Newly discovered lacustrine strata suggest that the most significant episode of stream capture in the upper Colorado River system (western USA), namely the abandonment of Unaweep Canyon, probably involved a combination of headward erosion and lake spillover. The abandonment of Unaweep Canyon occurred in two stages. The first stage was marked by the capture of the Colorado River, after which time the Gunnison River continued to incise. Continued incision by the Gunnison River created a wall of rock on the east side of the Gunnison River valley in Cactus Park and left the abandoned Colorado River bed well above the Gunnison River. The second stage involved two blockages, one created by the thick fill within Unaweep Canyon and one at the south end of Cactus Park, which led to the creation of a lake within Cactus Park. The lake level rose until it flooded the abandoned Colorado River bed and spilled over at the point where the Colorado River had been captured earlier, during the first stage of abandonment. Present-day East Creek was created by re-incision of the abandoned Colorado River course, which explains why the course of East Creek has a northeastward trend that is completely anomalous with respect to all other tributaries draining this area of the Uncompahgre Plateau. The rapid incision created a large quantity of debris that deflected the Gunnison River eastward at the mouth of East Creek. The evidence suggests that the abandonment of Cactus Park and Unaweep Canyon by the Gunnison River and the creation and destruction of Cactus Park lake all likely occurred ca. 800 ka and shortly thereafter. The Unaweep Canyon classic example of stream piracy illustrates how piracy alone can dramatically influence landscape development even in the absence of significant tectonic and climatic influences.


One of the outstanding questions regarding the evolution of the Colorado River (western USA) is the degree to which the river system has developed as a result of progressive headward erosion (bottom up) rather than through catastrophic lake spillover (top down) (House, 2008; Pederson, 2008; Dickinson, 2013). Stream piracy is also a related aspect of river integration, although the conditions that lead to stream piracy are often poorly documented because of their low preservation potential. Unaweep Canyon is an outstanding example of stream piracy in the upper Colorado River (Hunt, 1956). However, newly recognized lacustrine deposits suggest that the abandonment of Unaweep Canyon also involved one or more phases of lake spillover.

Unaweep Canyon is a spectacular wind gap that crosses the Uncompahgre Plateau in Mesa County, western Colorado (Fig. 1). The canyon is ∼42 km long, extending from Cactus Park (∼15 km southwest of Whitewater, Colorado) to ∼7 km northeast of Gateway, Colorado (Fig. 1). It has a maximum width of ∼1.5 km and is 1 km deep at the crest of the Uncompahgre Plateau. For most of its extent the canyon is incised into Precambrian igneous and metamorphic rocks. Debris flows and alluvial fans occupy much of the canyon floor and create a drainage divide, Unaweep Divide, within it. Two underfit streams, East Creek and West Creek, flow in opposite directions from near the divide. East Creek has excavated a landslide-filled canyon as it flows down the east side of the Uncompahgre Plateau to join the Gunnison River.

Cactus Park is a northwest-southeast–oriented valley that is tributary to Unaweep Canyon at its easternmost end (Fig. 1). Numerous small deposits of well-rounded gravel found within Cactus Park indicate that it was once occupied by a significant river. The valley is ∼7 km long and varies from 1 to 2 km wide. The eastern side of Cactus Park is a bluff ∼150 m high composed of Dakota Sandstone and Burro Canyon Formation sandstones overlying Morrison Formation mudstone and sandstone. The western side of the valley is a long irregular slope composed mainly of eastward-dipping Mesozoic sandstones. The northern three-fourths of Cactus Park is drained by Gibbler Creek, which flows into the valley from the west, then turns northwest and joins East Creek at the northern end of the valley. A drainage divide, informally called Cactus Park divide, is located a short distance south of where Gibbler Creek enters Cactus Park. South of this divide, the valley is drained by an unnamed ravine that is extending northward through headward erosion. An arc-shaped scallop that contains abundant landslide deposits occurs at the south end of the valley.

The purpose of this paper is to present new information that bears upon the events that caused abandonment of Unaweep Canyon and Cactus Park and processes that contributed to the reorganization of the upper Colorado River system.

Origin of Unaweep Canyon

The history of thought on the origin of Unaweep Canyon was recently reviewed in depth (Hood, 2011) and is only mentioned briefly here. The discussion of which river or rivers carved Unaweep Canyon began with the Hayden Survey of 1875, when Peale (1877) suggested that the Gunnison River carved the canyon and Gannett (1882) thought it was the Colorado River. Subsequent investigators have sided with either the Colorado River argument (Cater, 1955; Hunt, 1956, 1969; Lohman, 1961, 1965, 1981; Yeend, 1969; Betton et al., 2005; Scott et al., 2001, 2002a; Steven, 2002; Hood et al., 2002; Hood, 2011) or the Gunnison River interpretation (Dane, 1935; Cater, 1966; Sinnock, 1978, 1981; Perry, 1989; Perry and Annis, 1990; Perry and Young, 2003; Kaplan et al., 2005; Kaplan, 2006; Soreghan et al., 2007; Marra et al., 2010). A more recent controversy involves whether the canyon existed in late Paleozoic time (Soreghan et al., 2007, 2008, 2009a, 2009b; Hood, 2009; Hood et al., 2009). However, this controversy has no bearing on the abandonment of Unaweep Canyon.

