In the Colorado Rocky Mountains, the association of high topography and low seismic velocity in the underlying mantle suggests that recent changes in lithospheric buoyancy may have been associated with surface uplift of the range. This paper examines the relationships among late Cenozoic fluvial incision, channel steepness, and mantle velocity domains along the western slope of the northern Colorado Rockies. New 40Ar/39Ar ages on basalts capping the Tertiary Browns Park Formation range from ca. 11 to 6 Ma and provide markers from which we reconstruct incision along the White, Yampa, and Little Snake rivers. The magnitude of post–10 Ma incision varies systematically from north to south, increasing from ∼500 m along the Little Snake River to ∼1500 m along the Colorado River. Spatial variations in the amount of late Cenozoic incision are matched by metrics of channel steepness; the upper Colorado River and its tributaries (e.g., Gunnison and Dolores rivers) are two to three times steeper than the Yampa and White rivers, and these variations are independent of both discharge and lithologic substrate. The coincidence of steep river profiles with deep incision suggests that the fluvial systems are dynamically adjusting to an external forcing but is not readily explained by a putative increase in erosivity associated with late Cenozoic climate change. Rather, channel steepness correlates with the position of the channels relative to low-velocity mantle. We suggest that the history of late Miocene–present incision and channel adjustment reflects long-wavelength tilting across the western slope of the Rocky Mountains.

One of the outstanding tectonic questions in western North America regards the development and support of high topography (Fig. 1). It has long been recognized that correlations exist among high topography (Gregory and Chase, 1994), low-seismic-velocity mantle (Grand, 1994; Schmandt and Humphreys, 2010), high heat flow (Sass et al., 1971; Reiter, 2008), relatively thin crust (Sheehan et al., 1995; Hansen et al., 2013), and extrusive volcanism (Larson et al., 1975; Kunk et al., 2002). Although these data point to a role for mantle buoyancy in support of high topography, questions remain about when and how such buoyancy was established. A variety of potential mechanisms have been proposed, including: hydration of lithospheric mantle (Humphreys et al., 2003) and/or thermal re-equilibration following removal of the Laramide slab (Roy et al., 2004, 2009), delamination and/or removal of lithospheric mantle (Elkins-Tanton, 2005; Levander et al., 2011), and changes in the mantle flow field (Moucha et al., 2008; van Wijk et al., 2010; Forte et al., 2010; Liu and Gurnis, 2010).

Recent geophysical studies focused on the Colorado Rockies (Aster et al., 2009; Schmandt and Humphreys, 2010) reveal a prominent region of anomalously slow P- and S-wave speeds (Coblentz et al., 2011; Karlstrom et al., 2012) that resides in the upper mantle beneath the region of highest topography (Fig. 2). This observation reaffirms previous conclusions that support high topography in Colorado largely residing in the upper mantle (Grand, 1994; Sheehan et al., 1995). In fact, the Colorado Rockies exhibit some of the thinnest crust along the range, and a negative correlation between crustal thickness and high topography also favors mantle support for high topography (Hansen et al., 2013). The timing of when this buoyancy was established, however, is not known directly.

The timing and patterns of incision along fluvial systems within and adjacent to the Rocky Mountains suggest a possible role for differential uplift of the range relative to the Colorado Plateau and Great Plains. In the northern Colorado Rockies, the onset of fluvial incision appears to coincide with the cessation of late Tertiary deposition in intermontane basins (Larson et al., 1975; Buffler, 2003; McMillan et al., 2006). Along the eastern flank of the range, incision postdates deposition of the ca. 18–6 Ma Ogallala Formation (McMillan et al., 2002, 2006). Notably, reconstruction of paleo-transport gradients (McMillan et al., 2002; Duller et al., 2012) in these deposits argues for long-wavelength tilting in excess of that expected for a simple isostatic response to exhumation (e.g., Leonard, 2002). Thus, some conclude that tilting must have been, in part, driven by surface uplift within the Rockies (McMillan et al., 2002; Riihimaki et al., 2007; Duller et al., 2012; Nereson et al., 2013), but others argue that most, if not all, recent incision may reflect climatically modulated changes in erosive efficiency (e.g., Anderson et al., 2006; Wobus et al., 2010).

Along the western slope of the range, fluvial incision also appears to have initiated in the past ca. 10 Ma (Kunk et al., 2002; Aslan et al., 2008; Berlin, 2009; Aslan et al., 2010; Karlstrom et al., 2012), but the mechanisms driving incision remain uncertain. In particular, the possibility that incision along the western slope reflects upstream migration of a wave of transient incision in response to drainage integration along the Colorado and Green rivers (e.g., Pederson et al., 2002, 2013) presents an additional complication. In an effort to determine whether late Tertiary incision along the western slope reflects differential rock uplift associated with changes in mantle buoyancy (Aslan et al., 2010; Darling et al., 2012; Karlstrom et al., 2012), we examine the White, Yampa, and Little Snake rivers in Colorado (Fig. 2). Recent analyses of the regional patterns of channel steepness (ksn, a measure of channel slope normalized for contributing drainage area; Kirby and Whipple, 2012) reveal spatial differences that appear to correspond to the position of rivers relative to low-velocity mantle beneath the range (Karlstrom et al., 2012) and do not reflect spatial differences in discharge (Pederson and Tressler, 2012). In this paper we present a detailed analysis of river longitudinal profiles and their relationship to substrate lithology and combine this analysis with new 40Ar/39Ar ages of late Cenozoic basalts that provide new constraints on the timing and magnitude of fluvial incision. Collectively, these observations reveal spatial patterns in both channel steepness and in the magnitude of post–10 Ma incision that help deconvolve the relative roles of climate change, drainage integration, and/or differential rock uplift along the western flank of the Rockies.

Background: Previous Work on Late Cenozoic Incision

Colorado River System

Much of the evidence for late Cenozoic tectonism in the Rocky Mountains relies on the history of incision along major drainages (e.g., McMillan et al., 2006; Riihimaki et al., 2007). An extensive body of work over the past two decades indicates that the Colorado River has incised ∼1100–1500 m across the western slope of the Rockies during the past 10 Ma (e.g., Larson et al., 1975; Kunk et al., 2002; Aslan et al., 2010). We briefly summarize these constraints below; relevant data are compiled in Table 1 and shown for reference on Figure 3. Following Kunk et al. (2002), we exclude sites from within regions known to have experienced collapse during evaporite dissolution.

Most of the key markers used to reconstruct fluvial incision along the main stem of the Colorado River rely on associations of fluvial gravels representing ancestral river deposits with basalt flows (Table 1). The westernmost of these is located at Grand Mesa, just upstream from Grand Junction, Colorado (Fig. 3), where the basal basalt flow (10.8 ± 0.2 Ma; Kunk et al., 2002) overlies river gravels at ∼1500 m above the present-day river (Aslan et al., 2010). Farther upstream, the Colorado River flows between Battlement Mesa and Mount Callahan (Fig. 3). Here scattered basalt boulders on the southern flank of Mount Callahan overlie ancestral Colorado River gravels at ∼1100 m above the modern river (Berlin, 2009). Boulders from the deposit are similar in age (ca. 9.17 Ma; Berlin, 2009) to flows on Battlement Mesa (ca. 9.3 Ma; Berlin, 2009) and are interpreted to represent debris-flow deposits derived from these units and shed northward into the ancient Colorado River valley (Berlin, 2009). Because these deposits have been transported across the axis of the canyon, ∼1100 m represents a minimum value of incision (Berlin, 2009). The average modern transport slopes of debris-flow fans along the northern flank of Battlement Mesa (∼0.07; Berlin, 2009) and the distance from Mount Callahan to the present-day position of the Colorado River (∼4–5 km) imply that there may have been up to ∼280–350 m of additional relief. Thus, it seems likely that incision in the vicinity of Mount Callahan and Battlement Mesa is in the range of ∼1380–1450 m. This value is consistent with estimates (∼1400–1500 m) derived from the projection of Tertiary strata across the canyon from Battlement Mesa to the Roan Plateau (e.g., Bostick and Freeman, 1984). Collectively, these observations imply that an ancestral Colorado River was established across the western slope of the Rockies by ca. 10 Ma (e.g., Aslan et al., 2010) and that the river has subsequently incised ∼1400–1500 m since that time.

