The southwesterly course of the probably pre–early Miocene and possibly Oligocene Crooked Ridge River can be traced continuously for 48 km and discontinuously for 91 km in northern Arizona (United States). The course is visible today in inverted relief. Pebbles in the river gravel came from at least as far northeast as the San Juan Mountains (Colorado). The river valley was carved out of easily eroded Jurassic and Cretaceous rocks whose debris overloaded the river with abundant detritus, probably steepening the gradient. After the river became inactive, the regional drainage network was rearranged three times, and the nearby Four Corners region was lowered 1–2 km by erosion. The river provides constraints on the early evolution of the Colorado River and Grand Canyon. Continuation of this river into lakes in Arizona or Utah is unlikely, as is integration through Grand Canyon by lake spillover. The downstream course of the river probably was across the Kaibab arch in a valley roughly coincident with the present eastern Grand Canyon. Beyond this point, the course may have continued to the drainage basin of the Sacramento River, or to the proto–Snake River drainage. Crooked Ridge River was beheaded by the developing San Juan River, which pirated its waters and probably was tributary to a proto–Colorado River, flowing roughly along its present course west of the Monument upwarp.


Debate on how and when the Colorado River and Grand Canyon (southwestern United States) came into being as we know them today has continued in the 140 years since J.W. Powell’s epic journey of discovery on the Green and Colorado Rivers. The debate has illuminated not only the history of the Colorado River itself, but of rivers and canyons in general. Of especial scientific and popular interest is the question: How did this canyon, the Grand, come to be?

Views on the issue are in two main groups: One holds that the river is old (possibly as old as Eocene) and has always had approximately its present configuration; the other holds that the river has achieved its present course only relatively recently (generally near the end of the Miocene) through one of several processes of integration, including headward erosion and stream capture, lake spillover, subterranean piping, and reactivation of pre-existing canyons, possibly of early Tertiary age.

Most of the early workers from Powell on (Powell, 1875; Dutton, 1882; Davis, 1901) were proponents of the first notion, as was Hunt (1956, 1969), who painstakingly assembled a host of information about ancient courses of the Colorado and San Juan Rivers on the Colorado Plateau. More recent work generally is part of the second group.

Studies in the early to mid-1900s in the Basin and Range province along the present course of the lower Colorado River (Lee, 1908; Blackwelder, 1934; Longwell, 1936, 1946) brought about the transition from earlier views to the later ones. These studies showed that areas now traversed by the Colorado are filled with middle and late Miocene interior-basin deposits. This was later confirmed in detail for the Pierce Ferry area at the mouth of Grand Canyon (Fig. 1) by Lucchitta (1966, 1967, 1972, 2013). The conclusion was that no Colorado River could have existed in its present Basin and Range course until after 5–6 Ma.

The question then arose: What was the course of the mid-Miocene or earlier Colorado River on the western Colorado Plateau for the 10+ million years when it did not flow into the Gulf of California? Hunt (1969) suggested that the river flowed southward through Peach Springs Canyon (Fig. 1) until late Miocene time, when it established its present course in western Grand Canyon by some subterranean piping mechanism. The problem was not solved, however, because the widespread interior-basin deposits would have blocked this course as well. Such deposits are ubiquitous in the Basin and Range province, not just along the lower Colorado River area. Furthermore, Young’s (1979, 1982) work showed that fluvial deposits in Peach Springs Canyon were deposited by streams flowing north, not south as proposed by Hunt.

The stage was set for the notion that an ancient upper river and a much younger lower one were integrated after 5–6 Ma into a single river with the present course. This idea was strongly influenced by the then-new discovery that the upper Gulf of California, into which the Colorado flows, had opened only in late Miocene to early Pliocene time (Durham and Allison, 1960).

McKee et al. (1967) argued that the old upper river reached the east side of the Kaibab Plateau (Fig. 1), which was seen as an insurmountable barrier, and then flowed southeast along the present alignment of the Little Colorado River into the Rio Grande. In latest Miocene time this ancestral river was captured east of the Kaibab Plateau and diverted into its present course through Grand Canyon by a vigorous young stream that propagated itself from the Gulf of California by headward erosion. However, the course into the Rio Grande was not supported by available data, so Lucchitta (1975, 1984, 1989, 2013), making use of improved understanding of the paleogeology and paleotopography of the region, proposed instead that the ancient river had crossed the Kaibab Plateau in an arcuate strike valley controlled by the south-plunging part of the Kaibab dome. This valley followed the present alignment of eastern Grand Canyon (see Babenroth and Strahler, 1945), then continued northwest along regional strike, bypassing western Grand Canyon and the lower Colorado River area. Subsequent capture and diversion by the younger stream would have been as per McKee et al. (1967), but west of the Kaibab Plateau.

These proposals contain two major novelties: One is that the Colorado River in its present course through Grand Canyon is no older than 5–6 Ma; the other is that river systems are not immutable, but are part of drainage networks that change with time through a quasi-Darwinian competition in response to external circumstances such as tectonism and by means such as headward erosion and capture. Drainages with the steepest gradient survive and expand their drainage area by capturing the water of lower-gradient and therefore less aggressive drainages, which become inactive. There is no “beginning” and no “end” to most rivers, only changes in the connections and configuration of the drainage network.

The last quarter century has seen an explosion of theories and ingenious ideas regarding the history of the Colorado River and its integration into the present course through Grand Canyon. Some have ignored the evidence for Muddy Creek (Miocene) interior-basin deposition at the mouth of Grand Canyon (e.g., Robert et al., 2011), or have discounted it (Wernicke, 2011; Flowers and Farley, 2012); others proposed that this was never a constraint because some sort of canyon already existed in Muddy Creek time or even earlier (Faulds et al., 2001; Wallace et al., 2005; Young, 2008; Wernicke, 2011; Flowers and Farley, 2012). Subterranean piping is a popular theory (Hill et al., 2008; Pederson, 2008), as is lake spillover (Blackwelder, 1934; Meek and Douglass, 2001; Scarborough, 2001; Spencer and Pearthree, 2001). Another idea is that parts of Grand Canyon are old and were occupied by the Colorado River but were choked by debris during interior-basin deposition to the west, so the river became inactive (Elston and Young, 1991), or were formerly occupied by rivers that flowed in directions (generally north or northeast) other than the present one (e.g., Scarborough, 2001; Potochnik, 2001; Hill and Ranney, 2008).

