Hypotheses for the age of the western Grand Canyon (WGC) range from younger than 6 Ma to older than 70 Ma. We study the relationships among topography, geology, and available erosion rates in space and time to place constraints on plausible canyon incision histories. Evidence suggests that lateral retreat of the Shivwits Plateau escarpment left a lithologically controlled bench on the Sanup Plateau, but the Hualapai Plateau is beveled indiscriminately across rock types of the Paleozoic stratigraphic section. A period of accelerated base-level fall in the Tertiary is implicated by the canyon incised into the beveled Hualapai Plateau surface, consistent with higher erosion rates observed in canyons than on the surrounding plateau. Streams draining the Hualapai Plateau preserve relict headwater segments that were equilibrated with a slower base-level-fall rate before canyon incision. These relict segments are now separated from the Grand Canyon by knickpoints indicative of a transient landscape. Relief production since the beveling of the Hualapai Plateau is ∼1000 m in the WGC. Comparison of hillslope and channel morphologies between the Grand Wash Cliffs and the WGC provides a test to distinguish hypothesized ages of canyon incision. The data strongly suggest that carving of the WGC is younger than relief production due to slip on the Grand Wash fault ca. 18–12 Ma. Thus the geomorphic data are only fully consistent with the late Tertiary, transient incision model of canyon incision beginning at integration after 6 Ma.


The age of the Colorado River drainage system and timing of Grand Canyon incision have been debated since the pioneering studies of John Wesley Powell (1875). Although Powell interpreted the Colorado River as antecedent to Laramide structures and the uplift of the Colorado Plateau, since the turn of the twentieth century most geologists have interpreted a mid- to late Tertiary age of integration of the superimposed Colorado River and carving of the Grand Canyon (Blackwelder, 1934; Davis, 1901; Longwell, 1946; Lucchitta, 1972; Young and Brennan, 1974), and much of the evidence points to a ca. 6 Ma river integration event (House et al., 2008), contemporaneous with a significant fraction of canyon incision (Karlstrom et al., 2008, 2007; Pederson et al., 2002, 2006). Some evidence, however, has been suggested to support the carving of the western Grand Canyon (WGC) by 70 Ma (Wernicke, 2011; Flowers and Farley, 2012), or perhaps since ca. 17 Ma (Polyak et al., 2008; Young, 2008), reinvigorating the debate over estimates of the canyon age. The hypothesized 17 Ma precursor WGC cut by a moderate-sized pre–Colorado River drainage (Polyak et al., 2008; Young, 2008) is disputed (Karlstrom et al., 2008; Pearthree et al., 2008; Pederson et al., 2008; Karlstrom et al., 2014; Crossey et al., 2015). The idea of WGC formation by 70 Ma is even more controversial (Karlstrom et al., 2013; Lucchitta, 2013). Wernicke (2011) and Flowers and Farley (2012) argued on the basis of thermochronologic data that the WGC had been cut to within a few hundred meters of its current depth by 70 Ma and maintained at the surface (not buried) since then. The interpretation of the thermochronologic data is not simple, however, and has been challenged (Fox and Shuster, 2014; Karlstrom et al., 2014). In addition, Young and Crow (2014) reiterated, clarified, and augmented geologic and qualitative geomorphic evidence of the Tertiary cutting and infilling of smaller paleocanyons that is incompatible with the formation and maintenance of a deep canyon in the WGC area prior to ca. 19 Ma. Furthermore, the Muddy Creek Formation of the Grand Wash trough (GWT) has long been interpreted to prohibit the existence of a large river at the terminus of the WGC (Longwell, 1946; Lucchitta, 1966; Pederson, 2008) and the Muddy Creek constraint has been further strengthened by detrital zircon analysis of the rocks of Grand Wash trough at the mouth of Grand Canyon (Crossey et al., 2015; Ingersoll et al., 2013). Beyond sorting out the local geologic history, resolving this debate presents an opportunity for refining our understanding of both thermochronologic and geomorphic records of river incision and relief production. In this paper we analyze constraints on the timing of the incision of the WGC from landscape morphology and existing incision rate estimates.

Classic interpretation of a young WGC is consistent with the steep-walled, narrow, high-relief inner gorge morphology that is immediately apparent on topographic maps (Fig. 1) and in the field (Fig. 2). The range of circumstances that could produce and maintain such a landscape, however, has not been thoroughly explored. Moreover, given the broad attention this controversy has garnered, and given the apparent challenge of the “old canyon” evidence to this classical interpretation of a “youthful” landscape (e.g., Davis, 1901), an independent quantitative assessment of the geomorphic constraints on the antiquity of the WGC is needed. Three main hypotheses for the age of the WGC (older than 70 Ma, younger than 17 Ma, and largely younger than 6 Ma) are the focus of debate, and each must be consistent with the known paleocanyons preserved on the Hualapai Plateau (Young, 1979, 2008; Young and Crow, 2014). Each hypothesis constitutes a set of testable predictions for both landscape morphology and the spatiotemporal pattern of erosion. We analyze details of landscape morphology and available constraints on erosion rates to test the viability of each hypothesis on geomorphic grounds.

