A recent study posits that much of the 240-m-deep inner gorge of Grand Canyon was carved between 500 and 400 ka via passage of a migrating knickzone with incision rates of ∼1600 m/Ma during that time period; this was based on dating of a ca. 500 ka travertine deposit perched on the rim of the inner gorge, near Hermit Rapid, and a ca. 400 ka travertine drape that extends to within 60 m of river level nearby. However, a new U/Th age of 517 ± 13 ka on the same travertine drape challenges this model of a migrating knickzone and punctuated incision. The presence of ca. 500 ka travertine just 95 m above river level requires that most of the inner gorge was carved before that time. The resulting maximum bedrock incision rate of 230 m/Ma is consistent with independent results from sites up and downstream and with models for semi-steady Quaternary bedrock incision and dispels problems with the transient incision model. Downstream from the Hermit Rapid area, dikes present on both sides of the canyon have been used to support the migrating knickzone model. We report a new 40Ar/39Ar age of 517 ± 16 ka on one of these dikes, but argue that they don’t necessarily gauge incision.

Field observations suggest that the discontinuous travertine deposits, near Hermit Rapid, were deposited by springs that emanated from the Redwall-Muav aquifer, mantled the Tonto Platform, and locally built downwards into the inner gorge and tributary canyons. The range of U/Th ages from ca. 10–600 ka suggests these were long-lived spring systems. The travertine cements predominantly angular to subrounded locally derived clasts consistent with deposition on hillslopes and by tributaries. Well-rounded gravels are exceedingly rare but have been used to suggest that the Colorado River was at the rim of the inner gorge at ca. 500 ka. No exotic Colorado River clasts, derived from the area outside of Grand Canyon, were observed by us. In-place gravel from the main stem or tributaries (e.g., from paleo–Hermit Creek) within the travertine deposits can be reconciled with existing data, if: (1) travertine was deposited at ca. 2 Ma, which is approximately when the steady incision model suggests the inner gorge began to incise; (2) a 500 ka lava dam in the Lava Falls Rapid area, 140 km downstream, backed water and sediment up to the rim of the inner gorge in the Hermit area; or (3) regional climate-driven aggradation took place at 500 ka.


Recent efforts to better understand the controls on river profile form and incision of the Colorado River have focused on the processes and timing of river integration (Blackwelder, 1934; Longwell, 1936; Karlstrom et al., 2008; Polyak et al., 2008; Young, 2008; Cook et al., 2009; Pelletier, 2010; Hill and Polyak, 2014), the role of preexisting canyons (Flowers et al., 2008; Lee et al., 2011; Wernicke, 2011; Karlstrom et al., 2014), the effects of tectonics (Pederson et al., 2002; Karlstrom et al., 2007; Crow et al., 2014), regional denudation (Karlstrom et al., 2011; Pederson et al., 2013b), glacial cycles (Anders et al., 2005; Pederson et al., 2006), and substrate erodibility (Mackley, 2005; Cook et al., 2009; Pederson and Tressler, 2012; Pederson et al., 2013a; Bursztyn et al., 2015). Many of the models proposed by these studies suggest unique spatial and temporal patterns of incision and can be tested by quantifying variations in incision rates through time and space. Work to date has yielded incision constraints throughout much of Grand Canyon (Pederson et al., 2002; Pederson et al., 2006; Karlstrom et al., 2007; Pederson et al., 2013b, Crow et al., 2014; Abbott et al., 2015); these constraints were calculated primarily by dating of material associated with perched gravel. The geochronology used to calculate these rates comes primarily from 40Ar/39Ar dating of basalts and U/Th dating of travertine but also includes optically stimulated luminescence, U/Pb, cosmogenic burial, cosmogenic exposure, and 234U model ages. These data have been interpreted by some to suggest either temporally steady but spatially variable incision as a result of spatially differential uplift (Karlstrom et al., 2007; Karlstrom et al., 2008; Crow et al., 2014) or alternatively that a transient wave of incision passed through central Grand Canyon between 500 ka and 400 ka (Abbott et al., 2015; Abbott et al., 2016). The former models don’t discount knickzone migration, but suggest that it could only have occurred prior to 4 Ma in western Grand Canyon and prior to 650 ka in eastern Grand Canyon (Crow et al., 2014).