Upstream of Glenwood Canyon (Fig. 3), extensive preservation of ca. 10 Ma basalt flows at similar elevations (3000–3400 m) along the Colorado River suggests the presence of a broad, low-relief erosional and/or transport surface prior to ca. 10 Ma (e.g., Larson et al., 1975; Kunk et al., 2002). Incision into this surface was probably ongoing by ca. 8 Ma, as suggested by relationships between basalt flows and fluvial gravel at Spruce Ridge and Little Grand Mesa (Kunk et al., 2002). Moreover, Kunk et al. (2002) suggest that the presence of a young, 3.03 ± 0.02 Ma, high-elevation basalt at Gobbler’s Knob (Fig. 3), ∼730 m above the modern Colorado River, records an increase in the rate of incision during the past ca. 3 Ma. However, the base of the basalt flow at Gobbler’s Knob is unexposed and is not known to be associated with river gravels (Kunk et al., 2002). Thus, the flow may have been emplaced significantly above the ancestral Colorado River ca. 3 Ma and may not directly constrain incision (Aslan et al., 2010). Irrespective of this debate over the pace of incision through time, it is clear that incision in the upper Colorado River near Glenwood Canyon postdates ca. 10 Ma, similar to the river near Grand Junction. The total amount of incision, however, may be somewhat lower, with estimates ranging from ∼750 m to perhaps ∼1200 m (Table 1).

One of the primary tributaries of the upper Colorado River, the Gunnison River, displays a pronounced knickzone in the Black Canyon region (Sandoval, 2007; Aslan et al., 2008; Darling et al., 2009; Donahue et al., 2013). Abundant exposures of a strath terrace level that contain the ca. 0.64 Ma Lava Creek B tephra (Lanphere et al., 2002) reveal spatial differences in incision rate across this knickzone. Downstream of the Black Canyon, incision rates are ∼150 m/Ma (Sandoval, 2007; Aslan et al., 2008; Darling et al., 2009). These rates increase within the Black Canyon to ∼400–550 m/Ma (Sandoval, 2007; Aslan et al., 2008) but decrease again upstream to ∼90 m/Ma above the knickzone (Sandoval, 2007; Aslan et al., 2008). Thus, the Black Canyon knickzone is clearly a prominent expression of transient incision along this system that may be related to the abandonment of Unaweep Canyon by capture of the Gunnison River (e.g., Hansen, 1987; Donahue et al., 2013; Aslan et al., 2014).

Green River System

In contrast to the reasonably well understood history of incision along the upper Colorado River, relatively little is known regarding the timing and magnitude of incision along the western slope of the Rockies in northern Colorado. Here, the White, Yampa, and Little Snake rivers are not entrenched in narrow canyons for long reaches, and deposits of ancient fluvial gravels are exceedingly rare. However, the region was the locus of sediment accumulation during Oligocene through Miocene time (Kucera, 1962; Buffler, 1967; Izett, 1975; Larson et al., 1975; Buffler, 2003; McMillan et al., 2006), and these deposits, collectively referred to as the Browns Park Formation (Fig. 4) (originally described by Powell, 1876 and summarized by Kucera, 1962 and Buffler, 2003), have been deeply incised and eroded by the modern drainage system. Thus, the degree of preservation of basin sediments allows for a minimum estimate of both the timing and magnitude of mass removed by fluvial activity (e.g., McMillan et al., 2006).

Regionally, the Browns Park Formation is exposed in the Elkhead Mountains in the northeast, the Flat Tops in the south, and along the Browns Park graben in the west (Fig. 4). There are two, informally defined, members of the Browns Park Formation—a lower basal conglomerate that rests unconformably on older strata and an upper sandstone (Buffler, 2003). The basal conglomerate is generally thin (<100 m) but thickens and becomes coarser grained toward the margins of the basin; this unit is interpreted to represent alluvial fans being shed westward from the Park and Sierra Madre ranges toward the Sand Wash Basin (Buffler, 2003) and may be correlative with deposits elsewhere referred to as the Bishop Conglomerate (Boraas and Aslan, 2013). The upper sandstone of the Browns Park Formation, in contrast, ranges up to ∼670 m thick and consists of sandstones of both eolian and alluvial origin (Buffler, 2003). Paleocurrent indicators in these sandstones suggest transport directions toward the NNE (Buffler, 1967, 2003).

The age range of the Browns Park Formation is relatively well known from intercalated tuffaceous deposits; these range from ca. 24.8 Ma near the base of the sandstone member to ca. 8.2 Ma near the top of present exposure (Izett, 1975; Luft, 1985; summarized by Buffler, 2003). At City Mountain (Fig. 4, Locality 1), a latite porphyry intruding the formation has been dated to 7.6 ± 0.4 Ma (Buffler, 1967). Additionally, a volcanic tuff near the top of the Browns Parks along Vermillion Creek (Fig. 4, Locality 2) has been dated at 9.8 ± 0.4 Ma (Naeser et al., 1980). Collectively, these data suggest that sediment accumulation in the region continued from ca. 24 to 8 Ma.

Of particular relevance to this study are basalt flows that cap mesas and buttes throughout the region and often overlie thick sections of Browns Park Formation (∼400–600 m; Buffler, 1967, 2003). The age of the uppermost Browns Park Formation is similar to the flows themselves (K-Ar ages ranging from 9.5 ± 0.5 Ma to 10.7 ± 0.5 Ma; Buffler, 1967, 2003). Because these flows overlie the Browns Park Formation, they are broadly consistent with a minimum age for the formation of ca. 8–10 Ma (Buffler, 2003). Field relationships suggest, however, that local relief generation during fluvial incision likely postdates basalt emplacement, and thus we pursued refined chronology from selected localities in the region.

New Constraints on Late Miocene Incision in Northern Colorado

In order to refine our understanding of the switch from deposition of the Browns Park Formation to incision along modern rivers, we supplement existing chronology with new 40Ar/39Ar ages from basalt flows. Localities were carefully chosen where local relationships between the timing of deposition and emplacement between volcanic units allowed us to reconstruct the magnitude of incision along primary rivers or their tributaries. Generally, these localities are characterized by basalt flows that cap mesas and represent a formerly continuous flow or sequence of flows that has been dissected by incision along modern rivers (Fig. 5). In a few cases, where flows are absent, we use the exposed thickness of the Browns Park Formation where the uppermost strata are well dated by interbedded tuffs or intrusions that place bounds on the position of the ancestral land surface. Because ancestral river gravels are not preserved in these localities, our results do not constitute a measure of fluvial incision sensu stricto (Burbank and Anderson, 2011). Rather, they provide conservative estimates for the amount of relief generated in the landscape during fluvial incision.

The region has experienced extensional faulting in late Miocene time (e.g., Kucera, 1962; Buffler, 1967). Although fault slip is generally limited to a few hundred meters, displacement could have led to disruption of formerly continuous basalt flows. Therefore, we confine our analysis to markers of incision within a given fault block. At each locality, we compare our results to the local thickness of preserved Miocene basin-fill sediments (Table 2). Because the upper member of the Browns Park Formation is typically subhorizontal, the exposed vertical thickness of the Browns Park Formation provides a minimum bound on fluvial incision. Our analyses utilize U.S. Geological Survey (USGS) 1° × 2° quadrangles (Tweto, 1976), the geologic map of Wyoming (Love and Christiansen, 1985), and modern National Elevation Data set topographic data. A summary of results is shown in Table 2, and detailed 40Ar/39Ar methods, data, and results can be found in the Supplemental File1.

Elkhead Mountains Region

The Elkhead Mountains represent a significant area of late Tertiary volcanism and comprise high topography near the Colorado-Wyoming border (Fig. 4). The northern flanks of the range are drained primarily by the Little Snake River, whereas the southern portions of the range lie within the Yampa River watershed. Late Tertiary volcanics of the Elkhead Mountains intrude and overlie the Browns Park Formation (Buffler, 2003) and form elevated mesas ideal for reconstructing the amount of post-incision relief. Of importance to this study, late Cenozoic extensional faulting is documented in the region, and displacement across graben-bounding faults (Fig. 6) may be on the order of ∼300–600 m (Buffler, 1967).

Battle Mountain, Squaw Mountain, and Bible Back Mountain. Basalt flows cap the Browns Park Formation in three locations north and south of the Little Snake River (Fig. 6). Atop Battle Mountain, the basal contact of these flows is exposed in a recent landside; the underlying Tertiary strata contain two thin, ∼0.5-m-thick, tuffaceous layers. The elevation of the flow base is ∼2680 m and stands ∼650 m above the elevation of the Little Snake River. We determined a 40Ar/39Ar age of 11.46 ± 0.04 Ma of the basalt flow, which is consistent with the older K-Ar age of ca. 11 Ma (Buffler, 2003).