Landscape morphology alone is not diagnostic of the cause or mode of canyon formation. Narrow, steep-walled canyons incised into low-relief plateau surfaces can form in response to either (1) an acceleration in main-stem river incision (as implied in the 17 Ma and 6 Ma models) or (2) approximately steady incision into a subhorizontal stratigraphic succession with weak, easily eroded rocks or sediments overlying a notably stronger, erosionally resistant package of rocks (which could be consistent with the 70 Ma model). In the latter case, the low-relief erosion surface surrounding the canyon forms as weak rocks are eroded along the contact between weak and strong rocks (the surrounding bench postdates rather than predates canyon incision) and canyon formation need not reflect an increase in the rate of relative base-level fall (Fig. 3). Both scenarios involve relief production resulting from more rapid erosion within the canyon than on the surrounding low-relief bench; in one case the erosion rate within the canyon increases in response to base-level fall, in the other case the erosion rate in the surrounding landscape decreases in response to the formation of a lithologically controlled erosional bench. Consequently, the cause and timing of canyon formation must be considered separately. It is fortunate that, for either mode of canyon formation (accelerated incision versus incision into stronger rocks), there are clear differences in the expected evolution of both canyon side walls and tributary valley profiles that will be diagnostic of whether incision of the main-stem valley was early (e.g. 70 Ma) and persisted to recent times, or whether the landscape has undergone a significant period of reduced rate of base-level fall, or even stability, after canyon formation (as required by the 70 Ma hypothesis).

Approach and Scope

We do not address the incision history of all of the Grand Canyon, but rather focus on the WGC west of the Hurricane fault (Fig. 1). In our study area we determine (1) if the WGC is more consistent with an increase in incision rate or simply a lithologic control on landscape morphology; (2) how much relief production is recorded by landscape morphology; and (3) the range of plausible incision ages. The first two goals encompass determining the cause of canyon formation and the amount of relief production involved. Both are accomplished through an analysis of two-dimensional landform morphology (topographic cross sections, analysis of modern tributary river profiles, and reconstruction of pre-incision river profiles) and a qualitative evaluation of the correlation between rock strength and landscape morphology. As shown here, details of tributary canyon morphology can be used to differentiate between models allowing sustained river incision to present (e.g., the 17 Ma and 6 Ma models) and models requiring negligible incision following initial canyon cutting (the 70 Ma model). Constraining the age of the canyon is independently accomplished through (1) a quantitative comparison between canyon walls in the WGC and the escarpment along the Grand Wash fault (GWF; most active 18–12 Ma), and (2) a compilation of incision rates over a range of time scales in tributaries and on the surrounding plateau. Given the common climate, lithology, and base level, our comparative morphological analysis can readily gauge the relative timing of incision of the WGC and its tributaries and relief production related to slip on the GWF.

Background and Methods of River Profile Analysis

We employ river profile analysis and evaluate the relations between landscape morphology and geology to achieve our goals. United States Geological Survey (USGS) 30 m resolution digital elevation models (DEMs) are used for all topographic analyses. Pioneered and advanced by Hack (1957, 1975), analysis of river profiles can provide significant insight into the history of relative base-level fall in a region. As the methods and foundational conceptual background are not as well known as more standard analyses of river incision rates from dated terraces, mean catchment hillslope gradients, or topographic relief, we include here a brief primer tailored to aspects of river profile evolution that have diagnostic power. (For more detailed reviews of river profile evolution patterns and methods of profile analysis, see Wobus et al., 2006; Kirby and Whipple, 2012; Whittaker, 2012; Lague, 2014.)

River channels tend toward smooth concave-up river profiles that are well described by Flint’s Law (Flint, 1974; Hack, 1957):
where S is local channel gradient, A is upstream area (a proxy for water and sediment discharge), ks is the channel steepness index, and θ is the concavity index (usually between 0.4 and 0.6) (Whipple and Tucker, 2002). The most common deviations from this expected equilibrium form are either (1) channels broken into segments marked by slope-break knickpoints that separate reaches of distinct steepness but similar concavity, or (2) channels with smooth profiles but generally high concavity (θ > 0.6). Segmented channel profiles marked by slope-break knickpoints are expected to form in response to temporal changes in river incision rate or spatial patterns in rock uplift rate, substrate properties, or climatic conditions (e.g., Kirby and Whipple, 2012; Whittaker, 2012; Lague, 2014). Because the concavity index of discrete segments varies little and channel steepness and concavity determined from river profiles are strongly correlated, a normalized channel steepness index, ksn, is used to quantify the relative steepness of channels (e.g., Wobus et al., 2006):
where θref is a reference concavity often assumed to be ∼0.5 (here we use 0.45 for computing regional maps of ksn patterns, in keeping with most published estimates of ksn). Typically, as here, the assigned θref represents local determinations of the concavity of well-adjusted channel segments. High values of ksn are associated with rapid incision rates, hard rocks, low erosivity climates (few large floods), and coarse bed material (e.g., Kirby and Whipple, 2012). Convex-up knickpoints are thus associated with either a temporal increase in the rate of relative base-level fall or a downstream increase in rock strength, as might be caused by incision into subhorizontal stratigraphy with higher average strength in lower rock layers. Concave-up knickpoints, conversely, are associated with either a decrease in the rate of relative base-level fall or a downstream decrease in rock strength. Slope-break knickpoints associated with temporal changes in the rate of base-level fall are expected to occur at nearly constant elevations in a catchment subjected to uniform relative base-level fall (Niemann et al., 2001; Wobus et al., 2006). Conversely, knickpoints associated with rock strength controls are expected to coincide with lithologic contacts.
We include a complementary analysis useful for visualizing perturbations to the long profile known as the integral method (Fig. 4C; Harkins et al., 2007; Perron and Royden, 2013). The integration of equation two yields:
where χ(x) is determined directly from numerical integration of drainage area data. Any segments along the channel profile that are well described by θref will be straight lines on a plot of elevation versus χ (a chi plot; Fig. 4C), with slope equal to ksn.