Three current concepts are of particular interest in the context of the present paper. The first is that an ancestral upper Colorado River emptied into “Hopi Lake” (in which was deposited the Pliocene and Miocene Bidahochi Formation) (Fig. 1) during the interval when the river did not flow in its present lower course west of the Colorado Plateau. The second is that “Lake Hopi” at one time drained northward along the present alignment of Marble and perhaps Glen Canyons. The third is that the Colorado River became integrated in its course through Grand Canyon and Basin and Range reaches by means of cascading spillovers—Hopi Lake into Hualapai Lake (Pierce Ferry area; Fig. 1), Hualapai Lake into Bouse Lakes (at the border between Arizona and California), and Bouse Lakes into the Gulf of California to the south. This is a top-down mechanism, as opposed to the bottom-up mechanism involved in headward erosion and stream capture.

The data from Crooked Ridge River provide useful new constraints on these concepts. (Crooked Ridge River is a paleoriver whose course is now visible through inverted relief as Crooked Ridge.) Scattered exposures of gravel in northern Arizona have long been known (e.g., Cooley et al., 1969; Hunt, 1969; R. Hereford, 1975, personal commun.), and in some cases the gravel was thought to be derived from the San Juan Mountains in Colorado (Cooley et al., 1969; Hunt, 1969). However, integration of these observations into a coherent picture of the drainage system only became possible when the ridge could be seen as a distinct topographic entity with the advent of detailed topographic maps in the 1980s, and especially of satellite and digital elevation model (DEM) data. Such images presented here are composites of Landsat and shaded-relief DEM data with 10–30 m resolution, obtained from the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center’s seamless-data server (viewer.nationalmap.gov/viewer).


Crooked Ridge extends continuously across the Kaibito Plateau of northern Arizona from the eastern edge of White Mesa westward to The Gap, a large erosional gash carved into the Jurassic Navajo Sandstone at the Echo Cliffs (Figs. 2, 3, and 4). Kaibito Plateau is an area of little organized drainage and abundant sand dunes, excepting at the northern end, where canyons extend southward from Glen Canyon into the plateau. The lack of drainage and the sand dunes have contributed to the preservation of the ancient Crooked Ridge.

Ridges heading toward Crooked Ridge on the Kaibito Plateau, and gravel on the north side of White Mesa, suggest possible tributaries to Crooked Ridge River. These features were not examined during this study. The ridge is 48 km long as the crow flies, and 55 km long along its trace. An isolated remnant of river deposits with distinctive clasts similar to those on Crooked Ridge is present near the northwest corner of Black Mesa, ∼43 km from the nearest exposures on White Mesa and approximately on the same gradient and trend (Figs. 2 and 5). The discontinuous river course can thus be traced for 91 km. The sinuous Crooked Ridge is an example of inverted relief that came about because deposits in the floodplain of an ancient river were protected by a 1–2 m cap of massive pedogenic stage V calcrete, whereas the bedrock outboard of the river deposits was not so protected and has been preferentially lowered by erosion (Figs. 3, 4, 6, and 7).

Remnants of the river deposits are now found along about a quarter of the ridge’s length westward from White Mesa and about a third of the length eastward from The Gap (Figs. 3, 5, and 6). The intervening part has either fragmentary river deposits or none. The capped parts of the ridge rise as much as 110 m above the adjacent landscape, whereas the eroded bedrock parts are as much as 50–80 m above it. On satellite images, the river deposits are as much as ∼1000 m wide, though generally less.

Along most of the Kaibito Plateau reach, the bedrock for the Crooked Ridge River valley is the relatively weak upper part of the Navajo Sandstone and the easily eroded Carmel Formation at the base of the San Rafael Group. These rocks allowed formation of a wide river valley. Evidence for this is visible on satellite images, where the north edge of the old valley is marked by a south-facing 150-m-high scarp that parallels the trend of Crooked Ridge and truncates the south-southeast–trending Mormon Ridges (Figs. 3, 4, and 6). The south edge of the valley is ill defined, but probably was along the northern end of bedrock prominences such as Preston Mesa (Fig. 3) and the southern part of White Mesa. This alignment parallels both Crooked Ridge and the northern edge of the valley. If this alignment is taken as the south edge, the valley may have been as much as 5–10 km wide over most of the Kaibito Plateau. This width, however, is only that of the preserved (and lowest) part of the ancient valley. The width of the upper and now-eroded part of the valley probably was much greater.

At The Gap (Figs. 2, 3, and 4), the valley crosses the Echo Cliffs ridge, formed by the resistant Navajo Sandstone upturned along the Echo Cliffs monocline. This large erosional gap is visible today in cross section. The present rim-to-rim width of the gap is 3.4 km and its depth is 290 m, as measured on satellite images and topographic maps. There are no other breaks comparable in size to The Gap in the erosional escarpment of the Echo Cliffs for the ∼62 km from The Gap northward to the mouth of the Paria River at Lees Ferry. This shows that no drainages comparable to Crooked Ridge River crossed the escarpment in this section. Southward, the crest of the Echo Cliffs today is above or just below 1700 m (the lowest elevation seen on Crooked Ridge) for ∼12 km from The Gap, but gradient would have brought the level of the river to below 1700 m in this area. Therefore, the cliffs projected above the level of the river even then. In these 12 km, there is just one smaller break in the cliffs, which we interpret as being formed by a tributary to Crooked Ridge River. The conclusion is that Crooked Ridge River was the master stream for the entire Kaibito Plateau area.


Between Black Mesa and White Mesa, Klethla Valley (Fig. 2) crosses the ancient drainage at a low angle. This is a strike valley along the south continuation of the Tsegi–Comb Ridge system of monoclinal flexures, a part of a major structural trend that continues northeastward for ∼250 km along Comb Ridge and separates the Monument Upwarp to the northwest from the structurally low Black Mesa basin to the southeast (Figs. 2, 8, and 9).

On Kaibito Plateau, Crooked Ridge crosses several down-to-the-east monoclinal flexures (Fig. 8). West of the Echo Cliffs, the Kaibab Plateau is a north- and south-plunging dome (Figs. 1, 2, and 8) bounded on the east by the East Kaibab monocline, which also has down-to-the-east displacement. The Colorado River in eastern Grand Canyon forms a great bend that follows strike of the strata around the south-plunging part of the dome. The effect of the various monoclines is to expose strata successively lower in the section toward the west and downstream.

