Incision history of the Black Canyon of Gunnison, Colorado, over the past ∼1 Ma inferred from dating of fluvial gravel deposits

Three major orogenic plateaus on Earth include the Tibetan Plateau, the Altiplano-Puna Plateau of South America, and the Colorado Plateau of the western United States. In tectonically active regions such as Tibet and the Andes, the complex interactions between dynamically uplifting plateaus and geomorphic evolution are recorded in deep bedrock gorges commonly located at the edges of these uplifted plateaus (Schildgen et al., 2007). Rivers on the eastern margin of the Tibetan Plateau such as the Yalong, Dadu, and Yangtze rivers and tributaries (Kirby et al., 2003; Ouimet et al., 2010; Kirby and Ouimet, 2011) and exiting the Altiplano (Schildgen et al., 2007) are interpreted to be adjusting to tectonic forcings because of the youthfulness of these mountain belts that give rise to the dramatic landscapes. Spectacular bedrock gorges also exist in the western U.S. associated with the Sierra Nevada (House et al., 2001; Ducea et al., 2003), Hells Canyon and associated Wallowa Mountains uplift (Hales et al., 2005), the edge of the Colorado Plateau in Grand Canyon (Karlstrom et al., 2008), and the Rocky Mountains (Karlstrom et al., 2012). However, in the case of the western U.S., there is ongoing debate about the extent to which these rugged gorges and landscapes are due to active tectonics (Epis and Chapin, 1975; McMillan et al., 2002, 2006; Karlstrom et al., 2008, 2012; Eaton 2009; Aslan et al., 2010) and/or are influenced more by climatic forcing and isostatic adjustments to denudation (Gregory and Chase, 1992, 1994; Pederson et al., 2002; Molnar, 2004; Roy et al., 2004, 2009).

Figure 1 shows comparative profiles of deep bedrock canyons from the areas mentioned above. Compared to other bedrock canyons in the western U.S. and other orogenic plateaus, Black Canyon of the Gunnison is not the widest or deepest, but its depth-to-width ratio is among the highest of the world’s bedrock gorges. In the western U.S., Grand Canyon and Hells Canyon, both carved mainly in the last 5–6 Ma due to mantle-driven uplift (Hales et al., 2005; Karlstrom et al., 2008), are much wider overall, but their inner bedrock gorges are shallower and less steeply walled. Yosemite, primarily a glacial valley, is a much broader canyon. Incision rates are anomalously high on rivers that exit the three main orogenic plateaus of the world: Tibetan river gorges incise at rates of 300–500 m/Ma or more (Ouimet et al., 2010; Dortch et al., 2011) and incision in the Bolivian Andes has been measured at 1350 m/Ma (Safran et al., 2005). In the western U.S., Black Canyon incision rates are locally 400–550 m/Ma (Aslan et al., 2008; this paper). These are high incision rates in spite of the fact that the river is carving through resistant Precambrian basement rocks, including the unfoliated and sparsely jointed Vernal Mesa Granite (Hansen, 1965).

The goal of this paper is to integrate new geomorphic, geochronologic, and geophysical data and methodologies to understand the incision history of Black Canyon of the Gunnison. This work builds on the work of Wallace Hansen (1965, 1967, 1971, 1987), who pioneered the study of many of the geologic and geomorphic relationships explored here. To this geologic context, we bring several new analyses to this geologic problem, including: (1) analysis of longitudinal profiles of the Gunnison River system (e.g., Wobus et al., 2006; Aslan et al., 2008; Darling et al., 2009a, 2009b; Kirby and Whipple, 2012); (2) synthesis of new and previously published river incision rate data on the Gunnison River and its tributaries; (3) sampling and characterization of a key Lava Creek B ash locality for tephrochronologic identification and confirmation of ages; and (4) mapping and cosmogenic ²⁶Al and ¹⁰Be dating of terrace deposits associated with the Gunnison River. Collectively, these new observations and data allow us to evaluate possible driving mechanisms for incision of Black Canyon.