Figure 5 illustrates how canyon side walls and tributary channel profiles can be expected to evolve as a function of different base-level histories following the onset of canyon incision, assuming for simplicity no strong lithologic control on channel steepness below the canyon rim (the canyon rim may be either lithologically controlled or reflect an acceleration in main-stem incision rate in either case [Fig. 3]; canyon morphology is not diagnostic of the underlying cause of canyon formation): (1) ongoing, roughly steady incision (plausibly consistent with both 17 Ma and 6 Ma models), and (2) early incision followed by a substantial reduction or even cessation of incision (consistent with the 70 Ma model). Ongoing incision (scenario 1) maintains steep hillslopes and channels, with morphology set by incision rate, climate, and rock strength. Lithologic effects could be overprinted on this pattern, potentially inducing segmented profiles below the canyon rim with higher ksn through stronger rock layers. Substantial reduction or cessation of incision (scenario 2) triggers a progressive relaxation of initially steep canyon walls and tributaries even as slope-break knickpoints defining the canyon rim continue to migrate upstream. A sustained period of postincision base-level stability (or significantly reduced base-level-fall rate) is expected to manifest in tributary channel profiles as a distinct flattening of downstream reaches (below concave-up knickpoints) that can be anticipated to be most pronounced in larger tributaries (see Gasparini and Whipple, 2014), to be mimicked in the morphology of surrounding hillslopes, and to not coincide with lithologic layering. Thus scenarios involving recently active, quasi-steady incision can be readily distinguished from those with base-level stability over 106–108 yr time scales after canyon formation (e.g., the 70 Ma model), even in the presence of variable rock strength.

Figure 5 also illustrates how reconstruction of pre-incision river profiles (profiles prior to either main-stem incision into harder rocks or at an accelerated rate) can be used to quantify the amount of relief production, and, if erosional lowering of the surrounding landscape can be quantified, the total amount of main-stem river incision during the period of relief production. The method involved was used successfully by Schoenbohm et al. (2004), Harkins et al. (2007), and DiBiase et al. (2015), and is illustrated in Figure 4 for the example of Jeff Canyon on the south rim of the WGC (location in Fig. 1). In our analysis, profile reconstruction and projection differ by an estimate of erosional lowering of the channel profile upstream of the slope-break knickpoint over the period of interest (Fig. 4). Projection begins with locating the major slope-break knickpoints on the long profile. Upstream of many knickpoints, subtly oversteepened zones truncate well-graded upstream stretches and are readily recognized on profiles and slope-area diagrams (e.g., Fig. 4). Such oversteepened reaches are thought to result from stream flow acceleration and increased erosion rate that propagates upstream, sometimes tens of kilometers, over time (e.g., Haviv et al., 2006) and are not representative of pre-incision river profiles. The upstream limit of these oversteepened reaches is selected as the anchor point for profile projections. The average channel steepness and concavity of well-graded sections upstream of the anchor point are found by regression following the methods outlined in Wobus et al. (2006) and Kirby and Whipple (2012) (see Table 1). Slope-area data are smoothed using a 250 m window along the profile and a 40 ft (12.192 m) contour interval consistent with USGS elevation source data. Assuming that drainage area and channel network geometry have remained invariant over the time period of interest, and that upstream reaches are reflective of channel form prior to canyon formation, the best-fit channel steepness and concavity upstream of anchor points are used in Equation 1 to model the existing upstream profile and project the downstream profile as a function of drainage area. The projection from modern stream data is used as a baseline to which incision estimates can be added to allow reconstruction of the paleostream profile (Fig. 4).