Ideas Concerning the Abandonment of Unaweep Canyon

The earliest investigators thought that the Uncompahgre Plateau was recently uplifted and that the uplift caused abandonment of Unaweep Canyon (Gannett, 1882; Peale, 1877). Lohman (1961, 1965, 1981) presented a scenario for how stream capture might have happened. His idea was that the combined Colorado-Gunnison River originally occupied Unaweep Canyon. Uplift of the Uncompahgre Plateau first permitted capture of the Colorado River by its own tributary that was draining the northern end of the Uncompahgre Plateau, thereby removing the Colorado River from Unaweep Canyon. Continued uplift exceeded the Gunnison River’s incision rate in the canyon and allowed the river to be captured by a tributary of the Colorado River on the east side of the Uncompahgre Plateau. This resulted in the two rivers being rejoined in their modern configuration. Scott et al. (2002b) modified Lohman’s (1961, 1965, 1981) idea to take into account the presence of a Pleistocene lake in Cactus Park. In Hood (2011), evidence was presented for a two-step abandonment of the canyon, but the cause was not addressed.

The idea that uplift caused abandonment of Unaweep Canyon was first seriously called into question by Oesleby (1978, 1983, 2005a), who did geophysical studies that suggested that there is ∼300 m of fill in the canyon at Unaweep Divide. This study indicated that Unaweep Divide was not a structural feature caused by uplift, but rather was a post-abandonment alluvial fill feature. However, the geophysical evidence was ambiguous and the question remained open. Soreghan et al. (2007) drilled a hole in western Unaweep Canyon that penetrated 329 m of unconsolidated sediment before reaching bedrock, thereby confirming Oesleby’s (1978, 1983, 2005a) contention that there is thick fill in the canyon and showing that Cenozoic arching of the Uncompahgre Plateau has not been significant. This does not negate Lohman’s (1961, 1965, 1981) ideas about step-wise stream capture, but it does disprove the hypothesis that it was caused by late Cenozoic arching of the plateau. Because there is no evidence for significant uplift of the Uncompahgre Plateau, stream piracy remains the probable cause of abandonment.

Oesleby (2005b) presented evidence of a large landslide in the west end of Unaweep Canyon and hypothesized that this would have dammed the Gunnison River and caused a lake; he further suggested that when the lake level became high enough, the river could have overtopped a low divide within the Mancos Shale on the eastern side of the Uncompahgre Plateau and allowed the diversion of the Gunnison River to the north, its present direction of flow. Soreghan et al. (2007) reported the presence of lacustrine sediments in their drill hole in western Unaweep Canyon; they suggested that blockage of the river resulted in partial backfilling and ultimately to the abandonment of the canyon. Marra et al. (2010) suggested that abandonment likely occurred as a result of a landslide in the narrow westernmost end of the canyon. Balco et al. (2013) also proposed that a landslide at the western end of the canyon contributed to abandonment. There are reasons to question whether a landslide at the distal end of the canyon would have been the cause of abandonment. This will be addressed in the discussion section.

Previous Idea Concerning the Origin of East Creek

The only person to present a hypothesis about the origin of East Creek was Lohman (1981), whose hypothesis was based on his belief, common at the time, that the Uncompahgre Plateau was a recent uplift and was rising while the rivers flowed across it. Lohman (1981) thought that uplift resulted in abandonment of Unaweep Canyon first by the Colorado and then the Gunnison River. After the Gunnison River was captured, continued arching reversed the gradient of the abandoned river bed so that it sloped eastward from Unaweep Divide. Water flowing down the abandoned river bed created ancestral East Creek. Lohman (1981) suggested that at Cactus Park, the ancestral East Creek turned and flowed southeast through the valley and eventually joined the Gunnison River, reworking some of the gravel deposits in the process. Meanwhile, a small stream that occupied the abandoned bed of the Colorado River, which also would have been sloping to the east as a result of the arching of the Plateau, eventually captured ancestral East Creek, creating the course of modern East Creek.

Colorado River after Abandonment of Unaweep Canyon

Maroon Formation clasts, which are common in the Colorado River gravels but essentially absent in Gunnison River gravels, are present in some terrace gravels on the west side of the Uncompahgre Plateau (Hood, 2011). This provides evidence that the ancestral Colorado River was a component of the river that carved Unaweep Canyon. It entered the area from the northeast, joined with the ancestral Gunnison River at the mouth of Cactus Park (Fig. 2), and the combined river continued through Unaweep Canyon where it joined with the ancestral Dolores River at Gateway (Aslan et al., 2008). Terrace gravels in the Colorado River valley near Grand Junction show that the river has migrated northward as much as 6 km to reach its present position (Carrara, 2001; Scott et al., 2002a). The highest known Colorado River terrace remnants in this area are ∼170 m above the modern rivers. These are shown as red dots (not to scale) in Figure 2 (Scott et al., 2002a). These terraces have not been dated, but Berlin et al. (2008) dated a Colorado River terrace ∼170 m above the modern river near Rifle, Colorado, ∼100 km upstream, and obtained a cosmogenic burial age of ca. 1.8 Ma. If the two terraces are roughly correlative in age, it would provide a minimum age for the capture of the Colorado River.

Projection of the alignment of the 170 m Colorado River terrace remnants southeastward to the presumed path of the ancestral Colorado suggests that the capture point of the river by a “thief creek” might have been somewhere in the vicinity of the point marked x in Figure 2. Capture of the river anywhere near this point would have left an abandoned segment of the Colorado River bed sloping southwestward between the capture point and the confluence with the ancestral Gunnison River at Cactus Park. After the capture the ancestral Gunnison alone would have continued through Unaweep Canyon. Between the Colorado River and Gunnison River captures, the Colorado River would have continued to deepen its valley, giving perhaps as much as 300 m difference in elevation between it and the ancestral Gunnison River before the Gunnison was captured (Aslan et al., 2014).