The Gunnison River follows the crest of the Laramide anticlinal Gunnison uplift as it flows to the west-northwest across the western slope of the Colorado Rockies (Hansen, 1965). This Laramide uplift is within the broader Pennsylvanian-age Uncompahgre uplift of the Ancestral Rockies. During the Late Cretaceous, these Precambrian-cored uplifts were mantled by ∼2 km of sedimentary rocks that became broadly flexed and uplifted into the Gunnison uplift starting ca. 70 Ma (Hansen, 1965). There is little record of the paths of early post-Laramide rivers, but modern river courses seem unaffected by Laramide structures (e.g., Uinta, Uncompahgre, Gunnison uplifts), suggesting that river courses meandered across low-relief surfaces at higher stratigraphic levels and became superimposed on the uplifts during regional exhumation (e.g., Hansen, 1965; Hunt, 1969; Douglas et al., 2009). In contrast, modern rivers flow around Tertiary laccoliths, suggesting the influence of mid-Tertiary volcanic constructional topography (e.g., San Juan volcanic field, Elk and West Elk laccoliths; Hunt, 1969). Hence, the modern drainage system likely initiated with magmatic and tectonic events associated with the Oligocene ignimbrite flare-up (Lipman, 2007) resulting from the detachment of the Farallon flat slab from the base of North America (Humphreys et al., 2003).

A paleo–Gunnison River in about the same location as the modern river emerged in the Oligocene from ca. 35 to 29 Ma (Hansen, 1965), when drainage was constrained between volcanic topography of the West Elk Mountains to the north and the San Juan volcanic field to the south (Fig. 2A; Lipman, 2007). Eruptive sheets from these two volcanic centers thicken into an east-west paleovalley in approximately the same location as the modern Gunnison River course, pinning drainage between the two constructional edifices (Hansen, 1965). In particular, the thickening of the 28.5 Ma Blue Mesa Tuff in the paleovalley axis, with its base contoured by Hansen (1971), indicates the paleo-Gunnison flowed into an east-west–oriented valley between these volcanic edifices. Figure 3 shows two paleosurfaces for the Oligocene timeframe: the upper profile, modified from Hansen (1971), is based on structure contours at the base of the Oligocene Blue Mesa Tuff dipping upstream (28.5 Ma; Lipman, 2007); the lower profile is created from the elevations of Oligocene West Elk Breccia (29–34 Ma; Coven et al., 1999) at its closest and lowest proximity to the current Gunnison River. Both show that the Oligocene paleo-Gunnison was potentially incising into Precambrian bedrock at modern elevations of 7000–9000 ft (2134–2744 m).

The longitudinal profiles of the Gunnison River and its tributaries are shown in Figure 2B. The modern profile, from headwaters to confluence with the Colorado River, displays a striking convexity associated with the Painted Wall reach of the Black Canyon (Fig. 3). This knickzone is indicative of a disequilibrium river profile (Mackin, 1948) and separates two generally concave portions of the river that have different incision histories. The main features of the Gunnison profile relative to rock type are shown in Figure 3: (1) a low-gradient (1.25 m/km) western reach in erodible Cretaceous and Jurassic rocks, (2) a slight steepening of gradient within basement rocks for ∼30 km until the Red Rocks fault zone, (3) a concave river profile inflection marking where the river leaves the trace of the Red Rocks fault zone and transitions into, (4) the high-gradient (10 m/km) knickpoint in the Painted Wall reach that is partly within the resistant (sparsely jointed and unfoliated) Vernal Mesa Granite (pink stripe in Fig. 3). Above this, the river returns to a lower gradient (5 m/km) above Blue Mesa reservoir while still in Precambrian basement. Finally, the river steepens into the basement-cored Sawatch Range in its high headwaters (as the East River). In summary, while the shallowest reaches are floored by soft rocks (Mancos Shale) and the steepest partly in basement (Vernal Mesa Granite), there are both steep and gentle reaches that do not correspond to these rock types. Thus, we infer that the profile geometry is only partly attributable to erodibility of the substrate, in keeping with incision rate evidence that the steepest reach is rapidly incising and is likely to be a transient knickpoint.