Uncertainty in the profile reconstruction includes restoration of the profile based on incision rate estimates and the projection uncertainty. Projection uncertainties are: (1) any unrecognized modification of channel slope above the knickpoint such that the upper channel segment is not reflective of pre-incision conditions, (2) possible changes in drainage area or network geometry, (3) uncertainty in the channel concavity, and (4) uncertainty in the channel steepness index. Error sources 1 and 2 cannot be accurately quantified and may be responsible for outliers. Error sources 3 and 4 are constrained by regression analyses and used to represent the range of uncertainties in profile reconstructions (Fig. 4). Anchor points are at the downstream end of the slope-area data used in the regression. Projection uncertainty is quantified by fixing θref close to the regional mean and using the associated channel steepness uncertainty. Particular streams require slightly different concavity index (0.4, 0.45, or 0.5; Table 2), determined by comparing the projection and actual profile upstream of the anchor point. If the maximum and minimum projections are not below and above the headwater, respectively, a different θref may be used within the uncertainty of the regional mean and ksn (see Table 2 for θref and ksn of each projection). The projection is calculated from the mean and 2σ standard deviation (95% confidence interval) of the channel steepness index, ksn. The projection and reconstruction terminate at the confluence with the Colorado River. The difference in elevation between the Colorado River at this point and the projected elevation is reported as the height of the projection, an estimate of relief production. The reconstructed height (as opposed to the projected height) above the river is the incision of the river. The incision magnitude is the projected height plus the restored incision of the upper tributary segment since initiation of canyon cutting (Fig. 4). Without the reconstruction, incision estimates from the projection are minimum estimates of erosion. We use pre-dam Colorado River surface elevations (USGS 1:24,000 topographic scale maps) when calculating heights for tributaries that enter the Lake Mead reservoir.

The WGC is a deep, narrow, steep-walled gorge inset into a broad low-relief plateau with tributaries that head on the low-relief surface but are deeply incised below abrupt slope-break knickpoints near the main canyon (Figs. 1 and 3–6). This morphology implies 3 possibilities: (1) recent acceleration of incision, (2) reentrenchment of a previously buried, ancient canyon, or (3) stripping of weak rocks to form a low-relief surface cut on a hard, lithologically controlled bench surrounding the canyon. The first two scenarios could produce identical morphologies and would have to be distinguished based on geologic evidence for base-level rise and regional aggradation. Because neither the geology nor the thermochronology supports reentrenchment of a previously buried canyon for the WGC, we do not consider this scenario further except in discussion of the partially reexhumed Laramide-age paleocanyons on the eastern Hualapai Plateau (Elston and Young, 1991; Young and Brennan, 1974; Young and Crow, 2014; Young and McKee, 1978). The third scenario has been discussed in terms of the evolution of Grand Canyon and river knickpoints (Cook et al., 2009; Pederson and Tressler, 2012), and can produce landforms identical to those in the other scenarios, although cases with steadily falling versus stable base-level conditions following initial canyon cutting would have different expressions in tributary channel profiles and spatial incision rate patterns (Figs. 3 and 5).

Hillslope profiles in the Grand Canyon and surrounding Colorado Plateau are replete with classic examples of lithologic control on landform morphology with alternating strong cliff-forming and weak slope-forming rock layers (Selby, 1993). The topographic expression of different rock layers is a great aid in geologic mapping in the region. These familiar hillslope expressions of variable rock strength have been recently corroborated with direct laboratory measurement of rock tensile and compressive strength and associated with variations in river slope and channel width (Pederson and Tressler, 2012). It is ironic that the simple geology of only moderately deformed subhorizontal sedimentary rocks greatly complicates the geomorphology, making it difficult in much of the Grand Canyon to differentiate between lithologic and base-level controls on canyon morphology. In the WGC the Sanup Plateau north of the Colorado appears to be continuous with the Esplanade surface because it generally coincides with the top of the resistant Esplanade Sandstone (Fig. 7). The Sanup Plateau appears to be a lithologically controlled bench associated with the erosional retreat of the Shivwits escarpment along the contact between the easily eroded Hermit Formation shale and the resistant Esplanade Sandstone (Fig. 8; Lucchitta, 1966; Young and Crow, 2014). This configuration and the prevalence of cliff-forming limestones of the Redwall, Temple Butte, and Muav Limestones and Proterozoic basement exposed in canyon walls are suggestive of, but not necessarily diagnostic of, lithologic control of the narrow, steep-walled WGC. It is plausible that the morphology of the WGC records an accelerated river incision rate that is coincidental with incision below the level of the Esplanade Sandstone on the Sanup Plateau. Fortunately, Laramide-age deformation in the area of the WGC created a setting that allows us to resolve this.

Inspection of the geology, geomorphology, and Cenozoic stratigraphy of the Hualapai Plateau on the south side of the Colorado River (Figs. 1 and 7) reveals that it is not a lithologically controlled bench, but rather a product of a long period of erosion under a relatively stable base level. The critical observation is that the Hualapai Plateau is beveled indiscriminately across rock types ranging from Supai Group shales to Muav Limestone, consistent with arguments by Young and Brennan (1974), Young and McKee (1978), and Young and Crow (2014) that this is an ancient and long-lived low-relief surface (Figs. 7 and 8). It is important that the Hualapai Plateau is a low-relief erosional surface beveled to the same elevation as the Sanup Plateau; the two plateaus are part of a common surface incised by the Colorado River and its tributaries, suggesting a common paleo–base-level control overprinted by lithologically controlled retreat of the Shivwits Plateau escarpment to the north. Thus, we infer that the morphology of the WGC reflects a period of accelerated incision into a preexisting low-relief landscape (outlined in more detail in the following).

