Unaweep Canyon (Colorado, USA) is a large, Precambrian-cored gorge that bisects the Uncompahgre Plateau of the northeastern Colorado Plateau, but has no through-flowing axial stream; it is drained by two underfit creeks (East and West Creek) that head at a divide within the canyon. The history of the canyon and its role in drainage evolution of the Colorado River system remain controversial. New mapping of both bedrock and Quaternary units as well as analyses of Quaternary deposits in and near the canyon shed light on its late Cenozoic history, and call into question whether the canyon was incised by a Cenozoic river, or merely exhumed by one. Gravels near the western mouth of Unaweep Canyon (Gateway, Colorado) exhibit a distinctive intermediate volcanic provenance recording the presence of an ancestral Gunnison River; the youngest gravels are dated to 1.46 ± 0.33 Ma. Previously documented coring within the canyon reveals a thick (locally >330 m) fill that includes a lacustrine succession (∼140 m thick), dated to 1.4–1.3 Ma, overlain by stacked paleosols and a thick (∼160 m) conglomeratic unit emplaced between 1.3 Ma and the present, in addition to a basal unit of possible late Paleozoic age. Lake formation reflects catastrophic mass wasting in western Unaweep Canyon that blocked the ancestral Gunnison River, causing partial backfilling of the canyon, and forcing the river to seek a lower elevation exit eastward by breaching the Mesozoic rim at the northeast end of Cactus Park (Mesa County, Colorado). Ultimately, the ancestral Gunnison River joined the lower elevation Colorado River near Grand Junction by 1.3 Ma, incising the East Creek of Unaweep Canyon during the overspilling event.

Thermochronological data from Precambrian basement within Unaweep Canyon and Permian strata at the western mouth of the canyon indicate onset of incision in latest Miocene time (ca. 6–5 Ma), at a time-averaged rate of ∼210–275 m/m.y. Onset of canyon occupation and rapid incision by the ancestral Gunnison River coincided with the timing of integration of the lower Colorado River system to the Gulf of California. The synchroneity of this incision across the Colorado Plateau supports the inference of an ultimate tectonic or epeirorogenic driver for this widespread incision and ultimate drainage integration.

Several aspects of this data set support the previously published hypothesis that the ancestral Gunnison River exhumed a paleovalley. New mapping at the western mouth of the canyon documents a paleovalley filled with Permian strata that leads into the modern Precambrian-hosted gorge of Unaweep Canyon. In addition, the ancestral Gunnison River paralleled the Uncompahgre Plateau before making a 90° turn to bisect the structural axis in a manner that opposes both the northwestward plunge of the uplift and the northeastward dip of its northern flank. The rate of incision of Unaweep Canyon exceeds regional time-averaged incision rates, consistent with removal of sedimentary fill rather than incision of crystalline basement. This hypothesis implies that very ancient landforms can influence drainage evolution in even tectonically active landscapes.

Unaweep Canyon (western Colorado, USA) bisects the northwestward-trending Uncompahgre Plateau on the northern Colorado Plateau and is the only major canyon in the Colorado River drainage not occupied by a river. It penetrates Mesozoic strata deep into Precambrian crystalline basement, but paradoxically hosts two underfit drainages, East Creek and West Creek, which flow in opposite directions from a gentle divide within the canyon (Figs. 1 and 2). For this reason, Unaweep Canyon is said to be the only canyon in the world with two mouths.

Originally described by the Hayden Survey of the late 1800s (Peale, 1877; Gannett, 1882), the origin of Unaweep Canyon remains debated. All agree that the creeks currently occupying the canyon are markedly underfit. Most have hypothesized that the canyon originated from fluvial erosion related to an ancestral Gunnison and/or Colorado River, which subsequently abandoned the canyon; but its odd geomorphology has aroused speculations of a possible glacial influence. Soreghan et al. (2007) proposed a late Paleozoic age for the Precambrian-hosted inner gorge of Unaweep Canyon; Soreghan et al. (2008, 2014) further suggested an initial (Paleozoic) glacial origin followed by Cenozoic fluvial exhumation of the ancient valley. Both the Paleozoic age and glacial origin remain controversial (Aslan et al., 2008, 2014; Hood, 2009; Hood et al., 2009).

Unaweep Canyon would remain a geomorphic oddity of primarily local curiosity were it not positioned atop the Colorado Plateau and connected to the history of the greater Colorado River system, and possibly a deeper time history of the region. Resolving the geologic history of Unaweep Canyon, including the timing and processes of incision and abandonment of the canyon, contributes to our understanding of the tectonic, geodynamic, and geomorphic forces driving Cenozoic integration of the Colorado River system across the Colorado Plateau, a primary theme of this special issue. Moreover, if the Cenozoic history of Unaweep Canyon proves to be one of exhumation, rather than incision, it implies preservation (through burial) of a paleolandform of excessive antiquity. Note that here we use the geomorphic definition of exhumation as referring to the exposure through erosion of a formerly buried landscape. Preservation of landscapes of great antiquity (e.g., Mesozoic, Paleozoic) is well known from, e.g., cratonal regions of the Gondwanan continents (e.g., Twidale, 1998, 2003), but less appreciated for tectonically active regions, although debate continues on the possible role of paleocanyons in shaping the lower Colorado River system (Flowers et al., 2008; Wernicke, 2011; Flowers and Farley, 2012, 2013).

In this paper we assemble unpublished mapping, provenance, and paleocurrent data, and draw upon previously published sedimentologic, cosmogenic nuclide, and thermochronological data from units within and near western Unaweep Canyon to (1) document the late Cenozoic occupation and abandonment of Unaweep Canyon by a large tributary of the upper Colorado River system and identify that tributary, and (2) assess the hypothesis that this late Cenozoic fluvial history was one of rapid exhumation of a pre-Cenozoic landform, rather than primary incision. If valid, this scenario implies a strong role for a form of historical contingency in landscape evolution. These data bear on long-debated issues regarding the identity of the river or rivers that occupied Unaweep Canyon, the cause and timing of river abandonment, the postabandonment history, the odd orientation that crosscuts regional drainage patterns, and broader implications for identifying the drivers of drainage integration across the Colorado Plateau. This work also highlights the possible role of paleolandforms in shaping landscape evolution of even tectonically active regions.

Colorado and Uncompahgre Plateaus

The Colorado and Uncompahgre Plateaus form part of the greater Rocky Mountain orogenic plateau, a large region of high elevation in the United States (McMillan et al., 2006). The area has undergone multiple episodes of uplift, including the Pennsylvanian to early Permian rise of the Ancestral Rocky Mountains, which included the Uncompahgre uplift, a precursor to the modern Uncompahgre Plateau (Kluth and Coney, 1981). The southwestern edge of the Uncompahgre Plateau coincides with the southwestern margin of the late Paleozoic Uncompahgre uplift, defined by the Uncompahgre fault zone, a buried reverse fault system inferred largely from subsurface data (Frahme and Vaughn, 1983; White and Jacobson, 1983; Moore et al., 2008). In late Paleozoic time, the Uncompahgre uplift formed the highland bordering the Paradox Basin, which accumulated thousands of meters of sediments. In the proximal part of the Paradox Basin, this thick sedimentary succession includes the Permian Cutler Formation, which onlaps the western mouth of Unaweep Canyon (Gateway, Colorado; Plate 1), and buries the Uncompahgre fault zone (Cater, 1970; Moore et al., 2008; Soreghan et al., 2012).

Marine strata present across the Colorado Plateau, including the Uncompahgre Plateau, document the ultimate return of the region to sea level by Cretaceous time (e.g., Hunt, 1969; Pederson et al., 2002). Tectonic activity then resumed during the Laramide orogeny (ca. 80–40 Ma). Across the Colorado Plateau Laramide deformation is limited to a series of monoclines that include the Uncompahgre Plateau (Kelly, 1955; Williams, 1964; Bump and Davis, 2003; Davis and Bump, 2009), which exhibits as much as ∼1000 m of structural relief. Moreover, the Colorado Plateau has undergone epeirogenic uplift resulting in a modern average elevation of ∼2.2 km (McQuarrie and Chase, 2000; Pederson et al., 2002). Details of the timing, amount, and mechanism of uplift of the Colorado Plateau remain debated, with estimates of uplift timing ranging in age from Late Cretaceous–Paleocene (Laramide) to late Cenozoic (e.g., Hunt, 1956; Morgan and Swanberg, 1985; Spencer, 1996; McQuarrie and Chase, 2000; Pederson et al., 2002,; Sahagian et al., 2002: Morgan, 2003; Karlstrom et al., 2007, 2008, 2012b; Huntington et al., 2010; Liu and Gurnis, 2010; van Wijk et al., 2010). Incised landscapes characterized by deep canyons are common on the Colorado Plateau and linked to multiple events (Laramide through Neogene) of plateau uplift, as well as possible roles of geomorphology and climate in ultimate integration of the Colorado River drainage system (see overview in Karlstrom et al., 2012a).