Squaw Mountain sits directly across the Little Snake River southeast of Battle Mountain (Fig. 6). Here, basalts also cap the mesa, but their base is not exposed, complicating the interpretation of whether these outcrops represent extrusive flows. Outcrops are non-vesiculated and free of significant phenocrysts, and evidence for an intrusive or extrusive origin is equivocal. However, exposed just below the base of the outcrop are deposits of a volcanic breccia that is typically associated with extrusive flows elsewhere in the region (Buffler, 1967). These volcanic breccia deposits suggest a local surface vent, and we follow Buffler (1967) in considering the deposits atop Squaw Mountain as extrusive. The exposed thickness of the probable flow atop Squaw Mountain is ∼20 m. We obtained a new 40Ar/39Ar age on the lowest exposure found of 11.45 ± 0.04 Ma, which overlaps in age with the age of the flow at Battle Mountain. The lowest exposure is at an elevation of ∼2550 m and sits ∼520 m above the modern elevation of the Little Snake River.

Overall, the basalt flows at Battle Mountain and Squaw Mountain lie directly across the Little Snake River from one another (Fig. 6), are of essentially identical age, and are at broadly similar elevations. The relationship of these two basalt flows to the Little Snake River thus provides an opportunity for estimating the magnitude of fluvial incision along the Little Snake directly. Here, we assume that the ca. 11.5 Ma land surface extended between Battle Mountain and Squaw Mountain. Taking the average elevation of the two flow bases, ∼2600 m, above the modern elevation of the Little Snake, ∼2030 m, yields an estimate of fluvial incision of ∼580 m since ca. 11.5 Ma. This direct reconstruction of fluvial incision is similar to the exposed thickness of Browns Park Formation along the Little Snake and Yampa rivers.

At Bible Back Mountain (Fig. 6), the base of a ∼10-m-thick, columnar-jointed flow is exposed on the southern flank of the peak. Here, it appears that there may be a second flow of similar thickness above this outcrop, but the nature of the exposure made this upper outcrop inaccessible. We obtained a new 40Ar/39Ar age of the basal flow outcrop of 11.46 ± 0.04 Ma (Table 2). The elevation of the flow base is ∼2550 m and sits ∼550 m above the modern Little Snake River. Map relations suggest that volcanic material is present at lower elevation toward the northwest, as mapped by Buffler (1967); these deposits are discontinuous remnants and probably represent debris downslope of the unit. The similarity of the amount of incision (∼550 m) to that determined between Squaw and Battle mountains above lends confidence that this is a relatively robust measure of the amount of relief generated during Miocene–Pliocene incision.

Black Mountain and Mount Welba. Geologic relationships between basalt flows in the southwestern Elkhead Mountains (Fig. 6) show a markedly different relationship between the local thickness of Browns Park Formation and their elevation above the modern river. At Black Mountain, extensive deposits of vesiculated, basaltic debris cover the area adjacent to and directly below the mesa-shaped peak, but exposures are rare, and the base of the flow (or sequence of flows) is not exposed. We sampled an outcrop on the northeast end of the main ridge and determined a 40Ar/39Ar age of 10.92 ± 0.16 Ma (Table 2), similar to ages from the eastern Elkhead Mountains presented above. The lowest exposure of the flow is at an elevation of ∼3160 m.

Nearby at Mount Welba (Fig. 6), exposures are also poor and difficult to access. There are three topographic peaks in the vicinity of Mount Welba. Outcrops of volcanic deposits on the southernmost point, Mount Oliphant, do not display definitive flow textures. However, at Mount Welba itself, we discovered outcrops of weathered, vesiculated basalt inferred to represent an upper flow surface. A sample from this exposure yielded a new 40Ar/39Ar age of 12.60 ± 0.06 Ma (Table 2). The lowest exposure of the flow is at an elevation of ∼3150 m.

The flows at Black Mountain and Mount Welba are ∼500 m higher in elevation than Battle Mountain yet sit atop a slightly thinner section of Browns Park Formation. If we project these elevations to the main valley of the Little Snake River, this would predict ∼1170–1180 m of relief, far in excess of the ∼350–400 m thickness of Browns Park Formation exposed at these localities (Fig. 6). However, the flows at Black Mountain and Mount Welba sit in the footwall block of a NW-trending normal fault system (Fig. 6), and the possibility of syn- or postdepositional displacement along this structure (Buffler, 1967, 2003) makes projection to the Little Snake River subject to significant uncertainty. Rather, we take a more conservative approach of projecting to the nearest tributary within the same fault block, Slater Creek and Elkhead Creek, respectively (Fig. 6); both with headwater elevations at ∼2500 m. This yields local estimates of incision that are 660 m and 650 m from Black Mountain and Mount Welba, respectively. The similarity of these values to the exposed vertical thickness of Browns Park Formation suggests these are a likely measure of relief generation during fluvial incision.

Sand Mountain. A thick (>500 m) section of Browns Park Formation is mapped in the southeastern Elkhead Mountains (Snyder, 1980). The upper ∼300 m of the formation is well exposed in a landslide scar along the eastern flank of Sand Mountain (Fig. 6). Here, a sequence of tuffaceous deposits was dated by (Snyder, 1980); ages range from ca. 12 Ma near the base of the section to 9.2 ± 1.7 Ma at the top. The section is capped by andesitic deposits that form the mesa-like summit of Sand Mountain proper; portions of these deposits have been alternatively interpreted as extrusive (Buffler, 1967) and intrusive (Snyder, 1980).

We re-evaluated these relationships along the eastern flanks of Sand Mountain and observed local relationships that support both interpretations. Beneath the summit, andesite is found at similar elevations to horizontal strata of the upper Browns Park Formation on either sides of a steep gully, suggestive of a subvertical, intrusive contact. But, we also discovered outcrops of porphyritic andesite with weak flow banding that overlie the section on the northeastern shoulder of the peak. These relationships lead us to conclude that the andesite is likely a shallow intrusion that has extrusive facies along the flanks of Sand Mountain. We dated a population of 15 individual sanidine crystals concentrated from a sample of the extrusive facies. These exhibited individual ages ranging from ca. 28 to 9 Ma (see Supplemental File [see footnote 1]). The youngest three samples cluster around 9 Ma; a weighted mean from these three crystals is 9.05 ± 0.04 Ma (Supplemental Fig. 5 [see footnote 1]). We consider this a best estimate for the age of the volcanic deposit because the older crystals were likely xenocrystic and entrained during emplacement and/or flow of the andesite.

This age places a minimum bound on the age of the Browns Park Formation at Sand Mountain. Our results are consistent with the older fission-track age of the uppermost tephra in the deposit (9.2 ± 1.7 Ma; Snyder, 1980) but provide a more precise age. Notably, the Browns Park Formation must have been present for the intrusive relationships described above. However, we consider it likely that parts of the andesite were extruded on top of the Tertiary strata, and, thus, that the present exposures of the Browns Park Formation represent most of the pre-incision thickness. Locally, these inferences imply that fluvial incision and erosion into the Sand Wash basin did not begin until sometime subsequent to ca. 9 Ma. The exposed thickness of Browns Park sediments in the region implies that exhumation of material from this portion of the Sand Wash basin was at least 500–600 m, consistent with our estimates of incision from other parts of the Elkhead Mountains.

Flat Tops Region

Near the headwaters of the Yampa and White rivers (Fig. 4), a laterally expansive sequence of at least 27 stacked basalt flows make up the large, high-elevation mesas for which the Flat Tops Range is named (Larson et al., 1975). Here, basalt flows comprise an overall thickness of ∼470 m and range in age from ca. 24 to 9.6 Ma (Larson et al., 1975; Kunk et al., 2002). Individual flows range in thickness from 3 m to ∼60 m where locally ponded against paleotopography (Larson et al., 1975). In the southwest of the range, most of the stratigraphy is composed of superposed flows, which become increasingly intercalated with the Browns Park Formation toward the northeast (Fig. 7), in the direction of the Yampa River valley and the Park Range (Fig. 4). Overall, the sequence of stacked basalt flows is relatively conformable and lies within several hundred meters elevation from one another, despite the wide range in age from ca. 24 to 10 Ma (Larson et al., 1975). This relationship suggests that basalts were likely extruded onto a low-relief surface that persisted in the Flat Tops region until ca. 10 Ma. Thus, we follow Larson et al. (1975) in inferring that present-day canyons that dissect formerly continuous flows provide a measure of incision subsequent to that time.

We estimate the amount of fluvial incision in the uppermost headwaters of the Yampa and White rivers by averaging the highest elevation of the basalt surface on both sides of the modern valley and subtracting the elevation of the modern river channel. Across most of the Flat Tops region, the highest interfluves are capped by ca. 20 Ma basalt flows (Larson et al., 1975), but a few mapped flows that cap the highest peaks (Derby Peak, W Mountain, and Sugarloaf Mountain; Fig. 7) range from ca. 15 Ma to as young as 9.6 ± 0.5 Ma (Larson et al., 1975). Although the former extent of all of these flows is uncertain, their presence on the flanks of the volcanic pile that comprises the Flat Tops (Fig. 7) suggests that the present-day relief must have developed subsequent to their deposition. Thus, we consider ca. 10 Ma as a reasonable bound on the timing of local relief generation in the upper tributaries of the White and Yampa rivers.