Modern tributaries to the Gunnison River include the Uncompahgre, North Fork Gunnison, Lake Fork Gunnison, and Cimarron rivers and Cebolla Creek (Figs. 2A, 2B). The Uncompahgre River drains the San Juan Mountains from the vicinity of Ouray, Colorado, and has formed a broad, low-gradient canyon in the Cretaceous rocks flanking the Gunnison uplift. North Fork Gunnison River is the major northern tributary to the Gunnison River, and drains the West Elk Mountains. It is of similar slope to the Uncompahgre River, and upon exiting volcanic strata flows largely on Cretaceous bedrock. The Lake Fork Gunnison and Cimarron rivers and Cebolla Creek are steep tributaries draining the San Juan Mountains. The paleo–Bostwick River, also shown on Figure 2A, is defined by a series of high, grass-covered, low-relief parks underlain by fluvial gravels marking the location of this north-flowing paleotributary to the Gunnison River that had its confluence with the Black Canyon of the Gunnison in Red Rock Canyon (Hansen, 1987).

Our conceptual framework for drainage evolution in the Black Canyon region utilizes modern drainages as analogs for past drainages. Paleo-tributaries document drainage reorganization due to tectonic (Hansen, 1965; Steven, 1975) and/or geomorphic stream piracy events (Pederson et al., 2002; Schneeflock et al., 2002; Aslan et al., 2008). Stream piracy is considered to be the main mechanism here for rapid abandonment of river reaches and drainage reorganization. Grizzly Gulch Creek (Fig. 2A) is presumed to have been abandoned after headwater capture by Iron Creek (Hansen, 1965), while the paleo–Bostwick River was captured by Cedar Creek of the Uncompahgre system (Figs. 2A, 4A). The modern Lake Fork Gunnison and Cimarron rivers and Cebolla Creek (Fig. 2B), which drain the San Juan volcanic field and have confluences with the Gunnison River above the knickzone within Black Canyon, are used as proxies for pre-knickpoint gradients. We use these gradients to reconstruct the paleo-Bostwick drainage as a constraint on the former gradient of the Gunnison River.

Knickpoints and convexities in river profiles are often attributed to relative base-level lowering within the stream system (Gardner, 1983; Burbank and Anderson, 2001; Whipple, 2004) and/or fluvial response to variable rock strengths or erodibility (Burbank and Anderson, 2001; Berlin and Anderson, 2007). Incision rate data across knickpoints provide an avenue to evaluate the extent to which convexities and knickzones may be due to resistant bedrock versus transient incision pulses that may be impeded by hard bedrock. If knickpoints are primarily a response of the fluvial profile to resistant bedrock, incision rates should be broadly similar below, within, and above the knickpoint (Gilbert, 1877; Stock and Montgomery, 1999; Phillips et al., 2003; Duvall et al., 2004). Alternatively, in transient knickzones, rates are expected to be highest within the knickzone as the river adjusts and strives to reacquire equilibrium after a change in local base level. Such disequilibrium conditions may be prevalent at the edges of uplifted plateaus or following a change in uplift rate within an active region (Bishop et al., 2005; Goldrick and Bishop, 2007; Whipple, 2004), near a tectonic drop in base level (Karlstrom et al., 2008), or due to drainage reorganization such as stream capture or even bedrock meander cutoff events at small scales (Finnegan and Dietrich, 2011).

Figure 3 (vertical arrows) shows the best-constrained Gunnison River–region incision rates displayed on a longitudinal profile that extends from the confluence with the Colorado River at Grand Junction to the headwaters of the Gunnison River. A summary of incision rates used in paleoprofile reconstruction is listed in Table 1, including previously published as well as new data points (new data points in bold).

The best-constrained terrace ages are where the 0.64 Ma Lava Creek B ash has been found in association with strath terraces along the Gunnison River and tributaries (Figs. 2 and 3). Below the Black Canyon knickzone, in the downstream reaches of the Gunnison River, 0.64 Ma Lava Creek B ash is found on erosional pediments near Grand Junction (Figs. 2 and 3, points A, B), at Kelso Gulch (Figs. 2 and 3, point C), and near the modern river at Sawmill Mesa (Figs. 2 and 3, point D). Incision rates from strath terraces constrained by these ash locations are ∼130–150 m/Ma, indicating relatively consistent incision throughout this section of the river since ca. 0.64 Ma.