Cactus Park Valley

Numerous gravel deposits on the east flank of the Uncompahgre Plateau show that an ancient river flowed northward and then turned westward where it joined the ancestral Colorado River at Unaweep Canyon (Aslan et al., 2008). The gravel deposits indicate that the river migrated northeastward and downward along the contact between the erosion-resistant Dakota Sandstone and the easily eroded Mancos Shale, moving down along the contact to keep pace with the incision of Unaweep Canyon (Aslan et al., 2008). When the river reached the position of Cactus Park, it began flowing on the Cactus Park–Bridgeport fault zone, which is oriented subparallel to the strike of the rocks (Fig. 3). We suggest that when it reached the fault zone, the river incised into the shattered rock of the fault zone and, once entrenched, could only cut downward rather than follow the dipping Dakota Sandstone–Mancos Shale contact as before. Continued incision eventually created a bluff capped by Dakota Sandstone and Burro Canyon Formation sandstones on the east side of Cactus Park valley. The top of the bluff rises as much as 150 m above the modern valley floor. At the south end of the valley an arc-shaped escarpment makes up the western side of the valley (Fig. 4). The base of the escarpment is in the Morrison Formation, which is prone to landslides, and abundant landslide deposits are present between the escarpment and the bottom of the valley.

Several exposures of well-rounded stream gravel are present within Cactus Park valley, ranging in elevation from 1880 to 1960 m. The lowest and most significant exposure is in an abandoned gravel pit 2 km south of the entrance to the valley. The gravel pit exposes well-rounded older stream gravel partly overlain by angular to subround, poorly sorted local gravel derived from nearby outcrops of Jurassic and Cretaceous strata. The older gravel contains interbeds of sand and is inset against Jurassic Entrada Sandstone. The well-rounded stream gravel has been interpreted to be Gunnison River gravel (Lohman, 1981). Modern Gunnison River gravels contain ∼67% of intermediate volcanic clasts, 17% metamorphic rocks of various types, ∼5% felsic volcanic rocks and granites, and the remaining 11% other rock types (Aslan et al., 2005). Our clast counts indicate that the gravel in the pit matches these values reasonably well, with 65% intermediate volcanic clasts, 14% metamorphic, 14% felsic volcanic and granites, and 7% other rock types. The age distribution of detrital zircons is also similar to the modern Gunnison River, adding more evidence that these gravels are ancestral Gunnison River gravels (Aslan et al., 2014). Caskey (2008) also examined several outcrops of Cactus Park gravel, comparing the composition of Cactus Park gravel to that of the Colorado, Uncompahgre, and Gunnison Rivers, and also concluded that the Cactus Park gravel originated in the Gunnison River. We interpret these gravel deposits to be the remnants of a downstepping set of terraces created as the Gunnison River eroded Cactus Park.

South of Cactus Park, other Gunnison River gravel deposits delineate the ancient river course. Deposits that appear to be in place occur on the top of Triangle Mesa, on the top and west flank of Star Mesa, and in what appears to be an ancient channel on Good Point ridge (Fig. 4). The base of the Gunnison River gravel on Triangle Mesa occurs on Dakota Sandstone at an elevation of 1907 m, about the same elevation as the gravel deposits at the south end of Cactus Park. The similarity in elevations indicates that the Triangle Mesa gravel and Cactus Park gravel are nearly contemporaneous and means that the river was flowing on what is now Triangle Mesa before entering Cactus Park. As shown in Figure 4, Big Dominguez Creek and the unnamed gully that drains southern Cactus Park make up the south and north sides of Triangle Mesa, with the modern Gunnison River making the third side of the triangle. It is likely that the ancestral river followed a path that was approximately coincident with these drainages. From Triangle Mesa and any distance eastward, the river would have been on Mancos Shale (Fig. 5) and would have been in a position where it could be captured by a tributary to the Colorado River.

At Star Mesa, gravel occurs on top of the mesa, down its western flank and at the western base, where it is in a saddle between Big and Little Dominguez Creeks (Fig. 4). This is a critical occurrence because the lowest gravel is at an elevation of 1850 m, 30 m below the lowest elevation of gravel in Cactus Park. We interpret this to mean that the ancestral Gunnison River continued to incise this area subsequent to its abandonment of Cactus Park, and therefore the capture point had to be between Star Mesa and the south end of Cactus Park. The gravel deposit at Good Point likely represents the position of the river south of Cactus Park at about the time the river abandoned Cactus Park. Our reason for suggesting this is as follows. The straight-line distance between the Gunnison gravel deposit at the north end of Cactus Park and the position of the ancient channel on Good Point to the south is 20.2 km. Add to this ∼7 km for the eastward excursion of the river at Triangle Mesa, and the river distance would be ∼27.2 km. The gradient of the modern Gunnison River in the canyons between Delta and Grand Junction, Colorado, is 1.15 m/km. Making the assumption that the former gradient of the Gunnison River when it was flowing through Cactus Park was similar to the modern gradient (Aslan et al., 2005) and calculating the elevation rise from the northernmost gravel deposit in Cactus Park, the elevation of the river at Good Point projects to ∼1910 m. This is only 10 m below the actual elevation (1920 m). Considering the uncertainties in the distance of the eastern excursion of the river and the paleogradient, this is a decent fit and leads us to conclude that this was the location of the channel while the river was flowing in Cactus Park, and allows the reconstruction of the ancestral Gunnison River for several kilometers south of Cactus Park, as shown in Figure 4.

Cactus Park Lacustrine Sediments

From Cactus Park divide southward for ∼6 km, a light yellow, thinly planar-bedded, clay-rich silt and clay deposit is exposed in ravines. A small amount of sand is dispersed within the finer grained material. Irregularly shaped carbonate nodules, gypsum nodules, and rare gypsum crystals are present in the weathered outcrop materials. This deposit directly overlies the river gravels where the two are juxtaposed in outcrop and is interpreted to be lacustrine in origin (Scott et al., 2002b; Aslan et al., 2008). Based on drill hole and outcrops, the maximum thickness of the lacustrine deposits is ∼67 m.