The remnant of fluvial deposits on Black Mesa rests on the Wepo Formation in the middle of the erosionally resistant Upper Cretaceous Mesaverde Group. The Wepo thus formed the bottom of the valley, whose sides are not preserved, but must have been composed of Upper Cretaceous rocks above the Wepo Formation. These strata, which are subhorizontal here, include the Yale Point Sandstone of the Mesaverde Formation (Page and Repenning, 1958; Cooley et al., 1969), which is at most 190 m thick on Black Mesa. Inasmuch as the ancient valley was much deeper than 190 m on the Kaibito Plateau, it is likely that the valley walls on Black Mesa, which were carved in resistant rocks, were also high, so would have been in Upper Cretaceous strata above the Yale Point, and possibly even in overlying Tertiary strata.

The sides of the old river valley between Black Mesa and White Mesa, in the area now occupied by Klethla Valley, were in the Mesaverde Group and the underlying Mancos Shale and Dakota Sandstone of Cretaceous age as well as the Morrison Formation and Entrada Sandstone of Jurassic age (Fig. 8).

White Mesa is capped today by the Entrada Sandstone and remnants of the Cretaceous Dakota Sandstone (Cooley et al., 1969; F. Peterson, 2010, personal commun.). The reach between White Mesa and The Gap is rimmed by the upper part of the Jurassic Navajo Sandstone and the Carmel Formation, capped locally by small remnants of Entrada. In Crooked Ridge River time, Jurassic and Cretaceous rocks most likely rimmed all these parts of the valley (see below). In summary, the valley of Crooked Ridge River was carved mostly in easily eroded and clay-rich Upper Cretaceous rocks.


A thickness of about 30 m of fluvial material is exposed on Crooked Ridge. Good exposures show that the bulk of the deposits consist of sand containing stringers and interbeds of gravel, mud, and clay. Gravel is more abundant near the top of several exposures. Channels <1 to ∼10 m wide are common (Fig. 7), as is cross bedding that shows southwest flow direction (noted by us and by Hunt, 1969). The sand is weakly indurated, fine to medium grained but locally medium to coarse grained, and subrounded to subangular. A small percentage of grains are fine, very well rounded and frosted, and probably derived from Mesozoic eolianites. The scarcity of eolian sand suggests that the Oligocene Chuska erg of Cather et al. (2008) either was never present in the area traversed by Crooked Ridge River, or was eroded by the time the river was active.

Composition, sampling methods, sampling locations, and inferred sources of gravel are shown in Tables 1 and 2 and Figure 2. Clasts are considered distinctive or non-distinctive according to level of confidence in assessing provenance; both locally derived and far-traveled or exotic clasts are present.

Most locally derived clasts are quartz sandstones from the upper Mesozoic formations that formed the valley sides on the Kaibito Plateau and Black Mesa, and upstream from Black Mesa (Harshbarger et al., 1958; Page and Repenning, 1958; Beaumont and Dixon, 1965; Cooley et al., 1969).

Sandstone clasts showing eolian cross bedding and rounded and frosted quartz grains are derived from the eolian parts of the Jurassic Entrada Sandstone and the Cow Springs Sandstone; sandstone clasts without eolian features are from the Cretaceous Dakota and Mesaverde units. Petrified wood is from the Jurassic Morrison Formation, and limy sandstone containing abundant bivalve fossils is from the Cretaceous Dakota Sandstone. These clasts are found throughout the areas examined by us on Kaibito Plateau and White Mesa; we did not note areas of higher concentrations.

Typically, the clasts are subangular to subrounded; maximum size generally is ∼30 cm, but a few boulders, which probably rolled in from the valley sides, reach 100 cm in diameter. These clasts are important for determining what formations were present on the valley sides in Crooked Ridge time.

Selected exotic clasts are shown in Figure 10. The exotic clasts are subrounded to rounded, and mostly 1–3 cm in diameter, but reach 10 cm. They include many types of volcanic, hypabyssal, plutonic, and metamorphic lithologies; petrographic characteristics link them with specific sources northeast of Crooked Ridge and Black Mesa at least as far as the San Juan Mountains (Table 2). Minettes came from the late Oligocene to early Miocene dikes and diatremes of the Navajo volcanic field, most likely from the Monument Valley section (Laughlin et al., 1986, and references therein) (Fig. 9). Clasts of granite, pegmatite, and perthitic microcline also probably came from intrusions in Monument Valley, where these lithologies are especially common as xenoliths.

The intermediate porphyries are strikingly similar to rocks reported in laccolithic centers of the Colorado Plateau, including the Abajo, Carrizo, Henry, La Plata, La Sal, and Ute Mountains (Fig. 9) (Friedman and Huffman, 1998, and references therein). We have no isotopic ages for these clasts, but suspect that the ones incorporated in the Crooked Ridge sediments are from the Laramide-age centers (Carrizo, Ute, La Plata) because these are along the northeasterly trend of the river. Nevertheless, future determinations would be useful to sharpen interpretations of paleodrainage relations. Emplacement depth of these Laramide and middle Tertiary intrusions is uncertain. Published estimates are 1.9–6 km below the surface at the time of intrusion (Ross, 1998), and ∼3 km (Jackson, 1998). Erosion of Upper Cretaceous and possibly lower Tertiary strata unroofed these distinctive porphyries in the Crooked Ridge River drainage basin and fed pebbles into the river.

All the distinctive clasts of metamorphic rocks can be matched with Proterozoic rocks mapped in the Needle Mountains (Figs. 1 and 9), where today they crop out at elevations as high as 4000 m. The quartz metaconglomerate and metawacke are comparable with some of the lithologies in the Vallecito Conglomerate and the Irving Formation, and the fine quartzofeldspathic gneiss matches the Twilight Gneiss (Larson and Cross, 1956; Barker, 1969). Felsic lava flows, welded tuffs, andesites, latites, and hydrothermal mineral deposits are widespread and abundant in the late Eocene to early Miocene San Juan volcanic field (Lipman et al., 1978). The distinctive clasts of these lithologies, as well as most if not all the non-distinctive clasts of the same type, are likely to have come from this region. The pebbles of crystal tuff and altered rhyolite and andesite are not likely to be reworked from older gravel deposits because the tuffs are friable and the altered rhyolites are soft. Pebbles of rhyolitic vitrophyre are not likely to be reworked from the Chinle and Morrison Formations because they are glassy and not devitrified as are those from the Mesozoic formations (Cadigan, 1972; Thordarson et al., 1972; Dodge, 1973). Notably absent from the samples collected at Crooked Ridge and Black Mesa are clasts of limestone similar to the Kaibab Formation on the Kaibab Plateau, red sandstone similar to that of Triassic rocks on the Colorado Plateau, and monchiquite, a volcanic rock common in the Hopi Buttes to the south (Figs. 1 and 2) and also present at Wildcat Peak, not far south of Crooked Ridge on the Kaibito Plateau (Fig. 3).