In the headwaters of the White River (A–A′, Fig. 7) from Lost Lakes Peak to Sable Point, it appears that there has been ∼900 m of fluvial incision in the past 9.6 ± 0.5 Ma. In the headwaters of the Yampa River (B–B′, Fig. 7) from Orno Peak to Flat Top Mountain, the magnitude of incision appears to be somewhat less, ∼700 m, but still greater than observed in the Elkhead Range.

Yampa River Valley

The third region we studied is in the headwaters of the Yampa River, north and east of the Flat Tops Range (Fig. 4). Near the town of Yampa, Colorado, the river flows north in a fault- bounded valley before making a series of sharp bends; east toward Woodchuck Mountain, north parallel to the flank of the Park Range (Fig. 8), and eventually west at Steamboat Springs, Colorado (Fig. 4). Along much of its course, the river flows in Cretaceous Mancos Shale and the overlying Browns Park Formation, both of which have been intruded by young dikes and volcanic plugs (e.g., Kucera, 1962).

Lone Spring Butte. In the western half graben, a ∼10-m-thick, porphyritic, flat-lying basalt flow with moderately well-developed flow banding is exposed atop Lone Spring Butte (Fig. 8). In hand sample, the basalt has phenocrysts of olivine, plagioclase, and mafic accessory minerals. The base of the flow is at an elevation of ∼3090 m, ∼640 m above the modern Yampa River. This flow unconformably overlies gently dipping, coarse boulder conglomerates of the basal Browns Park Formation. Boulders up to ∼1 m in diameter are composed of crystalline gneisses and granites, similar to those exposed in the Park Range east of the valley (Fig. 8; Kucera, 1962). Bedding within the deposit dips ∼20°–25° west and appears to have been tilted in the footwall of an east-dipping normal fault, which defines the Yampa Valley half graben (Fig. 8). Volcanic ash from a thin Browns Park deposit overlying the basal conglomerates has a zircon fission-track age of 23.5 ± 2.5 Ma (Izett, 1975; Luft, 1985), confirming that the underlying conglomerate represents the base of the formation.

Deposits of volcanic breccia, previously described by Kucera (1962) and Buffler (1967), are also exposed along the flank of Lone Spring Butte, ∼300–400 m below the base of the basalt flow. Similar deposits are present locally throughout the Yampa River valley and were termed the Crowner Formation by Kucera (1962); herein we simply refer to these as “Crowner deposits.” At Lone Spring Butte, these deposits consist of poorly sorted, subangular to angular, cobbles of volcanics mixed with lithic fragments of Browns Park Formation, Mancos Shale, and granitic clasts derived from the Browns Park basal conglomerate. Crowner deposits are thin to thick bedded, and individual beds are on the order of a meter thick. The bedding is generally horizontal planar, although there is a minor amount of small-scale cross bedding in sandier facies. Cobble- to pebble-rich facies are poorly sorted and massive. Crowner beds dip concentrically inward in a ring-like geometry. Collectively, these observations suggest that the Crowner deposits represent maar deposits developed during phreatomagmatic interaction of volcanic intrusions into groundwater-saturated Browns Park Formation sandstones (Buffler, 1967). Thus, it is possible that these units were deposited close to the position of the ancestral land surface along the flank of Lone Spring Butte.

We sampled several of these volcanic units for 40Ar/39Ar chronology. A sample from the basalt flow capping the mesa of Lone Spring Butte yielded a 40Ar/39Ar age of 6.15 ± 0.03 Ma (Table 2). The relatively thin exposure of Browns Park Formation (∼80 m) preserved between the tuff (ca. 23.5 Ma) and the basalt flow (ca. 6 Ma) seems to suggest that a significant amount of sediment was removed by erosion prior to the emplacement of the basalt flow atop Lone Spring Butte.

We also dated samples that constrain the age of the Crowner deposits at Lone Spring Butte. A basaltic clast, contained within bedded Crowner deposits, yielded an 40Ar/39Ar age of 7.0 ± 0.4 Ma, consistent with the eruptive age of the basalt flow. We also obtained a younger age of 4.62 ± 0.05 Ma from an intrusive dike that crosscuts bedded Crowner deposits. Notably, all three of these ages attest to a significant episode of volcanism at ca. 7–5 Ma in the present-day Yampa River valley, consistent with recent age determinations on relict volcanic necks in the region (Cosca et al., 2014).

Relationships between deposits at Lone Spring Butte and the underlying Browns Park Formation make determination of the timing and amount of fluvial erosion difficult in this locality. The base of the 6.15 ± 0.03 Ma flow atop Lone Spring Butte sits ∼630 m above the Yampa River (Fig. 8), and a simple interpretation would suggest that all of this relief postdates ca. 6 Ma. However, the presence of the angular unconformity between the base of the flow and the underlying Browns Park Formation suggests that there may have been significant erosion and removal of the upper Browns Park prior to ca. 6 Ma. Notably, if the Crowner deposits on the flank of Lone Spring Butte indeed represent a paleo-land surface, their present-day position below the summit implies that a minimum of ∼300–400 m of relief existed by ca. 6 Ma. Thus, although it is possible that incision did not begin until after 6 Ma in this locality, the relationships observed between the basalt flow atop Lone Spring Butte, the underlying ash, and the Crowner deposits make it seem likely that some erosion of the Browns Park began prior to ca. 6 Ma in the Yampa River valley.

Woodchuck Mountain. Toward the northeast, the Yampa River makes a sharp turn to the east and enters a second half graben along the flank of the Park Range (Fig. 8). Basalt flows are poorly exposed atop another butte named Woodchuck Mountain (Fig. 8) but appear to be at least ∼50 m thick. At the top of Woodchuck Mountain, the topography is expansive and approximately flat, suggesting the top of a flow surface. Here, a sample from a rubbly outcrop yielded a 40Ar/39Ar age of 6.04 ± 0.04 Ma (Table 2). A second sample was collected from dark basalt outcrop with moderately developed flow banding ∼65 m lower in elevation (∼2620 m). This sample yielded a similar age of 5.97 ± 0.06 Ma. The proximity of Woodchuck Mountain to the Yampa River and the presence of Browns Park Formation beneath the flow make this a robust site to estimate that ∼460 m of relief has developed following basalt emplacement at ca. 6 Ma.

Summary: Mio-Pliocene Differential Incision along the Western Slope

Local relationships between volcanic deposits dated with new 40Ar/39Ar ages (Table 2) and the Browns Park Formation provide new constraints on the timing and magnitude of incision along northern rivers draining the western slope of the Rockies (White, Yampa, and Little Snake rivers). Regionally, basalt flows capping the Browns Park Formation in the northern and western Elkhead Mountains require that fluvial incision along the Little Snake River began sometime after ca. 11 Ma. Given that the youngest ages obtained from the uppermost strata in the Browns Park Formation are ca. 9 Ma at Sand Mountain (Snyder, 1980; Luft, 1985) and ca. 8.5–8.2 Ma in Browns Park proper (Izett, 1975; Naeser et al., 1980; Luft, 1985), it seems likely that incision began shortly after ca. 9 Ma. Similarly, the presence of ca. 10 Ma volcanic deposits atop modern interfluves in the Flat Tops Range (headwaters of White and Yampa rivers) suggest that incision postdates ca. 10 Ma.

In the Yampa River valley proper, geologic relationships regarding the timing of incision are somewhat more complicated. The hiatus in time associated with the unconformity below Lone Spring Butte (ca. 23 Ma to 6 Ma) implies that a significant, but unknown, amount of material could have been removed, perhaps related to tilting during extensional faulting (Buffler, 2003). However, whether this erosion occurred between ca. 9 and 6 Ma, as might be inferred from relationships described above in the Elkhead Mountains, or whether it occurred farther back in the Miocene, is unknown. As noted above, the presence of ca. 7 Ma clasts within the Crowner deposits that were transported at the surface implies that some topographic relief was present during the eruption of 5–7 Ma volcanics in the Yampa River valley (e.g., Cosca et al., 2014). Unfortunately, we are unable to be place quantitative estimates on the amount of relief. Geologic relationships at Woodchuck clearly imply >400 m of post–ca. 6 Ma incision. Thus, although it seems likely that the onset of incision across the region occurred prior to 6 Ma, it is also possible that incision did not initiate until as recently as ca. 6 Ma.