A drilling project by the Grand Junction Geological Society in 2006 was undertaken to sample these beds and the underlying gravel (Aslan et al., 2008). Two hollow-stem auger holes were drilled without reaching bedrock due to equipment limitations, but a third hole, slightly offsetting the first at 38.84235N, 108.45962W at a surface elevation of 1930 m, reached underlying river gravel and bedrock. Surficial material composed the uppermost 1 m, followed by lake beds to a depth of ∼50 m. River gravel was reached at a depth of 50.0 m, bedrock at 53.3 m, and total depth was 54.3 m. The uppermost 15.7 m of the deposit are oxidized to a light yellow color similar to the weathered outcrops, whereas the deeper clay-rich layers are dark gray (Fig. 6). A few rare silty layers are light gray. In physical appearance, the colors strongly resemble weathered and unweathered Mancos Shale. No megascopic plant or animal remains were found in either the cuttings or the core.

X-ray diffraction of the clay-size fraction shows that the clay mineralogy is dominated by mixed-layer illite/smectite with lesser amounts of kaolinite. A small amount of discrete smectite is present in a few samples. This assemblage is very similar to that of the Mancos Shale (Fig. 7) except for the discrete smectite, which was probably derived from the Brushy Basin Member of the Morrison Formation. Examination of the fine sand and/or silt residue from the clay study resulted in the discovery of Mancos Shale fossils, including rod-shaped Inoceramus(?) shell fragments and late Cretaceous foraminifera, including Heterohelix and possibly Archaeoglobigerina. Sparse pollen and coccolith fragments are also present.

The presence of a large Mancos Shale component in the lake sediments is consistent with a source from either the Gunnison River, which would have traversed a long stretch of Mancos Shale prior to reaching Cactus Park, or a source higher on the Uncompahgre Plateau west of Cactus Park. Several lines of evidence suggest that the Uncompahgre Plateau was the source, and not the Gunnison River. First, no river gravel has been found either within or above the lake beds. The lack of river gravels indicates that the Gunnison River did not occupy the valley during or after deposition of the lake sediments. Second, if the lake that deposited these clays and silts had been formed by damming the Gunnison River downstream from Cactus Park, the deposit should be sand and silt rich, at least near its base due to deposition by turbidity currents. This is not observed; instead, clay-rich layers directly overlie river gravel. Third, the heavy minerals recovered from lake sediments are largely distinct from those of the underlying stream deposits (Fig. 8; Schoepfer and Benage, 2008) and the geochemical composition of the lake beds is distinct from sediments carried by the modern Gunnison River (Fig. 9; Schoepfer and Benage, 2008). Taken together, all of these bits of evidence strongly suggest that the lake formed after the Gunnison River had completely abandoned Cactus Park and that the lacustrine sediments were sourced from the higher parts of the Uncompahgre Plateau to the west.

East Creek

From Cactus Park almost to its confluence with the Gunnison River, East Creek does not flow parallel to the dip of the Uncompahgre Plateau, but rather makes about a 30° angle to it (Fig. 1). Many streams on the eastern side of the Plateau make small divergences from the dip direction as they descend, but East Creek does it for >10 km. Another anomalous feature of East Creek is that from the point where it leaves Cactus Park to near its confluence with the Gunnison River, it is very much an underfit stream. No gauging stations exist on the creek, but for much of the year under present climatic conditions it is either dry or has a minimal flow. However, the canyon of this segment of East Creek suggests that the canyon was formed by a much larger stream. Where East Creek breaches the hill that makes up the eastern wall of Cactus Park (Fig. 2), the canyon is 1.4 km wide. It widens a bit downstream before narrowing sharply where it cuts through the Dakota Sandstone and Burro Canyon Formation near its confluence with the Gunnison River. At its deepest point the canyon is ∼200 m deep and the walls of the canyon and floor of the creek expose the Morrison Formation. The Morrison Formation is prone to slope failure and the interior of the canyon is covered with landslide debris for much of its length.


Landscape before East Creek

The following interpretation is based on the evidence that the combined Colorado-Gunnison River once flowed through Unaweep Canyon and that the Colorado River abandoned the canyon first (Scott et al., 2001; Aslan et al., 2008; Hood, 2011). This would have left a southwestward-sloping segment of the abandoned Colorado River bed between the capture point and the former confluence with the ancestral Gunnison River (Fig. 2). This is the opposite of modern East Creek, which flows steeply to the northeast. Residual gravel deposits indicate that prior to the abandonment of Cactus Park and Unaweep Canyon, the Gunnison River was flowing northward through Cactus Park valley to Unaweep Canyon, where it turned southwest and exited westward through the canyon. The highest lake bed that we have found is at an elevation of 1942 m, whereas the elevation of East Creek where it leaves Cactus Park is ∼1840 m, so East Creek could not have existed while Cactus Park lake was in existence. To determine what might have been in its place, we turn to evidence from the west side of the Uncompahgre Plateau.

Kaplan (2006) mapped terrace levels along West Creek after it leaves Unaweep Canyon on the west side of the Uncompahgre Plateau near Gateway, Colorado. Based on the presence of clasts of Maroon Formation that are present in the Colorado River but not in the Gunnison River, the highest terrace was interpreted (Hood, 2011) as having formed when the Colorado River was a component of the river that flowed through Unaweep Canyon. The lowest terrace has been interpreted to represent deposits of only the Gunnison River (Cater, 1966; Kaplan, 2006; Hood, 2011; Balco et al., 2013). The difference in elevation between these two terraces is more than 100 m. If the interpretation of the lowest terrace is correct, there was continued incision by the Gunnison River on the west side of the Uncompahgre Plateau after the Colorado River was captured and no longer present in Unaweep Canyon. If incision by the Gunnison River occurred on the western side of the Uncompahgre Plateau, it is reasonable to infer that incision would also have occurred on the eastern side in Cactus Park. The amount of incision that the ancestral Gunnison River made in Cactus Park and Unaweep Canyon after the Colorado River was captured is unknown, but the result of the continued incision would be that the abandoned bed of the Colorado River would be higher than the Gunnison River. The wall of rock that makes up the east side of Cactus Park would have extended across the area that is now East Creek. Any amount of incision greater than the ∼67 m maximum thickness of the lake beds plus a few meters to account for the depth of the lake would be enough to provide a wall of rock high enough to confine the lake. This is well below the amount of Gunnison River incision estimated from the western end of the canyon.