It is useful to estimate the height of the valley sides using the locally derived clasts in the fluvial sediments as described above. Because of structural complications, strata older than Cretaceous were not exposed upstream from White Mesa, so any clasts from these strata were necessarily derived from exposures in the valley sides on the Kaibito Plateau, White Mesa, or the monoclinal flexure now occupied by Klethla Valley (Fig. 8).

The predominant sandstone pebbles and cobbles are similar in lithology to those described from Middle and Upper Jurassic to Upper Cretaceous formations that crop out directly beneath or near the gravel deposits of Crooked Ridge on the Kaibito Plateau and on the flanks and top of Black Mesa (Harshbarger et al., 1958; Page and Repenning, 1958; Cooley et al., 1969). In particular, light gray, fine to coarse sand to granule sandstones that contain feldspar and mica probably were eroded from beds in the Mesaverde Group, and well-sorted fine sandstones with rounded and frosted grains compare well with the Cow Springs Sandstone (Table 2). Other specimens are comparable to rocks in the San Rafael Group, Morrison Formation, and Cretaceous Dakota Sandstone. Petrified wood is widely reported in the Morrison Formation, and is also known to occur in the Dakota Sandstone (Table 2).

The Dakota Sandstone, which is ∼60 m thick in this area, is preserved at the foot of the monocline along the western scarp of Black Mesa, where it typically overlies the Jurassic Morrison Formation (Fig. 11). On White Mesa, however, remnants of Dakota directly overlie the Cow Springs Sandstone because the Morrison has pinched out here (Beaumont and Dixon, 1965; Cooley et al., 1969) (Figs. 11 and 12). The Dakota at the foot of Black Mesa is 300 m below the level of Crooked Ridge River on White Mesa. Consequently, Dakota clasts could not have entered the river sediments from the Black Mesa area, but must have been derived from exposures on White Mesa and the Kaibito Plateau. This is confirmed by remnants of the Dakota at the east edge of White Mesa, where the Dakota directly overlies the Jurassic Cow Springs Sandstone (Cooley et al., 1969).

A good place to visualize the ancient river valley is on the Kaibito Plateau between Preston Mesa (Fig. 3) and two smaller mesas at the south edge of Mormon Ridges. All these mesas are capped by remnants of the Entrada Sandstone. There are no structural features here to complicate the picture.

In this area, Crooked Ridge River flowed at or near the top of the Jurassic Navajo Sandstone. This unit is overlain by the Jurassic Carmel Formation and then the Jurassic Entrada Sandstone, which together rise to a level ∼145 m above the river bed (Fig. 12).

The Entrada is overlain by the Jurassic Cow Springs Sandstone. Carmel, Entrada, and Cow Springs form the San Rafael Group, which is overlain by remnants of the Cretaceous Dakota Sandstone. According to Peterson and Pipiringos (1979), the thickness of the San Rafael Group at the north end of White Mesa is 270 m. Adding the ∼60 m of the Dakota to the 270 m of the San Rafael Group brings the height of the ancient valley sides to 330 m. However, this is a minimum figure because the probable Mesaverde clasts in the river deposits suggest that the Mancos and Mesaverde units were also present along the valley sides.

Cooley et al. (1969) gave a thickness of ∼450 m for the Cretaceous rocks on the nearby flank of Black Mesa, which brings the probable total height of the valley sides to ∼720 m (Fig. 12). These calculations are supported by the observation that the retreating Mancos-Mesaverde scarp is directly above the outcrop of the Dakota on all sides of Black Mesa. Furthermore, Holm (2001), Lucchitta (1975), and Lucchitta and Jeanne (2001) calculated rates of retreat for Mesozoic-rock scarps of 4–8 km/m.y. Even using a rate in the middle of this range (5 km/m.y.), and a very modest age of 10 Ma for Crooked Ridge River, we find that the Cretaceous rocks would have extended to well northwest and west of the Crooked Ridge River valley at that time, so would have been exposed on the valley sides. In this case, the sides of the valley would have been ∼720 m high, and the terrain surrounding the valley would have been at an elevation of ∼2600 m. The denudation implied by these figures since Crooked Ridge time is consonant with the denudation in the Monument Upwarp area since late Miocene time proposed by Hoffman et al. (2011) on the basis of apatite thermochronology.

Clearly, the valley of Crooked Ridge River would have been substantial, and the time required to carve it considerable. Furthermore, the elevation of the area surrounding the river, probably a regional plateau, would have been higher than the surfaces discussed by Cooley (1962; Cooley et al., 1969) and R. Hereford (2013, personal commun.). These lower surfaces therefore should be younger than Crooked Ridge River and would have formed after the river became inactive.


Crooked Ridge River is typical of rivers for which it is difficult, if not impossible, to determine with reasonable certainty when and how the river came into being. The best that can be done is to estimate when the river might have ceased to exist. This is because clasts (including detrital zircons) of some specific age may date the deposit that contains the clasts but not the river itself, which may have existed long before these clasts became available to the river.

The clast assemblage of Crooked Ridge River includes a variety of rock types from the San Juan volcanic field, which is of late Eocene to early Miocene age. The youngest detrital zircons described by Price et al. (2012) are 24–20 Ma, which fits into this age range. These zircons clearly were derived from the San Juan volcanic field.

The clasts we sampled and the zircons sampled by Price et al. (2012) were all obtained from the fluvial sediments that are exposed today. These sediments are the last material deposited by the river before its demise. Therefore, the demise (but not the birth) of the river could have happened at any time in the interval between 24 and 0 Ma because clasts can only be incorporated into river deposits after the source rocks for the clasts have been emplaced and exposed to erosion.

The massive calcrete that caps the deposits can constrain the time of demise. Massive calcretes like this one often take several million years to form. U-Pb dating of the calcrete would provide this age, but such a determination has not been made yet. Therefore, we must resort to indirect and imprecise methods.

