Rejection of the lake spillover model for initial incision of the Grand Canyon, and discussion of alternatives

The reason the Grand Canyon was cut by the Colorado River is disputed (Lucchitta, 1984, 1989, 1990; Young and Spamer, 2001; Ranney, 2005; Flowers et al., 2008; Pelletier, 2010; Wernicke, 2011; Beard et al., 2011; Douglass, 2011; Karlstrom et al., 2011). This paper evaluates the lake spillover model, which holds that (1) the Miocene Bidahochi Formation of northeastern Arizona contains lakebeds that are a vestigial record of a once-deep Hopi Lake, filled with waters from the Colorado River in Utah, that ponded east of the Kaibab-Coconino Plateau until it spilled over the crest of the plateau dam; and (2) lake spillover near the Miocene-Pliocene time boundary initiated incision of the Grand Canyon below the lake outlet. Bidahochi paleogeography does not support the lake spillover model without ancillary paleotopographic hypotheses that are difficult to sustain (Dickinson, 2011). Consideration of the upstream morphology of the Colorado River drainage system (Fig. 1) suggests constraints for alternate hypotheses.

Blackwelder (1934) rejected pre-Laramide antecedence for the course of the Colorado River through Laramide uplifts like the Kaibab uplift transected by the Grand Canyon. He suggested instead that the “haphazard” course of the river through multiple Laramide uplifts originated from successive spillover of waters from lake basins on a semiarid Colorado Plateau of low relief, before incision of the Grand Canyon promoted dissection of the plateau surface upstream by headward erosion. Scarborough (1989) concluded that lakebeds within the Miocene Bidahochi Formation were deposited within Hopi Lake (Williams, 1936), which in his view covered ∼30 × 10³ km² in northeastern Arizona east of the Kaibab-Coconino Plateau transected by the modern Grand Canyon. Scarborough (1989, p. 522) initially envisioned that Hopi Lake was drained when headward erosion upstream through the Grand Canyon breached the Kaibab-Coconino barrier that dammed the lake basin on the west, but later suggested that the lake level might have overtopped the Kaibab-Coconino barrier near a locale where previous Laramide erosion had cut a drainage notch through the confining uplands (Scarborough, 2001, p. 212). Meek and Douglass (2001) took the postulate of lake spillover a logical step further by inferring that incision of the Grand Canyon was triggered when Hopi Lake spilled across the elevated Kaibab-Coconino tract without benefit of a pre-existing paleocanyon as a guide for water flow.

Studies of sedimentation along the course of the lower Colorado River downstream from the Grand Canyon (Fig. 1) have documented that the Colorado River did not flow through the Grand Canyon until near the Miocene-Pliocene time boundary (Spencer and Patchett, 1997; Faulds et al., 2001a; Spencer et al., 2001; Patchett and Spencer, 2001; House et al., 2005; Roskowski et al., 2010). The inception of water flow through the canyon initially formed a chain of downstream lakes within which the lacustrine Bouse Formation was deposited along the modern river course (Fig. 1). The best current estimate for the entry of Colorado River water into desert basins below the Grand Canyon is ca. 4.9 Ma (Early Pliocene) based on geochemical correlation of distal ashfall tuff within the Bouse Formation with the 4.83 Ma Lawlor Tuff (Sarna-Wojcicki et al., 2011) erupted near San Francisco Bay, California (Spencer et al., 2011a, 2011b). That tephrochronology is compatible both with the age (5.6 Ma) of a sub-Bouse tuff (House et al., 2005) and with the age (6 Ma) of the youngest dated tuff interbedded with lacustrine Hualapai Limestone, which was deposited before the arrival of Colorado River water to interconnected desert basins north of the Bouse lake chain (Fig. 1) but still downstream from the mouth of the Grand Canyon (Spencer et al., 2001). Post-Bouse aggradation of lower Colorado River deposits culminated during the Pliocene interval of 4.1–3.3 Ma (House et al., 2005), but older Pliocene Colorado River gravels are interbedded with 4.7 Ma and 4.4 Ma basalts near the Grand Wash Cliffs farther north just below the mouth of the Grand Canyon (Howard and Bohannon, 2001).

The Bouse Formation records rapid inundation of desert basins along the Bouse lake chain. The abrupt arrival of Colorado River water downstream from the Grand Canyon has been taken to support the model of Hopi Lake spillover to initiate incision of the Grand Canyon (Spencer and Pearthree, 2001, 2005; Spencer et al., 2008a; House et al., 2008). Evidence for episodic integration of the lower Colorado River by progressive “spilldown” from successive Bouse lake basins (Spencer et al., 2008b; House et al., 2011) is unable, however, to define the mode of drainage evolution upstream from the Grand Canyon that integrated upper and lower courses of the Colorado River into a unified trunk stream passing through the Grand Canyon. The likelihood of spillover from Hopi Lake to initiate river flow through the Grand Canyon can be addressed by examining the sedimentary record of the Bidahochi Formation, which contains the only known lacustrine deposits thought to record Hopi Lake sedimentation.

The distribution of the Bidahochi Formation and modern elevations at its base are shown by Figure 2. The unit includes three lithologic members (Fig. 3B), at least in part lateral facies of a composite depositional system. Lower lacustrine and middle volcanic members are shown jointly on Figure 2, and the upper fluvial member locally includes interbedded eolian deposits (Love, 1989). Tuffs in the lower (lacustrine) member (100–150 m thick) have yielded Middle Miocene ⁴⁰Ar/³⁹Ar ages of 15.5 Ma and 13.7 Ma (Dallegge et al., 2001, 2003), postdating the initiation of extensional faulting along the Grand Wash Cliffs at the mouth of the Grand Canyon (Faulds et al., 2001b, 2010) but predating deposition of the Hualapai Limestone adjacent to the cliffs (Spencer et al., 2001; Faulds et al., 2001a; Lopez Pearce et al., 2011). The middle (volcanic) member (0–100 m thick), composed of mafic lava, tuff, and volcaniclastic sandstone, has yielded Late Miocene K-Ar and ⁴⁰Ar/³⁹Ar ages of 8.5–6.0 Ma (Damon and Spencer, 2001; Dallegge et al., 2001, 2003), leaving a span of ∼5 Myr undated isotopically between the lower and middle members. A tuff near the base of the upper member has been correlated by tephrochronology with the Blacktail Creek ash with an ⁴⁰Ar/³⁹Ar age of 6.6 Ma (Dallegge et al., 2001, 2003). The isotopic age of ca. 6 Ma for the base of the upper fluvial member is compatible with a late Hemphillian mammalian fauna from the Bidahochi Formation dated at 7–5 Ma from the magnetostratigraphy of Lindsay et al. (1984) as recalibrated by Cande and Kent (1995). The evolution of the Bidahochi depositional system from lacustrine to fluvial environments predated Colorado River flow through the Grand Canyon but coincided approximately with deposition of the youngest tuff (6 Ma) within the Hualapai Limestone (Fig. 3A).

Volcaniclastic strata of the Bidahochi Formation occur at multiple horizons intercalated locally with lacustrine strata (Love, 1989; Ort et al., 1998), and basal horizons of the fluvial upper member intertongue with and grade laterally into lacustrine successions of the lower member (Repenning et al., 1958) where the volcanic middle member is absent (Repenning and Irwin, 1954). Lateral equivalence of horizons mapped within the lower and upper members is demonstrated by intertonguing of fluvial and lacustrine strata within a deltaic complex exposed along the flanks of mesas west of Pueblo Colorado Wash between Ganado and Greasewood Springs (Fig. 2) near a topographic feature known locally as Sand Gap. In time, the upper fluvial member prograded across the lower and middle members, and also oversteps the older members up-paleoflow to the north and east to rest unconformably on pre-Tertiary strata along the fringes of Bidahochi exposures (Fig. 2). Outcrops of the Bidahochi lacustrine facies nowhere reach elevations in excess of ∼1950 m.