Regardless of the exact timing (6–9 Ma), our results suggest that the total amount of post–ca. 10 Ma incision varies from north to south across the study area. Relationships in the Elkhead Mountains clearly indicate that incision post–10 Ma was limited to 550–650 m. In the Yampa River valley, adjacent to the Park Range, we see similar values (Fig. 9). However, the amount of post–10 Ma incision appears to be somewhat greater in the Flat Tops, ranging up to ∼900 m (Fig. 9). All of these estimates are significantly lower than the ∼1200–1500 m of incision known to have occurred along the upper Colorado River system during broadly the same time period (Fig. 9).


Channel Profiles as a Guide to Landscape Forcing

Analysis and interpretation of longitudinal profiles of bedrock channels that are actively incising into mountainous landscapes (e.g., Whipple, 2004) has become a relatively common tool to guide the interpretation of landscape evolution in erosional settings. Although these analyses are typically conducted in convergent mountain ranges where differential rock uplift is associated with permanent deformation of the crust (e.g., Seeber and Gornitz, 1983; Merritts and Vincent, 1989; Snyder et al., 2000; Kirby and Whipple, 2001; Kirby et al., 2003; Duvall et al., 2004; Safran et al., 2005; Wobus et al., 2006; Harkins et al., 2007; Ouimet et al., 2009; Kirby and Whipple, 2012), recent studies export these techniques to regions of long-wavelength, epiorogenic uplift (e.g., Karlstrom et al., 2012). Here, we use channel profile analysis to examine possible drivers of Miocene exhumation related to possible epiorogenic uplift along the western flank of the Rocky Mountains. We provide only a brief introduction to the techniques below, and the reader is directed to several reviews of the subject for a more comprehensive examination of this technique (Whipple, 2004; Kirby and Whipple, 2012; Whipple et al., 2012).

Channel profile analysis exploits the empirical scaling relation between the local channel gradient (S) and the contributing drainage area upstream (A). In graded channel profiles (Mackin, 1948) from mountain ranges around the world, channel slope follows an empirical relationship of the form,
where ks is a measure of the relative channel steepness, termed the “channel steepness index,” and θ is the “concavity index,” a measure of how rapidly slope varies with changes in contributing drainage area (e.g., Flint, 1974; Snyder et al., 2000). In practice, the steepness index (ks) and concavity index (θ) can be determined by linear regression of slope (S) against drainage area (A) in log-log space. However, small uncertainties in the slope of this regression (θ) yield large variations in the regression intercept (ks) (Wobus et al., 2006). Thus, several methods for determining a normalized gradient index have been proposed to surmount this influence (e.g., Sklar and Dietrich, 1998; Wobus et al., 2006; Perron and Royden, 2013; Royden and Perron, 2013). Here we follow a large body of work (e.g., Kirby and Whipple, 2012) that determines a normalized channel steepness (ksn) by using a fixed reference concavity (θref); this method has been shown to provide a reasonable comparison of channels with widely different contributing drainage areas (Kirby et al., 2003; Wobus et al., 2006).

Over the past decade, numerous studies demonstrate that the normalized channel steepness index (ksn) co-varies with erosion rate in landscapes at or near steady state (see review in Kirby and Whipple, 2012). Early in the development of the metric, studies were limited to steady-state landscapes where uplift rates were known from independent geomorphic markers (e.g., Snyder et al., 2000; Kirby and Whipple, 2001; Duvall et al., 2004). These results supported theoretical predictions (e.g., Whipple and Tucker, 1999) that the normalized channel steepness (ksn) scales monotonically with rock uplift and/or erosion rate, but that the concavity index (θ) is relatively insensitive to rock uplift and/or erosion rate, provided that rock uplift, substrate properties, and climate were spatially uniform (e.g., Kirby and Whipple, 2001). The success of early studies bolstered the use of channel profile analysis as a tool to determine spatial patterns of rock uplift (Wobus et al., 2006). In recent years, the application of cosmogenic isotopic inventories in modern sediment to measure basin-averaged erosion rates (e.g., Bierman and Steig, 1996; Granger et al., 1996) has enabled comparisons of channel steepness (ksn) and catchment-scale erosion rates (e.g., Safran et al., 2005; Harkins et al., 2007; Ouimet et al., 2009; Cyr et al., 2010; DiBiase et al., 2010; Bookhagen and Strecker, 2012). Thus, all other factors being equal, normalized channel steepness can provide a first-order measure of spatial patterns in differential rock uplift (Kirby and Whipple, 2012).

In practice, numerous additional factors influence the adjustment of river profile gradient to erosion rate. These include: variably resistant lithology (Moglen and Bras, 1995; Duvall et al., 2004; Pederson and Tressler, 2012), climatically forced spatial variations in discharge (Roe et al., 2002; Bookhagen and Strecker, 2012), the role of thresholds and temporal distributions of discharge events (Snyder et al., 2003; Tucker, 2004; Lague et al., 2005; DiBiase and Whipple, 2011), and adjustments in channel hydraulic geometry (Duvall et al., 2004; Finnegan et al., 2005; Wobus et al., 2008). All of these factors may result in a nonlinear scaling between channel steepness and erosion rate (Lague et al., 2005). Although global data compilations (Kirby and Whipple, 2012) suggest that variability among field sites likely reflects differences in substrate lithology and climate (DiBiase and Whipple, 2011), within a given setting, it seems clear that channels experiencing higher rates of erosion and/or rock uplift exhibit greater channel steepness (ksn).

These scaling relationships also provide a means to interpret transient responses to perturbations in base level, either through drainage reorganization or variable uplift rate (e.g., Howard, 1994; Whipple and Tucker, 1999, 2002; Whittaker et al., 2007). Transient river profiles have been recognized in tectonically active landscapes around the world (e.g., Crosby and Whipple, 2006; Wobus et al., 2006; Berlin and Anderson, 2007; Harkins et al., 2007; Kirby et al., 2007; Whittaker et al., 2007, 2008; Cook et al., 2009; Morell et al., 2012; Olivetti et al., 2012). Interpretation of such landscapes can be guided by channel profile analysis. We follow Haviv et al. (2010) and Kirby and Whipple (2012) in distinguishing between “vertical-step” knickpoints—those that form an isolated, steepened reach of a river profile—from “slope-break” knickpoints—those that separate two distinct reaches of a profile with different ksn values. The distinction is that the latter is expected to form in response to a sustained perturbation in forcing (Wobus et al., 2006), whereas the former is often an indication of features that are anchored to the river profile (i.e., a steepened reach across resistant substrate).

Channel Steepness along the Western Slope of the Rocky Mountains

Previous analysis of modern channel profiles draining the western slope of the Colorado Rockies provides motivation for the present study. In a regional-scale analysis, Karlstrom et al. (2012) showed that channels in the upper Colorado River watershed that drain high topography above low-velocity mantle have higher normalized steepness indices (ksn) than those that drain topography developed above mantle with higher seismic wave speeds in the Green River watershed (see fig. 3 of Karlstrom et al., 2012). Notably, this signal does not appear to reflect climatically induced variations in mean annual discharge; the scaling between discharge and drainage area in the upper reaches of the Colorado and Green River watersheds are quite similar (Darling et al., 2012). In fact, re-analysis of these channels by Pederson and Tressler (2012) utilizing historic discharge records shows effectively the same pattern (see fig. 5 of Pederson and Tressler, 2012). Thus, variations in channel steepness along the western slope are not simply an artifact of differences in discharge.

In the second part of our study, we seek to evaluate potential explanations for these variations in channel steepness. One explanation may involve differences in lithology; Pederson and Tressler (2012) suggest that variably resistant substrate is the dominant influence on the position of knickpoints along the Green-Colorado river system. They argue that knickpoints and knickzones are anchored to resistant substrate and act to “decouple” topography from proposed loci of uplift (e.g., along the western edge of the Colorado Plateau; van Wijk et al., 2010). A second explanation may involve differences in the history of relative base-level fall, as upstream migration of knickpoints reflecting integration of the lower Colorado River (Cook et al., 2009; Darling et al., 2012; Pederson et al., 2013) may have influenced both patterns of incision and channel steepness across portions of the drainage network. Because these rivers may not be in steady state (e.g., Berlin and Anderson, 2007), we seek to identify transients in the system that may be associated with variations in channel steepness and distinguish these from knickpoints that are anchored to locally resistant substrate (e.g., Pederson and Tressler, 2012).

Finally, we compare patterns of channel steepness to the spatial distribution of post–10 Ma incision across the western slope of the Colorado Rockies. We ask whether the observed patterns are consistent with those expected by an increase in erosivity (e.g., Wobus et al., 2010) or a change in base level (e.g., Pederson et al., 2013), or whether regional patterns require a component of tilting associated with buoyant mantle beneath the Colorado Rockies.