The Colorado River and its upstream tributaries, including the Gunnison, Uncompahgre, Dolores, and San Miguel Rivers, drain the Colorado Plateau (Fig. 1). Evidence from the southwestern Colorado Plateau indicates that drainages there initially flowed to the northeast following eastward tilting in early Cenozoic (Laramide) time, and that structural inversion associated with Basin and Range extension caused a drainage reversal in Miocene time, with at least the lower Colorado River system integrated and flowing to the Gulf of California by ca. 6–5.3 Ma (e.g., McKee and McKee, 1972; Young and McKee, 1978; Young, 1982; Potochnik, 1989, 2001; Spencer et al., 2001; Young and Spamer, 2001; Pederson et al., 2002; Lucchitta, 2003; Dorsey et al., 2007, 2011; House et al., 2005, 2008; Karlstrom et al., 2012b). Data from the upper Colorado River system are less well constrained, but indicate that regional exhumation and associated major fluvial incision accelerated ca. 11–6 Ma (e.g., Czapla and Aslan, 2009; Aslan et al., 2010; Karlstrom et al., 2012b; Donahue et al., 2013); in Thomson et al. (2012), the onset was constrained to ca. 6–5 Ma.

Unaweep Canyon

Unaweep Canyon perpendicularly bisects the northwest-southeast–trending Uncompahgre Plateau, exposing Proterozoic crystalline (igneous and metamorphic) basement beneath a Mesozoic sedimentary carapace, and forms a 70-km-long wind gap extending from Whitewater, Colorado (elevation 1420 m) in the northeast to Gateway, Colorado (elevation 1400 m) in the southwest (Figs. 1 and 2). The modern canyon is as deep as 1 km (>400 m in the inner Precambrian gorge), and as wide as 6 km (3 km in inner gorge; Lohman, 1981; Cole and Young, 1983; Soreghan et al., 2007). Two small creeks, East Creek and West Creek, currently occupy Unaweep Canyon, draining from the nearly imperceptible Unaweep Divide at an elevation of 2148 m. Moreover, this divide is incongruously offset by ∼20 km to the northeast from the axial crest (drainage divide) of the Uncompahgre Plateau (Figs. 1 and 2). East Creek and West Creek flow through the canyon in opposite directions to join, respectively, the Gunnison River near Whitewater, Colorado, and the Dolores River near Gateway, Colorado (Figs. 1 and 2).

Core recovered from within Unaweep Canyon in 2004–2006 demonstrates that the canyon contains a relatively thick fill (∼330 m; summarized in Soreghan et al., 2007) at least locally. Where cored, this fill consists of four stratigraphic intervals (Table 1). On the basis of cosmogenic dating (Balco et al., 2013), the upper three units are of Pleistocene age. In Soreghan et al. (2007) a late Paleozoic age was posited for the basal unit, but this hypothesis remains contested (Aslan et al., 2008, 2014; Hood, 2009; Hood et al., 2009). (For additional details of methods and results of coring, see Marra, 2008.)

The anomalous course, size, and longitudinal profile of Unaweep Canyon, and recognition of the inability of East and West Creeks to carve a canyon of such magnitude, has led to three primary hypotheses for its formation: (1) late Cenozoic fluvial incision, (2) late Cenozoic (Pleistocene) glacial incision, and (3) late Paleozoic glacial incision and subsequent burial followed by late Cenozoic fluvial exhumation of the formerly buried landscape. See Hood (2011) for a comprehensive overview of the history of thought on Unaweep Canyon.

The prevailing fluvial hypothesis posits original canyon formation by the ancestral Gunnison River (Peale, 1877; Cater, 1966, 1970; Sinnock, 1981), the ancestral Colorado River, or a combined Colorado-Gunnison (Gannett, 1882; Stokes, 1948; Shoemaker, 1954; Cater, 1955a; Hunt, 1956; Lohman, 1961, 1981; Steven, 2002; Aslan et al., 2008, 2010, 2014; Hood, 2011; Hood et al., 2014). Some suggested that the Dolores River flowed northeastward to carve Unaweep Canyon (Peale, 1877; Hunt, 1956). The eventual abandonment of the canyon by a large river was attributed by some to neotectonic warping of the Uncompahgre Plateau, although the prevailing idea posits stream piracy (Cater, 1966; Lohman, 1961, 1965, 1981; Sinnock 1978, 1981; Scott et al., 2001; Steven, 2002; Oesleby, 1978, 1983; Aslan et al., 2005, 2008, 2014). Speculations on the origin and evolution of Unaweep Divide have included river incision in response to hypothesized differential neotectonic uplift of the Uncompahgre Plateau (e.g., Cater, 1966; Lohman, 1965; Hunt, 1969; Scott et al., 2001), differential erosion (Sinnock, 1981), and valley-fill sedimentation (Oesleby, 1978, 1983).

Straightforward evidence for the former presence of a large river in Unaweep Canyon remained elusive until recently. High terrace remnants with fluvial gravels are known from both the eastern (Cactus Park, Mesa County, Colorado; Lohman, 1961, 1965, 1981; Aslan et al., 2005, 2008, 2010, 2014) and western (Gateway, Colorado; Cater, 1955a, 1966, 1970) mouths of the canyon (Fig. 2), suggesting that a river connected these two locations, yet no correlative terraces or deposits are exposed in Unaweep Canyon. Furthermore, although the eastern (Cactus Park) gravels have been long known, those in the Gateway region were mentioned only cursorily by Cater (1955a, 1966, 1970), but never described or mapped until one of us (Kaplan, 2006) rediscovered and formally documented them (details herein).

The odd geomorphology of Unaweep Canyon inspired the Pleistocene glacial hypothesis. Lohman (1981) highlighted the prominent U-shaped cross section of Unaweep Canyon, and Cole and Young (1983) cited the many apparent glacial features (e.g., inferred cirques, truncated spurs) of the Precambrian basement gorge. The U-shaped cross section can now be attributed to the presence of a thick fill that obscures the true form of the basement surface, whether V or U shaped. The cirques refer to the numerous amphitheater-shaped (steep headwalls and stubby planform) tributaries of the inner gorge (Fig. 3). These are somewhat reminiscent of those found across the Colorado Plateau, commonly attributed to seepage erosion, a process well documented in valleys formed in loose sediment and sedimentary rock (e.g., Laity and Malin, 1985; Howard and McLane, 1988; Howard et al., 1988; Baker, 1990; Schumm et al., 1995; Abrams et al., 2009). Formation of amphitheater shapes has also been observed in stratified basalt, attributed to megafloods and mass wasting (Baker, 1990; O’Connor, 1993; Lamb et al., 2006, 2007, 2008, 2014). It is odd that the amphitheater tributaries lining the inner gorge of Unaweep Canyon occur entirely in Precambrian igneous and metamorphic basement with no obvious competency contrasts (Plate 1) and no evidence for megafloods or mass wasting of sufficient scale.

The Pleistocene glacial hypothesis has languished owing to the relatively low elevation of Unaweep Canyon. The highest point within Unaweep Canyon (Unaweep Divide, 2148 m) is well below most Pleistocene ice accumulations of the region (Richmond, 1962; Yeend, 1969; Sinnock, 1981), as is the Uncompahgre Plateau, and no Quaternary glacial deposits have ever been identified within or proximal to the canyon or on the plateau (Lohman, 1981; Scott et al., 2001; Soreghan et al., 2007), effectively refuting this hypothesis.