Tributary channel profiles have all the hallmarks of a disequilibrium landscape, i.e., deeply incised, steep-walled canyons in the vicinity of the main stem that are marked by significant slope-break knickpoints where they cross the canyon rim (Figs. 1 and 5–7). Either an acceleration of incision rate or a lithologically induced pattern of differential erosion could be the driver behind the transient evolution of landscape morphology implied by this topography (Fig. 3). In Figure 6, downstream of major knickpoints some tributaries show strong lithologic control of channel steepness, while others do not. Smaller tributaries and those lacking a significant source of gravel in headwater reaches appear to be detachment limited (e.g., Johnson et al., 2009) and have segmented channel profiles with ksn > 400 m0.9 through the Paleozoic limestones (Redwall, Temple Butte, and Muav Limestones), moderate ksn (∼150–250 m0.9) through the Bright Angel Shale, and steepen somewhat where cutting through basement rocks. Larger tributaries and those with significant headwater gravel sources (such as most of the streams draining the Shivwits and Sanup Plateaus) appear to be transport limited, showing no lithologic control on channel profiles (e.g., Johnson et al., 2009) that have the same steepness as segments in the Bright Angel Shale on both sides of the canyon (Fig. 6). However, these remote observations require significant field testing. Additional complexities reflect young travertine dams in some tributaries (e.g., near the outlet of Jeff and Quartermaster Canyons, and at 800 m elevation in Meriwhitica canyon; Figs. 1, 4, and 6). Variable elevation ranges of the prominent high ksn segments well expressed in Figure 6 are inconsistent with temporal changes in the rate of base-level fall following the initiation of canyon cutting (Fig. 5) (e.g., Wobus et al., 2006; Kirby and Whipple, 2012) and coincide with outcrop patterns of the sequence of Paleozoic limestones. However, the lack of similar lithologic control on profiles of channels draining the Shivwits and Sanup Plateaus suggests that canyon formation and the prominent slope-break knickpoints that define the canyon rim cannot be attributed to lithology. The surrounding low-relief surface (Hualapai and Sanup Plateaus) extends from the so-called Esplanade surface (the top of this sandstone) to cut across the Muav, Temple Butte, and Redwall Limestones, and the shales and sandstones of the lower Supai Group and Esplanade Sandstone (Figs. 7 and 8). Far from forming lithologically controlled benches, the resistant Paleozoic limestones are beveled to the same level as the much weaker units of the lower Supai Group. Long-term stability of erosional base level at the elevation of the Hualapai Plateau (∼1400 m elevation at present) is required for such effective erosional beveling of erosionally resistant cliff-forming limestones (e.g., Baldwin et al., 2003). These morphological observations are consistent with interpretation of landscape evolution based on the nature, distribution, provenance, and paleoflow directions of Tertiary sediments preserved on the Hualapai Plateau (e.g., Elston and Young, 1991; Young and Brennan, 1974; Young and Crow, 2014; Young and McKee, 1978). The morphological observations summarized here, however, speak primarily to the cause of canyon formation (accelerated incision dominating over lithologic controls), not to its timing. However, concave-up knickpoints expected to record any extended period of post-incision base-level stability are not seen in the WGC landscape.

The amount of river incision is the relative change in elevation of the main-stem river. Relief production associated with canyon cutting differs from total river incision; it is the difference between river incision above a knickpoint and river incision below that knickpoint (Fig. 3). Relief production can be measured as the difference in elevation of a projected stream and the modern stream (the projected stream height; Table 2). A simple measure of the vertical drop from canyon rim to river level will overestimate relief production because the rim is higher in elevation than the upper streams incised into the low-relief plateau. Determining the paleoelevation of the main-stem river involves two steps. First, relief production is measured from the elevation projected from relict upper channel segments above oversteepened zones down to the tributary confluence with the main-stem river (Fig. 5). Second, estimates of the lowering rate of the surrounding low-relief landscape can be multiplied by proposed or dated surface age to restore the lowering of relict channels over the time period of interest (Figs. 4 and 5). The sum of these provides an estimate of total river incision, such as might be recorded by thermochronometers.

Figure 9 summarizes the results of the profile projections, i.e., relief production. We analyzed 38 streams (slope area data available in Supplemental File 11) to determine if projections to paleo–base level could be made. We attempt profile projections only where we have grounds to believe the headwater reach of the profile reflects pre-incision channel morphology plausibly graded to paleo–base level. Thus channels that clearly express lithologic controls above the slope-break knickpoints that define the canyon rim are excluded. Conversely, streams that flow on the Tertiary angular unconformity and/or Tertiary sedimentary deposits are ideal (Figs. 7 and 8). Moreover, if these headwater stream segments preserve relict conditions reflecting a stable base level for an extended period they should all have similar slope-area relationships. We identify 12 streams most suitable for analysis. In Table 1, these 12 representative streams (red circles in Fig. 1) have a mean concavity of 0.44 ± 0.02 (1 standard error) and mean channel steepness of 27.5 ± 9.6 (1 standard error), a narrow range of channel profile morphology. Our analysis of surrounding streams yielded 1 stream, informally called Bat Cave, that matches slope-area metrics and may represent the same paleosurface but is not associated with independent geologic evidence of a stable base level. These 13 projectable streams yielded an estimate of relief production in the WGC of 1050 ± 30 m (1 standard error) (Fig. 9). As expected, this is a minimum relative to the simple estimate of the elevation difference between the Colorado River (∼350 m) and the lip of the Hualapai Plateau (>1400 m).