Channel Profile Analysis

We determine normalized channel steepness values (ksn) for six of the major rivers draining the western flank of the Rockies: the Colorado, Gunnison, and Dolores rivers and the White, Yampa, and Little Snake rivers upstream of their respective confluences with the Green River. Extraction of channel profiles and determination of channel steepness values follow the methods of Wobus et al. (2006); codes are available at Topographic data and upstream drainage area were extracted from a USGS 30 m digital elevation model (DEM). To reduce noise associated with the pixel-to-pixel channel slope, elevation data were smoothed using a moving-average window of 1 km and channel slopes calculated over a fixed vertical interval of 12.192 m (equivalent to the 40 ft contour interval of the original data used to generate the DEM).

Topographic data along the Colorado, Gunnison, and Dolores rivers contain artifacts that represent man-made reservoirs, the largest of which significantly influence local slope-area relationships along channel profiles (e.g., Karlstrom et al., 2012). The locations of these reservoirs were verified against a USGS database and were manually removed by linear interpolation of the channel elevation just upstream and downstream of each reservoir.

We analyzed topographic data along all six channels on log(S)-log(A) plots and used linear regression to determine values of ksn along each channel (cf. Wobus et al., 2006). A reference concavity (θref) of 0.45 was used for all ksn analyses in this study. We calculated steepness indices (ksn) across a fixed interval along each channel of 0.5 km. We binned these measurements every 10 km and calculated the mean and standard deviation. The average ksn value for each bin then provides a measure of “local normalized channel steepness” or “local ksn” at a spacing of 10 km, and the standard deviation provides an estimate of the error for each bin. This approach allowed for an objective measure of channel steepness that is not tied to a choice of regression interval (e.g., Kirby and Ouimet, 2011) and facilitated comparison to reaches of the channels underlain by variable lithology.

Bedrock geology along rivers in the study area was extracted from the digital geologic maps of Colorado (Tweto, 1979; Green, 1992), Utah (Hintze, 1980; Hintze et al., 2000), and Wyoming (Love and Christiansen, 1985; Green and Drouillard, 1994) and divided into the map units shown in Figure 10. These allowed us to examine whether streamwise variations in channel steepness were tied to lithologic variations along the channel at length scales >10 km (Figs. 11 and 12). To compare differences among channels, we evaluate the mean normalized steepness (ksn) of reaches that are underlain by substrate with similar mechanical characteristics. We focus on two primary rock types—Tertiary sandstones, which include the Wasatch and Uinta formations, as well as the Browns Park Formation, and Cretaceous shales (Lewis and Mancos formations). In a recent study of rock strength, Tressler (2011) found that variations in compressive strength among the former group are minimal. Compressive strength of Cretaceous shales was unable to be determined, due to the overall mechanical weakness of these units (Tressler, 2011), but we assume that variations across the study area are minimal. Therefore, comparison of channel steepness indices along these reaches should reflect adjustment of channel profiles to external forcing, rather than differences in substrate erodibility.

Results of Channel Profile Analysis

Colorado, Gunnison, and Dolores Rivers

The profile of the Colorado River exhibits a broad increase in channel steepness along the central portion of the profile (Fig. 11A). Generally, the lowest values of ksn (∼20–40 m0.9) are observed immediately upstream of the confluence of the Green River; ksn then increases toward values of ∼90–100 m0.9 just downstream of Glenwood Canyon (Fig. 11A). The uppermost ∼200 km of the profile are again gentler, with ksn ∼60–70 m0.9. Superimposed on this general trend, three locally elevated regions of ksn correlate with the position of distinct knickzones along the Colorado River at Westwater Canyon, Glenwood Canyon, and Gore Canyon (Fig. 11A). The association of these knickzones with crystalline basement rocks and their limited spatial extent suggest that these steep reaches are likely anchored to the underlying bedrock lithology, consistent with the interpretations of Pederson and Tressler (2012). However, these local features do not explain the broader signal of steep reaches along the central ∼300 km of the profile (Fig. 11A).

In contrast to the Colorado, the channel profile of the Gunnison River is characterized by a prominent knickzone within the Black Canyon of the Gunnison, ∼400–500 km upstream from the Colorado-Green confluence (Fig. 11B). The reach of the river below the knickzone exhibits local ksn values of ∼60 m0.9, which are consistent with the Colorado River downstream (Fig. 11B). However, local ksn values within the knickzone are much greater, ranging up to ∼770 m0.9 (Fig. 11B). Although the steep reach within Black Canyon of the Gunnison is developed within Precambrian crystalline rocks, similar to those along the Colorado River, recent analysis of incision rates along this portion of the channel network suggest that this knickpoint is associated with spatial differences in incision rate that suggest that this feature represents an upstream migrating wave of incision (Sandoval, 2007; Darling et al., 2009; Donahue et al., 2013). Although it is possible that the knickpoint is linked to autogenic drainage reorganization along the Colorado River network (Aslan et al., 2014), it probably also reflects the influence of resistant lithology in retarding regional incision. Because this knickpoint complicates interpretation of ksn values, we do not attempt a direct comparison with channel steepness along other rivers.

Directly upstream from its confluence with the Colorado River, the Dolores River displays variable but still relatively high values of local ksn (Fig. 11C). These high values of local ksn near the confluence may suggest adjustment of the Dolores River to base-level lowering along the Colorado River or the influence of variable substrate (the Dolores flows through the Permian Cutler sandstone and the Morrison Formation through this section). Much of the profile, however, exhibits local ksn values between ∼30 and 60 m0.9. A prominent knickpoint occurs in the headwaters ∼450 km above the confluence with the Green River (Fig. 11C). Because data on the timing and magnitude of incision are sparse along the Dolores River, we are unable to evaluate whether this feature is transient, similar to the Black Canyon of the Gunnison, or whether this has developed above resistant Paleozoic–Mesozoic substrate (Figs. 10 and 11C). For these reasons, we exclude the Dolores from further discussion.

White, Yampa, and Little Snake Rivers

The White, Yampa, and Little Snake rivers are all tributaries of the Green River that drain the western slope of the northern Colorado Rockies (Fig. 2). Because the Little Snake River is itself a tributary of the Yampa, we discuss these profiles together. The lower reach of the Yampa River, between the confluence of the Little Snake and the Green rivers, coincides with Dinosaur Canyon (Fig. 11E), where the river flows through the eastern tip of the Uinta block (Hansen, 1986). Within this reach, values of local ksn are generally high (100–150 m0.9) and exhibit rather substantial scatter (Fig. 11E). Rocks of the Uinta Mountain Group are typically quite resistant and probably contribute to the steepening of the river profile, either directly (Pederson and Tressler, 2012) or through the input of coarse debris from canyon walls (Grams and Schmidt, 1999). In addition, this region is a locus of late Cenozoic faulting (Hansen, 1986), and it is possible that the profile may be influenced by young or ongoing deformation. Alternatively, the Dinosaur Canyon knickzone may represent a transient feature associated with integration of the Green River into the Colorado River watershed, an event that is thought to have occurred between ca. 8 Ma and ca. 2 Ma (Hansen, 1986; Darling et al., 2012).

The Little Snake River joins the Yampa River just upstream of Dinosaur Canyon, and along most of its reach the profile is characterized by relatively uniform values of normalized channel steepness (Fig. 11D). A singular exception to this occurs in the headwaters, approximately ∼310 km above the confluence with the Green River, where a locally steep reach occurs in crystalline basement (Fig. 11D). Similar to knickpoints along the Colorado River, this knickpoint is characterized by a localized steepening of the profile, and ksn values both upstream and downstream are similar (Fig. 11D). Thus, we interpret this feature as anchored to resistant bedrock. Overall, the morphology of the Little Snake profile above Dinosaur Canyon appears to be consistent with a graded, or equilibrium, profile.

Upstream of Dinosaur Canyon, the Yampa River displays also relatively uniform values of local ksn along most of its profile. There is, however, a notable increase in ksn toward the headwater reaches of the river (Fig. 11E). There are two possible explanations for this increase in channel steepness in the headwaters. First, the river heads in the basalt fields that comprise the Flat Tops, and it seems possible that profile steepening may be associated with abundant coarse debris shed from this range (Larson et al., 1975). However, the headwater region of the Yampa also overlies the western flank of the region of anomalously low P-wave velocity (Fig. 2), and so it is also possible that these steepened reaches reflect long-wavelength tilting associated with this feature.

The White River exhibits a remarkably smooth profile with no obvious knickpoints (Fig. 11F). Although local ksn remains relatively uniform for ∼200 km upstream of the junction with the Green River, ksn values broadly increase toward the uppermost headwaters of the White River (Fig. 11F) from values ∼40 m0.9 to nearly ∼100 m0.9. Again, whether this steepening is associated with coarse debris being shed off of the Flat Tops or whether it is a signal of differential uplift between the headwaters and the Green River remains uncertain. We will address this question further in the regional discussion below.