In Soreghan et al. (2008, 2009a, 2009b, 2014), it was proposed that Unaweep Canyon was carved in the Permian–Pennsylvanian by upland glaciation when this region was at elevation atop the ancient Uncompahgre uplift, and that this paleolandscape was exhumed by late Cenozoic fluvial action associated with drainage evolution across the northern Colorado Plateau of the upper Colorado River system. This hypothesis has been criticized, and remains controversial (for discussions and replies, see Soreghan et al., 2007, 2008, 2009a, 2009b; Aslan et al., 2008; Hood, 2009; Hood et al., 2009).

Given the continued controversy surrounding Unaweep Canyon, this paper aims in part to assess evidence that could bear on distinguishing between (1) primary Cenozoic incision and (2) Cenozoic exhumation of a buried landform. If the canyon is entirely Cenozoic in age, then it should contain strata of entirely Cenozoic age, should follow a planform course that generally records migration down the structural plunges and dips of emerging Colorado Plateau uplifts, and should reflect incision rates consistent with time-averaged rates of comparable systems across the upper Colorado River drainage system. Alternatively, if the canyon reflects in part exhumation of a preexisting landform, then we might expect to detect evidence of burial of the landform by pre-Mesozoic strata, a planform that possibly ignores structural plunges and dips of uplifts, and incision rates that may exceed regional averages, reflecting less resistance to erosion of weak sedimentary fill relative to stronger crystalline bedrock (Sklar and Dietrich, 2001).

Field Mapping

Field mapping to detail the Quaternary deposits includes ∼30 km2 proximal to the western mouth of Unaweep Canyon: the Gateway area, just west of the western canyon mouth, and the Unaweep Seep area, just east (inside) of the western canyon mouth (Fig. 2; Plate 1). In addition, bedrock mapping was conducted of the entire region surrounding Unaweep Canyon (Plate 1; Eccles, 2013). Previous geologic mapping of the area encompassing Unaweep Canyon consisted of two quadrangle maps (Cater, 1955a, 1955b) capturing parts of the western mouth and southwestern rim, and a 1:250,000-scale compilation based on unpublished photogeologic mapping (Williams, 1964). Abundant Cenozoic (Quaternary) deposits occur adjacent to the western mouth of Unaweep Canyon, deposited unconformably on the Permian Cutler Formation. Cater (1955a) subdivided these into the following units: a fanglomerate (Qfg) capping low ridges in the valley, gravel beds (Qg) capping terraces along West Creek and the Dolores River, landslide deposits (Qls), and alluvium (Qal).

We mapped the study area using 7.5 min U.S. Geological Survey quadrangle topographic maps as base maps in the field, augmented with aerial photos and a handheld global positioning system. Using cobble clast provenance data and the map distribution of the Quaternary, we distinguished three units within what Cater (1955a) classified as Quaternary fanglomerate (Qfg) in the Gateway area, herein termed the Gateway gravels, Palisade gravels, and West Creek gravels (Plate 1; Fig. 4), detailed in the following. For each of these units, we characterized the sedimentology (grain size, rounding, texture, bedding, thickness) and logged several vertical exposures. Using similar techniques, we distinguished two additional upper Cenozoic units in the Unaweep Seep area (Unaweep gravels and Quaternary talus; Plate 1). Where possible, clast imbrications were measured to determine paleocurrent directions.

Provenance Analysis

Cobble clast counts were conducted on the mapped gravel deposits by point counting on outcrops using a 0.5 m × 0.5 m square grid and identifying clasts at intersection points spaced at ∼8 cm intervals. Clasts >2 cm diameter were counted as framework grains (average n = 118), and distinguished as Precambrian crystalline basement, sedimentary, or volcanic lithologic types.

Additional provenance data were obtained from the sand-sized fraction of gravel deposits in the mapped area, from sand recovered from subsurface strata (detailed in the following), and from 1.2 to 0.96 Ma Colorado River terraces mapped (Carrara, 2001) northeast of the Uncompahgre Plateau (Fig. 1; Supplemental Tables 11 and 22). Several samples of unconsolidated matrix (material <2 cm) were collected and sieved to isolate (where possible) the medium to coarse sand fraction (cf. Critelli et al., 1997), and counts of 400–500 framework grains were performed on thin sections stained for potassium feldspar (for additional details see Kaplan, 2006; Marra, 2008). Grains were analyzed using both the traditional and the Gazzi-Dickinson methods. In the Gazzi-Dickinson method, sand-sized grains and crystals within larger fragments are assigned to the category of the grain or crystal, as opposed to the category of the larger fragment as is done in the traditional method (Ingersoll et al., 1984). For comparison of the (ancestral) Gunnison and (ancestral) Colorado gravels in particular, we focused on identification of the various types of volcanic lithic fragments (vitric, felsitic, microlitic, lathwork; Supplemental Table 2 [see footnote 2]) to distinguish mafic (basaltic) from intermediate (andesitic) sources following Marsaglia (1993).

Geochronologic and Thermochronologic Analyses

We measured cosmogenic 26Al and 10Be on samples from both the lowest terrace of the Gateway gravels in the Gateway map area, and from several horizons of the cored interval (for details, see Balco et al., 2013). The Gateway gravels occur beneath a thick overburden of younger gravels (Palisade gravels) at both of these sites. Within the core, samples were analyzed from several horizons representing inferred lacustrine, pedogenic, and colluvial deposits (for details of sample preparation and laboratory analyses for cosmogenic-nuclide dating, see Balco et al., 2013).

Apatite fission track results have been obtained from several samples of the Precambrian basement in Unaweep Canyon, as well as on Cutler Formation sediments in the Gateway area (Thomson et al., 2012). These data were collected to constrain more fully the longer term Mesozoic burial and Cenozoic erosion and incision history of the canyon and the surrounding Uncompahgre Plateau. Experimental details and thermal history modeling were summarized in Thomson et al. (2012).

Precambrian–Permian Relations at the Western Mouth of Unaweep Canyon

The Permian Cutler Formation onlaps Precambrian basement along the southwestern margin of the Uncompahgre Plateau near the western mouth of Unaweep Canyon. This onlap of the uppermost (exposed) Cutler Formation onto Precambrian basement of the (paleo) Uncompahgre uplift records Permian syndepositional subsidence of the Uncompahgre highland, a realization first expressed by Cater (1970; see also Soreghan et al., 2012) on the basis of his original mapping (Cater, 1955a). Cater (1970, p. 68) stated, “After the highland attained its maximum height and while the Cutler was being deposited, the highland began sinking—at least along its southwest flank.”

Our new mapping (Plate 1; Eccles, 2013) of the Cutler-Precambrian contact here extends and further clarifies the remarkable onlap contact first depicted by Cater (1955a). At the Cutler-Precambrian contact near the western mouth of Unaweep Canyon, the Cutler Formation buries ∼520 m of paleorelief preserved on Precambrian basement (palinspastically restored for Laramide-age faulting; Soreghan et al., 2012). The new map of Eccles (2013) modifies Cater’s (1955a) original map by extending the large reentrant of Cutler Formation, which marks the presence of a paleovalley extending north-northwest toward modern Unaweep Canyon (Plate 1), ∼90° offset (south and east) from the modern exit of West Creek, and filled with undeformed and poorly consolidated Cutler Formation (Fig. 5). This paleovalley is not inferred, but observable directly in outcrop (Fig. 5).

Cenozoic Gravel Units in and Adjacent to Western Unaweep Canyon

Within the Gateway area, we subdivided the Quaternary fanglomerate recognized by Cater (1955a) into three mappable units, informally named as follows: (1) the Gateway gravels, named for the town of Gateway, Colorado; (2) the Palisade gravels, named for the eponymous erosional remnant towering above Gateway; and (3) the West Creek gravels, named for West Creek, the modern drainage (Plate 1; Fig. 4). Figure 6 shows the diagrammatic stratigraphy of these three units. Within the Unaweep Seep area of western Unaweep Canyon, we distinguished two additional deposits, referred to as the Unaweep gravels (Qug) and talus material (Qt) (Kaplan, 2006; Plate 1).

Gateway Map Area

Gateway gravels. The Gateway gravels occur as a flight of downward-stepping terraces, with terrace treads at ∼1615, 1555, and 1514–1524 m (Qgg, Plate 1). Gravel thicknesses are 1.5–13.0 m, with terraces confined to a narrow belt within ∼1 km of the axis of West Creek. These gravels unconformably overlie relatively planar straths carved into the upper Paleozoic Cutler Formation (Figs. 4 and 7).