The age range of the upper member is not controlled by isotopic dating but the fluvial succession is 75–85 m thick in measured surface sections and reaches thicknesses of 135–140 m in subsurface wellbores. Deposition of fluvial Bidahochi strata may have continued into Pliocene time after integration of lower and upper courses of the Colorado River through the Grand Canyon (Fig. 3A), but basalt lava that locally mantles outcrops of the Bidahochi Formation has yielded an ⁴⁰Ar/³⁹Ar age of 2.4 Ma (McIntosh and Cather, 1994), precluding continuation of Bidahochi sedimentation into Pleistocene time.

Volcanological analysis of the middle volcanic member forming the Hopi Buttes (White, 1989, 1990, 1991; Ort et al., 1998) indicates that Hopi Lake was never a deep body of water during the Late Miocene time frame (8.5–6.0 Ma), when its level must ultimately have risen were it eventually to achieve lake spillover to initiate incision of the Grand Canyon. Phreatomagmatic eruptions gave rise to multiple diatremes (tephra-filled volcanic necks) and maars (shallow craters formed by explosive eruptions). Subaerial scoria cones formed locally within the maars. Hydrovolcanism stemmed from contamination of magma with waterlogged sublacustrine sediment, rather than from eruption into lake waters. Water ponded within some maar craters, but maar rims were nowhere overtopped by lake surfaces. Outflow pyroclastic aprons that spread laterally from volcanic centers include dry base-surge deposits and subaerial pyroclastic flows emplaced locally on muds with desiccation cracks. At the time of the Late Miocene eruptions, Hopi Lake within the area of Bidahochi sedimentation was a playa-like body, and perhaps ephemeral rather than perennial (White, 1990; Ort et al., 1998; Dallegge et al., 2001, 2003). A rise in the elevation of the base of the Bidahochi Formation in its westernmost exposures (Fig. 2) suggests that the lacustrine facies formed within a local depression closed on the west as well as on the east.

The configuration of the basal contact of the upper fluvial member shows that fluvial Bidahochi strata occupy paleovalleys of ancestral Pueblo Colorado Wash, the Puerco River, and Carrizo Wash tributary to the Little Colorado River (Fig. 2). The upper fluvial member also spread, presumably by progressive aggradation, over gently sloping (15 ± 3 m/km) pediment-like surfaces carved across interfluves separating those three paleovalleys. The thalwegs of the floors of fluvial Bidahochi paleodrainages, as defined by contours on the base of the Bidahochi fluvial facies (Fig. 2), have slopes of ∼0.008 (8 m/km). By contrast, the slopes of the modern Pueblo Colorado and Carrizo valleys are only 0.004–0.005 (4–5 m/km) in the area of Bidahochi exposures, and the slope of the modern Puerco valley is even lower at ∼0.002 (2 m/km), comparable to the slope of the modern Little Colorado River valley between St. Johns and Holbrook south of Bidahochi exposures (Fig. 2). The differences in paleovalley and modern-valley slopes, coupled with the close coincidence of paleovalleys and modern valleys in plan view, suggests that Bidahochi paleodrainages served as piedmont feeders to an ancestral Little Colorado River as fluvial aggradation advanced over the Bidahochi lacustrine facies. Dissection of the Bidahochi Formation and gradual reduction in valley slope can be attributed to post-Miocene headward erosion by Little Colorado tributaries as the elevation of the Little Colorado River thalweg was progressively lowered over time from an initial elevation higher than the Bidahochi lacustrine facies buried beneath the Bidahochi fluvial facies.

The alluvial Fence Lake and Quemado Formations of westernmost New Mexico east of St. Johns (Fig. 2) have been regarded provisionally as lateral extensions of Bidahochi strata (Cather and McIntosh, 1994) but are excluded here from the Bidahochi depositional system for the following reasons.

The Fence Lake Formation (McLellan et al., 1982) is equivalent in age (14.5–6.9 Ma) to the lacustrine and volcanic facies of the Bidahochi Formation (Lucas and Anderson, 1994) but its basal contact slopes to the east-northeast (McIntosh and Cather, 1994), away from the Bidahochi depocenter. The net gradient of the basal contact is ∼0.003 (3 m/km) from Table Mountain, ∼50 km south of St. Johns (Fig. 2), to the vicinity of Quemado, 75 km distant in New Mexico. Near the Arizona–New Mexico border east of St. Johns, the base of the Fence Lake Formation is perched at elevations 150–300 m above the base of the Bidahochi fluvial facies. The elevation and configuration of the basal contact suggest that Fence Lake streams were tributary to the ancestral Rio Grande (Pazzaglia and Hawley, 2004), rather than to the ancestral Little Colorado River. Post-Miocene headward erosion by tributaries of the Little Colorado River has transferred exposures of the Fence Lake Formation into the Little Colorado drainage basin by shifting the continental divide to the east.

The Quemado Formation was deposited by streams tributary, like Bidahochi paleodrainages, to the ancestral Little Colorado River but is apparently younger than the Bidahochi Formation. Quemado strata are exposed in five local sub-basins or paleovalleys inset erosionally by ∼100 m into mesas capped by Upper Miocene basalts (6.8–5.2 Ma) that are correlative with the fluvial facies of the Bidahochi Formation, and interfinger locally with Pleistocene basalts ≤1 Ma in age (Cather and McIntosh, 1994).

Initial ⁸⁷Sr/⁸⁶Sr ratios for carbonate materials in the Bidahochi Formation are comparable to ratios in modern river waters on the Colorado Plateau (Fig. 4). The Bidahochi samples include ten of marl from the lower (lacustrine) member, three of molluscan fossils from the middle (volcanic) member, and two of tufa collected from western outcrops of the fluvial (upper) member and interpreted as lake-margin deposits (Gross et al., 2001). Strontium isotope ratios for the lacustrine member (Fig. 3B) closely match ratios for the modern San Juan River draining into the Colorado River from the east, but not ratios for the modern Little Colorado River closer to Bidahochi exposures. The significance of the isotopic difference between modern San Juan and modern Little Colorado water is uncertain with present information. Bidahochi strontium isotope ratios overlap with but tend to be somewhat lower than those of the Bouse Formation from the river-fed lake system that formed along the California-Arizona border immediately after the integrated Colorado River first flowed through the Grand Canyon (Fig. 4). Strontium isotope ratios from the pre–Grand Canyon Hualapai Limestone deposited in local desert basins north of the Bouse paleolakes (Fig. 1) are systematically higher than those from the modern Colorado River, Bouse carbonates, and the Bidahochi Formation (Fig. 4).

Speculative coupling of Colorado River flow with Bidahochi sedimentation (Gross et al., 2001) is apparently denied by the ages of Bidahochi detrital zircons. Bidahochi detrital zircon populations closely resemble those in sands from the modern Little Colorado River but differ substantially from detrital zircon populations in modern sands from the Colorado River and its major upstream tributaries, including the Green, Grand (upper Colorado), and San Juan Rivers (Kimbrough et al., 2011). The Bidahochi depositional system evidently received no sediment from a postulated ancestral Colorado River flowing southeast from Utah into Hopi Lake.