The hypothesis of 500–400 ka knickzone migration is based on a recent study that conducted U/Th dating on travertine deposits that occur at and below the rim of the inner gorge of Grand Canyon near Hermit Rapid (river mile [measured downstream from Lees Ferry (Stevens, 1983)] 95–96) (Abbott et al., 2015) (Fig. 1). The travertine occurs in a series of isolated deposits that rest primarily on a wide erosional bench in the Bright Angel Shale (Tonto Platform). That bench is supported by the resistant Tapeats Sandstone and the underlying Proterozoic basement rocks, which define the top of the sharp 240-m-deep inner gorge of Grand Canyon. Their conclusion of recent knickzone migration was based on ca. 600–500 ka dates on rim travertines that mantle what they interpreted to be side-stream gravel (paleo–Hermit Creek and perhaps some main-stem gravel) on the Tonto Platform. In contrast, inner-canyon travertine near the river level yielded a ca. 400 ka age. They concluded that incision rates had increased to ∼1600 m/Ma between 500 ka and 400 ka, then decreased to <210 m/Ma over the past 400 ka due to passage of a transient wave of incision. They also based their interpretation on ca. 500 ka basaltic dikes (40Ar/39Ar data presented below) present on both sides of the river to a height of 400–450 m above river level at river mile 159, the 159-mile dikes.

Crow et al. (2015a) suggested alternative explanations for the data including: (1) that the rim gravel deposition could have taken place much earlier and that ca. 600–500 ka U/Th ages record secondary travertine infillings; or (2) that the perched deposits, if due to side-stream aggradation, might have been a response to changes in local tributary base level associated with the formation of a lake behind a down-stream dam. Abbott et al. (2016) rejected the first hypothesis based on textural observations and the fact that they considered the dated travertine outcrops to be hydrologically isolated. They rejected the second hypothesis based on a lack of lacustrine deposits. In this paper, we will investigate these and other alternative hypotheses in the light of new U/Th dates from the same deposits and new geologic mapping. We conclude that the ages obtained by Abbott et al. (2015) are reliable but that their interpretations of the implication for Colorado River incision need revision. Our new geochronology and geologic mapping support semi-steady incision rather than passage of a migrating knickzone.


During the summers of 2008 and 2015, we collected a suite of samples from the Hermit travertine deposits, including deposits on the Tonto Platform and deposits that drape across the Tapeats Sandstone and extend into the inner gorge (Figs. 2 and 3). Our sampling was focused on the largest travertine mound perched on the Tonto platform (P1; see Figs. 2 and 3), where Abbott et al. (2015) obtained a 506 ± 33 ka U/Th date, and on a lower travertine drape that extends from near the Tonto Platform into the inner gorge to within 60 m of the river (Figs. 2 and 3; their I2 deposit or the Schist Camp drape), where Abbott et al. (2015) obtained a 394 ± 32 ka U/Th date from the toe of the deposit. We collected both travertine infillings around gravel clasts and detrital travertine clasts from these sites. Dating both infillings and detrital clasts from the upper (P1) deposits would bracket the age of gravel deposition and test whether the gravel was deposited at 500 ka as suggested by Abbott et al. (2015) or potentially earlier as suggested by us (Crow et al., 2015a). The lower travertine drape samples were collected to test the hypothesis that the Colorado River had already incised to its level (well into the inner gorge) by ca. 500 ka.

In 2016 and 2017, we re-mapped all the travertine deposits. The goals were to characterize each deposit, examine the internal bedding and travertine structures in each, and characterize the rounding and lithology of clasts within the travertine. We did not repeat the rappel described by Abbott et al. (2015), but we were able to walk completely around the deposits (including P1) to characterize interbedded gravels. The result was a more detailed geologic map that recognized new deposits and refined the geometry of previously mapped deposits (Fig. 3). We have visited the 159-mile dikes repeatedly; in 2006, the attitudes of dikes were measured, and the newly dated sample was collected in 2010.

Travertine samples were slabbed on a rock saw with a dedicated blade, and the slabs were washed in deionized water for ∼10 min in an ultrasonic bath. Discrete portions of the slabs were then subsampled with either a tungsten carbide dental bit or a diamond-impregnated microcore bit; in both cases, the material from the surface of the slab was discarded. In the case of the microcore bit, the core was etched with weak HNO3 before being crushed inside of plastic bags in an agate mortar and pestle; only a few of the resulting pieces were analyzed. Forty to 105 mg of material were then dissolved in HNO3 and mixed with a 229Th-233U-236U spike. U and Th were separated using conventional anion exchange chromatography. U and Th isotopes were measured using a Thermo Neptune multicollector–inductively coupled plasma mass spectrometer (MC-ICPMS) at the University of New Mexico; the instrument was optimized for U-series analytical work as described by Asmerom et al. (2006). 234U was measured on a secondary electron multiplier (EM) with high abundance filter or on the center Faraday cup, while the other isotopes of uranium were measured on Faraday cups. Mass fractionation was monitored using the 238U/235U ratio, while EM/Faraday gain was set using sample standard bracketing and uranium standard NBL-112. A similar procedure was used for Th-isotope measurements with 230Th measured in the EM or on the center Faraday cup, and 229Th and 232Th were measured in Faraday cups. An in-house 230Th standard was used to measure the EM/Faraday gain, and mass fractionation was corrected using 238U/235U or 236U/233U. All analyses used new half-lives for 234U and 230Th from Cheng et al. (2013).