The course of the river has no relation to the present drainage network. Instead, three distinct episodes of drainage arrangement have occurred since Crooked Ridge River time. The first is the development of the ancestral San Juan River drainage system along the south flank of the San Juan Mountains. This probably is what brought about the demise of Crooked Ridge River by beheading it and pirating its waters (see below). The second is the development of Klethla Valley (Figs. 2 and 8), which cuts across the course of the ancient river. Klethla Valley is broad and mature and typical of many such old valleys on the Colorado Plateau. This valley in turn is being beheaded by canyons of the canyon-cutting phase that are tributary to the Colorado and current San Juan Rivers. These three major rearrangements of the drainage network since Crooked Ridge time require an indeterminate, but surely substantial, time interval.

Another indirect method involves Wildcat Peak (Fig. 3), a monchiquite intrusive that is part of the Tuba volcanic field (Akers et al., 1971). Wildcat Peak has the same lithology as volcanic rocks in the Hopi Buttes, which have been dated at 6–8.5 Ma (Damon and Spencer, 2001). Many of these volcanoes are composed of tuff breccias capped by lava flows. Recently, Wildcat Peak itself has been dated at 19.5 ± 0.10 Ma (Peters, 2011).

The present-day exposure at Wildcat Peak consists of the roots of a volcano that are made up of a monchiquite and tuff-breccia neck and a dike swarm of monchiquite. The neck reaches an elevation of ∼2000 m. Delaney and Pollard (1981) suggested a minimum of 750 m of overburden for Ship Rock, another eroded volcano in nearby New Mexico that is now preserved as a neck-and-dike complex. McGetchin and Silver (1972) estimated 1000 m of overburden for Moses Rock dike near Monument Valley. Adding 1000 m to the top of Wildcat Peak would bring the volcano to an elevation of ∼3000 m, which is in decent agreement with our estimate of the elevation of the Crooked Ridge River valley sides. This is also well above the 1950–2000 m elevation of the Crooked Ridge River thalweg nearest Wildcat Peak. Therefore, one can infer substantial topographic relief between Wildcat Peak volcano and the valley of Crooked Ridge River, which probably resulted in a slope toward the valley. Furthermore, by analogy with monchiquite necks and lava flows in the Hopi Buttes, the volcanic edifice at the surface probably was much more extensive than the roots that are preserved today (Williams, 1936, his figure 4), thus providing abundant material to be eroded and transported toward Crooked Ridge River.

No monchiquite clasts have been found in the gravel even though Wildcat Peak is less than 13 km from Crooked Ridge. We cannot exclude that the clasts are indeed present in the gravel but were just not seen by us, in spite of our careful examination. Nevertheless, we can reasonably infer that the Crooked Ridge River drainage system became inactive before the intrusive was emplaced or unroofed in earliest Miocene time.

Removal of ∼1 km of strata from the area of Wildcat Peak is consonant with removal of 1–2 km of strata from the area of the laccolithic intrusives since they were first unroofed. Such deep erosion is in agreement with results obtained by other techniques (e.g., Hoffman et al., 2011; Pederson et al., 2013), and presumably happened over a substantial time interval.

On the basis of these arguments, we believe that the demise of Crooked Ridge River can plausibly be assigned to the interval between 24 Ma, or possibly 20 Ma (zircon), and 19 Ma (Wildcat Peak), or early Miocene. The river itself would be older.

To arrive at a time when the river might have been born we must add to the time of demise the time needed to carve a valley many hundreds of meters deep, starting from an original position on top of Cretaceous (and possibly lower Cenozoic) strata. This must have taken a considerable time, but we have no constraints on just how long this time might have been. It is reasonable to postulate that the river was born while the San Juan volcanic field was active between late Eocene and early Miocene time because the high elevation of these mountains would have started to provide abundant precipitation to feed this and other rivers.

R. Hereford (2011 and 2013, personal commun.) has suggested that Crooked Ridge River was not a major river but a young local wash that obtained its exotic clasts by recycling them from a pediment or alluvial apron that extended over a large area south of the San Juan volcanic field. We do not believe this notion is tenable for the following reasons:

(1) The Crooked Ridge deposits contain 24–20 Ma detrital zircon. The pediment deposits, proposed by Cather et al. (2003, 2008) as part of the Chuska units, were deposited ca. 35–33.5 Ma, and the youngest of the Chuska units is 33.5–25.2 Ma, while the volcanic Conejos Formation (in the southern extension of the San Juan Mountains), allegedly the source of the clasts in the pediment, is ca. 35–30 Ma. Therefore, deposition of all these units ended well before 24–20 Ma, the age of the detrital zircons in Crooked Ridge River, so the zircons, and presumably the fluvial deposits that contain them, could not be derived from the Chuska material.

(2) The interpretation of a large pediment is based on a few exposures of sediments, all on the Chuska Mountains, that are interpreted to be derived from the Conejos Formation, some 200 km away. However, the Conejos Formation is mineralogically and petrologically different from the “pediment” clasts in the Chuskas, so the correlation does not hold.

(3) The “pediment” deposits are nowhere preserved outside the Chuska Mountains, they are not documented, nor have they been shown to have ever existed.

(4) The Crooked Ridge clasts contain lithologies that would not survive recycling.

(5) The extensive Crooked Ridge clast assemblage is composed of rocks exposed well and extensively to the northeast and north of Crooked Ridge. It is more logical and likely that this assemblage was derived directly from the source areas rather than circuitously from an undocumented pediment deposit.

In conclusion, the age of Crooked Ridge River is not known with any degree of precision. All we know is that the demise of the river occurred sometime between 24 and 0 Ma. There are no data to tell us when the river came into being. However, the circumstantial evidence discussed above leads us to think that the river is of the same age as the San Juan volcanic field, i.e., late Eocene to early Miocene, and that the demise of the river occurred between ca. 24 and 19 Ma.


The farthest upstream exposure of river deposits is on Black Mesa at 2240 m elevation; the farthest downstream exposure is near The Gap at ∼1700 m. Over the 91 km that Crooked Ridge River can be traced from Black Mesa to The Gap, the river deposits drop 530 m, giving an average present-day gradient of 5.8 m/km. The gradient of individual reaches varies considerably (Fig. 5), probably reflecting differences in the rocks into which the valley was carved, structural features such as monoclines and faults, and constrictions such as The Gap in the river’s path. The relatively high gradient can plausibly be ascribed to overloading by sediment derived from the easily eroded Jurassic and Cretaceous rocks that formed the valley sides upstream from The Gap. Internal structures of the river sediments and the upward coarsening suggest a braided stream and support the overloaded-stream interpretation.