The level of Hopi Lake could not have risen higher than ∼2000 m without leaving a record of lacustrine transgression, which is not observed within the fluvial facies tract of the Bidahochi Formation (Fig. 2). Contours on the present landscape (Fig. 5) show that a lake surface ≤2000 m could not have overtopped the Kaibab-Coconino Plateau in its present configuration. Moreover, the Grand Canyon transects the highest rather than the lowest segment of residual plateau uplands that extend northwest across the Colorado River from the Mogollon Rim in central Arizona. The lowest modern topographic saddle in pre-Pliocene strata forming the uplands lies due west of Bidahochi exposures beneath the post–6 Ma San Francisco volcanic field (Figs. 3A and 5).

These relations jointly imply that spillover of a Bidahochi-related Hopi Lake to initiate incision of the Grand Canyon is not a realistic hypothesis unless (1) post-Bidahochi tectonic or isostatic deformation and associated erosion has substantially altered the morphology of northern Arizona to erect a Kaibab-Coconino topographic barrier that did not exist in its present form at Bidahochi time, (2) integration of the Colorado River through the Grand Canyon was facilitated by the presence of an older paleocanyon that provided a pathway for water flow from lake spillover through an erosional notch incised into the Kaibab-Coconino Plateau at an elevation lower than its crest, or (3) the water surface of Hopi Lake rose at least transiently to elevations at which no sedimentary record of lacustrine conditions has survived post–Grand Canyon erosion. None of these postulates is attractive.

Consideration of differential modern and Pliocene incision rates within the Grand Canyon, coupled with geodynamic modeling from mantle tomography, suggests that the southwestern rim of the Colorado Plateau has been uplifted by as much as 400 m over the past 6 Myr (Karlstrom et al., 2008). If the modern elevation of the top of the Kaibab-Coconino Plateau is restored downward by 400 m, the 6 Ma elevation of the plateau barrier damming postulated Hopi Lake would stand near the maximum elevation of the Bidahochi lacustrine facies. Spillover of even a shallow lake or an array of ponded wetlands across such a subdued Kaibab-Coconino barrier would be feasible.

The postulated tectonic welt continues, however, along the Mogollon Rim to the southeast of the Grand Canyon (Fig. 5) where present maximum elevations are approximately the same (2200–2300 m) as on the Kaibab-Coconino Plateau where transected by the Grand Canyon. Restoring the Mogollon Rim downward to the same elevation in Late Miocene time as the lacustrine facies of the Bidahochi Formation would conflict, however, with paleogeographic analysis implying a persistent regional slope to the northeast from the Mogollon Rim toward an ancestral Little Colorado River since Middle Miocene time (Holm, 2001a, 2001b).

Downward restoration of the Mogollon Rim for Miocene time would allow potential pre–Grand Canyon drainage of the Bidahochi depocenter toward the southwest across the present Mogollon Rim (Potochnik, 2011) and into the ancestral Salt River (Fig. 1), thereby precluding a Bidahochi outlet through the Grand Canyon. The Salt River paleocanyon was incised to a depth of nearly 1500 m (Potochnik and Faulds, 1998; Potochnik, 2001) southwest of Show Low (Fig. 5) by paleoflow to the northeast during Laramide time. Detrital zircons from a sample of the Mogollon Rim Formation collected ∼15 km due west of Show Low indicate that northward paleoflow through the paleocanyon continued at least into Early Miocene time (Potochnik et al., 2012). A cluster of the five youngest detrital zircon grains that overlap in age at 1σ yield a weighted mean average age, with both random and systematic errors taken into account, of 18.4 ± 2.9 Ma (at 2σ), a time frame immediately before lacustrine sedimentation was initiated in the Bidahochi Formation (Fig. 3). The cluster of young zircons was probably derived from the Apache Leap Tuff erupted at 18.6 Ma (McIntosh and Ferguson, 1998) near the headwaters of the Salt River paleocanyon. Drainage reversal, with subsequent streamflow to the southwest through the paleocanyon, occurred during the interval 15–12 Ma within the time frame of Bidahochi lacustrine sedimentation (Potochnik and Faulds, 1998; Potochnik, 2001). Drainage of a Bidahochi lake down the Salt River paleocanyon after drainage reversal is disfavored, however, by the observation that Bidahochi fluvial paleovalleys were ancestral to modern drainages tributary to the Little Colorado River rather than to the modern Salt River.

On the basis of thermochronometry, Wernicke (2011) concluded that a deep (∼1500 m) ancestral paleocanyon roughly the length and breadth of the modern Grand Canyon was incised mainly in Campanian time (80–70 Ma) by reverse paleoflow to the northeast along the course of the “California” River (or paleoriver). By his hypothesis, the Colorado River flowing to the southwest through the Grand Canyon played a major role in dissecting the interior of the Colorado Plateau but was not an important factor in the excavation of the Grand Canyon. Wernicke (2011) postulated that drainage reversal of the California paleoriver by Paleogene time gave birth to the “Arizona” River (or paleoriver), a precursor of the modern Colorado River in the Grand Canyon (Wernicke et al., 2012). His interpretations revive the notion of antecedence for the course of the Grand Canyon across the Laramide Kaibab uplift, although by paleoflow to the northeast rather than to the southwest. The uplift did not begin to grow until 75–80 Ma (Fig. 3A) and was not overlapped by post-Laramide strata of the Claron Formation (Dickinson et al., 2012) until ca. 55 Ma (Figs. 3A, 6). Most important for the evolution of Hopi Lake, his interpretations dictate that initial incision of the Grand Canyon predated all Bidahochi sedimentation by ∼50 Myr.

Other thermochronological analyses (Kelley et al., 2011) and geotectonic arguments (Karlstrom et al., 2012) do not support the existence of a pre-Neogene Grand Canyon, which four key geological relationships argue against:

1. Headwater tributaries of the Verde River, working headward into the Colorado Plateau along the northern fringe of the Gila River drainage, reach to within <5 km of the rim of the western Grand Canyon and to within ∼10 km of the Colorado River itself in the bottom of the canyon (Fig. 1). It seems unlikely that a deep canyon could persist as such a narrow feature for 55–75 Myr after its initial incision, yet otherwise Verde headwaters could not approach the Grand Canyon so closely if the latter originated in Cretaceous or Paleogene time.

2. A butte informally named “Separation Canyon hill” (Young, 2011), perched on the rim of the western Grand Canyon northwest of Peach Springs (Fig. 6), is composed of Oligocene Buck and Doe Conglomerate (Fig. 3A) overlain by a cap of 19 Ma basalt. The butte is now completely surrounded by gorges of the Grand Canyon and its local tributary canyons over the positions of which the strata composing the butte would have projected laterally before canyon incision. The strata of Buck and Doe Conglomerate exposed on the butte could not have reached their present position as part of the correlative gravel apron overlying the Hualapai Plateau (Fig. 6) at any time after the Grand Canyon was incised, nor could the lava capping the butte have reached the site from its source vent ∼8 km to the south after canyon incision (Young, 2011). The relations at “Separation Canyon hill” constrain the Grand Canyon to an origin post–19 Ma.

