear transect) are some terraces flat and some tilted?" (the profile crosses some terraces parallel to and some perpendicular to the contours, because of the concentric nature of the contour lines - see Cedar Creek fan).

FIGURE 2. Map of terraces and pediments at Jack Creek and representative cross-profiles.

MEASUREMENT

In a typical Geomorphology course, students are introduced to the concept of generating longitudinal profiles of surfaces to interpret bed gradient. Longitudinal profiles of all surfaces (Figure 3) were generated from 1:24,000 scale, 20'(6.1 m) contour interval topographic maps (1949 Missouri River Basin Project) equivalent to the 7.5' provisional quadrangles. All profiles are projected to the current stream position by rotation about the mouth of Jack Creek Canyon, which must have been the fixed source of the stream during Pleistocene time.

The longitudinal profiles of Jack Creek (Figure 3 and Table 1) reveal a strong parallelism among the surfaces, but significant differences between some surfaces. Specifically, T2 (upper part), T6, and T8 are steeper than the others, with T1, T9, and T10 (the modern floodplain) significantly gentler. Several terrace fragments (T4, T5,) are too short to have high confidence in the calculated gradients. The similarities attest to a generally consistent discharge/load relationship, but the differences suggest changes in that ratio across the period of terrace evolution. The terrace gradients vary between about 0.02 and 0.03, suggestive of differences in discharge of about 50%, assuming comparable load, or differences in load of about 50%, assuming comparable discharge. Both of these simplistic models assume comparable channel geometry. The assumptions might best be tested by sedimentologic studies.

Such differences could have resulted from extreme events through the period of terrace evolution (e.g., jokulhlaulps, floods or landslides; Thompson and Jones, 1986) or could represent extended periods of adjustment to changing conditions. Working models for the origin of the terraces, based largely on work elsewhere in the northern Rocky Mountains, include:

• I) they represent a sequence of episodes of glacial aggradation and interglacial degradation across the entire Pleistocene (e.g., Reheis, 1984),
• II) they represent minor climatic and load variability superimposed on a single rapid episode of postglacial degradation following glacial aggradation (e.g., Pierce and Scott, 1982; Adams and Locke, in press),
• III) they represent minor climatic and load variability superimposed on a gradually declining rate of downcutting (e.g., Thompson and Jones, 1986; Bull and Knuepfer, 1987),
• IV) they represent variability in the elevation of and distance to local base level, superimposed upon a single major episode of glacial aggradation and postglacial degradation, (Bearzi, 1987), and
• V) they represent aggradation following earthquakes, with net uplift (thus erosion) of the mountain block, and net subsidence (thus deposition) on the valley block. This latter hypothesis requires that downcutting exceeded deposition in order for terraces to be preserved (cf. Bull, 1984). If deposition exceeded erosion, then an alluvial fan should have formed.

For example, if terraces 1, 2, 4, and 5 represent discharge/load relationships similar to present, T3 and T6-T8 indicate a decreased discharge/load ratio, and T9 is again similar to present conditions (T10), this could represent:

1. an interglacial (T1-T5) / glacial (T6-T8) / interglacial (T9-T10) sequence (with   T3 perhaps reflecting an earlier glacial episode) according to Model I, or
2. a glacial (T1) to interglacial (T10) sequence with minor superimposed fluctuations according to Models II and III, or
3. migration of local base level (the Madison River) towards the Madison Range according to Model IV. Terraces 1-3 (glacial) clearly project well above the modern Madison River (local base level), T7-T10 project to or near present mainstream levels, and T4-T6 could conceivably project to the present local base level, but only at the far side of the present Madison Valley floor.

In a similar setting in central Idaho, but for smaller drainage basins, Pierce and Colman (1986) describe glacial-age fans with slopes of 5.5% and 8% incised by postglacial streams with gradients of 4% and 6% (Ramshorn Canyon and King Canyon, respectively). The similarity in the ratios of maximum to present steepness suggest that the variations seen on the Jack Creek terraces could represent glacial/interglacial transitions. However, resolution of these competing models clearly requires additional information, such as terrace sedimentology or ages.