An alternative explanation for the steep gradient of the deposits is post-depositional tilting of the channel due to crustal warping such as a “bullseye” of isostatic unloading to the north (Lazear et al., 2011; Pederson et al., 2013) or mantle dynamics (Robert et al., 2011; Moucha et al., 2009). However, the inferred bullseye is in the Canyonlands country north to northwest of Crooked Ridge River, so the river’s southwesterly course is essentially tangential to the uplift contours and should be little affected by this unloading.

The mantle-dynamics studies suggest northeast tilting for nearly all the Miocene and southwest tilting since then. Therefore, the river would have been tilted first northeast and then southwest. It is currently not possible to evaluate the net effect of these postulated tilting events on the gradient of Crooked Ridge River.

The erosional history of the Colorado Plateau in the eastern Grand Canyon region shows that erosion has progressed northeast with time, mostly through cliff retreat at 4–8 km/m.y. (Lucchitta, 1975; Holm, 2001; Lucchitta and Jeanne, 2001). As a result, Paleozoic rocks are exposed near the south margin of the Colorado Plateau, whereas uppermost Mesozoic strata are exposed in the inner part of the plateau to the northeast. This erosion would cause northeast tilting of the Kaibito Plateau region because of isostatic uplift of the denuded region to the southwest, and may have decreased the original gradient of Crooked Ridge River. In conclusion, the steep gradient of the river deposits is plausibly explained by overloading, although tectonic adjustments can be neither proved nor ruled out at this time.


It is useful to compare the channel characteristics and gradient of Crooked Ridge River with those of present-day rivers that are overloaded and have substantial base flow. This has a bearing on whether the gradient of Crooked Ridge River was primary or not, on whether the river was overloaded and braided, and on whether the river was a substantial river or merely a local wash. Two suitable streams in the region are a reach of the Animas River upstream from Durango, and the San Juan River upstream from Pagosa Springs, both in southwest Colorado. Data for these streams and for Crooked Ridge River are in Table 3.

The comparison shows that a steep primary gradient for the overloaded Crooked Ridge River is not unreasonable, and indicates that the overall channel and floodplain characteristics are in keeping with those of two substantial and overloaded rivers of the region. The comparison, together with the exotic clasts, are also permissive with the proposition that the ancient Crooked Ridge River was not a minor local wash, but a river of regional extent whose discharge was at least comparable to those of today’s Animas and upper San Juan Rivers.

It could be argued that comparison with the two streams in Table 3 is not valid because these streams are near their headwaters. However, proximity to headwaters is not necessarily correlated with overloading and a braided channel. Some streams near the headwaters are overloaded and braided, whereas others are not. However, overloading always results in a steep gradient and a braided channel. The effect of overloading even far from headwater regions is well illustrated by two rivers in northeast Italy, the Tagliamento (46°3.562′N, 12°54.886′E), and the Piave (45°51.191′N, 12°10.303′E). Both rivers emerge from the rugged Southern Alps and flow southward on a coastal plain to their mouths in the Adriatic Sea. The Alps contribute great quantities of carbonate debris to the rivers, which are overloaded and highly braided even on the coastal plain and at considerable distances from the headwaters (60–65 km from where the Tagliamento emerges from the mountains, 45 km in case of the Piave). The gradients of the rivers over tens of kilometers are 3.4 and 3.3 m/km, respectively.


The relative scarcity of porphyry clasts in Crooked Ridge gravel contrasts markedly with their abundance in even the oldest terraces of the Colorado River. We infer that the laccolithic intrusives were less exposed when Crooked Ridge River was active than today and that some were not exposed at all (Eckel et al., 1949). Today, a few of these porphyries are exposed at altitudes as high as 4000 m, and many are at 3000–3500 m; in Crooked Ridge time, only the highest would have been exposed, as implied by the scarcity of these clasts in the gravel. This suggests a topographic surface in the region of the intrusives in the 3500+ m range. Now, the region in southwest Colorado and southeast Utah near the intrusives is at 1500–2000 m, suggesting that 1–2 km of strata has been removed since Crooked Ridge time.


The middle Miocene to lowermost Pliocene(?) Bidahochi Formation figures prominently in hypotheses on the history and integration of the Colorado River on the Colorado Plateau. The formation was deposited in Bidahochi Basin, roughly centered on the Hopi Buttes (Figs. 1 and 2). Origin of the formation is controversial. Some (e.g., Spencer et al., 2008) proposed that it was deposited in a deep “Hopi Lake,” possibly large enough to extend into southern Utah. The lake would have been the sump of a large drainage system that crossed the Colorado Plateau and originated in the Snake River basin. These interpretations are based on the presence of fossil fish of Snake River affinities and adapted to large permanent aquatic habitats and swift-flowing rivers. The lake would eventually have spilled over into a drainage along the present eastern Grand Canyon, initiating a top-down cascade of spillovers that integrated the Colorado River in its present course through Grand Canyon and the lower Colorado region (e.g., Meek and Douglass, 2001; Scarborough, 2001).

Others (e.g., Ort et al., 1998; Dallege et al., 2001; Dickinson, 2011) argued that the formation was not deposited in one deep lake, but rather in shallow and ephemeral ones that at times probably were playas or marshes and were filled by fluvial aggradation.

Crooked Ridge River provides useful constraints on the Bidahochi Formation issue. The southwesterly course of the river from the San Juan Mountains to the Kaibab upwarp makes it impossible for rivers such as an ancestral Colorado to flow southward from western Colorado or eastern Utah into “Hopi Lake” or, conversely, northward from “Hopi Lake” (Fig. 13).

As the current evaluations of tilting are speculative, we use the 1700 m altitude of Crooked Ridge River sediments near The Gap as a preliminary constraint on possible continuations of Crooked Ridge River downstream, where no deposits are preserved (Fig. 13). Accordingly, the Bidahochi Basin is part of the excluded terrain, because the base of the Bidahochi Formation is at 1750–1700 m (Love, 1989; Cather et al., 2008; Dickinson, 2011). Thus, it is unlikely that Crooked Ridge River would have reached and filled the hypothetical Hopi Lake even in the absence of any gradient between The Gap and the Hopi Buttes area.