Creation of Cactus Park Lake

A satisfactory hypothesis to explain the presence of lake beds in Cactus Park must address several things. It must explain how the lake could have formed, given the modern presence of East Creek. It has to explain why the Cactus Park lake beds are unlike those of the western Unaweep Canyon lake (discussed herein). It must explain why the Gunnison River gravel deposit at Star Mesa is much lower than those in Cactus Park, which are downstream.

We have addressed the evidence that the segment of East Creek east of Cactus Park did not exist prior to the lake; now we address how the lake could have formed.

The simplest hypothesis would be that Unaweep Canyon was dammed by one or more landslides, creating a lake. This would be similar to modern Lake San Cristobal created by the Slumgullion Earthflow on the Lake Fork of the Gunnison River (Varnes and Savage, 1996). Certainly there is ample evidence of landslides and alluvial fan deposits within Unaweep Canyon, and Oesleby (1978, 1983, 2005a) presented evidence that the fill at Unaweep Divide is ∼300 m thick. Today’s elevation of the divide (2134 m) is more than adequate to account for the highest Cactus Park lake beds, which are at an elevation of ∼1940 m.

This hypothesis requires that the Cactus Park lake beds be sourced by the Gunnison River, which conflicts with the evidence presented earlier that the lake beds are not derived from the river and that the lake post-dates abandonment. A lake created in this fashion and deep enough to account for the thickness of the Cactus Park lake beds (∼67 m) would have extended at least 50 km south of Cactus Park. To date, lake beds have not been found south of Cactus Park, nor are there remnants of deltas where streams coming off of the Uncompahgre Plateau would have entered the lake. These points make it unlikely that a long-lasting Cactus Park lake was created by a landslide within Unaweep Canyon. It does not, however, preclude the idea that a short-lived lake could have been created and that overflow from such a lake could have resulted in diversion of the Gunnison River at a point south of Cactus Park.

A second hypothesis is based on the abundant landslide debris at the south end of Cactus Park (Fig. 4). Because there have been no lake beds found south of this area, it is reasonable to assume that this slope failure blocked the south end of Cactus Park valley and allowed the eventual formation of Cactus Park lake. A landslide in this area could also have blocked the Gunnison River and led to its diversion into its present position.

The hypothesis of a landslide at the south end of Cactus Park causing the diversion of the ancestral Gunnison River has two lines of evidence to support it. First, the eastward excursion of the river at Triangle Mesa would have put the river on the Dakota Sandstone–Mancos Shale contact (Fig. 5). At the time of capture, the deep valley of the modern Gunnison River did not exist. Everything lower than the ancestral Gunnison River was occupied by sedimentary rocks. Because the river would be the lowest elevation in the area, the area north of Triangle Mesa would have been composed of Mancos Shale at some unknown elevation higher than the river. Modern erosion has resulted in small drainages developing along the Dakota Sandstone–Mancos Shale contact parallel to the strike of the beds. It is very likely that some small tributary of the Colorado River was doing the same thing in the past, draining northwestward essentially parallel to the modern Gunnison River. We suggest that a short-lived lake created by a landslide at the south end of Cactus Park could have overtopped a low divide in the Mancos Shale near Triangle Mesa and overflowed into the Colorado tributary. Once this occurred, the spillway would have rapidly incised and diverted the river so that it would never again occupy Cactus Park. A second piece of evidence that supports the idea of the diversion taking place just south of Cactus Park is the gravel in the saddle on the west side of Star Mesa (Fig. 5). This gravel is ∼30 m lower than the lowest gravel in Cactus Park and indicates that the river was deepening its channel south of Cactus Park. A reasonable interpretation of this is that the river was diverted out of Cactus Park some place between the landslide at the south end of Cactus Park and Star Mesa. This makes the landslide at the south end of Cactus Park an attractive candidate as the cause of the diversion.

Neither of the hypotheses outlined above answers all the questions, especially concerning the river south of Cactus Park. If the gravels at Star Mesa and Good Point represent the position of the river when it was flowing through Cactus Park and for a time after it was captured, there had to be another, later capture much farther south to put the modern Gunnison River in its present position. Where and when did this occur? If, instead, the river south of Cactus Park had been flowing about in its present position (in map view) until it turned and entered Cactus Park and was then captured, how does one account for the gravel deposits at Star Mesa and Good Point Ridge? These questions are the subject of continued research.

The small unnamed gully that now drains the south end of Cactus Park also requires an explanation. If a landslide at the south end of Cactus Park created the blockage necessary to create Cactus Park lake, the abandoned Gunnison River channel between the landslide and the diversion point would have been the setting for a small lake. The channel would have been lower than the land on either side of it, and the elevation of the channel bottom would rise upstream from the landslide to the capture point. The modern river gradient is only 1.15 m/km, so this lake would not have been deep if the capture point was close to the position of the modern river. Seepage through the landslide, rainfall, and runoff from the adjacent hillside would supply the water in the lake. On those occasions when the lake was high enough to reach the capture point, the spillover would erode the escarpment in the same manner we propose for the origin of East Creek, but on a much smaller scale. The resulting drainage would trace the position of the ancestral Gunnison River back to the landslide. Eventually the gully breached the landslide and began eroding headward into Cactus Park, exposing the lake beds. The headward erosion continues today.