EXPLANATION

At the level of a graduate geomorphology or Quaternary stratigraphy course, students should be required to synthesize regional chronologies, geomorphic agents, and indicators of relative age. Age is particularly important because it may indicate the effects of climate change, which occurred at recognized periods in the past, relative to tectonic activity, which occurred at unknown rates and times. The relative ages of the surfaces can be estimated by degree of landform development (scarp evolution) and weathering (soil development).

If the terrace sequence is dependent on tectonism, it might show significant offset of earlier terraces by later ground rupture events. If it is dependent on base level change it should correlate with terraces of the Madison River (local base level). If it represents the effect of multiple glaciations it should show a marked increase in relative age upward across the sequence. If it reflects short-term stability within a single glacial-interglacial sequence it should show a very small range of relative ages. Finally, if it reflects changes in load and discharge, it should show significant variability in bed gradient (unless compensating changes in the other variable or in channel geometry occurred).

Regional Stratigraphies

Two paradigms exist for terrace sequences in the northern Rocky Mountains. One (e.g., Moss, 1974) equates terraces with episodes of glacial aggradation and subsequent interglacial degradation. In the Rock Creek valley near Red Lodge, MT, Reheis (1984) defined a terrace sequence consisting of a Holocene, Pinedale (last-glacial), Bull Lake (penultimate glacial), Boyd, and three older terraces, 5, 14, 27, and 66 m and more above the present stream channel and about 5000, 20,000, 120,000, 400,000 and more years old. The other extreme is exemplified by Pierce (1979) in the Yellowstone valley and Adams et al. (in press) in the West Yellowstone basin, where sequences of 5 or more terraces represent episodes of stabilization in post-Pinedale time.

Chronology - Methods

Terrace scarps.  Fifteen to thirty profiles were surveyed across each of the well-defined terrace scarps using a tape measure and Abney level. The operative assumptions are summarized by Pierce and Colman (1986), and include homogeneity of parent material, homogeneity of process, and recognition of the importance of initial scarp height, scarp aspect, and time in the evolution of scarps. Terrace soils.  From one to three pits per terrace were excavated by hand to a depth of 1 meter. Soil descriptions followed Soil Survey Staff (1985) and Birkeland (1984), samples were collected by horizon, and bulk densities were measured using a modified sand excavation method (after McLintock, 1959, and Cassidy, 1981). Soil carbonate was determined following Machette (1978, 1985). In addition, detailed (1:24,000) soils maps of the area have recently been issued (SCS, 1989).

Chronology - Results

Terrace scarps.

When the scarp data are plotted (after Bucknam and Anderson, 1979) as maximum slope angle versus log(scarp height), both the limitations and strengths of the data are apparent (Figure 4). Although the morphologic age of the scarp is that of abandonment of the lower bounding terrace, thus terraces of several heights may be bounded by scarps of a single age, the general parallelism of the terraces defines a restricted range of scarp heights for any given terrace. Three groups of consistent height/slope relationships are defined by the data, with both NE-facing and S-facing scarps less than 6.5 m tall (defined by terraces 5, 6, and 8) clustering distinct from two scarps, both S-facing, both greater than 6 m tall (defined by terraces 7 and 9), which differing from each other by a factor of two in maximum slope angle for a given scarp height. All of the measured scarps except those with T9 at their base are less steep for their height than late Pleistocene scarps of comparable height elsewhere in the West, including similar stream terraces in Idaho (Pierce and Colman, 1986).

FIGURE 4. Maximum scarp-slope angle as a function of scarp height for Jack Creek terrace scarps (numbered by lower bounding tread) and comparable stream and fault scarps.  Dashed line represents south-facing (comparable) postglacial scarps in Idaho (after Pierce and Colman, 1986); colored lines represent Holocene (Fish Springs and Drum Mountains), and mid-Pleistocene (?) fault scarps in Utah (after Bucknam and Anderson, 1979)

Terrace soils.