Clasts within the Gateway gravels are well rounded, with a distinctive component of (intermediate) volcanic clasts; these gravels exhibit a light gray color imparted by the abundance of Precambrian granite clasts. The color contrasts with the red-tan of other Cenozoic gravels in the area (Figs. 4 and 7) derived predominantly from local redbeds. Clast counts of framework grains average ∼60% Precambrian basement, 30% sedimentary, and 10% volcanic clasts of intermediate (andesitic) composition (Fig. 8). Point-count data of the sand-sized fraction also indicate abundant volcanic lithic fragments (Lv) of intermediate (andesitic) composition (Fig. 9; Supplemental Table 1 [see footnote 1]). Deposits of the Gateway gravels are clast supported and generally massive, consisting primarily of well-rounded, poorly sorted, imbricated cobbles to boulders (maximum clast size ∼3.0 m) with minor sand matrix. Paleocurrent data measured from imbricated clasts in the Gateway gravels yield an average flow direction of ∼200° (to the south-southwest; Fig. 6).

The lowest-mapped terrace of the Gateway gravels (elevation ∼1514–1524 m) exhibits Stage IV calcic soil development (>500 k.y. development as defined by Machette, 1985; see also Mack, 1997; Retallack, 2001), with continuous coatings on clasts and a thick (∼30 cm) tabular accumulation of calcium carbonate at the top.

Palisade gravels. The Palisade gravels constitute the most voluminous and widespread gravel unit in the Gateway map area (Qpg; Plate 1), essentially including what Cater (1955a) mapped as Quaternary fanglomerate. The Palisade gravels cap most of the low ridges in the study area (Plate 1; Figs. 4, 6, and 7) and range in elevation from >1860 m near the Mesozoic escarpment to <1525 m near modern West Creek. Despite poor cementation, the Palisade gravels locally form vertical cliffs as much as ∼30 m thick and tall hoodoos (Fig. 4). The Palisade gravels are stratigraphically above the Gateway gravels (where present) or unconformably overlie straths cut in the Permian Cutler Formation (Figs. 6 and 7). The surface of the Palisade gravels consists of a heavily dissected, yet discernible, terrace tread that slopes from the Mesozoic escarpment toward West Creek (Fig. 7).

The Palisade gravels are distinguished by their relatively high percentage of sedimentary clasts (Figs. 4 and 8). Point-count data on the sand component indicate abundant quartz and sedimentary lithic fragments (see Kaplan, 2006; Supplemental Table 1 [see footnote 1]). The Palisade gravels consist of a very poorly sorted mix of angular to subangular pebbles to boulders (maximum clast size ∼4.0 m) with locally significant mud- to granule-sized matrix. Deposits of the Palisade gravels are generally clast supported and massive to locally crudely stratified. Clast imbrication is rare, thus paleocurrent data are sparse. However, terrace treads and straths slope ∼4° to the southeast and 5° to the northwest on the north and south sides, respectively, of West Creek.

Well-developed calcic soil profiles (Calcisols) are present on the top of the Palisade gravels. A typical 2-m-thick profile, located on top of the Palisade gravels, consists of coatings on basal surfaces of large clasts, continuous coatings enveloping sand grains, and indurated sheet-like calcium carbonate, representing Stage IV (>500 k.y.) development (cf. Machette, 1985; Retallack, 2001).

West Creek gravels. The West Creek gravels are within a narrow belt (<700 m wide) bordering West Creek and are well exposed in cuts along Highway 141 (Qwcg; Plate 1). The deposits locally reach ∼30 m in thickness and underlie a single terrace tread inset against older units, including the Palisade gravels and the Permian Cutler Formation (Fig. 4F). Only one terrace is evident (excluding the modern floodplain), and it occurs at an elevation of ∼1490–1550 m.

The West Creek gravels have a composition intermediate between the other two gravel deposits, consisting of nearly subequal proportions of Precambrian basement and sedimentary clasts, with trace intermediate volcanic clasts (Fig. 8). Point counts of the sand-sized fraction show that the matrix contains mostly monocrystalline quartz and feldspar grains (see Kaplan, 2006). Deposits of the West Creek gravels are generally clast supported and display crude stratification consisting of poorly sorted, rounded to angular, imbricated pebbles to boulders (maximum clast ∼3.0 m). Imbricated clasts are abundant in the West Creek gravels and yield an average flow direction of 198°, which coincides with the axis of modern West Creek (Fig. 7). Sandy layers occur locally within the coarser layers and commonly display decimeter-scale cross-stratification.

Unaweep Seep Map Area (Westernmost Unaweep Canyon)

Unaweep gravels. The Unaweep gravels and the talus material mapped in westernmost Unaweep Canyon exhibit similar compositions and sedimentological characteristics; however, we distinguish them primarily on the basis of slope characteristics.

The Unaweep gravels form coalescing debris aprons that mantle the flanks of Unaweep Canyon, forming the modern, highly vegetated surface (Qug; Plate 1; Fig. 10). Roadcuts indicate that these debris aprons are thicker than 20 m and core data confirm a local minimum of 330 m of fill within the canyon, including an uppermost unit of >160 m of mostly conglomerate essentially correlative to the Unaweep gravels (see Table 1). The Unaweep gravels consist generally of a very poorly sorted mix of subangular to rounded pebbles to boulders with a locally significant sandy granule matrix. Approximately subequal proportions of sedimentary and Precambrian basement clasts are present (Fig. 8). Monocrystalline quartz and feldspar grains predominate within the sand fraction (Fig. 9).

Few vertical exposures of the Unaweep gravels exist, but roadcuts and a gravel-pit exposure reveal a record of calcrete development on top of these debris aprons (Fig. 10). A 2.5 m section of a surficial exposure consists of alternating layers of conglomerate and granular sand exhibiting calcification, ranging from coatings on basal clast surfaces to veins and sheets in several discrete intervals separated by noncalcified intervals. The basal interval represents a Stage I carbonate accumulation (discontinuous clast coatings), whereas the middle and upper intervals record two intervals of Stage IV development (clast coatings and platy sheets; cf. Machette, 1985; Retallack, 2001), similar to that developed on the Palisades gravels of the Gateway area. In addition, all fan surfaces within Unaweep Canyon are heavily vegetated, reflecting long-term (Holocene) stabilization (Fig. 10).

Cosmogenic-Nuclide Burial Ages, Thermochronology

Seven 26Al-10Be burial ages from borehole sediments as well as an isochron burial age (e.g., Balco and Rovey, 2008) from the lowest terrace level of the Gateway gravels were reported in Balco et al. (2013). Three burial ages from the upper conglomeratic interval in the borehole range from 1.16 ± 0.25 to 0.76 ± 0.19 Ma; 2 from the transitional interval are 1.37 ± 0.17 and 1.27 ± 0.23 Ma; and two from the lower lacustrine unit are 1.69 ± 0.44 and 1.33 ± 0.18 Ma. The isochron age for the lowest Gateway gravels is 1.46 ± 0.33 Ma. As discussed in detail in Balco et al. (2013), fitting a piecewise linear age model to these data under the assumption that the lacustrine and conglomeratic units accumulated at constant rates yields a summary age estimate of 1.41 ± 0.19 and 1.34 ± 0.13 Ma for the base and top, respectively, of the lacustrine unit. In addition, this showed that the basal age of the lacustrine section in core was indistinguishable from the isochron age for the Gateway gravels in outcrop.

Comparison of apatite fission track results from Precambrian basement from the top and near the base of Unaweep Canyon (Thomson et al., 2012) shows contrasting thermal histories over the past ∼25 m.y. A basement sample from near the canyon rim (∼2644 m) has long track lengths indicative of slow erosion during this time. In contrast, samples from the base of the canyon (down to 1445 m) have shorter track lengths and slightly younger ages (ca. 20 Ma) that are best reproduced by a thermal history with rapid cooling beginning ca. 6–5 Ma, reflecting time of onset of canyon incision. This estimate is more precise and somewhat younger than previous estimates (11–6 Ma) for onset of regional exhumation in the upper Colorado River drainage (Aslan et al., 2010; Karlstrom et al., 2012b; Rosenberg et al., 2014).