In contrast, in the upstream portions of the Gunnison River and above the central knickzone, Lava Creek B ash is found at the Blue Mesa Dam (Figs. 2 and 3, point H) and the Lake Fork Gunnison–Gunnison River confluence (Figs. 2 and 3, point I). These two locations give relatively slower incision rates of ∼95 m/Ma. The interpretation of these low incision rates is that these reaches of the Gunnison River upstream from the modern knickzone have not yet experienced the base-level fall associated with knickpoint migration. A dearth of datable material within the middle concavity between Chasm View and Blue Mesa Dam (1900–2350 m elevation, Fig. 3) leaves incision rates within this reach unconstrained; this paper focuses on the upper and lower reaches of the Gunnison River.

Incision rates within the Black Canyon knickzone area have been inferred by previous workers (Hansen, 1987; Dethier, 2001; Aslan et al., 2008) to be among the highest rates in the Rockies (Dethier, 2001). Hansen (1987) quoted an incision rate based on projection of the Bostwick paleoriver to where it would have joined the Gunnison River near Red Rocks Canyon (400–500 m/Ma; Figs. 2 and 3, point G; Fig. 4C). His age constraints came from the presence of Lava Creek B ash beds overlying gravels of the paleo-Bostwick tributary (Hansen, 1987). The combined data led him to infer that Black Canyon has been carved in the last 1–2 Ma. We concur with this conclusion and refine it based on new data presented below. The elevation of the Chukar Ridge Lava Creek B locality (Table 1) provides the highest calculated incision rate in this study (640 m/Ma; Figs. 2 and 3, point F). This site is associated with river gravels but is difficult to correlate with other terrace heights.

One of the uncertainties in Hansen’s calculation of a 1–2 Ma age of Black Canyon stems from uncertainty of the age of post-abandonment deposits that overlie Bostwick paleoriver gravels in Bostwick Park (Fig. 4A). Dickinson (1966) reported two ashes at different stratigraphic levels in a road cut leading up to the south entrance to Black Canyon (Fig. 5). The upper ash was identified as the 0.64 Ma (Lanphere et al., 2002) Yellowstone Lava Creek B ash (Izett and Wilcox, 1982; Hansen, 1987). The lower ash (Fig. 5) was analyzed and reported by Izett and Wilcox (1982) as being similar to the 1.29 Ma Mesa Falls Yellowstone ash (Lanphere et al., 2002), but the correlation was uncertain. Given Hansen’s argument that Bostwick paleoriver gravels were graded to an elevation ∼350 m down into the depth of the modern Black Canyon, use of the Mesa Falls age as the minimum age on the gravels led to an inferred rate of 325 m/1.29 Ma = 252 m/Ma, necessitating ∼2.5 Ma to carve the entire canyon; whereas use of the Lava Creek B age yields a rate of 507 m/Ma and a duration of ∼1 Ma for the carving of Black Canyon, assuming steady rates.

We revisited the site and attempted to resample from the location shown in Figure 5. Although an extensive excavation was conducted, no ash layer was found in the stratigraphic position of the lower ash noted by Dickinson (1966). Instead, we obtained the original sample (courtesy of Dickinson) and reanalyzed it. Our reanalysis of the Dickinson “Mesa Falls (?)” ash (Table 2) indicates that while the Mesa Falls correlation is still permissive, there is not a definitive match between the Dickinson ash sample and the Mesa Falls ash (see the Supplemental File¹). However, there is a positive (>0.95 similarity coefficient) correlation with multiple Lava Creek ash bed samples. The Dickinson sample exhibited a lower iron concentration level that is more similar to Mesa Falls glass shards than Lava Creek B, as well as the moderate hydration typically seen in older samples. Given that Yellowstone tephra (Huckleberry Ridge, Mesa Falls, and Lava Creek) have overlapping chemistries, the original identification as Mesa Falls ash was likely based on these latter characteristics. Although a Mesa Falls correlation remains possible from the geochemical analysis, new field mapping has failed to reveal a marked unconformity in the few-meters-thick section of fill between the ashes; therefore, the correlation to Lava Creek B ash is supported by stratigraphic context (Aslan et al., 2008). Thus, based on the new analysis (Table 2) and detailed work on the geology at Bostwick Park that suggest post-abandonment deposits are likely close in age to the gravels (Sandoval, 2007; Aslan et al., 2008), we prefer the 0.64 Ma age as a minimum age on the paleo-Bostwick river gravels. We suggest these ashes represent differential erosion and re-deposition of reworked Lava Creek B tephra following abandonment of the paleo-Bostwick paleovalley.