Although the significance of soil data is limited because of only two pits per surface, the distribution of data on loess caps, "B" horizon thickness, and calcium carbonate content (Table 2) are consistent with three major groups of soil development, thus three distinctly different relative ages. The modern floodplain (T10) differs significantly from terraces 3-9, which in turn differ markedly from T1 and T2. Within the terrace sequence the degree of soil development does not increase consistently with height above the stream (thus relative age) suggesting that local differences in soil-forming factors exceed the effect of time. The consistency of the results suggests that increased precision from more sites and samples would yield the same result.

TABLE 2. Soils data from Jack Creek terraces

DISCUSSION

Despite concentrated effort to model the processes of scarp degradation, dating of scarps by their morphology yields at best a crude estimate. The morphology of scarps at Jack Creek, compared to those of known age from the Great Basin, central Idaho, and SW Montana (Figure 4), suggests a last-glacial (Pinedale) age only for T9. Based on their statistical similarity to a single line, subparallel to but below the lines defined by all last-glacial scarps from the region, T8, 7, 6, and 5 can be assigned to a pre-last-glacial (Bull Lake) age. Scarp data do not allow differentiation within that group, and provide no information on the ages of T4, 3, 2, and 1. Alternatively, of course, the rate of scarp degradation at Jack Creek may exceed that in the forested obsidian sand plain of West Yellowstone (Nash, 1984), in the high interior valleys of Idaho, and in the Great Basin. The climates are similar (e.g., Ennis, MT, Mean Annual Air Temperature = 6.2^oC, Mean Annual Precipitation = 30 cm; Arco and Mackay, ID, MAAT = 5.5^oC, MAP = 23 cm; Pierce and Colman, 1986) so that alternative is considered unlikely, but it is also possible that human uses (irrigated agriculture) have accelerated scarp evolution.

The soil data define a different subdivision of the terrace sequence. T1 and T2 have about 28 and 50 grams of secondary CaCO₃/cm², respectively, compared with an average of 14 + 3 g/cm² for T3-T9. Similarly, T1 and T2 average more than 50 cm of silt loam, interpreted as loess, over fragmental gravelly loam, whereas the lower terraces have between 0 and 20 cm of loess over sandy loams (similar to that of the modern floodplain), over fragmental gravelly loam. T3-T9 are statistically indistinguishable in degree of soil development thus presumably in age. Conversely, T1 and T2 are much older, and T10 (the modern floodplain), much younger. As T2 crosscuts T1, it is morphologically younger than T1, thus the large surplus of CaCO₃ in T2 relative to T1 is problematical. However, both surfaces are so different from the lower/younger surfaces that they must represent a fundamental difference in age.

The total carbonate data allow estimation of numerical as well as relative age, given the necessary assumptions of rate of accumulation. Machette (1985) has summarized quantitative studies of carbonate accumulation in the southwestern United States. In areas with climates warmer and somewhat drier than that of the Jack Creek area (Albuquerque, NM, MAAT = 13.1^oC, MAP = 20.5 cm; San Acacia, NM, MAAT = 14.7^oC, MAP = 21.2 cm; Las Cruces, NM, MAAT = 15.5^oC, MAP = 20.4 cm) total secondary carbonate contents of 16-18 g/cm² required 60 to 80,000 years to form, whereas secondary carbonate contents of 55-79 g/cm² required 250-320,000 years. By comparison, T3-T9, with about 14 g/cm² of secondary carbonate, should be about 55,000, T1 (28 g/cm²) about 110,000, and T2 (56 g/cm²) about 250,000 years old. Given the higher available moisture in the Jack Creek area, these age estimates should be minima: Reheis (1984) measured rates of CaCO₃ influx in the Rock Creek area of 0.3x10^-4 g/cm²/yr, which would increase the estimated ages by roughly an order of magnitude.