Summary: Lithologic Influences on Profile Steepness

One of the notable results of this study is that systematic changes in channel steepness along the western slope do not appear to be controlled by differences in lithology. The lower reach of the Colorado in the study area is relatively steep (ksn = ∼90–100 m0.9), whereas the Little Snake is significantly gentler (ksn = ∼40 m0.9). The White (ksn = ∼80 m0.9) and Yampa (ksn = ∼70 m0.9) rivers are intermediate in both geographic distribution and normalized steepness. These differences persist when we restrict our analysis to lithologies with broadly similar mechanical characteristics. The Colorado River exhibits relatively high values of ksn where it flows across Tertiary sandstones equivalent to the Browns Park (ksn = 81.6 ± 38.5 m0.9), within the Wasatch Formation (ksn = 107.3 ± 39.0 m0.9), and, notably, within the Mancos Shale (ksn = 82.6 ±14.8 m0.9) (Fig. 12). In contrast, the profile of the Little Snake River is approximately half as steep within the Browns Park and equivalent sediments (ksn = 36.8 ± 0.1 m0.9), and nearly three times less steep within the Wasatch Formation (ksn = 36.5 ± 6.6 m0.9). Although there is significant variability, channel steepness values along the White and Yampa rivers are intermediate between these end members. This analysis provides compelling evidence that substrate lithology is not the dominant control on variations in channel steepness across the study area. Rather, north-south variations in channel steepness appear to correlate strongly with the magnitude of late Cenozoic incision along the western slope (Fig. 12), a point that we address in our regional interpretations.

Exceptions to the absence of a regional correlation between steepness and lithology occur within reaches of crystalline Precambrian rocks, in the Flat Tops region, and within Dinosaur Canyon along the Yampa River. Along the Gunnison, Colorado, and Little Snake rivers, reaches underlain by crystalline bedrock often coincide with isolated knickpoints that are associated with locally elevated ksn values. As noted above, we interpret these correlations as indicative of locally resistant substrate (e.g., Tressler, 2011; Pederson and Tressler, 2012) and exclude them from our regional analysis. Likewise, the knickzone along the Yampa River through Dinosaur Canyon (Figs. 9 and 11E) was also excluded from regional comparison. Here, locally resistant substrate (e.g., Darling et al., 2009; Pederson and Tressler, 2012), input of coarse debris (Grams and Schmidt, 1999), or ongoing late Cenozoic faulting (Hansen, 1986) all have the potential to influence channel steepness along this reach. Finally, because of the potential for localized steepening associated with coarse debris being shed off of the Flat Tops (Larson et al., 1975), we consider the steep profiles along the uppermost ∼50 km of the Yampa and White rivers as uncertain in origin.

Late Cenozoic climate change (e.g., Wobus et al., 2010), base-level fall during drainage basin integration (Pederson et al., 2013), and differential rock uplift in the Rocky Mountain headwaters (Karlstrom et al., 2012) have all been proposed as possible drivers of late Miocene exhumation along the western slope of the Colorado Rockies. The combination of new constraints on the timing and magnitude of fluvial incision and channel profile analysis presented here demonstrates that (1) the onset of fluvial incision is broadly synchronous at ca. 6–9 Ma along tributaries of the Green and Colorado river systems; (2) channel profile steepness (ksn) of major river systems increases from north to south along the western slope (Fig. 12); (3) differences in profile steepness are independent of both average annual discharge (cf. Pederson and Tressler, 2012) and substrate lithology (Fig. 12); and (4) the steepest rivers have experienced the greatest amount of late Cenozoic incision (Fig. 12). In this section, we consider what potential driving mechanisms best explain the correspondence of steep channels and deep incision across the study area.

Enhanced Fluvial Incision in the Late Cenozoic

One of the potential explanations for late Cenozoic incision along the western slope of the Rockies is the possibility that climatic changes during the late Miocene enhanced the potential for fluvial transport, either through an increase in storminess (e.g., Molnar, 2001, 2004) or increased mean discharge from snowmelt (Pelletier, 2009). Apparent increases in global sedimentation rates between 3 and 5 Ma have often been cited as evidence for an increase in the efficacy of fluvial erosion (e.g., Zhang et al., 2001; Kuhlemann et al., 2002), although the global significance of these findings has recently been called into question based on isotopic archives (Willenbring and von Blanckenburg, 2010).

Along the western slope, evidence for an increase in Pliocene incision rate is limited, however. Along the Colorado River, the key marker often cited as evidence for an increase in Pliocene incision rates is the basalt flow at Gobbler’s Knob (Kunk et al., 2002). As argued previously by Aslan et al. (2010), the absence of fluvial gravels means that relationship of this flow to the position of the river is uncertain. In contrast, if one considers the ca. 640 ka Lava Creek B tephra and basalt flows of known association to the position of river gravels and inset fluvial terrace and fan complexes, incision rates along the Colorado River appear to be relatively constant with time (Aslan et al., 2010). Likewise, incision data from the Gunnison River permit semi-steady long-term differential incision over the past 10 Ma above and below the Black Canyon knickpoint (Donahue et al., 2013). Along northern rivers, markers of younger age are sparse, but the data admit the possibility of relatively constant incision during the past ca. 6–9 Ma. Although it is possible that slightly elevated rates of incision during the past ca. 640 ka (Dethier, 2001) reflect a climatic influence, these rates are only subtly different from post–ca. 10 Ma averages (Aslan et al., 2010). Thus, we consider that the question of whether incision rates increased during Pliocene time remains unanswered along the western slope of the Colorado Rockies.

Regional patterns in the magnitude of fluvial incision and channel steepness, however, argue strongly that climate change is not a primary driver of incision along the western slope. Nearly all models of river profile response to an increase in the efficiency of erosion (e.g., Wobus et al., 2010), regardless of whether this is associated with changes in mean discharge or storminess (e.g., Lague et al., 2005), are characterized by (1) a reduction of steady-state channel gradients that leads to (2) systematically greater incision in an upstream direction (Whipple and Tucker, 1999; Wobus et al., 2010). These expectations are not met along the western slope. The Colorado River has experienced the greatest amount of incision in the past ca.10 Ma, but remains the steepest of the rivers in our study area (Fig. 12). Moreover, it seems unlikely that climate change alone can explain spatial variations in the amount of incision observed along the western slope. It is difficult to envision a change in erosive efficiency that could simultaneously drive ∼1500 m of incision along the Colorado River while only resulting in ∼500 m of erosion along the Little Snake River. These rivers are only a few hundred kilometers apart, have headwaters at broadly similar elevations, and exhibit similar discharge-area relationships today. Overall, the correlation of channel steepness with synchronous, yet spatially variable, fluvial incision appears to rule out climate change as a significant driver of incision in western Colorado; some additional process is required to maintain steep gradients in the face of ongoing incision.

Transient Incision during Drainage Integration

Relative base-level fall during drainage integration has long been thought to be a primary driver of incision and canyon development across the Colorado Plateau (Hunt, 1956; Pederson et al., 2002). Although the present position of Grand Canyon may exploit an older paleocanyon (e.g., Flowers et al., 2007; Wernicke, 2011; Flowers and Farley, 2012), or segments of preexisting canyons (Karlstrom et al., 2014), it seems clear that final integration of the Colorado River through the Grand Canyon occurred between ca. 5 and 6 Ma (e.g., Lucchitta, 1990; Dorsey et al., 2007). Given that incision along the western slope appears to initiate prior to this time—shortly after ca. 10 Ma along the Colorado River (Aslan et al., 2010; Karlstrom et al., 2012) and ca. 6–9 Ma along tributaries of the Green River (this study)—transient incision associated with the final integration of Grand Canyon is unlikely to be the primary driver for the initiation of incision in the Colorado Rockies. Rather, the data presented here bolster the interpretation that transient incision associated with integration of the Colorado River through Grand Canyon is restricted to the middle reaches of the Colorado River (Wolkowinsky and Granger, 2004; Karlstrom et al., 2008; Cook et al., 2009; Darling et al., 2012).

Our results do not preclude the possibility of an older drainage integration event upstream of Lee’s Ferry, however. The presence of ∼1500 m of relief that developed between 35 Ma and 16 Ma in the southern Colorado Plateau (Flowers et al., 2007; Cather et al., 2008) suggests that a paleo-drainage divide may have existed somewhere to the south of the present-day Book Cliffs (Lazear et al., 2013). It is possible that breaching of that divide led to incision along the upper Colorado River and Green River systems, but importantly, this hypothetical event must have pre-dated final integration of the Colorado River through Grand Canyon at ca. 5–6 Ma. Thus, although data from this study seem to rule out incision driven by drainage integration through the Grand Canyon, they leave open the possibility that integration of the upper Colorado River was achieved through a protracted series of integration events.