Demise of Cactus Park Lake and Creation of East Creek

The area where East Creek leaves the Unaweep Canyon–Cactus Park area and flows northeast to the Gunnison River is the most likely place that the Colorado River could have entered Unaweep Canyon. The shortest distance between Unaweep Canyon and DeBeque Canyon several kilometers to the northeast is a line that closely coincides with the course of East Creek (Fig. 2). We argue that the orientation and position of East Creek reflect the position of the ancestral Colorado River as seen in map view, as first discussed by Gannett (1882). East Creek’s position is simply the downward projection of the ancestral Colorado River onto the present topography.

We suggest that the process began several million years ago as the combined Colorado and Gunnison Rivers excavated Unaweep Canyon. Before the Colorado River was captured, it crossed a long stretch of Mancos Shale between DeBeque Canyon and Unaweep Canyon. The Gunnison River joined it in the vicinity of the future valley of Cactus Park (Fig. 10A). Subsequently the Colorado River was captured by a tributary stream that had developed around the nose of the Uncompahgre Plateau, as suggested by Lohman (1961, 1965, 1981). The escarpment created by this capture was probably significant, because there is ∼300 m elevation difference between the floor of Cactus Park and the oldest Colorado River terrace east of the Uncompahgre Plateau (Fig. 10B). The Gunnison River kept eroding its channel in Cactus Park and Unaweep Canyon, incising its bed below that of the abandoned Colorado River where modern East Creek now exits Unaweep Canyon. Meanwhile, the Colorado River also deepened its new valley (Fig. 10C). Subsequently, landslides or alluvial fans blocked Unaweep Canyon and the southern end of Cactus Park, resulting in the capture of the ancestral Gunnison River and creating Cactus Park lake. (The landslides and river capture were not necessarily simultaneous.) As the water level in Cactus Park lake rose, it eventually reached the level of the abandoned Colorado River valley. Continued rise in the lake level allowed the lake to extend up the abandoned Colorado River valley until it reached the point where the river had been diverted some time before. When the lake level reached this point, water spilled over the escarpment and eroded the underlying Mancos Shale (Fig. 10D). The spillover from the lake incised into the old channel, creating a new drainage that followed the position of the former Colorado River channel (Fig. 10E). As the lake drained, the spillway would have deepened, eventually incising through the sandstones of the Dakota Sandstone and Burro Canyon Formations and into the Brushy Basin Member of the Morrison Formation, which is rich in expandable clays (Fig. 10F). This set the stage for the widening of the valley, which has taken place largely by slope failure in the Brushy Basin Member. When the lake was finally drained, Gibbler Creek, which flowed into Cactus Park lake from the west, turned northward through Cactus Park and joined the small stream that drained Unaweep Canyon east of Unaweep Divide. The combined small creeks flowed down the path of the old spillway to the Gunnison River, thereby creating modern East Creek. Over the course of the millennia after the lake drained, erosion by Gibbler Creek would have removed the lake beds from the northernmost part of Cactus Park, if any were ever present.

This scenario highlights how lake spillover probably contributed to the drainage reorganization surrounding the abandonment of Unaweep Canyon, and explains the anomalous orientation of East Creek relative to other streams that flow down the dip slope of the Uncompahgre Plateau. Once the lake spillway had cut into the abandoned Colorado River valley, the position of East Creek was fixed and, unlike most streams that drain the Plateau, it does not follow northeast-trending dip slopes down the flank of the Plateau.

Draining Cactus Park lake and creating East Creek in the manner described also provides an explanation for a curious feature of the Gunnison River. East Creek joins the Gunnison River at the only place for many kilometers where the Gunnison River is not in a canyon. Northward from near Delta, Colorado, the Gunnison River is in a northwest-trending canyon that is approximately parallel to the axis of the Uncompahgre Plateau. When it approaches the confluence with East Creek, the river turns to the northeast, emerges from the canyon, and flows on Mancos Shale. A kilometer beyond the confluence, the river turns and flows in a west-southwest direction for 4 km, then turns northwest and enters another canyon, which continues until its confluence with the Colorado River to the north (see the area near Whitewater; Fig. 1). A possible reason for this northeast excursion of the Gunnison River is the development of East Creek. After the Gunnison River had been diverted out of Cactus Park, terrace remnants on either side of East Creek near Whitewater show that the river was flowing along the eastward-dipping contact between the Dakota Sandstone and the Mancos Shale. When the lake level reached the spill point and flowed down the escarpment of Mancos Shale, incision of the spillway would have been rapid because the Mancos Shale has very little resistance to erosion. Because of this, the lake overflow would have been carrying a large quantity of Mancos mud and would have quickly built an alluvial fan where it reached the Gunnison River floodplain. The development of the fan deflected the Gunnison River toward the northeast, farther out onto the Mancos Shale and away from the east-dipping Dakota Sandstone–Mancos Shale contact. Much later, when a drop in base level somewhere downstream caused the Gunnison River to entrench, the East Creek confluence was far enough to the east that it remained on the Mancos Shale, whereas most of the river was on Dakota Sandstone and became locked in a canyon. These steps are shown in Figure 11.

Triangle Mesa area

Our suggestion that the point of capture of the Gunnison River was between the south end of Cactus Park and Star Mesa depends on the eastward excursion of the ancestral Gunnison River to provide a gap in the ridge of rock that constitutes the eastern wall of Cactus Park. The gravel deposit atop Triangle Mesa at an elevation of 1907 m is of the proper elevation to be nearly contemporaneous with the southeasternmost gravels in Cactus Park (elevation ∼1911 m) and provides good evidence that the eastward excursion existed while the river was occupying Cactus Park. This eastward excursion was probably generated by some mass movement event that traveled down ancient Big Dominguez Creek. For such an event to deflect the river, it would have had to occur when the river was generally flowing along the Dakota Sandstone–Mancos Shale contact and before any significant incision into the sandstones that make up the eastern wall of Cactus Park had occurred. Otherwise, there would have been a sandstone wall on the east side of the river that would have prevented the deflection. As the river eroded downward and created the valley of Cactus Park, the eastern loop required that it erode downward through the Cretaceous sandstones. This began creating Triangle Mesa while leaving the easternmost part of the loop positioned about at the Dakota Sandstone–Mancos Shale contact (Fig. 5). This put the ancestral Gunnison River into a position where it could be captured by a tributary of the Colorado River.