Identifying the Ancestral River of Unaweep Canyon

Debate continues on the identity of the river(s) that occupied Unaweep Canyon in Neogene time; both the ancestral Colorado and ancestral Gunnison have been suggested (Peale, 1877; Gannett, 1882; Stokes, 1948; Shoemaker, 1954; Hunt, 1956; Cater, 1955a, 1966, 1970; Lohman, 1961, 1981; Sinnock, 1981, 2002; Aslan et al., 2008, 2010, 2014; Hood, 2011; Hood et al., 2014). The composition and provenance of the ancient fluvial deposits provide one means to distinguish these options. Although both the Colorado and Gunnison Rivers traverse sedimentary, plutonic, and igneous bedrock types upstream of their confluence, the compositions of volcanic rocks in these drainages exhibit distinctive contrasts (Fig. 11). Specifically, modern gravels of both the Gunnison River and Uncompahgre River consist predominantly of intermediate (andesitic) volcanic and shallow intrusive rocks that originated from the Tertiary volcanic provinces of the San Juan Mountains and West Elk Mountains (Cater, 1966, 1970; Aslan et al., 2005, 2008; Fig. 1). Basalt clasts are rare in gravels of the Gunnison River (Lohman, 1961; Cater, 1966). In contrast, Colorado River gravels upstream from the Colorado and Gunnison confluence contain basalt, derived from the voluminous lava flows associated with the Grand Mesa and other upstream sources (Cater, 1966; Aslan et al., 2005). Therefore, volcanic and shallow (porphyritic) intrusive rocks of intermediate-felsic composition are rare in gravels of the Colorado River upstream of its confluence with the Gunnison River. This contrast provides a viable basis for identifying the river or rivers that formerly occupied Unaweep Canyon by examining the provenance of the river deposits in and proximal to the canyon.

The well-rounded-clast–supported and imbricated deposits mapped as the Gateway gravels near the western mouth of Unaweep Canyon record a major river flowing southwestward out of Unaweep Canyon toward the Dolores River. These gravels contain a significant contribution from intermediate volcanic clasts (Figs. 4, 8, and 9) that we interpret to indicate the presence of the ancestral Gunnison River, or more likely, combined Gunnison-Uncompahgre (hereafter referred to as simply the ancestral Gunnison River).

Others have suggested that the ancestral river of Unaweep Canyon was a combined Colorado-Gunnison river (e.g., Aslan et al., 2008, 2010, 2014; Hood, 2011; Hood et al., 2014). In particular, Hood (2011) cited the presence of minor (∼2%) red siltstone and sandstone clasts in terraces of the Gateway gravels (as much as 2.6% in the higher terrace) near the western mouth of Unaweep Canyon to infer a combined Colorado-Gunnison river in Unaweep Canyon. Both Aslan et al. (2014) and Hood et al. (2014) used this argument to agree with the two-stage history of abandonment of Unaweep Canyon, first by the ancestral Colorado River, and ultimately by the ancestral Gunnison River, that was initially proposed by Lohman (1961). Hood (2011) noted that the Colorado River terraces upstream of the Colorado-Gunnison confluence contain ∼8% red sedimentary clasts, and interpreted these clasts to derive from the Pennsylvanian Maroon Formation in the upper reaches of the Colorado River drainage basin. However, there are abundant possible sources of red siltstone and sandstone in the Mesozoic strata along the rims of Unaweep Canyon, and minor red silt and sandstone also occur within the Permian Cutler Formation near the mouth of Unaweep Canyon (Soreghan et al., 2009c), where the Gateway gravels crop out. Hood (2011) did not compare the red clasts from the gravels to red clasts from local Mesozoic and Permian units; the presence of red sedimentary clasts in distinguishing the ancestral Gunnison and Colorado rivers remains equivocal.

The sand provenance data provide additional support for a uniquely Gunnison River provenance for the Gateway gravels. Figure 9 illustrates sand framework composition from the (1) ancestral (1.2–0.96 Ma) Colorado River terraces upstream of the Gunnison River confluence (Figs. 1 and 2; Supplemental Table 2 [see footnote 2]), (2) Cactus Park (ancestral Gunnison gravels dated to ca. 0.8 ± 0.24 Ma; Aslan et al. 2014), and (3) the Gateway gravels (ca. 1.46 ± 0.33 Ma; Balco et al., 2013). If the Gateway gravels were deposited by a combined Colorado-Gunnison river, then there should be a discernible signal of the Colorado River, such as the presence of mafic volcanic lithic fragments (Fig. 9), but there is not. The Gateway gravels, however, exhibit a provenance most consistent with an exclusive Gunnison River occupation. This interpretation corroborates that of Aslan et al. (2014), who used detrital zircon data from the Gateway gravels to infer a Gunnison River and Uncompahgre River signal.

These provenance data are most consistent with a source from the Gunnison drainage system upstream of the Unaweep Canyon region. The youngest Gateway gravel deposit is ca. 1.46 ± 0.33 Ma, and there are no dates for the oldest Gateway gravel deposit. Therefore, these constraints exclude the occupation of Unaweep Canyon by an ancestral Colorado River over at least the past ∼1.5 m.y. Moreover, the hypothesis that the ancestral Colorado River ever occupied Unaweep Canyon (suggested by Hood, 2011; Aslan et al., 2014; Hood et al., 2014) is problematic since the only Cenozoic river deposits older than 1.5 Ma are the older (higher elevation) Gateway gravels that exhibit a Gunnison provenance. Additionally, Aslan et al. (2008) documented gravels of inferred ca. 7–2 Ma age on top of the Uncompahgre Plateau south of Unaweep Canyon that contain intermediate volcanic clasts consistent with derivation from the ancestral Gunnison and ancestral Uncompahgre Rivers, but no gravels bearing an ancestral Colorado River provenance. In addition, Price et al. (2012) used detrital zircon data to infer a Gunnison River source for gravels beneath the ca. 11 Ma basalt capping Grand Mesa, a gravel unit previously assumed to record the ancestral Colorado River (Czapla and Aslan, 2009).

Relationship of the Gateway Gravels to the Cactus Park Gravels

Gunnison River gravels are well known from Cactus Park, a northwest-southeast–trending inferred paleovalley of the Gunnison River located at the northeast end of Unaweep Canyon near East Creek (Fig. 2; Lohman, 1961, 1965; Aslan et al., 2005, 2008, 2014; Hood et al., 2014). The Cactus Park gravels are interpreted to represent a series of downstepping terraces at elevations ranging from ∼1870 to 1950 m, inset against the Jurassic Wingate Sandstone (Aslan et al., 2005, 2008). Gravels with a clear Gunnison River (intermediate volcanic) provenance range in thickness from 3 to 5 m, are overlain by gravels derived from Mesozoic strata, and are locally buried beneath inferred lake beds (Aslan et al., 2005, 2008, 2014; Hood et al., 2014). Cosmogenic burial ages of 0.8 ± 0.24 Ma for the youngest gravels (Aslan et al., 2014) below lake deposits suggest that these do not correlate directly with the Gateway (ancestral Gunnison) gravels mapped near Gateway. Aslan et al. (2014) suggested that this date (0.8 ± 0.24 Ma) records a minimum date of canyon abandonment, and that the river that flowed through Unaweep Canyon continued to do so until that time. In this view, the ultimate abandonment at 0.8 Ma is marked by the occurrence of the Cactus Park lakebeds, which do not exhibit a Gunnison provenance (Hood et al., 2014). We agree with Hood et al. (2014) that the cored lacustrine strata in western Unaweep Canyon (Table 1) cannot correlate with the lacustrine strata of Cactus Park; they are separated by a Precambrian basement high (1886–1905 m) in the canyon floor located just west of Cactus Park, which exceeds the elevation (1870 m; Aslan et al., 2014) of the 0.8 Ma Cactus Park gravels. Figure 12 shows the longitudinal profile of the modern floor of Unaweep Canyon and the projected 1.4 Ma profile of the ancestral Gunnison River through Unaweep Canyon; this profile uses terrace levels from the 1.46 Ma Gateway gravels, core data (from Soreghan et al., 2007; Marra, 2008; see Table 1), and new mapping (Eccles, 2013). Connecting the Cactus Park gravels to the Gateway gravels to produce a profile with a continuous gradient through Unaweep Canyon is untenable, owing to known points of Precambrian basement highs in East Creek, West Creek, and at the coring site (Fig. 12). Figure 12 shows the profile constrained with all known (outcrop and core) control points, and produces a slope that is anomalously flat in the western canyon. Regardless, these relationships indicate that the Gunnison River continued to occupy Cactus Park after it had abandoned Unaweep Canyon.