3. On the Hualapai Plateau south of the western Grand Canyon, the Peach Springs and Hindu-Milkweed paleovalleys (Fig. 6), with maximal paleorelief of 1225–1525 m, are partially backfilled with gravelly Paleogene sediment (Fig. 3A) displaying clast imbrication and other paleocurrent indicators recording paleoflow toward the northeast (Young, 1999, 2008). The paleovalleys are now backtilted to the southwest at a gentle angle owing to post–16 Ma collapse of the Laramide Kingman uplift into the Colorado River extensional corridor developed along the lower Colorado River west of the Colorado Plateau (Beard and Faulds, 2011). Wernicke (2011) inferred that the Peach Springs paleovalley descended into a Paleogene paleo–Grand Canyon to which it was tributary. That interpretation rested upon the estimate of a net reversed gradient from backtilt of 0.8° (downward to the southwest or up-paleocurrent) by Young (2001) based on the subsurface configuration of the Peach Springs paleovalley to the southwest of Peach Springs (Fig. 6). Bedding attitudes observed in the paleovalley fill near the head of Peach Springs Wash north of Peach Springs imply, however, that the amount of backtilt could have been greater northward from Peach Springs toward the Colorado River. Strata of the Buck and Doe Conglomerate and overlying 20 Ma basalt exposed in road cuts at the head of modern Peach Springs Wash near Peach Springs (Fig. 6) dip 2°–4° to the south. Moreover, the present elevation (1200 m) of the intersection of the Milkweed-Hindu Canyon paleotributary with the master Peach Springs paleovalley <10 km from the Colorado River (Young, 2001; Young and Hartman, 2011) argues against descent of the paleovalley into a deep ancestral Grand Canyon. Paleogeographic reconstructions (Graf et al., 1987; Young, 2001; Young and Hartman, 2011) suggest that the Peach Springs paleovalley flowed parallel to the Hurricane fault zone northward from Peach Springs to cross the site of the modern Grand Canyon ∼75 km north of Peach Springs (Fig. 6). Any suggested backtilt of the Peach Springs paleovalley would carry its thalweg hundreds of meters above the rim of the modern Grand Canyon where the Colorado River crosses the Hurricane fault. That conclusion is compatible with the longstanding interpretation of Young (1982, 1985, 1999; Graf et al., 1987) that the Paleogene paleodrainage system of the Hualapai Plateau south of Grand Canyon continued northward across the later site of an exclusively Neogene Grand Canyon.

4. The Wernicke (2011) model of a Cretaceous California paleoriver flowing northeast through a paleo–Grand Canyon, supplanted by a Paleogene Arizona paleoriver flowing to the southwest through the same canyon after drainage reversal, is difficult to reconcile with sedimentological and detrital zircon data for foreland stratigraphic units in Utah (Davis et al., 2010; Dickinson et al., 2011, 2012). Detrital zircons from the Mesozoic Cordilleran magmatic arc in the Mojave region were transported in variable proportions longitudinally from southwest to northeast along the keel of the Sevier foredeep beginning with the onset of Cretaceous marine sedimentation near the Early-Late Cretaceous time boundary, and continuing into Paleogene time during Laramide structural breakup of the foreland region until near the Early-Middle Eocene time boundary. There is no indication that transit of arc-derived detrital zircons across the Grand Canyon region was arrested near the Cretaceous-Tertiary time boundary as required by postulated drainage reversal of the California and Arizona paleorivers.

Of uncertain significance for the speculative development of a paleo–Grand Canyon is a gravel deposit associated with nonmarine limestone displaying columnar internal structure suggestive of lacustrine travertine or tufa exposed at an elevation of 1795 m near Cape Solitude (Scarborough, 2001, p. 209), the topographic point on the south rim of the Grand Canyon south of the junction of the Colorado and Little Colorado Rivers (Fig. 6). The gravel and limestone deposit, never dated or described in detail, stands ∼1000 m above the floor of the Grand Canyon, and has been suggested as possibly a vestige of the floor of a Laramide paleocanyon that incised a notch through the Kaibab-Coconino Plateau to serve later as a Neogene spillway for Hopi Lake (Scarborough 2001; Potochnik, 2011). However, the deposit lies topographically only ∼25 m above the lowest exposed base of the Bidahochi Formation and ∼150 m below the highest exposures of the Bidahochi lacustrine facies. Its substratum could not readily have formed part of a dam for Hopi Lake at any time during Bidahochi sedimentation. The suggestion that a Laramide paleocanyon developed along segments of the Grand Canyon west of the Kaibab-Coconino Plateau, and flowed northward into Utah without crossing the Kaibab uplift (Hill and Ranney, 2008), does not bear on the question of Hopi Lake spillover.

Speculation that Hopi Lake rose transiently to a level allowing Colorado River water to spill over the Kaibab-Coconino uplift without leaving any discernible sedimentary record of the lake highstand is inherently impossible to address with any direct geological evidence. The same problem plagues evaluation of suggested Miocene paleolakes located farther north in Utah at stratigraphic levels since removed by erosion (Hunt, 1969; Hill et al., 2008, 2011). In the absence of locally preserved geological evidence, the existence of either a deep Hopi Lake or analogous paleolakes farther north rests upon considerations of Colorado Plateau denudation and the geometry of the Colorado River and its tributaries. A brief summary of plateau denudation and river evolution is indicated to address the issue of potential Neogene lakes within the interior of the plateau.

Denudational episodes on the Colorado Plateau have been associated with erosional intervals that involved the stripping of sedimentary cover from the pre-Laramide Mogollon highlands, the beveling of Laramide uplifts, and post-Laramide development of the Colorado River drainage system.

The southwestern margin of the Colorado Plateau, including the site of the modern Grand Canyon, was stripped of much of its pre-Cretaceous Mesozoic cover by Upper Jurassic to Early Cretaceous erosion along the tilted northern flank of the Mogollon highlands, which formed the rift shoulder of the Bisbee Basin in the extensional border rift belt (Dickinson and Lawton, 2001; Stern and Dickinson, 2010). A map of the paleogeology (Fig. 7) beneath Upper Cretaceous (Dakota-Mancos) marine strata that onlapped the eroded flank of the Mogollon highlands reflects removal of ∼250 m of Jurassic strata from above the Grand Canyon during regional beveling of the tilted flank of the Mogollon highlands. The deposition of Upper Cretaceous strata directly on Aztec Sandstone of the Lower Jurassic Glen Canyon Group (Stewart, 1980) at the Valley of Fire west of the Grand Canyon (Fig. 7) is compatible with the paleogeologic relations shown. An estimated total of ∼2500 m of Mesozoic strata above the site of the Grand Canyon at the end of Cretaceous time is in agreement with stratigraphic assumptions made for thermochronological analyses of denudation over time at the Grand Canyon (Dumitru et al., 1994; Flowers et al., 2008; Kelley et al., 2011; Lee et al., 2011). The thermochronology suggests two intervals of significant additional denudation at the Grand Canyon, the first Laramide and the second Neogene.

The southern end of the Kaibab uplift near the Grand Canyon exposes no post-Paleozoic strata (Fig. 6), and thus provides no information about the timing of the Laramide deformation that formed the uplift. Farther north, however, Cretaceous–Paleogene strata of the Table Cliff syncline adjacent to the East Kaibab monocline, which bounds the uplift on the east, preserve a sedimentary record of the development of the uplift. Growth faults displacing units as young as the middle Campanian Wahweap Formation (Tindall et al., 2010) along the East Kaibab monocline show that deformation was under way by 80–78 Ma (Fig. 3A). In the Table Cliff syncline east of the Kaibab uplift, lower to middle Campanian Wahweap Formation is overlain conformably by middle to upper Campanian Kaiparowits Formation (ca. 75 Ma) composed of ∼1000 m of distal meander-belt and anastomosed fluvial facies derived from the west (Lawton et al., 2003; Roberts, 2007; Lawton and Bradford, 2011). The net Kaiparowits paleocurrent vector is N70E (Dickinson et al., 2012), reflecting ca. 75 Ma paleoflow across the crest of the Kaibab uplift, which then had incipient structural relief as a growth fold but not enough topographic relief to block sediment transport.