One Lake or Two?

Soreghan et al. (2007) reported 156 m of lacustrine sediment in the lower part of a hole that they drilled to probe the depth of canyon fill in western Unaweep Canyon. Soreghan et al. (2007, p. 476) described the lacustrine sediments in their drill hole as follows: “Below this, and extending to ∼315 m, the core consists of green-gray (5GY 6/1) to yellow-gray (5Y 72) well-sorted fine sand and clay with abundant dark, macerated leafy and woody carbonaceous debris; the upper 120 m exhibits a large upwardly coarsening succession (UCS) comprising well-sorted medium-fine sand in <1 – 3 m beds with irregular laminations and rare bioturbation, yielding below to very fine sand and clay, olive gray (5Y 5/2 or 5Y 4/1) to pale brown (5YR 5/2) in color, with abundant plant fragments and a fetid odor.”

Soreghan et al. (2007) further reported that the sand-size grains contain locally significant volcanic rock fragments that indicate a Gunnison River provenance (Marra et al., 2008, 2010); this indicates that the Gunnison River was still flowing into the canyon at that time.

The description of the western Unaweep Canyon lake beds is nothing like the Cactus Park lake beds, adding to the evidence that the Gunnison River was not the source of Cactus Park lake. Moreover, Soreghan et al. (2007) described an ancient soil at the top of their lake beds. If the lakes were the same age, one would expect a similar paleosol, but there is no obvious evidence for this feature in Cactus Park.

The top of the western Unaweep lake beds is at ∼1825 m (Balco et al., 2013), whereas the top of the Cactus Park lake beds is at ∼1942 m or a little higher. That would mean that the dam that created western Unaweep lake would need to be 117 m higher than the highest lake beds in the drill hole in order for Cactus Park lake to be the same lake. There is no evidence for such a high landslide. In addition, Balco et al. (2103) presented cosmogenic dating evidence that the sedimentation in the western lake ceased by 1.34 ± 0.13 Ma. This is a 0.5 m.y. older than the burial age of the Cactus Park gravels (800 ± 240 ka, discussed herein; also see Aslan et al., 2014). Based on the evidence that the Gunnison River did not contribute to Cactus Park lake, whereas it sourced the western Unaweep lake, the improbability that the western Unaweep lake level could have been high enough to reach the top of the Cactus Park lake beds, and the considerable difference in apparent ages of the two lakes, we conclude that the two lakes were distinct and that the western Unaweep lake is older than the Cactus Park lake.

Did a Landslide at the Western End of Unaweep Canyon Cause Abandonment of the Canyon?

Balco et al. (2013), Marra et al. (2010), and Oesleby (2005b) suggested that the cause of abandonment was a landslide in the western part of Unaweep Canyon that backed water up and ultimately diverted Gunnison River flow into the Colorado River valley. This hypothesis has some problems. Balco et al. (2013) estimated that the surface elevation of the western Unaweep lake was 1825 m. To divert the river south of Cactus Park, the water elevation of the lake would have needed to be at least as high as the gravel atop Triangle Mesa (1907 m). This difference in elevation (82 m) suggests that the Gunnison River flowing at ∼1907 m at Triangle Mesa would have been entirely unaffected by the landslide tens of kilometers downstream. Oesleby (2005b) postulated a much higher landslide (∼1926–1950 m), but this has two problems. There is no evidence of a landslide anywhere near that elevation in western Unaweep Canyon. Oesleby’s drawing, presented in Aslan et al. (2008), postulates the landslide to be the one that formed western Unaweep lake. The elevation that Oesleby (2005b) gave is 100 m higher than the lake elevation reported by Balco et al. (2013). The second problem is that if there was a landslide in the western part of the canyon that backed water up through Cactus Park, the sediment that would have accumulated in Cactus Park should have been as least as coarse as that in the western Unaweep lake, as Cactus Park would have been ∼30 km upstream. The Cactus Park lake beds should also have had a Gunnison River provenance. Because neither of these characteristics describes the Cactus Park lake sediment, we conclude that a landslide at the western end of Unaweep Canyon did not cause the abandonment of the canyon.

When Were Unaweep Canyon and Cactus Park Abandoned?

Balco et al. (2013, p. 156) dated the western Unaweep lake beds as well as the lowest Gunnison River gravel just west of Unaweep Canyon, and concluded “…canyon blockage, initiation of sediment accumulation, and presumed river capture took place 1.41 ± 0.19 Ma. Lacustrine sedimentation ceased 1.34 ± 0.13 Ma.” Although these numbers appear reasonable for the age of the western Unaweep lake, they do not necessarily indicate the time of abandonment of the canyon by the Gunnison River. Balco et al. (2013) calculated that the duration of the lake was likely <100 k.y. and perhaps as few as 15 k.y. Blue Mesa Reservoir, upstream on the Gunnison River, filled to full pool depth of 100 m <3 yr after the dam was completed (data from U.S. Bureau of Reclamation website, www.usbr.gov). This is only ∼20 m less that the thickness of the western Unaweep lacustrine sediments (Balco et al., 2013) and suggests that the western lake would have filled with water in a few years. The ancestral Gunnison River had to continue flowing into and through the canyon for a significant period of time after the landslide dam occurred in order to deposit 156 m of lake sediment. It is plausible that given thousands and perhaps tens of thousands of years of water flowing over it, the dam would have been eroded below its original level and allowed the river to establish a channel in the lake beds in some location not intersected by the drill hole. This would have left lake sediments that were higher than the channel exposed to form the soil zones observed in the drill hole. The date obtained on the uppermost lake sediment certainly relates to the time the sediment was buried, but does not necessarily indicate the time of canyon abandonment. Likewise, the Balco et al. (2013) age estimate for the ancestral Gunnison gravel near Gateway does not necessarily indicate the time of abandonment. It is well known that rivers typically incise their channels downstream from dams (Williams and Wolman, 1984), and the same phenomenon would be expected downstream from the landslide dam that created the western Unaweep lake. A water well in the West Creek valley 6 km downstream from the gravel location sampled by Balco et al. (2013) did not reach bedrock until depths of 22 m (well information from Himes Drilling Company, Grand Junction, Colorado, 2004). Water coming over the dam would not transport gravel, so the river gravel deposit probably does not mark the time of abandonment.