To the south of Jack Creek, in the West Yellowstone basin, Pierce (1979) reported loess thicknesses of 20-100+ cm on Bull Lake-age deposits and thicknesses of about 10 cm on Pinedale moraines. The implication is that in mountainous areas glaciation results in major loess deposition, whereas de- and post-glacial processes result in only minor loess deposition. The similarity in loess thicknesses on the Jack Creek terraces to those on surfaces estimated as of post-Bull Lake and post-Pinedale ages supports a similar age assignment here. By analogy, T1 and T2 precede the last (Pinedale) glaciation, and T3-T10 postdate it. As no buried soils were recognized in the loess cover on T1 and T2 (cf., Pierce and others, 1982), it cannot be proven if T1 and/or T2 predate Bull Lake glaciation as well.

Finally, the soil pit data show a marked difference between the 100+ cm of loam and sandy loam on the modern floodplain (1 pit) and the roughly 50 cm of silt, silt loam, loam, and sandy loam over fragmental gravelly loam and sand of the terraces. The observed dominance of coarse sediment in the terraces and fines in the modern floodplain parallels that of Pierce and Scott (1982) in the similar valleys of Idaho. They assign the fine sediments to Holocene times and processes and the coarse deposits to the Pleistocene. The 50 cm veneer of fines over coarse in the terraces may represent a depth of scour, in which case the veneer characterizes T3-T9 as post-glacial, or may represent waning stages of deposition of dominantly coarse sediment fills, in which case only T10 (the modern floodplain) is post-glacial.

Correlation should be possible between the Jack Creek terrace sequence and the Madison River terraces. Although the definitive study on the Madison River terraces is not yet complete, tentative correlation to the Madison River terraces of Lundstrom and Burke (1985) and Lundstrom (1986) 50 km south of Jack Creek makes the Madison River terrace which is about 6 m (20 ft) above the modern river at Ennis Pinedale-equivalent, the terrace which is about 24 m (80 ft) above the modern river Bull Lake-equivalent, and the Cameron Bench (80 m/260 ft above the river), pre-Bull Lake. Corresponding Jack Creek terraces would be T7 and T8 (Pinedale-equivalent), T3 (Bull Lake-equivalent), and T1 and T2 (pre-Bull Lake). The remaining terraces may represent interstades or the local effects of lateral migration of the Madison River. Westward migration of the Madison River would force aggradation, thus a gentler gradient (T4, T5, and T9), whereas eastward migration would initiate incision.

In contrast, Schneider (1990; and Locke, 1990) suggests that Madison River terraces less than about 30 m above the modern river are Pinedale and post-Pinedale in age, those less than about 55 m are Bull Lake and post-Bull Lake, and the Cameron Bench (at Jack Creek) is pre-Bull Lake. Corresponding Jack Creek terraces would be T7-9 = Holocene, T4-6 = Pinedale, T3 = Bull Lake, and T1-2 = pre-Bull Lake.

SUMMARY AND CONCLUSIONS

Terrace slope, soil, and scarp data independently characterize the terraces at Jack Creek: together they allow estimation of the ages of the terraces and evaluation of mechanisms of formation. Unfortunately, they do not lead to similar estimations / evaluations!  The application of multiple tests of relative age and climate to the Jack Creek terraces allows tentative assignments of age to the terraces (Table 3). The aggregate assignments are somewhat sensitive to the confidence in individual techniques. For example, correlation to the Madison River terrace interpretations of Schneider and Ritter (1987) would make T1 and T2 pre-Bull Lake-equivalent (thus the lower terraces probably no older than Bull Lake). Acceptance of an average western U.S. rate of CaCO₃ input allows a similar correlation, as does the subdued morphology of all but the most recent scarps. Given the similarity of loess thicknesses in the Jack Creek area to those in similar settings elsewhere in the region, however, Bull Lake (T1 and T2) and post-Bull Lake ages are possible. The definitive assignment of ages to the Jack Creek terraces awaits numerical age dating of those or comparable surfaces.

TABLE 3. Summary of tentative age indicators.

REFERENCES CITED