To further constrain the age of Bostwick gravels and attain additional new incision rate data, we used cosmogenic nuclides (¹⁰Be and ²⁶Al) to determine the depositional ages of gravels sampled from two locations: (1) Bostwick Quarry, and (2) terrace Qt7 (223 m above modern river level) at the North Fork Gunnison–Gunnison River confluence (see Figs. 2 and 4 for locations). We used the isochron burial dating method (Balco and Rovey, 2008) at Bostwick Quarry, and profile dating at terrace Qt7.

The depositional history of these gravels was described by Hansen (1987), Sandoval (2007), and Aslan et al. (2008). About 10 m of coarse gravel was deposited in this Gunnison River paleo-tributary, which had its confluence with the Gunnison through Red Rock Canyon (Fig. 4A). Headward cutting of Cedar Creek, a tributary to Uncompahgre River, then captured the paleo-Bostwick drainage, leaving a largely abandoned canyon with an underfit, ephemeral stream draining through Red Rock Canyon into Black Canyon. Abandonment triggered the relatively rapid infilling of the canyon with 5–15 m of locally derived “yellow” alluvium that includes proximal deposits composed of reworked Mancos Shale from nearby hill slopes. As mentioned above, the locally derived sediment interfingers with lenses of reworked Lava Creek B ash that directly overlie the gravel deposit at Bostwick Quarry (Fig. 4B), constraining Bostwick river gravels to be older than 0.64 Ma (Aslan et al., 2008).

Cosmogenic nuclide burial dating provides a more direct constraint on the age of these gravels. We obtained a sample from the base of a >10-m-thick gravel exposure in the recently excavated Bostwick Quarry (Fig. 4B), ∼1 km east of the road outcrops shown in Figure 5. Significant shielding (>10 m) in this location allows the burial age to be interpreted as an accurate (although imprecise) depositional age of the gravels.

The cosmogenic burial analysis was from four quartzite clasts that were collected from the base of the quarry face ∼10 m stratigraphically below Lava Creek B ash. Post-burial shielding from cosmic radiation was thus at least 10 m (thickness of the gravel) and post-abandonment deposits added an additional 3 m of post–0.64 Ma shielding. Cosmogenic ¹⁰Be and ²⁶Al were determined for each clast and analyzed using the isochron method for burial dating (Balco and Rovey, 2008). This method is best executed with clasts collected from the same stratigraphic horizon such that each clast has the same post-burial cosmogenic production history due to similar depth of burial. Each clast has a high probability of having experienced a different exhumation and transport history prior to deposition than the other clasts, such that the initial concentrations of ¹⁰Be and ²⁶Al vary. Evolution of these different initial ratios due to radioactive decay after deposition forms an isochron that has an age indicated by its slope and post-burial production indicated indirectly by the y-intercept. Since concentrations of ¹⁰Be and ²⁶Al need to be corrected for the post-burial production experienced due to spallation reactions and muons at depth, age determination is necessarily an iterative process (Balco and Rovey, 2008). The isochron date for deposition of the gravel (Fig. 6) is constrained primarily by one data point, and could be improved with further analyses. The best-fit line yielded an age of 870 ± 220 ka (Fig. 6; Darling et al., 2012), compatible stratigraphically with both the Lava Creek B ash minimum age for the gravels and field relationships that suggest the gravel to be older than but close in age to the Lava Creek B ash. Thus, the time of major depositional activity of Bostwick paleoriver gravels is constrained to between ca. 0.87 and 0.64 Ma.