Despite repeated field work at the 159-mile dikes site, unaltered material suitable for 40Ar/39Ar dating was not discovered until the summer of 2010. On both sides of the river, the dikes typically have thick cooling rinds and show evidence for alteration, including the replacement of olivine phenocrysts by “iddingsite” and green groundmass consistent with alteration to chlorite. Pristine material with unaltered olivine phenocrysts and groundmass has only been found ∼100 m above river level on the north side of the river. The dated sample was collected from a piece of angular float tens of meters below an outcrop of the dike. The angular morphology of the basalt piece and the lack of upstream sources preclude significant transport distances to the site. It seems likely that the sample came from a less altered outcrop of the dike inaccessible without rock climbing.

The 40Ar/39Ar sample was crushed in a disk mill. Phenocrysts were isolated from groundmass by magnetic susceptibility and then handpicked to remove remaining phenocrysts and any visibly altered grains. The resulting groundmass concentrate was treated in an ultrasonic bath for 40 min in 10% HCl and then rinsed multiple times in deionized H20. The treated groundmass concentrate was then loaded into 12-hole machined aluminum disks and irradiated for 1 h in the U.S. Geological Survey (USGS) TRIGA reactor, near Denver, Colorado. 40Ar/39Ar analysis using the step-heating age-spectrum method was conducted at the New Mexico Geochronology Research Laboratory with a Mass Analyzer 215-50 and a Synrad CO2 laser. Evolved sample gas was cleaned of reactive gases within an automated all metal extraction line using a combination of various getters and a “cold finger” operated at −140 °C. Fish Canyon Tuff sanidine was used as a neutron flux monitor with the assigned age of 28.201 Ma (Kuiper et al., 2008). J-factors were determined by fusion of approximately six flux monitor single crystals from equally spaced radial positions around the irradiation tray to a precision of 0.5% or better. Correction factors for interfering reactions were determined from analysis of K-glass and CaF2 from relatively long irradiations, and values used here represent average values determined from long-term monitoring of the reactors. The sample was analyzed along with a series of lava dam remnants from western Grand Canyon. See the supplemental file from a manuscript on that topic (Crow et al., 2015b) for additional details about the 40Ar/39Ar methods used. Heights used in incision rate calculations were measured in the field using a Tru Pulse™ laser rangefinder with an accuracy of ±1 m.


U-Series Dating Results

Two U-series samples from the Schist Camp drape (I2) have been analyzed for U/Th age. The lowest sample (RC15-96-2; Fig. 4) was collected near the base of the deposit, deep in the inner gorge, and was dated three times to assess reproducibility. Three aliquots (one powder and two pieces of the same microcore) of the sample were successfully analyzed; when possible, fractions of the same Th solution were analyzed in the EM and the Faraday cups. This resulted in five dates that all overlap at 2σ and give a weighted mean age of 517 ± 13 ka (see Table S11). Results from an additional run with unsteady gain values and known software issues were discarded but would have indicated an older age just outside of U/Th range. A stratigraphically higher sample (RC15-96-3) from the top of the Schist Camp drape gave an age of 246 ± 13 ka, which likely represents the last time this drape was actively depositing travertine. The older age from the base of the deposit and the sample’s height of 95 m above river level indicate a maximum bedrock incision rate of 229 m/Ma, assuming a depth to bedrock of 23.6 m below the modern river (Crow et al., 2014). Using the height of the drape’s toe (∼60 m) instead of the sample height would result in a rate of 162 m/Ma, using the same depth to bedrock value. Because the 517 ka age is on an infilling around colluvium and because no river gravel is present in or under the deposit, the resulting incision rate is interpreted as a maximum rate.