In the Table Cliff syncline, the Kaiparowits Formation is overlain unconformably with an angularity of ∼10° (Bowers, 1972) by the conglomeratic Canaan Peak Formation (80–140 m) of Maastrichtian to Early Paleocene age (Fig. 3) deposited on a coarse proximal braidplain (Schmitt et al., 1991; Larsen et al., 2010) by northeasterly paleoflow (N55E) off the adjacent Kaibab uplift (Dickinson et al., 2012). Continued Laramide deformation produced a closed basin of interior drainage in the Table Cliff syncline where the Lower Paleocene to Lower Eocene Pine Hollow Formation (Fig. 3), which overlies the Canaan Peak Formation with an angularity of 5°–10°, is a cyclic succession of alluvial fan and playa lake deposits 80–120 m thick (Larsen et al., 2010). Centripetal Pine Hollow paleocurrent indicators record inward paleoflow into the Table Cliff synclinal basin, with a weak resultant vector (R = 0.32) southward subparallel to the axis of the syncline (Dickinson et al., 2012).

Following cessation of Laramide deformation in Eocene time, the Claron Formation (Fig. 3) of lacustrine and associated fluvial deposits was deposited as an overlap succession across the Kaibab uplift, East Kaibab monocline, and Table Cliff syncline (Fig. 6). Beneath the Claron overlap, only local remnants of the Kaiparowits Formation ≤25 m thick were preserved from Laramide erosion along the crest of the Kaibab uplift on the present Paunsagunt and Markagunt Plateaus (Bowers, 1991; Moore and Straub, 2001; Eaton et al., 2001). Exposures of the Canaan Peak and Pine Hollow Formations are confined to the Table Cliff syncline east of the Kaibab uplift (Fig. 6). These stratigraphic relationships across the East Kaibab monocline north of the Grand Canyon imply that Laramide deformation giving rise to the Kaibab uplift began in Late Cretaceous time, reached some maximum rate in Paleocene time, and waned during Eocene time.

Laramide removal of all or much of the Mesozoic section once present above Paleozoic strata forming the core of the uplift presumably contributed to the evidence from thermochronology for an episode of Laramide denudation at the Grand Canyon (Dumitru et al., 1994; Naeser et al., 2001; Flowers et al., 2008; Kelley et al., 2011; Lee et al., 2011).

The effects of Laramide erosion are also shown by stratigraphic relations of Tertiary strata in northeastern Arizona east of the Grand Canyon (Fig. 8). The Oligocene Chuska Sandstone unconformably overlaps the East Defiance monocline of Laramide age, and buries paleo-cuestas developed in Cretaceous strata along the trend of the monocline during Eocene time (Wright, 1956). The Miocene Bidahochi Formation overlaps Permian through Cretaceous strata along the western flank of the Defiance uplift where beds dip regionally westward into the Black Mesa Basin, a downfold of Laramide age. Significant Late Oligocene to Middle Miocene denudation intervened, however, between the twin episodes of Chuska and Bidahochi sedimentation (Cather et al., 2008; Cather, 2011).

The surface of the pre-Tertiary substratum in northeastern Arizona is ∼500 m lower beneath the Miocene Bidahochi Formation than beneath the Oligocene Chuska Sandstone (Fig. 9). The concave-upward configuration of the base of the Bidahochi fluvial facies suggests that Bidahochi streams draining toward the Bidahochi lacustrine tract were working headward into the flank of the Chuska erg, which reached a maximum thickness of >500 m (Cather et al., 2008). Areally restricted exposures of Chuska Sandstone capping the present Chuska Mountains are a surviving erosional remnant of the erg (Fig. 1). The uppermost preserved eolian strata of the Chuska Sandstone reach an elevation of ∼3000 m at the crest of the Chuska Mountains (Cather et al., 2008), implying that the local landscape in the Bidahochi depocenter was eroded to an estimated depth of 1000–1250 m between Oligocene and Middle Miocene time. Even after deposition of nearly 300 m of Bidahochi strata, the Miocene ground surface over large areas of northeastern Arizona stood nearly 1000 m below the Oligocene landscape atop the Chuska erg. Analogous mid-Tertiary erosion farther west in the Grand Canyon region is implied by thermochronology (Kelley et al., 2011; Lee et al., 2011).

The modern Little Colorado River flows at an elevation of 1475–1525 m only 25–50 km to the southwest of Bidahochi exposures (Fig. 2), and Bidahochi lakebeds are exposed at elevations as low as ∼1750 m. Estimated post-Bidahochi erosion has thus been ∼500 m and net post-Bidahochi denudation of sub-Tertiary strata has been ∼250 m in the vicinity of Bidahochi exposures. Those figures are ∼50% of comparable figures for erosion and net denudation between Chuska and Bidahochi sedimentation. Approximately two-thirds of the net post-Laramide lowering of the landscape in northeastern Arizona, and perhaps across much of the southern Colorado Plateau (Cather, 2011), evidently occurred before post-Miocene river flow through the present Grand Canyon.

The Crooked Ridge paleochannel located not far east of the Grand Canyon and of probable intra-Miocene age (Lucchitta et al., 2011a, 2011b) underscores the regional scope of Miocene denudation on the southern Colorado Plateau. Its net thalweg slope of 0.006 (6 m/km) toward the southwest over a preserved length of ∼50 km implies flow into an ancestral Colorado River near the site of the present confluence with the Little Colorado River (Fig. 6). The paleochannel elevation of ∼1700 m at the southwestern end of its exposures is within 50–100 m of the base of the Bidahochi Formation, suggesting that development of the Crooked Ridge paleodrainage postdated dissection of the Chuska erg in northeastern Arizona. That inference is supported by the presence of clasts within the paleochannel derived from basement and volcanic rocks in the San Juan Mountains of southwestern Colorado. The inferred original extent of the Chuska erg (Fig. 1) would have blocked drainage off the San Juan Mountains from reaching the Crooked Ridge paleochannel prior to erosional stripping of the erg from much of its original extent.

Detrital zircons as young as ca. 24 Ma extracted from fluvial sediment in the Crooked Ridge paleochannel (Price et al., 2012) confirm that the Crooked Ridge stream was flowing in Miocene time or later (Price et al., 2012). As the Crooked River paleochannel descends from an elevation of 1825 m near Kaibito on the northeast (Fig. 6) to 1700 m at its southwesternmost exposure at The Gap, its existence during the Miocene would preclude delivery of Colorado River water to the Bidahochi depocenter. The paleochannel lies athwart any logical pathway leading from the upper Colorado River in Utah toward Bidahochi lakebeds that reach higher elevations of 1850–1900 m (Fig. 2).

The history of erosion and denudation on the northern Colorado Plateau contrasts strongly with relations in northeastern Arizona. Near Grand Junction, Colorado, at the confluence of the Colorado (Grand) and Gunnison Rivers, Gunnison River gravels (Price et al., 2012) underlying Upper Miocene (9.5–11.0 Ma) basalt capping Grand Mesa (Fig. 10) stand at an elevation 1500+ m above modern river valleys (Aslan et al., 2011; Cole, 2011), showing that the major erosional episode affecting the northern plateau postdated eruption of the basalt. Similarly late denudation of the Colorado Plateau over much of Utah is implied by thermochronology (Hoffman et al., 2011) indicating an onset of cooling after ca. 10 Ma through removal of 1000–2000 m of overburden. The last 845 m of Neogene erosion on the central Colorado Plateau has produced 640 m of isostatic rock uplift (Pederson et al., 2002), implying that surface elevations on the plateau were reduced by only 25% of the erosional denudation, allowing for persistence of a high-standing plateau as denudation proceeded. Geometric patterns of the Colorado River and its tributaries may offer alternatives to Hopi Lake spillover for the integration of upper and lower courses of the Colorado River through the Grand Canyon.