Abandonment of Unaweep Canyon

Previous models for the abandonment of Unaweep Canyon by the Gunnison River (or Gunnison-Colorado) differ in the timing of the event (ca. 1.4 Ma versus 0.8 Ma), and in the proposed mechanism, i.e., landslide blockage (Marra, 2008; Balco et al., 2013) versus stream piracy from Cactus Park (Lohman 1961, 1965, 1981; Aslan et al., 2014; Hood et al., 2014). Any model for abandonment must honor the following observations: (1) establishment of a lake in western Unaweep Canyon (Table 1; Soreghan et al., 2007; Marra, 2008) ca. 1.4 Ma that existed for a few millennia, evinced by the ages for the base and top of the lacustrine section within the core (Balco et al., 2013), and (2) the existence of a Precambrian basement high (1890–1905 m above sea level; Eccles, 2013) just west of Cactus Park at an elevation that exceeds that (1870 m; Aslan et al., 2014) of the 0.8 Ma Cactus Park gravels.

As noted in Soreghan et al. (2007) and Marra (2008), the presence of lacustrine sediments at the Unaweep coring sites (Massey #1 and Massey #2; Plate 1), indicates that a lake occupied the western canyon. The lowest terraces of the Gateway gravels near the western mouth of Unaweep Canyon archive the youngest known record of the ancestral Gunnison River through Unaweep Canyon and yield cosmogenic-nuclide burial ages of ca. 1.46 ± 0.33 Ma (Balco et al., 2013), indistinguishable from ages obtained from the lacustrine (lower clayey-sandy) interval of the core. Furthermore, the provenance of the lake sediments, which include a distinctive intermediate volcanic component (to 8%; Soreghan et al., 2007; Marra, 2008), link the Gateway gravels and the basal lake sediments to the same ancestral Gunnison source. Thus, the thick lacustrine unit present in the core within the canyon records damming of the ancestral Gunnison River near the western mouth of Unaweep Canyon (Figs. 12 and 13), followed by the creation of a lake.

Water supplied from the ancestral Gunnison River would have filled the available lake volume (liberally ∼135 km3, assuming 1.5 km depth, 3.5 km width, and 30 km length) in years to decades, assuming mean annual discharge of the ancient Gunnison River was on the same order as that of the modern (Qm = 73 m3/s; Pitlick et al., 1999). In contrast, the lacustrine sedimentation probably records a few millennia (Balco et al., 2013). These data suggest one of two options: (1) the Gunnison River abandoned the canyon once water elevation rose sufficiently (∼1950–1975 m; Fig. 12) to breach the lower elevation rim of Mesozoic strata at the northeast end of the canyon after the damming event (and initial sediment progradation); or (2) the Gunnison River delta prograded completely to fill the lake, subsequently flowed across the filled lake basin during the time recorded by paleosol development in the core, and then over the top of the damming landslide without any incision and resultant breaching of the dam for millennia before abandoning the canyon as a result of stream capture from the east. The former option is more tenable, for the following reasons: (1) the lacustrine section exhibits incomplete progradation in that the uppermost part is relatively fine grained (fine to lower medium) sand (Marra, 2008), (2) the lacustrine sediments exhibit a clear Gunnison provenance (volcanic lithic component) primarily in the lower part of the core (Soreghan et al., 2007; Marra, 2008), (3) no evidence exists in the core for coarse river gravels of volcanic provenance above the lacustrine section, and (4) it seems unlikely that the landslide dam at the southwestern end would remain intact for a protracted interval while being overtopped by a large river.

For Lake Unaweep to exist and receive sediment from the ancestral Gunnison River (early) and Unaweep tributaries (later), for a few millennia, we argue that the Gunnison River had to abandon Unaweep Canyon between 1.4 and 1.3 Ma (Fig. 13). However, the presence of the 0.8 Ma Gunnison gravels at 1870 m elevation (Aslan, et al., 2014) indicates that the Gunnison River continued to occupy and incise Cactus Park. Geomorphic and limited outcrop data suggest that the lake-forming blockage occurred near the southwestern mouth of Unaweep Canyon (Figs. 2, 12, and 13), where the canyon narrows markedly and is surrounded by steep cliffs. A roadcut here at ∼1829 m elevation exposes a deposit of relatively fresh, monolithologic basement boulders derived from the surrounding steep canyon walls, but no lacustrine unit. The top of the lower lacustrine and overlying pedogenic successions in the core (Table 1) also occurs at ∼1829 m elevation, suggesting that river blockage most likely occurred northeast of the roadcut here, and that the elevation of the dam exceeded 1829 m. Subsequent landscape modification within the canyon precludes determination of the maximum elevation of the proposed landslide, but the mapping of Quaternary units (Kaplan, 2006; Oesleby, 2005); illustrated in Aslan et al., 2008) is consistent with a landslide-induced blockage of the ancestral Gunnison River in this general locality of western Unaweep Canyon.

In this blockage-induced abandonment model, the landslide dam elevation was sufficiently high (∼1950–1975 m) to cause the water to seek a lower exit toward the east, resulting in the ancestral Gunnison River flowing out what is now East Creek. This interpretation implies that the currently underfit canyon now occupied by East Creek was carved by the ancestral Gunnison River upon abandonment of western Unaweep Canyon. Once Lake Unaweep’s surface reached the eastern spillway elevation, lake levels would have lowered as incision occurred along the modern East Creek canyon. Eventually, the lake level would have fallen below ∼1890 m, at which point the basement sill just west of Cactus Park would be above the Gunnison River in Cactus Park. Thereafter, Lake Unaweep became effectively isolated from Cactus Park and the local drainages were the only water and sediment inputs (shown as purple in Fig. 14). This scenario is consistent with the upward decrease in lithic volcanic fragments noted in the cored lacustrine interval (Soreghan et al., 2007), indicating that the Gunnison-fed delta failed to prograde as far as the coring site before river abandonment occurred.

This model contrasts with that of Aslan et al. (2014) and Hood et al. (2014) in the (1) timing of abandonment of Unaweep Canyon (1.4 Ma versus 0.8 Ma), (2) identity of the river or rivers that occupied the canyon, and (3) age and manner of incision of East Creek. To summarize, we suggest that abandonment occurred ca. 1.4 Ma by the ancestral Gunnison River, as constrained by the cosmogenic ages on the lake deposit and the provenance data on both the Gateway gravels and older gravels documented in the region. The data do not support continued occupation of Unaweep Canyon after 1.4 Ma by a large river entering from the east, because the 0.8 Ma gravels in Cactus Park occur below the elevation of the Precambrian basement high within the floor of Unaweep Canyon west of Cactus Park (Fig. 12). We propose that East Creek was carved during abandonment of Unaweep Canyon as the ancestral Gunnison River sought a lower elevation exit eastward.

Postabandonment History

By ca. 1.3 Ma, the canyon floor formed a low-gradient surface at ∼1829 m elevation near the western Unaweep Canyon coring site, and was subject to relatively slow aggradation, evinced by the occurrence of stacked paleosols in the interval above the lake deposits (172–164 m). The upper conglomeratic unit records an influx of sidewall canyon debris into the inner gorge, inferred to reflect deposition within shallow lacustrine and/or paludal, alluvial, and colluvial environments (Marra, 2008). The age data within this upper conglomeratic interval (Marra, 2008; Balco et al., 2013) indicate that this unit accumulated during the mid-late Pleistocene (ca. 0.9 Ma), a time of marked glacial-interglacial climate fluctuation and associated precipitation variation conducive to mass-wasting processes. Following deposition of >160 m of the upper conglomeratic interval documented at the coring site, the surface of Unaweep Canyon stabilized, recorded by the heavily vegetated fan surfaces and well-developed Calcisols (Fig. 10), as the slope progressively decreased and the canyon walls were increasingly buried in debris.

Outside the western mouth of Unaweep Canyon, material shed from the retreating Mesozoic cliffs accumulated by mass wasting and ephemeral streams. Eventually, these deposits, the Palisade gravels, buried the remnants of the ancestral Gunnison River. Dissection of the surface of the Palisade gravels (Fig. 7) then began as the Dolores River continued to lower base level, forcing incision. In contrast, the surface of the Unaweep gravels inside the inner gorge of Unaweep Canyon remains minimally dissected (Fig. 10), with one or more late episodes of aggradation and intervening stability, marked by the multiple Calcisols, following the major fan-building episode. Inside Unaweep Canyon, fluvial base level is relatively fixed by the Precambrian bedrock floor at the western narrows canyon outlet, which somewhat isolates the canyon-interior landscape from major base-level–induced dissection common outside the western mouth of the canyon.