Our infilling and detrital travertine samples from the higher perched mound (P1) have yet to be analyzed. Figure 3 shows all dated travertines: six ages reported by Abbott et al. (2015) for P1 and other platform deposits range from 489 to 591 ka; ten ages from the inner canyon and tributary deposits range from 12 to 394 ka (two additional samples from the inner gorge gave ages of 468 ± 166 ka and 562 ± 177 ka but were not favored by Abbott et al. [2015] because of their large errors); and seven samples from P1, P4, and I2 showed evidence for open-system behavior (Abbott et al., 2015) but are plausibly older than the upper limit of U/Th dating (i.e., >650 ka).

Ar-Ar Dating Results

Step heating of the crystalline 159-mile dike sample with the Synrad CO2 laser yielded a well-defined plateau that includes six of the ten heating steps and >75% of the 39Ar released (Fig. 5, Table S2 [footnote 1]). The resulting weighted mean age of the steps in the plateau is 534 ± 16 ka with a mean square of weighted deviates (MSWD) of 1.33. A reverse isochron including all the heating steps yields an age of 517 ± 16 ka with a 40Ar/36Ar intercept of 307 ± 3 (Fig. 5). Despite the higher than expected MSWD of 5.5, the isochron age is favored because the 40Ar/36Ar intercept is significantly greater than atmosphere indicating a trapped component not accounted for in the plateau age.

Field Observations

Hermit Area

Figure 3 shows results of new mapping of the travertine deposits. For clarity, we maintain the nomenclature and numbering of Abbott et al. (2015) for platform deposits that developed on the Tonto platform, inner gorge drapes that flowed into the steep inner gorge, and tributary deposits that are in modern tributaries. However, in detail, this distinction breaks down for P4 and I1, which are travertine drapes that flowed from the Tonto Platform into the inner gorge. Our mapping reveals that most if not all of the mounds are drapes that mantle the underlying bedrock and mimic the angle of the slopes that they rest on (Figs. 6I–6L). In addition, all of them have conical shapes and slopes that project up to what we postulate to have been spring sources at the base of the Redwall-Muav aquifer (black stars in Fig. 3). The cliff retreat concept of Abbott et al. (2015) based on beheaded spring mounds seems to apply to all of them. These deposits are analogous to many other active and inactive spring and travertine systems in Grand Canyon such as the area from Kwagunt to the Little Colorado River (RM 56–61, river left), the Elves Chasm area (RM 115–117, river left), along the Hurricane fault (RM 189–191, river left), near Quartermaster Canyon (RM 260, river left), Travertine slot and Travertine bluff deposits (RM 268–276, river right). In all of these cases, travertines formed in areas where carbonic springs discharge from the base of the Redwall-Muav aquifer coalesced on the Tonto Platform forming mounds there or draping farther into the canyon (Crossey and Karlstrom, 2012). Unlike the Hermit examples, some of these drapes extend downwards and cement unambiguous Colorado River gravels above bedrock straths, and it is these occurrences that give the most reliable incision rates (e.g., Crow et al., 2014). Carbonic springs can also vent into tributary side canyons and contribute to the tributary’s base flow. Havasu Creek and Travertine Grotto are examples where travertine is actively being deposited along those tributaries and has cement side-stream gravels. Modern Hermit Creek water is only weakly carbonic (Crossey et al., 2006; Crossey et al., 2009) and not associated with major travertine deposition.

The internal stratigraphy of the mounds was described by Abbott et al. (2015) to involve predominately horizontal fluvial bedding and lenses of side-stream gravel (see their figures 6 and 10). Our observations indicate both sloping internal stratification that parallels the top and base on the deposits and subhorizontal layering. Travertine drapes, into which horizontal layering is often inset, demonstrate at least some paleotopography on a slope (Fig. 6N). A preponderance of angular to subrounded clasts within the deposits suggests both hillslope and tributary depositional environments (Abbott et al., 2015). Abbott et al. (2015) reported isolated well-rounded clasts that were suggestive of main-stem Colorado River gravel at a single location in the perched (P1) outcrop (see their figure 13 and Fig. 6N from this study). The outcrop is inaccessible and the lithology of the clasts could not be determined; perhaps a few hundred well-rounded clasts are present. A nearby out-of-place block (Fig. 6B) contains similarly well-rounded clasts that are locally derived. To our knowledge, these are the only locations in any of the Hermit area travertines where well-rounded gravels are present. Identification of exotic, far-traveled clasts, foreign to Grand Canyon, would confirm the main-stem Colorado River provenance. The rarity of well-rounded gravels suggests at most minor sediment input from the Colorado River.