An alternative explanation is that Crooked Ridge River is younger than the basal Bidahochi Formation because its deposits at The Gap are lower than those beds. Thus, the river could have been graded to a fluvial system that included the ca. 4–5 Ma upper Bidahochi fluvial deposits and could have been of that age. We consider this possibility extremely unlikely in light of the arguments presented above that point to a much greater age for Crooked Ridge.

Figure 13 shows that the only possible continuation beyond The Gap is southward along the alignment of the Echo Cliffs, then westward to near the present-day confluence of the Colorado and Little Colorado Rivers. From this point, the river could flow either north along the alignment of the present Marble Canyon, or west along the alignment of the present eastern Grand Canyon.

If the ancient river flowed north, possibly to a hypothetical and undocumented “Glen Lake” (Hill and Ranney, 2008), its current average gradient would place it at an elevation of 920 m at Lees Ferry (Fig. 2), which is below even the present Colorado River elevation there. If instead the gradient had been only 3 m/km, the river would have been ∼350 m above present river grade. However, the present Colorado River and its valley are much lower than they were in Crooked Ridge time because the Colorado and the bedrock underneath it were ∼200 m higher just 525–600 k.y. ago (Lucchitta et al., 2000, 2001). This gives an incision rate of 380–330 m/m.y. Therefore, even at 1 Ma the Colorado River grade would have been at, or higher than, the elevation of Crooked Ridge River at The Gap. The incision is the product of vigorous post–5 Ma downcutting by the Colorado River. Projecting the rate over 5 m.y. would bring river and bedrock elevation at Lees Ferry to 1600–1900 m above the present 940 m river grade. This would place them at ∼1 km above the elevation of Crooked Ridge River at The Gap. Neither distant nor more local warping is likely to overcome such a topographic disparity.

We conclude that Crooked Ridge River was unlikely to have flowed northward from the area of The Gap. On the other hand, our data do not preclude an ancient river such as the ancestral Colorado River flowing southwestward along approximately its present course but at a higher elevation, to join Crooked Ridge River in the area of the present-day confluence with the Little Colorado River.

A course westward is possible along the alignment of the present eastern Grand Canyon (Figs. 1, 13, and 14). This potential route was proposed long ago on geologic grounds by Babenroth and Strahler (1945) and Lucchitta (1975, 1989) and, more recently, by Scarborough (2001). Flowers et al. (2008), Lee et al. (2013), and Flowers and Farley (2012) have confirmed these findings using apatite thermochronology.

The greater width and complexity of eastern Grand Canyon supports an older age than for other parts of the canyon. Most likely, the old course was in a broad valley that followed the curving strike of strata around the south-plunging part of the Kaibab arch. Valley rims at the time were in Mesozoic rocks and the floor was incised some hundreds of meters below the top of the Kaibab Limestone.

Regarding the continuation of Crooked Ridge–Colorado River beyond the Kaibab Plateau, a course along western Grand Canyon and the Lake Mead area is precluded by widespread and well-documented Miocene interior-basin deposits at the mouth of the canyon (Longwell, 1936; Lucchitta, 1966, 1967, 1972, 1989, 2013). We offer instead the hypothesis, first proposed by Lucchitta (1975), that the river flowed northwestward (Fig. 13) to somewhere in the general vicinity of Hurricane and St. George, Utah. Possible continuations are unknown at this time and purely speculative. A possible constraint is given by the fossil fish of Snake River affinities that are found in the Bidahochi strata (Spencer et al., 2008). These fossil fish are found in the drainage basins of the Snake and Sacramento Rivers, suggesting a connection with either of these areas. The fish must have utilized the pre–Grand Canyon drainage network to reach the Bidahochi area.


It is difficult to reconcile the existence of a large Hopi Lake near (let alone across) the well-established Crooked Ridge River, which had cut down to a lower elevation than that of the lake and presumably had a well-developed regional drainage network that was tributary to it. On the other hand, shallow and ephemeral lakes and marshes, possibly produced by aggradation of a stream flowing through the area (Ort et al., 1998; Dallege et al., 2001; Dickinson, 2011) are compatible with the Crooked Ridge River data.

It is even more difficult to envision a sudden, catastrophic, and canyon-forming spillover of the lake in an area where the valley of Crooked Ridge River already existed. Even had a spillover taken place, the resulting flow would have been along the course of Crooked Ridge River (and paleo–Colorado River?), not along some new path. Consequently, a spillover could not be the cause of integration of the Colorado River in western Grand Canyon and the adjacent upper Lake Mead area.


Evidence documented herein supports the idea that the source of Crooked Ridge River was northeast of the Kaibito Plateau and Black Mesa. Salient data include: distinctive volcanic, hypabyssal, plutonic, and metamorphic pebbles in the gravel deposit; northeast projection of the river’s course; southwest gradient and current indicators; and late Oligocene to early Miocene detrital zircon sand grains. In the northeast direction, the San Juan Mountains are the most likely host of the headwaters (Fig. 9).

We can confidently identify pebbles of minette and diverse lithologies of hornblende-plagioclase porphyry (Figs. 8G and 8P) in the gravel. This proves that, at a minimum, the river system crossed the Four Corners area through the Navajo volcanic field and passed near one or more laccolith centers (Figs. 1 and 11). The straight-line distance from The Gap to the center of the San Juan Mountains is ∼400 km, and the distance to the Carrizo Mountains (the nearest center) is ∼225 km, so ∼56% of the course of the river can be confidently identified. The distances from The Gap to the Ute Mountains laccolith center (274 km) would increase the percentage to 72%.

Volcanism in the San Juan Mountains began ca. 36 Ma with construction of numerous large andesitic stratovolcanoes scattered across the nascent volcanic field (Lipman et al., 1970); this period of volcanism lasted 7–8 m.y. before large-volume, caldera-related ignimbrite eruptions greatly increased the size of the field ca. 29–26 Ma (Lipman, 2007). Post-caldera silicic magmatism continued to ca. 19 Ma (Lipman, 2007). In the Needle Mountains area (Figs. 1 and 9), the volcanic edifices were built on top of the Laramide San Juan uplift that was probably already more than 3 km high (Larson and Cross, 1956).