One of the objectives of the Grand Junction Geological Society drilling project in Cactus Park was to collect a sample of buried river gravel for cosmogenic dating to determine a minimum age of abandonment of Unaweep Canyon and the maximum age of the lake beds. A quartz sample, purified from the underlying gravel, was submitted to the PRIME lab (Purdue Rare Isotope Measurement Laboratory, Purdue University) for 26Al/10Be burial dating. This gave an age of 920 ± 310 ka. A second sample, of the same material, was submitted later and returned an age of 620 ± 390 ka. The average of these two ages is 800 ± 240 ka (Aslan et al., 2014). Although the cosmogenic age indicates when the gravel was buried and not necessarily the age of abandonment by the river, there is some evidence that suggests that the lake beds closely post-date abandonment. About a meter of colluvium and wind-blown sand occurs on top of the lake beds. Similar material is very common throughout the Cactus Park area. If there had been a large gap in time between the abandonment of the valley and the deposition of the lake beds, there should be some colluvium and/or wind-blown sand at the base of the lake beds. This was not observed in the drill cuttings. Exposed terraces show some signs of reworking by local streams and include some local gravel, but the length of time represented by these features is unknown. One other line of evidence suggests that the time of abandonment was ca. 800 ka or a little earlier. There are Gunnison River gravels on the east flank of the Uncompahgre Plateau on either side of East Creek, the highest (and therefore oldest) of which are ∼124 m higher than the Gunnison River. Using the 142 m/m.y. regional long-term incision rate to approximate the age (Aslan et al., 2008), these deposits would be ca. 870 ka. This is a little older than the cosmogenic date for the burial of the Cactus Park gravel, but considering the uncertainties of the two methods the age estimates support the time of abandonment as close to 800 ka.

Because the Gunnison River was still flowing into the canyon at least until 1.34 ± 0.13 Ma, we consider this age to be the oldest date that the canyon would have been abandoned. Our evidence suggests that the latest it would have been abandoned was 800 ± 240 ka. Based on the evidence that there were two distinct lakes and that the Cactus Park lake appears to have formed soon after the canyon was abandoned, we suggest that it is probable that the 800 ka date is closer to the actual time of abandonment than the 1.34 Ma date.

If the 800 ± 240 ka is correct and Cactus Park lake formed soon after abandonment, there are two implications. The lake would have been in existence during a cold episode of the Pleistocene and could have existed in the present period of normal magnetic polarity (Fig. 12). An unpublished U.S. Geological Survey (USGS) study involved collecting oriented samples of the lowest, middle, and uppermost exposed lake beds and analyzing them at the USGS paleomagnetic laboratory (Denver, Colorado). All three samples showed normal magnetic polarity (R.B. Scott, 2003, personal commun.). This is consistent with the lake beds being 776 ka or younger, the age of the Matuyma-Brunhes magnetic reversal (Coe et al., 2004).


Abandonment of Unaweep Canyon was caused by stream piracy, but not necessarily in response to Pleistocene uplift of the Uncompahgre Plateau, as was once thought, or by a landslide dam that produced a lake filling the entirety of Unaweep Canyon. It is more likely that the stream piracy involved one or more small landslide dams and the creation of a localized lake, which ultimately drained via spillover. The abandonment and subsequent partial filling of Unaweep Canyon, the lake beds in Cactus Park, the creation of East Creek, and the deflection of a part of the Gunnison River are all interconnected and form a coherent story. The stepwise abandonment of Unaweep Canyon first by the Colorado River and later by the Gunnison River allowed the Gunnison River to incise well below the abandoned Colorado River bed. This left a wall of rock where East Creek now exists. The evidence suggests that the Gunnison River had abandoned Unaweep Canyon and Cactus Park by ca. 800 ka as a result of a landslide either within Unaweep Canyon or at the south end of Cactus Park. A second landslide at the other of these two places made it possible for a lake to form within Cactus Park. This lake covered the abandoned channel of the Gunnison River by ca. 776 ka. The lake level eventually reached the elevation of the abandoned channel of the Colorado River and that channel became the spillway of the lake. The spillway incised the abandoned channel, creating the course of modern East Creek as a downward projection of ancestral Colorado River. The Unaweep Canyon–Cactus Park area has a complex history involving stream captures, landslides, headward erosion, and lake spillover that have all combined to create the drainage pattern that we see today.

We thank Verner Johnson for his assistance with Figure 1, Mary Benage and Shane Schoepfer for helping with the collection of the mineralogical and geochemical data of the Cactus Park lake beds, and Robert Scott for permission to use his unpublished map of the south end of Cactus Park. The Grand Junction Geological Society plus several individual members of the society financed the drilling in Cactus Park. Darryl Granger of Purdue University conducted the cosmogenic analyses. Comments by two unidentified reviewers and Sue Beard helped the manuscript, and we thank them for their thoughtful comments.