According to the Abbott et al. (2015) model, the travertine deposits were synchronous with gravel deposition that occurred at ca. 600–500 ka by a paleo–Hermit Creek that joined the Colorado River at a height similar to that of the Tonto Platform today. In that model, paleo–Hermit Creek would have flowed slightly west of its current location to P1. This is consistent with newly mapped paleo–Hermit Creek deposits (labeled QTa on Fig. 3). From P1, the Abbott et al. (2015) model suggests that paleo–Hermit Creek would have flowed subparallel to the Colorado River to P2 (Fig. 3). Although subrounded gravel is less common in P2, P3, or P4 than P1, Abbott et al. (2015) have documented subrounded clasts and sedimentary features (see their figure 11) that are plausibly consistent with deposition by a tributary in P2. The tributary paleo–Hermit Creek geometry to link P1 to P2 is unlike all modern tributaries to the Colorado River in Grand Canyon, which all join the river at high angles. Abbott et al. (2015) suggest they may have formed in a >1-km-long tributary fan that flanked the paleo–Colorado River. Modern fans of similar dimensions exist but contain Colorado River deposits throughout, especially at their downstream end. Likely, Colorado River–derived material has only been found at the upstream-most P1 deposit.

159-Mile Dikes

Two dikes at river mile 159 intrude the Muav Limestone through the Supai Group on both sides of the river to near heights of 400–450 m above river level (Fig. 7). In the lower Supai Group on the south side of the river, welded pyroclastic tuff (Wenrich et al., 1997) is present along the strike of the downstream-most dike. The dikes are roughly aligned but not coplanar with the Yumtheska vents to the southeast and the Cork to the northeast. The Yumtheska West vent was dated by the K-Ar method at 780 ± 15 ka, and the Cork was dated at 410 ± 70 ka (Wenrich et al., 1995); both of these pyroclastic cones were established at a slightly higher stratigraphic level on the Esplanade Sandstone. This map pattern and outcrop-scale field observations (Fig. 4B) suggest that the dikes were injected along en echelon fractures or preexisting joints. A similarly oriented joint set is present both locally and throughout the region (Fig. 8), although oblique joints are also present and well developed in the immediate vicinity of the dikes (Abbott et al., 2016).


Hermit Travertine Incision Constraints

In the Hermit Rapid area, the migrating knickzone hypothesis of Abbott et al. (2015) suggests that the Colorado River incised through bedrock from the level of the perched travertine mound on the Tonto Platform (P1) at ca. 500 ka to at least the toe of the Schist Camp drape (I2) at ca. 400 ka. This elevation difference is ∼180 m, which would result in an incision rate of ∼1600 m/Ma (upper blue arrow in Fig. 9); however, this hypothesis is challenged by the 517 ± 13 ka age at the base of the I2 drape, within 95 m of river level. Our new data require that approximately two-thirds of the inner gorge had been carved when the ca. 500 ka travertine at the foot of the perched P1 travertine mound (i.e., at the rim of the inner gorge) was deposited. This provides new maximum incision rates of 162–229 m/Ma (green arrows on Fig. 9) that are consistent with rates of 160 m/Ma ∼35 km (25 river miles) upstream from the site and rates of ∼100 m/Ma ∼30 km (20 river miles) downstream at locations where unambiguous perched Colorado River gravel has been dated (Crow et al., 2014). The new age on the Schist Camp drape of 517 ± 13 ka and the 506 ± 33 ka age on the P1 deposit overlap within error. Considering the 2 sigma errors, it is possible that the P1 deposit is between 35 and 9 ka older than the Schist Camp drape (it is also possible that it is younger). This allows for the possibility of bedrock incision rates of ∼5000 m/Ma between 539 ka and 504 ka or ∼25,000 m/Ma between 539 ka and 530 ka. Such bedrock incision rates over large river reaches are very high globally and likely unrealistic geologically for the southwest United States.