Relatively little is known about the timing of breeching across the Book Cliffs and the integration of the Green River into the Colorado watershed. It has been hypothesized, however, that the Green River was relatively recently integrated into the Colorado watershed across the Uinta Mountains (Hansen, 1986). Recent dating of high terraces in the Green River basin, downstream of this point, suggests this event occurred before ca. 1.2 Ma (Darling et al., 2012) and sometime after ca. 8 Ma (Hansen, 1986). It seems probable that this integration event explains the ∼100–200 m of relief across the knickzone along the Yampa River through Dinosaur Canyon (Fig. 11E). However, the fact that this knickzone appears to be confined to the lower reaches of the river implies that it is not responsible for the incision we reconstruct along the western slope tributaries.

Importantly, given the modern drainage configuration, the hypothesis that differences in the amount of incision along the Colorado River (∼1500 m) and the White, Yampa, and Little Snake rivers (∼500–900 m) reflect a wave of incision that has propagated upstream along the Colorado River but has not yet reached the northern tributaries (e.g., Pederson et al., 2013) requires that transient incision stalled across the knickzone along the Green River (Desolation and Gray canyons; Fig. 9). There are two problems with this hypothesis. First, the drop in elevation along the Green River through these canyons is <200 m, and thus there does not appear to be enough relief along the steepened reach of the profile to explain the observed difference in incision (∼600–1000 m). The second problem with the hypothesis that incision was driven only by base-level fall (Pederson et al., 2013; but cf. Karlstrom et al., 2013) is that it fails to explain nearly simultaneous incision in both the headwaters of the Colorado River as well as in the Little Snake River. As our results demonstrate, the best estimates of the onset of fluvial incision along both systems is between ca. 8 and 9 Ma, although it remains possible that much of the incision along the Yampa River took place post–ca. 6 Ma. Thus, if incision across the western slope is entirely a response to drainage integration through Grand Canyon, it would require a scenario in which nearly instantaneous propagation of an initial wave of incision made its way throughout the entire system. For unknown reasons, this wave of incision would have continued along the Colorado River but stalled along the Green River in Desolution and Gray canyons (Fig. 9). As we argue below, we find it more likely that incision was driven by local changes in channel gradient during tilting across the western slope.

Differential Rock Uplift and Tilting across the Western Slope

As argued above, neither climatically enhanced incision nor basin integration seem sufficient to explain the patterns of fluvial incision and channel steepness along the western slope of the Colorado Rockies, which appears to leave open the possibility of differential rock uplift between the Colorado Rockies and the Colorado Plateau (e.g., Darling et al., 2012; Karlstrom et al., 2012). The association of steep channels in regions of large-magnitude incision is consistent with this hypothesis, as we expect such relationships in systems adjusted to spatial variations in rock uplift (e.g., Kirby et al., 2003). In the Colorado Rockies, moreover, the spatial correspondence between steep, rapidly incising rivers and presumably buoyant, low-seismic-velocity mantle (Karlstrom et al., 2012) suggests the possibility of a genetic association between fluvial incision and low-velocity mantle beneath the central Colorado Rockies.

At a regional scale, spatial differences in channel steepness, normalized for lithology (Fig. 12), provide perhaps the strongest evidence for a tectonic component driving late Cenozoic incision. Without some forcing mechanism to drive channel steepening in the face of continuing incision, it is hard to explain why rivers would exhibit such systematic differences along the western slope. However, if low-velocity mantle beneath Colorado is associated with dynamic support of topography, our data suggest that the flanks of the anomaly could be (or have been) characterized by long-wavelength tilting between the central Rockies and the Colorado Plateau. Notably, the width of regions of elevated steepness along rivers appears to correspond roughly with the degree to which channels extend across the region of low-velocity mantle (Fig. 13). The Colorado River maintains a steep profile (ksn ∼80–120 m0.9) from Grand Junction to just below Gore Canyon (Fig. 11), where it crosses the axis of low-velocity mantle (Fig. 13). In contrast, the White and Yampa rivers only steepen in the upper ∼100 km of their profiles (Fig. 11), coincident with where they extend over the region of lowest seismic velocities (Fig. 13), and the Little Snake River exhibits relatively uniform steepness values along its entire length, consistent with its position off the flank of the anomaly. We suggest that these associations indicate that channel profiles are still responding to a pulse of uplift that began within the past 6–9 Ma; this adjustment may still be ongoing, as suggested by the knickpoint along the Gunnison River (Donahue et al., 2013).

Some of the apparent tilting and differential rock uplift inferred from the pattern of incision could be a consequence of rebound related to unloading of the lithosphere (e.g., Wager, 1937; Molnar and England, 1990; Small and Anderson, 1995; Pederson et al., 2013). Most attempts to estimate the magnitude and distribution of isostatic rebound across the Colorado Plateau rely on volumetric reconstruction of material eroded over the past 10–30 Ma (Pederson et al., 2002; McMillan et al., 2006; Lazear et al., 2013) and yield generally similar patterns with a locus of rebound in the central and southern Colorado Plateau. The most recent of these models (Lazear et al., 2013) makes refined predictions for the amount of rebound along the western slope of the Rockies. We rely on those predictions here as the current best estimate. In the vicinity of the Little Snake River, rebound is predicted to have been between 300 and 400 m (fig. 7 of Lazear et al., 2013), a value that could explain a sizable fraction of the 500–600 m of incision we observe. Predicted rebound increases toward the south but remains between 500–700 m along most of the Colorado River upstream of Grand Junction (fig. 7 of Lazear et al., 2013). Thus, although isostatic rebound in response to late Cenozoic exhumation has the potential to explain some of the observed incision along rivers draining the western slope, it does not appear to be sufficient to explain the full signal.

Overall, the results of our study appear to require late Cenozoic tilting along the western slope of the Colorado Rockies. Although a quantitative estimate remains beyond our ability to determine, it seems that patterns of incision require several hundred meters of differential rock uplift, in excess of isostatic adjustment, that range from ∼200 m in northern Colorado to perhaps as much as ∼700 m along the Colorado River. We note that these values are similar to the magnitude and wavelength observed along the eastern slope of the Rockies (e.g., Leonard, 2002; McMillan et al., 2002; Nereson et al., 2013), suggesting that both flanks of the range may be responding to changes in mantle buoyancy beneath central Colorado. We also note that extensional deformation (e.g., Buffler, 2003) and the presence of late Cenozoic alkalic volcanism in the Yampa region (Cosca et al., 2014; this study) are both consistent with the addition of buoyancy associated with continued modification of the mantle lithosphere beneath the range (e.g., Hansen et al., 2013). We suggest that long-wavelength tilting along the flanks of the range during the past 6–10 Ma has a tectonic origin associated with differences in the buoyancy of the mantle between the northern Rocky Mountains and adjacent regions.

New chronology of basalt flows in the headwaters of the White, Yampa, and Little Snake rivers allows estimates of the magnitude and timing of fluvial incision along the western slope of the Colorado Rockies. Combined with detailed analysis of the steepness of channel profiles (ksn), these data provide new insights into the history and potential drivers of late Cenozoic fluvial incision across the western slope of the Rocky Mountains and lead to the following conclusions:

  1. Incision along the White, Yampa, and Little Snake rivers postdates ca. 9–10 Ma and most likely predates 6 Ma. This is broadly synchronous with previous studies that infer post–8–10 Ma incision along the Colorado River.

  2. Channel profile steepness (ksn) of major river systems increases from north to south along the western slope, such that the Colorado River is two to three times as steep as the Little Snake River. These differences in profile steepness are independent of both discharge (e.g., Pederson and Tressler, 2012) and substrate lithology.

  3. Spatial variations in channel steepness coincide with apparent differences in the magnitude of late Cenozoic incision. Incision along the Colorado River approaches ∼1500 m, whereas incision along the White and Yampa rivers is less, ∼700–900 m, and incision along the Little Snake River is even lower, ∼550 m.

  4. Collectively, the association between steep channels, deep exhumation, and low- velocity mantle at depth appears to implicate differential rock uplift during the past ca. 10 Ma as the best explanation for late Miocene–present incision along the western slope of the Rockies.

This study was funded by the National Science Foundation–Colorado Rockies Experiment and Seismic Transects (NSF-CREST) project grant EAR-0607808. Additional support provided from grants EAR-0711546 (KEK) and EAR-1119635 (KEK and AA). Brandon Schmandt provided the EarthScope + CREST–derived tomographic images. We thank three anonymous reviewers and the Guest Associate Editor for comments that helped improve the manuscript.

1Supplemental File. 40Ar/39Ar analytical methods and results. If you are viewing the PDF of this paper or reading it offline, please visit or the full-text article on to view the Supplemental File.