North Fork Gunnison River is the major northern tributary to the Gunnison River, and joins the Gunnison River below Black Canyon (Figs. 2 and 7). This tributary junction has migrated down the dip slope of the underlying Paleozoic–Mesozoic strata as it has incised, leaving abandoned river drainages (e.g., Smith Fork and Red Canyon) and 11 remnant river terraces abandoned above the present profile (Fig. 7B; Hansen, 1965; Sandoval, 2007). These terrace deposits at the confluence are severely eroded (treads are poorly preserved) and relatively thin (less than several meters), implying they represent strath terraces (Sandoval, 2007). The highest terraces are more than 300 m above the modern Gunnison River (Fig. 7B) in this locality. The confluence of North Fork Gunnison River with the Gunnison River has apparently maintained its general geometry, while the North Fork and the confluence migrated down the dip of the underlying strata and the Cretaceous Dakota escarpment retreated northwards. This is similar to models for streams “sliding down” the dip slopes of the Uncompahgre uplift (Aslan et al., 2008; Darling et al., 2009a) and differs from other locations where rivers become entrenched during superposition across basement-cored uplifts, as was likely to be simultaneously occurring to the east on the Gunnison uplift (Hansen, 1987). The thickest and most widespread of these deposits, Qt7, with a strath at 223 m above the modern river (Fig. 7B), was chosen for trenching by backhoe and collection of a seven-part vertical profile for cosmogenic profile dating.

The trench into terrace Qt7/223 m was 6.1 m deep; sample depths ranged from 1.4 to 5.6 m. The stratigraphy included an upper 1.1 m of local fine-grained fill overlying pedogenic carbonate atop the gravels. The top of the gravel was at 1.4 m depth, and continued below the base of the excavation, which was limited by the reach of the backhoe. Gravel exposures on the edge of the terrace nearby were ∼6 m thick, suggesting that the bottom of the trench was within 1–2 m of the base of the gravel. The overall stratigraphy seen in the excavation at the confluence is similar to that of Bostwick Quarry, where river gravels are abruptly overlain by fine-grained, yellowish, locally derived colluvium. Five samples consisting of combined mixtures of sand and gravel were analyzed for ¹⁰Be; two of these also yielded meaningful ²⁶Al results. Cosmogenic data are given in Table 3 and Figure 7C.

The model fit to the data is shown in Figure 7C. The best-fit age is 0.98 (^+0.38/_–0.07) Ma for a local production rate of 12 at/g/yr (¹⁰Be) and 82 at/g/yr (²⁶Al). The knee in the profile is due to the inherited component that dominates at depth but is overprinted by much higher postdepositional production near the surface. The uncertainty in the fit is strongly asymmetric due to the possibility of terrace erosion that increases the model age. The best-fit value for inheritance corresponds to an erosion rate in the source area of 60 m/Ma. The model describes the data well, with a value for reduced χ² of 0.85.

Using the model profile age of 980 ka for terrace Qt7 yields an incision rate of 228 (⁺¹⁷/_–64) m/Ma. Assuming that the incision rate has been relatively steady, the lower terrace Qt5 (at an elevation of 190 m) probably correlates with the gravels at Bostwick Quarry and deposition of the Lava Creek B ash at 0.64 Ma. An inferred paleoprofile for this “Lava Creek B” age range is shown as the light purple swath in Figure 3. This relatively well-constrained paleoprofile mimics the modern river profile and suggests that a steep knickzone existed at 0.64 Ma within but farther west in Black Canyon, separating low-gradient upstream and downstream reaches.

An examination of the relationship of both the modern and the 0.64 Ma knickpoints to changes in bedrock substrate and to the fault-bounded Gunnison uplift supports the hypothesis that the knickpoint is transient, with an upstream propagation rate that has been slowed in the last 0.64 Ma as the knickzone entered basement rocks. The existence of the modern 10-km-long high-gradient reach may be partly explained by the presence of the Vernal Mesa Granite. However, the knickzone extends upstream and downstream beyond this resistant rock (Fig. 3), suggesting that bedrock is not the only controlling factor. Prior to 0.64 Ma, the knickzone is thought to have moved rapidly through Cretaceous rocks from the Unaweep Canyon area to the Gunnison uplift (Aslan et al., 2008). At 0.64 Ma, the river was just beginning to incise through basement rocks and the top of the steep knickpoint was likely near the Red Rock Canyon confluence.