The new data on the timing of carving of the inner gorge at Hermit Rapid dispel several problematic aspects of the migrating knickzone hypothesis of Abbott et al. (2015), chiefly the inconsistency with well-dated unambiguous Colorado River gravels both up and downstream from the Hermit site. In a comment to their paper (Crow et al., 2015a), we focused on seven incision points that are inconsistent with the Abbott et al. (2015) migrating knickzone hypothesis. In their reply, Abbott et al. (2016) refuted this claim focusing on the accuracy of two dates, the precision of three dates, and they concluded that the remaining two were consistent with their model. In Figure 10, we present a “strath-to-strath” plot similar to those shown by Crow et al. (2014). These types of plots show the height and age of deposits interpreted to be associated with past strath levels. Temporal variations in incision rate are noted by changes in the slope of the line connecting the past strath locations. Figure 10 shows incision constraints mainly from Crow et al. (2014) and Abbott et al. (2015) and corresponding envelopes showing bedrock strath heights and ages consistent with each model. Note that for ages older than ca. 400 ka, the envelopes suggested by Abbott et al. (2015) and Crow et al. (2014) don’t overlap. Additionally, all of the bold incision points, including the seven incision points focused on in the reply (Crow et al., 2015a), plot below the envelope for the Abbott et al. (2015) migrating knickzone hypothesis; this means that those data suggest less bedrock incision than predicted by their model. This remains true when the 2 sigma analytical errors are included; so the imprecision of some rates is not relevant.

Abbott et al. (2015) pointed out that two U/Th ages reported by Crow et al. (2014) are at the upper limit of the U-series dating method. Because they are close to secular equilibrium, even minor amounts of U or Th loss or gain could produce large changes in the apparent age. However, careful examination of the analytical results, using the same methods as Abbott et al. (2015), indicates no evidence for open-system behavior as implied by Abbott et al. (2015, 2016); those two samples (plus a third of a similar age) fall along evolution curves on both 234U/238U versus 230Th/238U and 234U/238U versus 230Th/234U plots (Fig. 11). Undetected open-system behavior is still possible but not indicated by the available data. Other studies have demonstrated that ages in this range can be precisely obtained. For example, Cheng et al. (2013) dated Chinese speleothems back to 640 ka; δ18O values for those speleothems are consistent with insolation values, suggesting that their chronology was accurate.

In summary, our new age on the Schist Camp drape (I2) indicates that approximately two-thirds of the inner gorge was carved when the ca. 500 ka travertine at the base of the P1 deposit, on the rim of the inner gorge, was deposited. This new age strongly suggests that the Abbott et al. (2015) conclusion of incision rates of ∼1600 m/Ma between 500 and 400 ka at the Hermit site is incorrect.

159-Mile Dikes Incision Constraints

In addition to the Hermit travertine deposits, Abbott et al. (2015) used the 159-mile dikes as evidence for rapid incision of the Muav Gorge after ca. 500 ka. The carving of the Muav gorge during that time frame would require an incision rate of ∼760 m/Ma based on our dating of the 517 ka 159-mile dike. Such a high rate is inconsistent with independent incision rate calculations from the Lava Falls and Surprise Valley areas, all based on the dating of unambiguous river gravel deposits. How could the Colorado River have incised to within 46 m of the modern river level at Elves Chasm (RM 116, upstream from the dikes) by ca. 650 ka (Crow et al., 2014), to within 42 m of the modern river level at RM 189 (downstream from the dikes) by ca. 830 ka (Crow et al,. 2015b), and be ∼400 m above river level at RM 159 at ca. 520 ka? Alternatives include: (1) the dikes vented into an existing canyon; (2) unrecognized differential incision might be attributed to fault slip; or (3) the Muav Gorge may have been filled with aggraded gravel or lake beds at the time of dike emplacement.