More recent incision near the western mouth of Unaweep Canyon by West Creek, coupled with dissection of the Palisade gravel surface, exposed the remnants of the ancestral Gunnison River that had long remained buried beneath the Palisade gravels. This incision has not been continuous, but rather included at least one period of aggradation, as recorded by the higher terrace underlain by West Creek gravels found along the axis of the modern valley near Gateway (Fig. 6). The West Creek gravels are derived from the older deposits in the area, including the Gateway gravels, the Palisade gravels, and material from Unaweep Canyon.

Analogous to the Gateway area, alluvial and colluvial aprons resulting from slope retreat ultimately buried the remnants of the ancestral Gunnison River in Unaweep Canyon. Furthermore, the modern longitudinal profile of Unaweep Canyon (Fig. 12), including enigmatic Unaweep Divide, bears no relationship to the profile of the ancestral Gunnison River through the canyon. Rather, Unaweep Divide formed as a result of aggradation of coalescing debris aprons within Unaweep Canyon after abandonment by the ancestral Gunnison River, coupled with recent headward erosion of East and West Creeks from their respective base levels at the modern Gunnison and Dolores Rivers, respectively, into the thick valley fill, with no need to appeal to differential neotectonic uplift (warping) to produce the divide (e.g., inferred by Lohman, 1961, 1981; Hunt, 1969; Cater, 1966; Sinnock, 1981; Scott et al., 2001; Steven, 2002; as also noted by Oesleby, 1978).

Duration of Gunnison Occupation of Unaweep Canyon, and Implications for Drainage Evolution and Drivers of Incision Across the Colorado Plateau

Rosenberg et al. (2014) compiled incision rates for the Colorado River system and noted that most estimates of long-term incision rates in the upper Colorado River system are between ∼100 and 150 m/m.y. (cf. Larson et al., 1975; Kunk et al., 2002; Aslan et al., 2010). Aslan et al. (2014) reported that, prior to abandonment of Unaweep Canyon, long-term incision rates averaged 100 m/m.y. Similarly, long-term incision rates reported for various parts of the Colorado and Gunnison Rivers across the Colorado Plateau typically are between ∼25 and 150 m/m.y. (e.g., Aslan et al., 2008, 2011; Karlstrom et al., 2008; Cole, 2011; Donahue et al., 2013).

Using new thermochronologic constraints for time of onset of incision ca. 6 Ma (Thomson et al., 2012), the elevation difference between the basement from the canyon rim and the borehole (∼990 m) and the duration of incision (4.7 m.y. using the abandonment age of 1.4–1.3 Ma) imply a time-averaged incision rate for the ancestral Gunnison River through Unaweep Canyon of ∼210–275 m/m.y. This rate is more than double most published long-term time-averaged rates for the upper Colorado River (Karlstrom et al., 2012b; Rosenberg et al., 2014; Aslan et al., 2014). Short-term rates for this region can range much higher (e.g., Aslan et al., 2014), but cannot be compared to long-term rates owing to the dependence of incision rates on measurement interval (Finnegan et al., 2014; Sadler and Jerolmack, 2014).

Our inference that erosion in Unaweep Canyon began in the latest Miocene is consistent with other work documenting probable late Miocene landscape evolution associated with the integration of the lower Colorado River system and inception of southwestward flow across the Colorado Plateau. Abundant data indicate that the lower Colorado River system was integrated and flowing to the Gulf of California by ca. 6–5.3 Ma (e.g., McKee and McKee, 1972; Young and McKee, 1978; Young, 1982; Potochnik, 1989, 2001; Spencer et al., 2001; Young and Spamer 2001; Pederson et al., 2002; Lucchitta, 2003; Dorsey et al., 2007, 2011; House et al., 2005, 2008; Karlstrom et al., 2012a, 2012b). Data from the upper Colorado River system have been less well constrained, with regional uplift and associated major fluvial incision assumed to have begun accelerating ca. 11–6 Ma (e.g., Aslan et al., 2010; Karlstrom et al., 2012b). The ca. 11 Ma date is based largely on the occurrence of river gravels below basal basalt flows dated to 10.8 ± 0.2 Ma on the Grand Mesa (Kunk et al., 2002; Aslan et al., 2010) or similar regions. Our new thermochronological results from Unaweep Canyon (Thomson et al., 2012), however, more precisely constrain major fluvial incision in this region to beginning ca. 6–5 Ma, approximating the values from the lower Colorado River system. This result supports the plausibility of an essentially synchronous onset of rapid incision across the entire Colorado Plateau and, by extension, the inference of a regional epeirogenic and/or isostatic driver for this rapid incision, with secondary geomorphic (propagation of headward erosion) or climatic drivers (cf. Karlstrom et al., 2012b; Rosenberg et al., 2014).

Late Cenozoic Incision, or Exhumation of Unaweep Canyon?

The culmination of research over the last century, including data presented here, has delineated the broad strokes of the late Cenozoic history of Unaweep Canyon: a large river (we posit the ancestral Gunnison alone) flowed from northeast to southwest through the canyon, beginning ca. 6–5 Ma (Thomson et al., 2012), and persisting until canyon abandonment ca. 1.4–1.3 Ma (Balco et al., 2013; this work), or ca. 0.8 Ma by other estimates (Aslan et al., 2014; Hood et al., 2014). Did this large river, however, incise a new canyon, or exhume and reoccupy a paleovalley? The latter hypothesis was proposed (in Soreghan et al., 2007, 2008), in part on the basis of the occurrence of strata of inferred late Paleozoic age in the basal interval of the Unaweep Canyon borehole (Table 1). The oldest age of the stratigraphic fill yields the minimum age of the landform. This paleolandform hypothesis is controversial (Aslan et al., 2005, 2008, 2014; Hood, 2009; Hood et al., 2009); we propose, however, that it remains viable because it explains aspects of the canyon that are otherwise perplexing. These aspects include the planform of Unaweep Canyon relative to the structure and regional drainage network of the Uncompahgre Plateau, together with the Cutler-Precambrian contact map relations, and the incision history.

The modern planform course of Unaweep Canyon remains problematic because it indicates that the ancestral Gunnison River flowed subparallel to the Uncompahgre Plateau before turning to cleave through Precambrian basement perpendicular to the plateau axis in a manner that ignores the regional drainage network (Fig. 14). Moreover, this crossing bisects the structural crest of the Uncompahgre Plateau, incising through the highest-elevation region (Figs. 1 and 14). Many rivers of the Colorado River system are superposed atop Laramide uplifts, likely reflecting the positions of ancestral stream courses that flowed through easily erodible strata in precursor lowlands (e.g., Hunt, 1969; Dickinson, 2013). Dickinson (2013) highlighted the example of the upper Colorado River where it cuts through Precambrian basement around the plunging nose of the northwestern Uncompahgre Plateau (Westwater Canyon; Figs. 1, 2, 13, and 14). Dickinson (2013) called upon processes first highlighted by Oberlander (1965, 1985), i.e., that this superposition reflects stream erosion into alternating layers of resistant and erodible strata progressively exposed as the stream migrated down structural plunge, ultimately forming a looping transit transverse to the structure, through basement-cored Westwater Canyon. Figure 14 illustrates how the drainage divide of the Uncompahgre Plateau trends northwestward south of Unaweep Canyon but curves sharply to the northeast north of the canyon. The Colorado River migrated from the northern margin of the northeast-trending drainage divide north of Unaweep Canyon, and continued down plunge toward Westwater Canyon, as Dickinson (2013) detailed. The course of the Colorado River follows the structural plunge of the base of the Dakota Sandstone around the nose of the Uncompahgre Plateau (Fig. 15). In this model, the resistant layer was the Dakota Sandstone, which crops out discontinuously at all elevations of the Uncompahgre Plateau. The softer retreating layer was likely the Mancos Shale and the Mesa Verde Formation, as they do not occur anywhere atop the Uncompahgre Plateau, but form retreating fronts on the outside loop of the modern Colorado River. Analogously, south of Unaweep Canyon, the Gunnison River (and Uncompahgre) flowed parallel to the spine of the Uncompahgre Plateau controlled by the northwest-trending drainage divide, and thus migrated northeastward along the top of the Dakota Sandstone, since that unit forms the youngest unit of the eastern dip slope (Williams, 1964), transecting the dipping flank of the plateau. Through Unaweep Canyon, however, the looping transect model breaks down; here, the Gunnison River made a nearly 90° turn to cut southwestward across the structural axis in a manner that opposes both the northwest plunge of the uplift, and the northeast dip of its flank (Figs. 13, 14, and 15). The longitudinal path of the Gunnison River where it parallels the axis of the Uncompahgre Plateau upstream of Unaweep Canyon indicates that this part of the river course was influenced by the northeast structural dip of the plateau flank, but its perpendicular path across the uplift does not follow the pattern of superposition outlined by Oberlander (1965, 1985) and Dickinson (2013).