The Colorado River and many of its major tributaries transect multiple Laramide uplifts of the Colorado Plateau and its margins, and cross Laramide basins without notable diversion (Fig. 10). The maximum structural relief of intra-plateau uplifts as inferred from structure contours is typically within the range of 1500–2000 m (Kelley, 1955; Hunt, 1956; Bump, 2004; Flowers et al., 2008). Given the Paleogene age of the structural uplifts and basins (Dickinson et al., 1988), the Neogene Colorado drainage network must reflect superposition, but not necessarily from a unitary Tertiary erosion surface uniformly capping all uplifts and basins simultaneously. The stratigraphy of the Colorado Plateau involves alternating successions of strata that are relatively resistant and nonresistant to erosion. The superposition of some streams across selected Laramide uplifts may stem from ancestral streams that once flowed across extensive lowlands developed within outcrop belts of less-resistant strata before downcutting lowered the stream courses into more-resistant strata occupying the cores of the uplifts. For example, Hunt (1969, p. 108) suggested that the incised meanders of the San Juan River where it crosses the Monument upwarp (Fig. 10) are inherited from meanders formed when the river was flowing through weak Mancos Shale, of Cretaceous age, at a much higher stratigraphic horizon than at present. Where preserved on the faces of high plateaus rimming the central Colorado Plateau, the Mancos Shale is commonly 1200–1500 m thick. If that thickness is projected across the site of the Monument upwarp, the Mancos Shale could once have blanketed a broad span of the uplift to mask much of the structural relief then hidden in more-resistant older strata.

Hunt (1956, p. 70–71; 1969, p. 103) noted that neither the Colorado River nor any of its major tributaries transect any of the laccolithic igneous centers of the Colorado Plateau (Figs. 1, 10). Paleotopography above the laccolithic centers evidently exerted strong control on the paths of plateau streams in the post-Laramide time frame during which the modern Colorado River drainage network evolved. Volcanic edifices may once have towered above the subvolcanic laccoliths and stocks, with volcaniclastic aprons >50 km in diameter (Lipman, 1989) influencing paleotopography. The San Juan and Dolores Rivers, however, swing around or pass between Cretaceous–Paleocene laccolithic centers of the Colorado Mineral Belt (Figs. 1, 10) just as surely as the Colorado River and other tributaries avoid Oligocene laccolithic centers farther north on the central Colorado Plateau (Fig. 3C). As no Paleocene volcanic edifices are likely to have persisted for 30–40 Myr into mid-Tertiary time, resistant flatirons along the flanks of local igneous-cored uplifts were perhaps sufficient to divert drainages away from the igneous centers. In effect, the laccoliths and accompanying stocks punched holes through the plateau stratigraphy, thereby breaking the lateral continuity of nonresistant stratigraphic horizons to prevent superposition of streams across the igneous-cored uplifts.

As late as Eocene time, plateau drainages were still flowing into Laramide basins (Fig. 10). Paleoflow in southeastern Utah, where the Colorado River now flows southward, was then northward into the Uinta Basin (Dickinson et al., 2012). There are only two fundamental mechanisms by which a southward-flowing Colorado River superimposed across the Laramide uplifts of Utah could have developed on the post-Laramide landscape: (1) headward erosion eating into the Colorado Plateau from the south to gradually capture drainages on the central Colorado Plateau, or (2) progradation of headwater streams across the plateau from the north and east to progressively prolong the drainage system southward. Reversal of central plateau drainage was centered on Oligocene time, but there are no Oligocene strata preserved in the interior of the plateau except for the dominantly eolian Chuska Sandstone. Both the Laramide and the mid-Tertiary laccolithic centers of the Colorado Plateau were emplaced by the end of Oligocene time, allowing the laccolithic structures to guide either successive stream captures upstream from the south or progressive elongation of headwater drainages downstream across the plateau from the north.

The Chuska erg (Fig. 1) occupied an area of ∼100 × 10³ km² on the southeastern Colorado Plateau, limited on the north by positive paleotopography associated with Oligocene laccolithic centers in Utah and the San Juan volcanic field in Colorado (Cather et al., 2008). The existence of such a large erg implies an arid paleoclimate for the Oligocene Colorado Plateau, suggesting that any through-going plateau streams were rare and of limited vigor, if present at all. There was evidently a time break between a previous regime of northward-flowing Laramide streams and a subsequent Neogene regime of southward-flowing streams. In effect, an arid Oligocene interval wiped the drainage slate clean on the central Colorado Plateau.

Chuska sand accumulation occupied a broad intramontane tract bounded on the east by residual Laramide highlands of the Rocky Mountains and on the west by the elevated Nevadaplano plateau (DeCelles, 2004) between the relict Sevier thrust belt and the Sierra Nevada. The Takla Makan dunefield of central Asia, although twice the reconstructed size of the Chuska erg, occupies a similar modern intramontane depression between the Kunlun and Tien Shan Mountains (Breed et al., 1979). Stream courses that issue from the enclosing high ranges dissipate into the Takla Makan dunefield as their waters seep into the sand. Perhaps there was no pre-Miocene Colorado River because no through-going trunk stream traversed the Colorado Plateau in any direction during Oligocene time. The Colorado River drainage system may have initially been integrated across Utah by gradual prolongation of streams issuing from highlands surrounding the Colorado Plateau as desert conditions within the interior of the plateau were ameliorated during Miocene time. The pattern of paleoflow on the Colorado Plateau north of the Chuska erg is uncertain, however, because the only Oligocene strata preserved in the region surrounding the Uinta Basin are alluvial deposits of the Bishop Conglomerate that flanked the core of the Uinta Mountains before deposition of the Miocene Browns Park Formation within an axial graben (Fig. 3C).

The Grand Canyon is not the only place where the Colorado River has eroded into a Laramide uplift to the depth of Precambrian basement. Farther north (Fig. 10), the upper Colorado (Grand) and Green Rivers incise the Uncompahgre and Uinta uplifts, respectively, to expose Precambrian rocks in Westwater and Lodore Canyons. The courses of the rivers follow the same pattern of deflection for each of the three uplift transects including the Grand Canyon, suggesting some commonality of origin. In each case, upstream reaches of the rivers flow directly toward the uplifts, the Kaibab in the case of the Grand Canyon. The river courses then turn abruptly at uplift margins to flow for 25–50 km subparallel to the trends of the uplifts, in each case in a direction down plunge to the configuration of uplift crests. Final passages of the rivers through canyons transecting the uplifts then follow looping paths with curvatures that are convex down plunge with respect to the plunges of uplift axes.

This pattern of river flow across folded structures has been attributed by Oberlander (1965, 1985), from studies in the Zagros Mountains in Iran, to the transverse superposition of streams across folds with an internal stratigraphy composed of alternating resistant and nonresistant stratigraphic intervals. When easily erodible strata are exposed at the surface, extensive lowland straths are developed across fold crests. Only when a stream erodes downward though the weak strata does it “find” underlying resistant strata previously hidden within the core of the fold. When the downcutting stream encounters exposures of the resistant strata, it migrates laterally down the plunge of the fold at the uppermost horizon of the resistant strata. A retreating escarpment of overlying strata bounds the stream valley on its down-plunge side. When the height of the retreating scarp is sufficient to deliver more debris to the stream than the latter can transport, lateral stream migration is arrested, and the stream is forced to incise downward to form a steep canyon in the resistant strata. The result of the process is a looping stream transit of the fold in a canyon that is convex down plunge in plan view with respect to the fold axis. The scenario can be repeated as stream erosion bites ever deeper into alternating resistant and nonresistant stratal intervals.