Historical observations indicate that the first hypothesis is at least plausible. For example, on the big island of Hawaii, dikes were injected across significant relief and volcanic craters during the 10 November 1973 eruption at Mauna Ulu without erupting at the crater bottom (Tilling et al., 1987). Abbott et al. (2015) suggested that this is a poor analogy because previous eruptions occurred at the crater bottom, because minor pooling occurred at the crater bottom (it did not fill to the level of the dikes) and because the dikes crossing the craters were not coplanar. However, dike outcrops across the river and the locations of vents along the general strike of the 159-mile dikes indicate a system of en echelon dikes that are also not coplanar (Fig. 7), and, instead, dikes are strongly controlled by regional joint sets (Fig. 8). In the second hypothesis, at least six sites downstream from the Toroweap fault record less than 56 m of incision (neglecting the depth to bedrock below the river’s surface) over a period greater than the age of the 159-mile dikes (Karlstrom et al., 2007; Crow et al., 2015b). This suggests that any migrating knickzone (e.g., in the Abbott et al. [2015] model) would have been stationary upstream from the Hurricane fault from 830 to 520 ka, or that it was generated by slip on a fault. Slip-rate changes on the Toroweap fault, at RM 179, cannot have been a major factor because the total Paleozoic separation on that fault is only 177–193 m in Grand Canyon (McKee and Schenk, 1942; Huntoon, 1977; Fenton et al., 2001), about half of the amplitude of the proposed knickzone of Abbott et al. (2015), and the post–500 ka slip is ∼50 m (Karlstrom et al., 2007). The third hypothesis suggests that the dike could have intruded aggraded gravel or lake beds. Analogous dikes intruding gravels and cinders are present in the Lava Falls area (Hamblin, 1994). A complete lack of verifiable lake deposits (Kaufman et al., 2002) casts some doubt on this hypothesis, but every study of Grand Canyon lava dams has concluded that the dams were stable enough to at least impound Colorado River water in large lakes (Powell, 1875; Hamblin, 1994; Fenton et al., 2004; Crow et al., 2015b). A general lack of features indicating that runout lava flows interacted with river water supports this. Based on the presence of monomictic basalt gravels overlying most lava dam remnants, Crow et al. (2015b) concluded that most of Grand Canyon’s lava dams failed before completely filling with sediment, casting some doubt on this model for dike intrusion. In summary, we reject the fault-slip hypothesis and remain ambivalent about whether or not the dikes could have intruded sediments that had aggraded within the canyon behind a lava dam. Our favored interpretation is that the dikes vented into an existing canyon. No preserved basalt has been found on canyon walls in this area, but this is not unexpected given the extreme relief in the area if the erupted volumes were small.

Alterative Interpretations of the P1 Travertine Deposit

Although the presence of ca. 500 ka travertine on the Tonto Platform and ∼145 m lower in the inner gorge indicates that the ca. 500 ka date at the base of the perched (P1) travertine mound does not constrain bedrock incision, it may inform other processes. Crow et al. (2015a) suggested two alternatives to the Abbott et al. (2015) model that could explain all the existing data.

The P1 Deposit May Be Much Older Than Originally Thought

The first alternative suggested by Crow et al. (2015a) was that the ca. 500 ka U/Th ages reported by Abbott et al. (2015) at the Hermit site may be a secondary travertine infillings that significantly postdate gravel deposition. Secondary infilling of carbonate in travertine systems is common throughout Grand Canyon and often does not have clear textural indications (Crow et al., 2014). Other samples analyzed from this deposit suggest open-system behavior (Abbott et al., 2015) but are plausibly outside of U/Th dating range (>650 ka) and would give 234U model ages of up to ca. 1.6 Ma, assuming a wide range in initial 234U/238U activities; analysis of similar material in Grand Canyon indicates initial 234U/238U activities between 1.1–8.9 (Pederson et al., 2002; Polyak et al., 2008; Crow et al., 2014; Abbott et al., 2015). However given the evidence for open-system behavior, this hypothesis is best tested by additional U/Th and U/Pb work on both infillings and detrital travertine material from the deposit, which should bracket the age of gravel deposition, assuming that unaltered material can be found. Extrapolating back incision rates of 100–160 m/Ma suggests that the deposit could be 1.6–2.6 Ma.

The P1 Deposit May Be Related to Local or Regional Aggradation

A second possible reconciliation of all data might involve the deposits being related to side-stream aggradation (Crow et al., 2015a), possibly controlled by a natural landslide or lava dams on the main stem or travertine accumulation on the side stream. Subrounded gravel present in P1 was likely deposited by paleo–Hermit Creek, and well-rounded gravels in the same deposit may indicate isolated and exceedingly rare deposition by the main stem as well (Abbott et al., 2015). The origin of isolated sub-rounded gravel documented in P2 is less clear as the paleo–Hermit Creek geometry required to move tributary gravel to that location without mixing with Colorado River deposits seems unlikely. Perhaps these gravels were deposited by a lower-order tributary or reworked from older deposits.