The stream named 234L is located near the remnants of the Tertiary paleocanyons and projects lower than the other streams. Our criteria eliminate the rest of the streams in the Peach Springs area for projection due to a lack of reliable stretches of slope-area data, but the knickpoint elevations and qualitative interpretation of these river profiles suggest that all the streams in this area are consistent with the relatively low projection of 234L (100–200 m lower than the rest of the Hualapai Plateau streams; Fig. 6).

Of the 26 other streams (Table 2), headwater reaches of 10 were too short to allow reliable projection. The remaining 16 either showed evidence of lithologic complications (13 of 16) or were associated with drainages that traverse the modern GWT, a now subareal region that was paleo–Lake Hualapai until ca. 6 Ma (Faulds et al., 2001; Lucchitta, 1979), and given the certainty of drainage network disruptions could not be projected with confidence.

The lithologic complications partly reflect the retreating Shivwits Plateau escarpment (Fig. 8), which significantly complicates interpretation of longitudinal profiles of drainages on the north side of the Colorado River. The hard (Kaibab and Toroweap Formations) over soft (Hermit Formation) lithology of the Shivwits escarpment implies that pre-incision channel profiles likely had lithologically controlled knickpoints perched above the base level of the Sanup and Hualapai Plateaus. Projection of the 13 streams thought to represent the pre-incision topography is contingent on the assumption that the Shivwits escarpment was already north of these streams when their headwater reaches equilibrated to their current form. The Tertiary sediments and beveled unconformity on the Hualapai Plateau support this assumption directly in the immediate area of most of these streams. However, those streams that now exist in the area between the modern Shivwits escarpment and the Tertiary sediments on the south rim are likely to have evolved from steep reaches draining off the retreating Shivwits escarpment. Such streams would have complex histories not conducive to simple profile projection and reconstruction.

In summary, although only 13 streams could be reliably projected and all but one rise on the Hualapai Plateau, this drainage-projection analysis makes it clear that carving of the WGC into the preserved older, low-relief landscape of the Hualapai and Sanup Plateaus involved ∼1050 m of relief production since incision rate increased during formation of the WGC. For comparison, we infer >700 m of incision from the base of the Tertiary sediments (∼1120 m elevation) exposed in the axes of Tertiary paleocanyons in the Peach Springs area. Considering the ∼400 m thickness of the Tertiary sediments (Young and Crow, 2014) gives ∼1100 m relief production, compared to our 1050 m for relief production. Previously noted Peach Springs area channels (234L) are modestly inset into the paleosurface (see Fig. 6 and Supplemental File 22), suggesting that fluvial erosion in the vicinity of the paleocanyons took advantage of topographic lows or possibly slightly higher erosion rates through the softer rock.

Flowers and Farley (2012) and Wernicke (2011) interpreted low-temperature thermochronometric data from river-level samples as requiring 70% to 80% of canyon incision since 70 Ma. Although no more than 70 m of incision is recorded directly by the interaction of the river and basaltic lava flows for the past 0.625 m.y. (Karlstrom et al., 2007) and <300 m is inferred from groundwater-table elevation inferred from speleothems since 3.87 Ma (Polyak et al., 2008), our result of ∼1050 m of relief production in the late Tertiary suggests that the understanding and interpretation of thermochronometric data (Flowers and Farley, 2012) is incomplete, as suggested by Fox and Shuster (2014) on the basis of models of thermal evolution and kinetics of damage annealing and helium diffusion.

Plateau Lowering Rate

We derive estimates of the lowering rate of the plateau from volcanic deposits (Wenrich et al., 1995; Young and Brennan, 1974) with incision amounts estimated as the relief between tops of hills capped by dated basalts to closest channel bottoms. Height measured from the tops of flows to channel bottoms results in rates that are local maxima. Ages range from 19 to 15 Ma, yielding a lowering rate on the low-relief surface of 2–11 m/m.y., with a mean of ∼6 m/m.y. (Table 3). The amount that incision estimates differ from relief production requires an estimate of the time of onset of incision (estimates in Table 3).

Incision Magnitude versus Relief Production

The three competing hypotheses for the incision of the WGC suggest incision starting ca. 6 Ma or ca. 17 Ma, or virtually complete by 70 Ma. A succession of sedimentary deposits from early Tertiary until ca. 19 Ma records net moderate aggradation (as much as ∼400 m in infilled paleocanyons) on the Hualapai Plateau over that interval (Elston and Young, 1991; Young, 1979, 2008; Young and Brennan, 1974; Young and Crow, 2014; Young and McKee, 1978; Young and Spamer, 2001). The plateau lowering rate estimate (∼6 m/m.y.) is a measure of the average rate of lowering since the ca. 19 Ma cessation of aggradation. It is important to acknowledge that slow headwaters erosion is not zero, so total incision during canyon cutting will exceed the relief production estimates presented here. In addition, the background erosion rate of the plateau surface is the only direct constraint on the pre-canyon erosion rates in the region. At 6 m/m.y., total Colorado River incision would be 36 m greater than relief production if canyon incision began at 6 Ma, and 100 m greater if canyon incision began at 17 Ma. The latter estimate is a maximum given the preserved thickness (typically <100 m) of aggradation that remains above low-relief modern stream segments (i.e., above projected streams). Thus our estimate of total incision since the onset of canyon formation ranges between 1050 and 1150 m, based on available measurements.