The geologic histories of Lodore Canyon (Hansen, 1986; Pederson and Hadden, 2005), Westwater Canyon (Aslan et al., 2011), and the Grand Canyon differ in detail. The common plan-view geometry of the trunk river courses for each of the three uplift transits nevertheless encourages speculation that the erosional process described by Oberlander (1965, 1985) was a fundamental constraint for the development of each of the three canyons.

Babenroth and Strahler (1945) and Strahler (1948) applied the logic of the later Oberlander thesis to elucidate the looping course of the Colorado River across the Kaibab uplift, evoking the Permian Kaibab Limestone, which forms the stripped stratigraphic surface of the Kaibab-Coconino Plateau (Fig. 6), as the controlling resistant layer for incision of the Grand Canyon. They regarded the overlying Triassic–Jurassic stratigraphic interval (Moenkopi, Chinle, Moenave, and Kayenta Formations) as the nonresistant stratal cover that once blanketed the resistant core of the Kaibab uplift beneath more-resistant Jurassic Navajo Sandstone. They envisioned north-facing cliffs of Navajo Sandstone, since removed by subsequent erosion, as the upper tier of a retreating escarpment that curved (convex to the south) across the plunging Kaibab uplift on the south side of an ancestral Colorado River valley before the river incised into Kaibab Limestone. Chase (2001) has focused renewed attention on the manner in which the half-circular segment of the Colorado River in the eastern Grand Canyon “sidehills” around the nose of the Kaibab uplift with respect to structure contours of the fold.

The Babenroth-Strahler hypothesis of stratal superposition to explain the course of the eastern Grand Canyon through the Kaibab uplift and the upper granite gorge (Fig. 10) is broadly compatible with the known structural geometry of the Kaibab uplift. The projected thickness of dominantly shaly Triassic–Jurassic strata once present over the eastern Grand Canyon between Kaibab Limestone below and Navajo Sandstone above is ∼900 m. The structural relief of the Kaibab uplift at the Colorado River is ∼750 m on its western limb and ∼1050 m on its eastern limb (Karlstrom et al., 2007). Before removal by erosion, the Triassic–Jurassic shaly strata could once have buried most if not all of the structural relief of the Kaibab uplift.

The concept of stratal superposition for the evolution of the Grand Canyon is reinforced by the similarly looping course of the western Grand Canyon through an anticlinal structure into which the Colorado River has incised the lower granite gorge (Fig. 10). The river and canyon describe a prominent southward loop, similar in geometry to the loop at the Kaibab uplift, in crossing a gentle anticline that trends north-south between the Hurricane fault and the mouth of the Grand Canyon (Fig. 6). Structural relief on the anticlinal structure is ∼750 m on the east and ∼550 m on the west (Karlstrom et al., 2007), comparable to the estimated ∼800 m thickness of shaly Triassic–Jurassic strata as projected speculatively over the western Grand Canyon. The course of the Colorado River through the Grand Canyon is thus partially adjusted to two separate anticlinal structures in the manner envisioned by Oberlander (1965, 1985) for the superposition of transverse drainages across structure.

The location of the Crooked Ridge paleochannel (Fig. 6) and the configurations of fluvial Bidahochi paleochannels ancestral to tributaries of the Little Colorado River (Fig. 2) suggest that both systems flowed into an ancestral Miocene Colorado River not far east of the Kaibab uplift near the present confluence of the Colorado and Little Colorado Rivers. The elevation of the downstream limit of the preserved Crooked Ridge paleochannel (∼1700 m) and the elevation of the downstream edge of the Bidahochi fluvial facies (∼1900 m) suggest that the termini of the Crooked Ridge and Bidahochi fluvial systems at an ancestral Colorado River were near an elevation of 1625 m. That elevation is calculated by projecting both paleoflow systems downstream at the net gradient (1.75 m/km) of the modern Little Colorado River valley along the 250 km of its course across lowlands of northern Arizona upstream from the mouth of Moenkopi Wash near Cameron (Fig. 6). The calculation suggests that the Miocene Colorado River had already breached the Kaibab Limestone on the Kaibab uplift, and was transecting the Kaibab-Coconino Plateau in an ancestral valley with its floor at ∼1600 m near the top of the Redwall Limestone (Karlstrom et al., 2007). That inference assumes that large areas of northeastern Arizona were denuded to levels within less than ∼500 m of the modern landscape by Middle Miocene time.

Multiple arguments thus jointly disfavor Hopi Lake spillover as the trigger for incision of the Grand Canyon through the Kaibab-Coconino Plateau, implying instead that an ancestral Colorado River with a large upstream drainage basin but an unknown volume of discharge crossed the Kaibab uplift before the modern Grand Canyon was cut. It remains a mystery where such a river exited the plateau west of the Kaibab uplift.

There is conclusive evidence that no such river passed through, or at some pre-canyon erosional level above, the present mouth of the Grand Canyon. The Upper Miocene Hualapai Limestone (Fig. 3), deposited within a lake system that extended upstream along the course of the modern lower Colorado River as far as the mouth of the Grand Canyon, displays strontium isotope ratios elevated above those of any modern or pre-modern waters of the Colorado River system (Fig. 4). Hualapai paleolake waters were apparently spring fed by local groundwater that emerged from Paleozoic limestones (Faulds et al., 2001a; Hill and Ranney, 2008; Hill et al., 2008; Crossey et al., 2011; Lopez Pearce et al., 2011). The detrital zircon populations in sandstones of the Hualapai Limestone depositional system are dominated by ca. 1400 Ma and ca. 1700 Ma grains derived from local sources (Lopez Pearce et al., 2011). Wholly missing are the Grenville, Neoproterozoic, and Paleozoic subpopulations of detrital zircons that are prominent in all modern Colorado River sediments collected both upstream and downstream from the Grand Canyon (Kimbrough et al., 2011).

If an ancestral Colorado River crossed the Kaibab uplift but did not follow the route of the western Grand Canyon, the most logical pathway for paleoflow lies to the west-northwest across the area of the northern Uinkaret and Shivwits Plateaus north of the Grand Canyon (Fig. 6), as postulated by Lucchitta (1984, 1989) and presumed by Pelletier (2010). Surface elevations on the plateaus north of 8.2–6.2 Ma volcanic edifices near the Grand Canyon (Lucchitta and Jeanne, 2001) are ≤1600 m, slightly less than the inferred elevation of the Miocene valley floor where the Colorado River crossed the Kaibab uplift. Given the regional northeasterly dip of strata in the Grand Canyon region (Fig. 6), the inferred pre–Grand Canyon river valley crossing the Uinkaret and Shivwits Plateaus would have been a strike valley (Lucchitta, 1990; Lucchitta and Jeanne, 2001) with a southeast-to-northwest alignment. Subparallel strike valleys of Late Oligocene–Early Miocene age also developed south of Flagstaff (Fig. 6) in the ancestral Gila River drainage below the nascent Mogollon Rim (Blakey et al., 2011).