The Paleozoic-age hypothesis can explain this odd map pattern. Consider the possibility that the Precambrian-hosted inner gorge of Unaweep Canyon existed as a paleovalley in late Paleozoic time, and was subsequently filled by sediment of late-early Permian age (Soreghan et al., 2012) prior to deposition of the superjacent strata that cover the plateau and the greater region. Later, differential compaction between the 1 km compactible sedimentary fill within the paleovalley and the noncompactible crystalline Precambrian basement should have resulted in a low valley in the overmass burying the ancient Uncompahgre uplift above the Precambrian-hosted gorge southwest of Cactus Park. The ancient Gunnison-Uncompahgre gravels on top of the Uncompahgre Plateau south of Unaweep Canyon (Aslan et al., 2008) record the positions of the ancestral Gunnison and Uncompahgre Rivers as they migrated northeastward down the dipslope of the Uncompahgre Plateau. In this interpretation, the topographic low created by the differential compaction above the ancestral gorge would have guided the combined Gunnison-Uncompahgre river in its 90° turn westward to bisect the Uncompahgre Plateau, ignoring the prevailing northeastward dip slope. East Creek east of Cactus Park was carved by the ancestral Gunnison River upon canyon abandonment. The existence of the Cutler Formation–filled paleovalley at the western mouth of Unaweep Canyon corroborates the hypothesis of several hundred meters of preserved paleorelief here (Soreghan et al., 2012). The rapid time-averaged incision rates documented from the thermochronological and geochronological data are consistent with the inference of erosion of a paleovalley filled by weak, more erodible sedimentary strata, rather than primary incision through crystalline bedrock, owing to the relationship between fluvial erosion and bedrock strength (Sklar and Dietrich, 2001).

Ultimately the ideas presented here imply preservation (through subsidence and burial), and subsequent exhumation of a very ancient landscape, a phenomenon well documented in cratonal regions (e.g., Gondwanan continents; Twidale, 1998, 2003), but generally dismissed for tectonically active landscapes, such as those of the western United States. Recent research on the southern Colorado Plateau, however, has posited the existence of paleocanyons (pre–6 Ma, 70 Ma) influencing the evolution of the lower Colorado River system and especially the iconic Grand Canyon (Flowers et al., 2008; Wernicke, 2011; Flowers and Farley, 2012, 2013), although this idea is debated (Dickinson, 2013; Karlstrom et al., 2013; Lee et al., 2013; Young and Crow, 2014). The example of Unaweep Canyon (and perhaps the Grand Canyon) highlights the potential persistence, through burial and exhumation, of paleolandforms in even noncratonal regions. Exploitation of paleolandforms by recent drainages may have played a greater role than commonly appreciated in the landscape evolution of the western United States.

  1. Composition and paleocurrent analyses of the Gateway gravels near Gateway, Colorado, comparisons with undisputed ancestral Gunnison River and Colorado River gravels, published provenance of older (10–2 Ma) gravels in the region, and other considerations show that the ancestral Gunnison River, and not the ancestral Colorado River, flowed to the southwest through Unaweep Canyon prior to canyon abandonment at ∼1.4 Ma.

  2. Previously documented coring within Unaweep Canyon confirms the presence of a thick (locally >330 m) fill that includes an ∼140 m lacustrine interval dated to ca. 1.3 Ma that exhibits, especially in its lower part, sands of Gunnison River provenance. Lake formation resulted from catastrophic mass wasting in the western narrows of Unaweep Canyon that blocked the ancestral Gunnison River, ultimately resulting in abandonment of Unaweep Canyon by the ancestral Gunnison River between 1.4 and 1.3 Ma, and carving of East Creek canyon as the river sought lower elevation toward the Grand Valley. The ancestral Gunnison River continued to occupy and incise Cactus Park even after the abandonment of Unaweep Canyon.

  3. Thermochronological data from Unaweep Canyon indicate that the ancestral Gunnison River began to occupy the Precambrian-hosted gorge of Unaweep Canyon in the latest Miocene (ca. 6–5 Ma). This date, coupled with abandonment estimates (1.4–1.3 Ma) produces a time-averaged incision rate of ∼210–275 m/m.y., more than double most long-term incision rates of the greater region.

  4. Onset of canyon occupation and rapid incision by the ancestral Gunnison River coincided with the timing of integration of the lower Colorado River system to the Gulf of California. The synchroneity of this incision across the Colorado Plateau supports the inference of an ultimate tectonic or epeirogenic driver for this widespread incision and ultimate drainage integration.

  5. Although the late Cenozoic history of Unaweep Canyon is one of fluvial occupation by the ancestral Gunnison River, several perplexing aspects of the canyon remain enigmatic, and are reconciled by considering the possibility that the Cenozoic river exploited and ultimately revealed a paleolandform. The anomalous course of the canyon across the highest-elevation Precambrian surface of the Uncompahgre Plateau in a manner that ignores the regional drainage network, new mapping demonstrating the existence of a paleovalley of substantial relief at the western mouth of Unaweep Canyon, and the rapid long-term incision rates documented by the combined geochronologic and thermochronologic data support the previously proposed hypothesis that the inner, Precambrian-hosted gorge of Unaweep Canyon was initially carved in the late Paleozoic, and was simply exhumed in the late Cenozoic.

Funding for this research was partially supported by grants from the National Science Foundation (EAR-0230332 and EAR-0934259) and from the U.S. Geological Survey EDMAP program (G10AC00329, G11AC20217, G12AC20266), including in-kind support for field mentorship and mapping from the Colorado Geological Survey (V. Matthews and D. Noe), and the Oklahoma Geological Survey (R. Keller, R. Standridge, and N. Suneson). Marra was partially funded by a student grant from the DOSECC Consortium (Drilling, Observation, Sampling of Earth’s Continental Crust). We thank Himes Drilling (Grand Junction, Colorado) for major in-kind support of drilling operations, and J. Stowell of Mt. Sopris Instruments for well logging. Marra’s work on Unaweep core material benefitted from Rock-Eval analyses provided by Humble Geochemical, tephra analyses by E. Wan (U.S. Geological Survey), and palynological analyses by D. Willard and J. O’Keefe. University of Oklahoma students D. Ambuehl, T. Foster, L. Keiser, K. Patrick, V. Priegnitz, A. Shock, and A. Sweet provided field and laboratory support. We thank the Powell family for their hospitality, the Massey Ranch for permission to drill, and the many kind landowners in the region for property access, including R. Beach, P. Bristol, B. Chesnick, J. Lewis, R. Tipping, the Larsens, and the Moores. We also thank M. Blum, C. Chase, and M. Soreghan for discussions of the geology of Unaweep Canyon over the years, and W. Bowen (http://geogdata.csun.edu) for creating perspective digital elevation models for our use. Page contributions for this publication were covered by the Maxey Professorship, University of Oklahoma. We greatly appreciate the considerable time and effort invested in constructive critiques of an earlier version of this manuscript by Guest Editor S. Beard and two anonymous reviewers.

1 Supplemental Table 1. Location, Provenance, and Paleocurrent Data for Gravels of Mapped Areas. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES01112.S2 or the full-text article on www.gsapubs.org to view Supplemental Table 1.
2 Supplemental Table 2. Provenance Data of Colorado River Terrace Material and Gunnison Gravels. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES01112.S3 or the full-text article on www.gsapubs.org to view Supplemental Table 2.