Here we develop geomorphologic constraints on the time scale over which incision of the Colorado River and its tributaries into the low-relief surface defined by the Hualapai and Sanup Plateaus in the WGC occurred. We take two approaches to this analysis. The first is based on a comparative analysis of landscape morphology in the WGC and nearby GWT. The second is based on measurements and estimates of erosion rates over a range of time scales within and around the WGC.

Morphological Analysis

Details of tributary river profile forms can be diagnostic, at least in general terms, of the history of base-level fall. Similar examples of this type of analysis were discussed in, for example, Whittaker (2012), Kirby and Whipple (2012), and Gallen et al. (2013). Comparison of all tributaries to the WGC (Figs. 4–7 and 9; Supplemental File 1 [see footnote 1]), with expectation from theory, strongly suggest that there is no indication of the morphology expected to result from a stable base level maintained for millions to tens of millions of years after canyon incision (see Fig. 5B). Tributary profiles are rather suggestive of relatively steady incision over the duration of canyon formation and persisting to present (Fig. 5), with variable degrees of lithologic control on below-rim channel steepness (Fig. 6), as recorded by lava flow remnants (Crow et al., 2008) and speleothems (Polyak et al., 2008).

The interpretation of the WGC tributary profile morphology is powerfully enforced by a morphological comparison of the WGC landscape with the landscape of the Grand Wash Cliffs (GWC). The history of the GWF provides a well-known timing of relief production in a setting where both climate and rock strength are nearly identical to the WGC. Slip on the GWF is well documented to have begun ca. 18 Ma, with most of its 5.5 km of offset accumulated by 12 Ma (Bohannon, 1984; Bohannon et al., 1993; Karlstrom et al., 2010; Lucchitta, 1966, 1972, 1979; Quigley et al., 2010; Umhoefer et al., 2010). Ensuing slip on the GWF was dramatically less, ∼300 m, from 12 to 6 Ma. It is possible there has been limited post–6 Ma slip on the GWF. While the history of the GWF is not fully known and is complicated, relief production is no older than 18 Ma and both hillslope and channel morphologies may reflect rejuvenation by younger episodes of slip and base-level fall associated with the ca. 6 Ma breaching of Lake Hualapai (Faulds et al., 2001; Lucchitta, 1979) and the subsequent incision through the Hualapai Limestone and associated basin fill sediments.

We use topographic metrics to determine directly whether incision in the WGC is older, younger, or similar in age to incision of the tributaries of the GWT in response to faulting on the GWF and base-level fall associated with post ca. 6 Ma incision. Tributaries to the GWT exhibit more subtle knickpoints at the edge of the Sanup Plateau, suggesting a period of slow incision or stable base level after initial relief production, but have nearly identical channel steepness and concavity to north-side tributaries in the WGC (Figs. 6 and 7). The 150–250 m0.9 mean steepness index (Fig. 10) is identical to channel segments not exhibiting lithologic control and strong oversteepening where cut on resistant Paleozoic limestones or basement rocks, suggesting a very similar timing and rate of base-level fall, but under transport-limited conditions (e.g., Johnson et al., 2009). Like channel steepness, hillslope gradients are a function of erosion rate, rock strength, and climate. Figure 11 shows a composite hillshade and slope map of the GWT and the WGC. In the WGC, hillslopes are clearly steep from the rim of the Hualapai and Sanup Plateaus down to the Colorado River, particularly where Paleozoic limestones and basement rocks crop out (Fig. 7). Canyon walls have not retreated laterally from rivers edge. However, the escarpment of the GWC has in places retreated 2–3 km from the trace of the GWF and is less steep and more subdued (Fig. 12). Average slopes of interfluvial ridges near 270L, Quartermaster, and Horse Flat canyons are 0.76, 0.44, and 0.49, respectively. Similar ridges along Pearce, Pigeon, and Squaw canyons of the GWT, show average hillslope gradients of 0.14, 0.21, and 0.17, respectively. The topographic comparisons of the WGC to the GWC show that the two regions have had significantly different base-level-fall histories. Minimal fault activity over 12 m.y. has resulted in several kilometers of retreat of the GWC and considerable relaxation of hillslope gradients. Morphologic analysis strongly suggests that relief production in the WGC is younger than along the GWC, where the bulk of relief production occurred from 18 to 12 Ma. The hypothesis that the narrow, steep-walled WGC is ≥70 m.y. old is untenable. However, the controversy over the timing of formation of the WGC is over a difference of ∼600 m of incision. Geologic and geomorphic constraints demand ∼1000 m of post–19 Ma incision, while initial thermochronometric interpretations allow for no more than ∼400 m of post–70 Ma incision (Flowers and Farley, 2012).