From model calculations, Pelletier (2010) concluded that a proto–western Grand Canyon working headward from the foot of the Grand Wash Cliffs, which formed by normal faulting between 16 Ma and 9 Ma (Fig. 3A), could by the end of Miocene time have captured the pre–Grand Canyon Colorado River somewhere west of the Kaibab uplift but east of the Shivwits Plateau at a point <150 km from the Grand Wash Cliffs. Estimated rates of headward erosion are not sufficient to allow a knickpoint to migrate upstream from the Grand Wash Cliffs far enough to propagate through the Kaibab uplift within the time frame (<6 Ma) available for canyon evolution (Spencer and Pearthree, 2001; Pelletier, 2010). The greater breadth of the western Grand Canyon as compared to the eastern Grand Canyon is interpreted by Pelletier (2010) to be a function of the time available for lateral cliff retreat begun after a knickpoint had migrated upstream, first from the Grand Wash Cliffs toward the capture point in the central Grand Canyon, and then past the capture point into the eastern Grand Canyon.

The role of groundwater sapping (Pederson, 2001b, 2008; Karlstrom et al., 2011) in the capture of the Colorado River in the Grand Canyon is uncertain, but subterranean flow of groundwater through karstic features which later collapsed may have played a major role in promoting incision of the eastern Grand Canyon through the Kaibab uplift (Hill and Ranney, 2008; Hill et al., 2008). From a regional perspective, the role of karstic erosion for either stream capture or knickpoint migration is a local detail that does not alter the overall evaluation of Grand Canyon timing.

No significant sediment delivery to the Hualapai Limestone depocenter at the mouth of the modern Grand Canyon would be expected before capture of the upper Colorado drainage near the Miocene-Pliocene time boundary. In that respect, stream capture in the central Grand Canyon would have abruptly initiated transit of water and sediment through the full length of the Grand Canyon for the first time near the Miocene-Pliocene time boundary just as surely as Hopi Lake spillover might. Continued deepening of the Grand Canyon would be expected to proceed apace once Colorado River flow was established through the full length of the canyon.

The evolution of the Grand Canyon outlined above involves a paradox because no exit from the plateau for an ancestral Shivwits-crossing Colorado River has yet been surely identified. Any such stream would have continued northwest toward St. George (Fig. 6) in the drainage of the modern Virgin River (Figs. 1, 10). The Virgin River now debouches from the plateau into desert lowlands to the west where the Muddy Creek Formation (Fig. 3A) accumulated during Miocene time in multiple sub-basins fringing the Hualapai paleolake system (Fig. 1) on the north. Precursors of the modern White River and Meadow Valley Wash, which rise in the Basin and Range province west of the Colorado Plateau (Fig. 1), contributed sediment to the Muddy Creek Formation, as did the paleo–Virgin River itself, but there is doubt that any Colorado River sediment is present in the Muddy Creek Formation (Pederson, 2001a). Muddy Creek sandstones contain volcaniclastic detritus derived from the Caliente caldera complex (Fig. 1) of eastern Nevada (Pederson, 2008), and perhaps also debris from the Lower Miocene Pine Valley laccolith forming high topography along the western flank of the Virgin River drainage (Figs. 3A, 6).

U-Pb age data for detrital zircons in the Muddy Creek Formation (Forrester, 2009; Swenberg and Hanson, 2010; Muntean and Hanson, 2010; Hanson and Forrester, 2010; Muntean, 2012) do not preclude the possibility that Muddy Creek basin fill is an amalgam of Colorado River sediment contaminated with Virgin River sediment and detritus from both the Caliente caldera complex (Fig. 1) and other sources within the Basin and Range province. The age spectra of pre-Cenozoic detrital zircon grains in Virgin River–Muddy Creek sands are not qualitatively different from the age spectra of modern Colorado River sand. The most prominent nine pre-Neogene U-Pb age peaks on composited age-probability plots (age-distribution curves) for 450 Virgin River–Muddy Creek detrital zircons (Forrester, 2009) and 327 Colorado delta detrital zircons (Kimbrough et al., 2011) are essentially indistinguishable (figures in Ma): 85–105, 165–175, 225–275, 350–475, 535–675, 1025–1225, 1375–1500, 1625–1850, 2700–2825.

The principal contrast between Muddy Creek and Colorado River detrital zircon populations is the consistent presence of Neogene grains (11–24 Ma) in the former, where they typically form ∼10% of the grain populations (Forrester, 2009). In nine Muddy Creek samples from near the Overton Arm of Lake Mead near the modern Colorado River, ∼10% of the detrital zircons are 18–21 Ma in age (Muntean, 2012), but most of the other detrital zircons are compatible with plateau derivation. Neogene detrital zircon grains (19–24 Ma) also form 6% of the total grain population in modern Virgin River sand (Forrester, 2009). The Neogene detrital zircons derive mainly from the Caliente caldera complex (Fig. 1) or the Pine Valley laccolith (Fig. 3A), with the latter source indicated for modern Virgin River sand. An ancestral Colorado River emerging from the Shivwits and Uinkaret Plateaus into a paleo–Virgin River drainage would have acquired a similar subpopulation of Neogene detrital zircons. Otherwise, headwaters of both the Colorado and Virgin Rivers tap essentially the same plateau succession of Triassic through Cretaceous strata capable of yielding analogous age spectra of detrital zircons.

Comparative detrital zircon ages from the Muddy Creek Formation and the Colorado River are thus provisionally consistent with the hypothesis that the Muddy Creek depocenter was the Miocene terminus of an ancestral pre–Grand Canyon Colorado River where riverine sand mixed in varying proportions with more locally derived sands delivered by the ancestral Virgin River and drainages still farther west. If so, the central conundrum of how the Colorado River was integrated through the Grand Canyon is resolved by the Pelletier (2010) postulate of stream capture at a stratigraphic horizon above the modern central Grand Canyon following headward erosion from the Grand Wash Cliffs through the western Grand Canyon, with subsequent deep incision of the eastern Grand Canyon through the Kaibab-Coconino Plateau. Hopi Lake spillover can then be viewed as not only an unlikely but also an unnecessary hypothesis. Further study of Muddy Creek sedimentation is indicated, and calculations of water and sediment balances would be helpful, although the latter are difficult to hindcast without better knowledge of the discharge and sediment load of the Miocene Colorado River (Pederson, 2001a). A sluggish Colorado River of Miocene age crossing a Colorado Plateau as yet unaffected by headward erosion upstream from the modern Grand Canyon may have had a different discharge and sediment load.

The paleogeography of the Bidahochi Formation and the elevation of Bidahochi exposures relative to the crest of the Kaibab-Coconino Plateau disfavor spillover of Hopi Lake to initiate incision of the Grand Canyon by integrating upper and lower segments of the Colorado River, without ancillary paleotopographic assumptions that are difficult to defend. The history of denudation on the Colorado Plateau and the geometry of the Colorado River drainage network suggest that a Miocene paleo–Colorado River had already transited the Kaibab uplift by the time of Bidahochi sedimentation. Stream capture at a point near the central Grand Canyon could have diverted river water into Bouse lakes downstream from the Grand Canyon as abruptly as lake spillover could have. The point of exit of the pre-capture Colorado River from the Colorado Plateau remains uncertain but a paleo–Virgin River and the Muddy Creek depositional basin are seemingly viable candidates for the exit point.

Comments by reviewer Ivo Lucchitta, an anonymous reviewer, and private reviewer Jon Spencer improved parts of the text, and editorial suggestions by Richard Young improved my treatment of the Peach Springs paleovalley. Jim Abbott of SciGraphics prepared the figures.