Even today the San Juans are a high and large mountain complex that reaches altitudes of more than 4300 m in the higher peaks. One of these peaks is Mount Sneffels, which is entirely composed of the hypabyssal porphyry of a subvolcanic stock. Therefore, the bulk of a large volcano would have overlain the present summit, bringing the altitude of the original volcano to maybe 7000–8000 m. P.W. Lipman (2011, personal commun.) felt that such altitudes for the original volcanic complex are about right.

A large and high mountain complex would certainly have given rise to many rivers flowing radially outward, as they do today. One or more of the rivers draining the southern and western slopes of the complex most likely flowed south or southwest; Crooked Ridge River would have been one of these streams and could have started collecting runoff as early as late Eocene to early Oligocene.


Erosion downstream from The Gap has removed all traces of Crooked Ridge River such as deposits, terraces, and channels. However, inferences can be made about possible river connections.

Aslan et al. (2011, 2012) reported Colorado River gravel buried beneath 11 Ma basalt at 2935 m on Grand Mesa, southwestern Colorado. Price et al. (2012) considered this gravel to be a deposit of the Gunnison River and not the Colorado River, but this difference is not important for this paper. Aslan et al. (2011, 2012) also reported gravel of the ancestral Gunnison and Uncompahgre Rivers at 2500–3000 m on the Uncompahgre Plateau in the same region. They proposed that the ancestral Colorado, Gunnison, and Uncompahgre Rivers joined near the northwest end of the Uncompahgre Plateau and then flowed west onto the Colorado Plateau ca. 11 Ma. The integration of these three rivers would have preceded 11 Ma by some period of time, possibly measured in millions of years.

The conclusion is that at ca. 11 Ma the Colorado River and its major tributaries in the Uncompahgre Plateau–Grand Mesa region were at elevations 1.6 km or more above the present grade of the Colorado River, and flowed onto the Colorado Plateau somewhere near the northwest end of the Uncompahgre Plateau. The further course of the paleo–Colorado River across the Colorado Plateau has not been identified, but one possibility is that it was approximately along the present southwesterly course on the west side of the Monument Upwarp (see also Cooley and Davidson, 1963). The river would have been at a considerably higher elevation along this course than the present elevation, but lower overall than the smaller and shorter Crooked Ridge River, which flowed on the east side of the upwarp. Crooked Ridge River could have been a tributary to this paleo–Colorado River and would have joined the master stream somewhere west or south of The Gap, possibly near the present confluence with the Little Colorado River (Figs. 2, 13, and 14). Both Crooked Ridge River and the paleo-Colorado River would have obtained their water from the very high terrain formed by the superposition of the San Juan volcanic mountains on the San Juan uplift.


Today, the course of Crooked Ridge River is truncated on the northeast by the valley of the San Juan River, showing that the developing proto–San Juan River was younger than Crooked Ridge River, whose waters it captured. The beheading and capture could only have happened if the San Juan were lower in elevation and more vigorous than the other streams, such as Crooked Ridge River, issuing from the south slope of the San Juan Mountains. This means that either the San Juan was the master stream of the region at the time, or that it was tributary to the master stream.

We propose that the proto–San Juan River came into being when the topographic surface of the region was still underlain by the upper part of the Mesozoic section or possibly even by early Tertiary rocks and was well above the present topographic surface. The proto–San Juan was likely a tributary to the proto–Colorado River, as was Crooked Ridge River. However, with time the proto–San Juan River became the master stream draining the southern San Juan Mountains, whereas Crooked Ridge River was captured and abandoned downstream from the capture point.

The San Juan River flows in a westerly direction through a low region bounded on the north by the Ute and La Plata Mountains, and on the south by the Carrizo Mountains (Fig. 9). The confluence of the San Juan River with the Colorado River is much upstream from where Crooked Ridge River once joined the proto–Colorado River in Marble Canyon. When the proto–San Juan River extended itself eastward from the confluence, it sequentially captured all the older south- and southwest-flowing drainages coming off the San Juan Mountains, including Crooked Ridge River. The San Juan River thus became the local master stream, and the added water allowed it to cut down vigorously. After capture, the downstream portion of Crooked Ridge River was left high and dry, forming the erosional remnant we see today.

Both Crooked Ridge River and the proto–Colorado River in this area would be of Oligocene to early Miocene age.


(1) Crooked Ridge River was a substantial stream of the region whose head was at least as far northeast as the San Juan Mountains. The demise of the river probably occurred in early Miocene time, between 24 and 19 Ma. Most of its history dates to before that time.

(2) The exotic clasts in the river deposits were not reworked from some pre-existing colluvial apron.

(3) Channel characteristics and gradient compare favorably with two present-day rivers in the region that have substantial base flow.

(4) The ancient river bore no relation to the present-day drainage network. On the contrary, three entirely different drainage configurations have developed since Crooked Ridge time.

(5) The southwesterly course of the river from the San Juan Mountains to the Kaibab upwarp makes it impossible for other rivers such as an ancestral Colorado to have flowed southward across Crooked Ridge River into Hopi Lake, or northward from the lake into western Colorado or eastern Utah.

(6) The river could not have flowed northward along the present course of Marble Canyon.

(7) A southwest- or south-flowing river (ancestral Colorado?), topographically much higher than the present Colorado River but lower than Crooked Ridge River, could have joined Crooked Ridge River somewhere near the present mouth of the Little Colorado River.

(8) A westward continuation of the river(s) along the alignment of the present eastern Grand Canyon around the Kaibab arch is possible and our favored alternative.

(9) The demise of Crooked Ridge River was caused by beheading and capture by the ancestral San Juan River as this river expanded its drainage basin eastward along the south flank of the San Juan Mountains.

(10) It is unlikely that a lake spillover would have integrated the Colorado River through western Grand Canyon because an older and well-developed river system was already in place that did not flow through western Grand Canyon into the upper Lake Mead area.

(11) After the river became inactive, the Four Corners region has been lowered erosionally by 1–2 km.

We thank Sue Beard, Richard Hereford, Fred Peterson, and Marith Reheis of the USGS, as well as two anonymous reviewers, for their helpful analyses. Their perceptive and occasionally pithy comments have caused us to re-examine and strengthen our data and conclusions with care, and to clarify our writing, which evidently was opaque in places. We also wish to thank Luke Blair and Trent Hare of the USGS for their indispensable help with the graphics. Finally, we must recognize gratefully those grand old masters, from J.W. Powell to Edwin D. McKee, who constructed the field upon which our games are now played.