A profile reconstructed from 10 Ma basalts at Grand Mesa and Flat Top Mountain near Almont, Colorado, have ancient gravels beneath them (Fig. 3), and suggest that the late Miocene was characterized by a low-relief landscape with an established westward-draining river system (Hansen, 1971). River gravels containing volcanic and basement clasts are found below and incorporated into the basal basalt flows of Grand Mesa (Aslan et al., 2010). Gravels found beneath the ca. 10 Ma Flat Top Mountain (north of Gunnison) also contain basement clasts. Grand Mesa and Flat Top Mountain, while similar in age and in their preservation of gravels from major paleorivers, have very different calculated long-term incision rates. The Grand Mesa basalt flow is 1503 m above the modern Gunnison–Colorado River confluence, while the Flat Top gravels are 639 m above the headwater regions of the Gunnison River. Thus, post–10 Ma incision by the Gunnison and Colorado rivers in the lower parts of these rivers has removed nearly three times the thickness of rock from the lower reaches of the Gunnison River, with a regional long-term incision rate of 140 m/Ma (Fig. 3, point A; Aslan et al., 2008; Darling et al., 2009a). In contrast, the high-elevation Rocky Mountain headwater regions of the Gunnison River have been incising through the action of small streams in steep, narrow canyons at 64 m/Ma (Fig. 3, point J) over the same time frame. The headwaters region is upstream from the modern knickpoint (Fig. 3) and is adjusted to a local base level of the low-gradient upper section of the Gunnison River. This river reach has yet to feel the effects of events that have modified the lower reaches of the river.

Unaweep Canyon, a major bedrock gorge now hosting two divergent, underfit streams (Aslan et al., 2008), has been postulated as a possible early path for the Gunnison River as it flowed northwest downstream of the present Black Canyon (Hansen, 1965, 1967). The abandonment of Unaweep Canyon at ca. 1 Ma (Aslan et al., 2008) shows clear evidence of drainage reorganization, probably through stream piracy rerouting the Gunnison River through the Grand Valley. One possibility is that this piracy set off an incision pulse that is now expressed as the Black Canyon knickzone. Darling et al. (2009b) noted that such a transient knickzone would have passed rapidly upstream through the weak Mancos Shale, and then more slowly once it encountered Precambrian bedrock of the Gunnison uplift.

Recent tomographic images indicate complex patterns in mantle velocity structure below the Gunnison River (Fig. 8; Karlstrom et al., 2012). Of particular interest, the transition from the low-velocity mantle domain beneath the headwaters to higher velocities in the lower reaches of the Gunnison River suggests that buoyant mantle may be driving differential uplift in the Rocky Mountains, and may also be influencing the longitudinal profile of the Gunnison River (Schmandt and Humphreys, 2010). The hypothesis of young uplift in the headwater region of the Gunnison River is supported by apatite fission-track and apatite helium cooling ages, which show <10 Ma AHe cooling ages in the San Juan Mountains to the south (McKeon, 2009) and the West Elk and Elk Mountain Ranges to the north (Garcia, 2011) of Black Canyon. This suggests that more than 1 km of material has been eroded off the headwater regions of the Gunnison River in the last 10 Ma. Our hypothesis is that low-velocity mantle under these recently exhumed areas may have driven headwater uplift resulting in higher gradients for the Gunnison River, which then facilitated drainage reorganization and upstream knickpoint migration.

An important and often overlooked component of the uplift story involves flexural isostatic rebound driven by erosional denudation of the Rocky Mountain–Colorado Plateau region. Lazear et al. (2013) presented calculations of isostatically driven rebound applied to the Gunnison River region that suggest that independent of any tectonic surface uplift, there has been ∼200 m of differential isostatically driven rock uplift of the Gunnison knickpoint region relative to both the Colorado-Gunnison confluence and the headwater regions, including Grand Mesa.