Building on the work of Hamblin (1994), Crow et al. (2015b) documented ∼17 different lava dams in Grand Canyon, downstream from the Hermit area. Seven of those overlap with the 506 ± 33 ka age at the base of the perched (P1) travertine mound. These include: lower and upper Black Ledge, lower and upper Prospect, Buried Canyon, 183.4-mile, and the Toroweap dams. The heights of the original edifices are estimated from the highest remaining dam remnant. Although effort was made to focus on intracanyon remnants related to damming, overlain by gravel, it is possible that some remnants are related to where lava cascaded into the canyon and not where it pooled at the canyon bottom. Conversely, relatively little of each dam remains; so it is also likely that the highest parts of dam are no longer preserved. Figure 12 shows our best estimates for dam heights. It also shows the heights of landslide dams in the Surprise Valley area based on the identification of Colorado River gravel on top of landslide deposits (Robertson, 2015). Of these landslide and lava dams, only the Toroweap and Prospect dams have estimated heights sufficient to inundate water to the level of the perched (P1) travertine mound. The Prospect Dam height is likely an overestimate because cascade remnants are likely convoluted with dam remnants (Crow et al., 2015b). The height of the Toroweap dam, based on intracanyon remnants in the Lava Falls area (RM 18.5), is more certain and very similar to the height of the perched (P1) travertine mound in the Hermit area. Although this does not prove that perched side-stream gravel deposition in the Hermit area was related to a lava dam, it is consistent with the hypothesis that some of the deposits could be related to base-level changes and delta progradation into a transient standing body of water that was rapidly filling with sediment at its head. The Toroweap dam would be the most likely culprit for creating that standing body of water. However, lacustrine deposits as proposed by Hamblin (1994) have yet to be found by us, Abbott et al. (2015), or others (Kaufman et al., 2002). Crow et al. (2015b) also concluded that most of Grand Canyon’s lava dams likely failed in tens to hundreds of years, before they completely filled with sediment.

Alternative scenarios involving aggradation to reconcile all of the data include regional climatically-driven aggradation and local side-stream aggradation behind spring mounds. Climatically-driven aggradation of 50–60 m has been documented on the main stem at ca. 65 ka, 110 ka, and 320 ka (Anders et al., 2005; Pederson et al., 2006). However, as of yet, to our knowledge, no fill terraces have been documented in the southwestern United States at ca. 500 ka.


Field observations and a new U/Th age of 517 ± 13 ka on the Schist Camp (I2) drape, in the Hermit Rapid area, cast serious doubt on the migrating knickzone hypothesis, which suggested the inner gorge of Grand Canyon in the Hermit area was carved rapidly between 500 and 400 ka (Abbott et al., 2015). Instead, our data require that approximately two thirds of the inner gorge were carved (to within at least 95 m of modern river level) by ca. 500 ka. This is consistent with observations that show that both perched travertine mounds (P1, at the rim of the inner gorge) and inner canyon drapes were deposited at least in part on existing slopes by spring mounds with a strong pulse of travertine deposition at ca. 600−500 ka. Our new ages yield maximum incision rates of 162–229 m/Ma that are consistent with temporally steady but spatially variable incision models (Karlstrom et al., 2007; Karlstrom et al., 2008; Crow et al., 2014).

Subrounded gravel in the P1 deposit is consistent with deposition by paleo–Hermit Creek, and well-rounded gravels may indicate some deposition by the main-stem Colorado River. Drapes within and at the base of travertine mounds also indicate hillslope topography at the time of deposition. Step pools on travertine mounds draping the Tonto platform and partially filling paleo–Hermit side canyon may have accumulated a mix of hillslope and tributary and/or river deposits, some of which may have been reworked. These observations and the new age constraints can be reconciled in several ways.

  • 1. The tributary gravels of P1 are actually older than 600–500 ka, and the U/Th dates are from secondary carbonate infillings.

  • 2. Main-stem aggradation could have occurred locally in response to delta progradation within a lake controlled by a downstream dam, and the tributary deposits were graded to the higher lake and/or river level.

  • 3. The main stem might have aggraded to the rim of the inner gorge at ca. 500 ka due to regional climatic fluctuations, and the tributary deposits may have been graded to this higher river level.

The first possibility is most likely in our opinion and is readily tested; it could be confirmed if travertine from the perched deposits gave U/Pb ages of 1–2 Ma. Attention should be focused on dating not only infillings but also detrital clasts to fully bracket the age of the reported tributary gravel deposition. The other alternate hypotheses might be supported by additional evidence for lake deposits, documentation of fill-terrace sequences in analogous Grand Canyon tributaries and in other southwestern river systems, and more detailed mapping and dating of the facies within the Hermit travertine deposits.


This manuscript benefited from helpful reviews from Brian Wernicke, Kyle House, and an anonymous reviewer. Long-term work in this area has been funded by multiple NSF projects and most recently by the USGS Mendenhall program and a USGS mapping project focused on the evolution of the Colorado River.

1Supplemental Files. Tables S1 and S2. Tables detail analytical results reported by this study. Please visit https://doi.org/10.1130/GES01562.S1 or the full-text article on www.gsapubs.org to view the Supplemental Files.
Science Editor: Raymond M. Russo
Associate Editor: Graham D.M. Andrews
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.