Erosion Rate Constraints

The hypotheses about the timing of WGC incision require different rates of tributary incision and cliff retreat in order to produce and maintain the observed landscape morphology for the duration of the different hypothesized incision histories. We use several methods to determine average incision rates on the Colorado River in the WGC. Separation Hill basalt (1500 m above river and 19 Ma, Fig. 1; Wenrich et al., 1995) yields an average incision rate of 60 m/m.y. over the past 19 m.y. Quaternary basalt flows along the Colorado River overlie river gravel yielding similar rates of ∼70 m/m.y. (Karlstrom et al., 2008) over the past 0.625 m.y. Recently determined cosmogenic erosion rates in the WGC yield rates of 61 ± 18 m/m.y. (Nichols et al., 2011) averaged over the past ∼10 k.y. Given the broad range of time scales of these data, it is possible that there are significant rate fluctuations between data points, but the available evidence implies that the incision rate since ca. 0.6 Ma is roughly equal to the average incision rate since 19 Ma.

The “old canyon” model suggests 70%–80% of Colorado River incision in 70 m.y. (Flowers and Farley, 2012). Using our estimate of incision magnitude of 1100 m, the 70 Ma model suggests at most 400 m of incision since 70 Ma. Basalt flows in the WGC are as much as ∼70 m above the Colorado River, so the remaining 330 m of allowable incision in the model imply an average incision rate of 4 m/m.y. prior to the time of WGC basaltic volcanism in the gorge to be consistent with the 70 Ma hypothesis. This rate is comparable with erosion on the flat plateau, but all independent measures of canyon incision rate are consistently and substantially higher than rates proposed in the hypothesis that a canyon of nearly modern depth existed by 70 Ma.

The 17 Ma canyon hypothesis implies that average incision has occurred at ∼65 m/m.y. (∼1000 m/17 m.y.), implying incision slightly slower on average than estimates in the past 0.6 m.y. The 6 Ma canyon hypothesis, however, implies that ∼1000 m of incision has occurred at an average rate of 167 m/m.y. (∼200 with correct significant figures) over the past 6 m.y. Although this appears to be a mismatch with independent constraints (faster than both the average since ca. 17 Ma and rates recorded in the past 0.6 and 0.01 m.y.), the difference is consistent with expected evolution of the WGC following initiation of incision. Integration of the Colorado River across the GWC would trigger a period of rapid incision that would decrease with time as a knickpoint swept upstream (Pederson et al., 2002; Cook et al., 2009; Pelletier et al., 2009). Available constraints on incision rate history allows a few million years over which incision may have been faster than ∼160 m/m.y. and then slowed to ∼70 m/m.y. Alternatively, this difference in rates could indicate that part of the incision of the WGC occurred between ca. 17 and ca. 6 Ma in response to base level fall associated with slip on the GWF (Polyak et al., 2008; Young, 2008; Young and Crow, 2014).

The primary observation that the Hualapai Plateau is a low-relief surface cut across numerous rock-types is a strong indication that it reflects erosional beveling during a long period of base-level stability. Paleo-base level of the region until at least 17 Ma was therefore directly associated with this preserved landscape. The fluvial geomorphic data from headwater portions of Hualapai Plateau streams is interpreted to be consistent with preservation of a relatively stable base level from 70 to at least 17 Ma prior to incision of Grand Canyon. Channel projections primarily off the Hualapai Plateau yield estimates of relief production magnitude near 1000 m throughout the WGC.

We have tested three hypotheses for the age of western Grand Canyon against geomorphic constraints. Our data are most consistent with the 6 Ma model for the timing of canyon cutting. The 70 Ma model requires erosion at an improbably low ∼4 m/m.y. for tens of millions of years while supporting significant amounts of relief. The 17 Ma model is plausible based on existing incision rates and is compatible with Tertiary deposits on the Hualapai Plateau. However, landscape morphology indicates that the erosion of the walls of the WGC started considerably more recently than the erosion of the GWC 18–12 m.y. ago. Therefore, of the three hypotheses tested, data and analyses reported here only fully support western Grand Canyon incision in the last 6 m.y.

We thank Joel Pederson, Karl W. Wegmann, and one anonymous reviewer for constructive reviews that greatly improved our analysis, and the editor and guest editor for their efforts. We would also like to thank Rich Rudow for field assistance, photography, and digital enhancement of his photographs published here. And, we would like to thank the National Science Foundation for funding this research (EAR-1324721), and the National Park Service and the University of New Mexico advanced field camp for providing access to the field area and sparking discussions that motivated us to complete and publish the analysis presented here.

1Supplemental File 1. All stream slope-area data, plotted and compiled for inspection. Please visit http://dx.doi.org/10.1130/GES01131.S1 or the full-text article on www.gsapubs.org to view Supplemental File 1.
2Supplemental File 2. All stream elevation-chi data, plotted and compiled for inspection. Please visit http://dx.doi.org/10.1130/GES01131.S2 or the full-text article on www.gsapubs.org to view Supplemental File 2.