Our data suggest a very young river system responding to multiple forcings, including downstream drainage reorganization, differential isostatic rock uplift, and possible mantle-driven headwater uplift. New data conclusions are as follows.

(1) Differential incision across a transient knickzone: Incision rates downstream from the modern knickpoint are higher (130–150 m/Ma over 0.64 Ma and 140 m/Ma over 10 Ma) in comparison to upstream rates (90–100 m/Ma over 0.64 Ma and ∼60 m/Ma over 10 Ma). Incision rates within the 10-km-long Painted Wall knickzone are strikingly higher at ∼500 m/Ma over 0.64 Ma (Fig. 3, Table 1). This variation in incision rate within a relatively small lateral distance (100 km) and within the hardest rocks in the region (Fig. 3) argues that this feature is the transient expression of a stream system adjusting to a downstream drainage reorganization and related regional base level fall and/or headwater uplift. Our model also indicates ∼30 km of upstream knickpoint migration through basement from 1.0 to 0.64 Ma (0.36 Ma). This documents the importance of bedrock erodibility on knickpoint propagation rates.

(2) Incision history of Black Canyon of the Gunnison: Projection of the abandoned paleo-Bostwick tributary, that deposited the 0.64 Ma Lava Creek B–associated river gravels, to its intersection with the Gunnison at Red Canyon is presently the best constraint on the age and rate of incision of Black Canyon. At this location, ∼350 m of the roughly 700 m total depth of Black Canyon has been incised in the last 0.64–0.87 Ma. We use the Cimarron gradient as a proxy for the paleo-Bostwick to extrapolate preserved and dated straths to their inferred confluence with the river at 350 m above the river. The result is that over the past 0.87–0.64 k.y., incision has occurred at the high average rate of 400–550 m/Ma. If one assumes steady incision rates, this implies the spectacular bedrock Black Canyon was carved within the last 1.3–1.75 Ma. However, given that these high rates are only likely to have been active in the Black Canyon reach once the incision transient from Unaweep abandonment arrived near Red Canyon after 1 Ma, rates may have been higher and carving of the entire bedrock canyon in the last ∼1 Ma seems most likely.

(3) New incision rates from tephrochronology and cosmogenic burial dating of gravels: Our new data suggest that the previously reported 1.2 Ma Mesa Falls tuff locality in the Shinn-Bostwick drainages is likely to be the 0.64 Ma Lava Creek B. The new age of the paleo-Bostwick gravels below the ash of 870 ± 220 ka precludes this ash being Mesa Falls. This reassignment is also supported by stratigraphic relations, indicating no major unconformity between the two exposed ashes. Cosmogenic dating indicates that terrace Qt7/223 m at the North Fork Gunnison–Gunnison River confluence is 0.98 (^+0.38/_–0.07) Ma, yielding an incision rate there of 228 (⁺¹⁷/_–64) m/Ma.

(3) Persistent differential incision in the early Gunnison River: Incision rates calculated from 10 Ma basalts show the same long-term differential incision as seen in the younger profiles and suggests that either (a) the knickpoint was already in existence at 10 Ma or (b) that post–3 Ma incision has created the differential incision rates (our favored hypothesis). The Oligocene profile is near the modern profile in the upper reaches of the Gunnison system, suggesting zero long-term incision—presumably due to a combination of aggradation and surface uplift following the construction of Oligocene-age volcanics. Erosionally driven isostatic rebound of the region may also have helped drive incision of the Black Canyon of the Gunnison.

This work was supported by the Colorado Rockies Experiment and Seismic Transects (CREST) experiment funded by the National Science Foundation (NSF) Continental Dynamics Program under award EAR-0607808. Thanks also to David C. Noe and anonymous reviewers for editorial comments. We also thank the Alfred P. Sloan Minority Ph.D. Program, and participants who were supported in 2006 by the Mesa State College Research Experience for Undergraduates (REU), which was supported by NSF REU Site Award EAR-0453264.