How indurated sediment and rock outcrops in the walls of some submarine canyons were exhumed is unclear; the flows traversing them were muddy sedimentary flows, so abrasion is unlikely to have been important in these cases. The answer may lie in the importance of quarrying or plucking. Observations of rivers in extreme floods suggest that such erosion processes begin to operate effectively on jointed bedrock when reach-scale bed shear stress exceeds 100 Pa and become increasingly rapid beyond that stress level, some floods producing >1000 Pa. Here, the sedimentary flow weights that would be needed to produce similar shear stresses are estimated for two canyons where observations from submersibles have revealed exposed bedrock: Monterey Canyon (California, offshore western USA) and Hendrickson Canyon (New Jersey, offshore eastern USA). Assuming that the dense portions of turbidity currents occupied Monterey Canyon to a height corresponding to a 100-m-high steep inner wall, the minimum flow-averaged excess density derived from the estimated flow weight is found to be only 5 kg m–3, whereas a 1000 Pa condition would suggest values 10 times larger. In contrast, muddy debrites and other mass transport deposits dominate the New Jersey slope and upper rise. The weight of debris flows occupying Hendrickson Canyon was computed using the stress constraints and converted to equivalent sediment thickness using upper slope in situ sediment deposit densities. These thicknesses were found to be 1–10 m, within the range of possible mobile sediment thicknesses suggested by relief of upper slope landslide embayments. The flow density (Monterey) and thicknesses (Hendrickson) are both modest, so bed shear stresses generated by sedimentary flows in continental slope canyons seem adequate to explain exhumation of bedrock by quarrying or plucking erosion mechanisms.


In the early twentieth century, the growing evidence from echo soundings of the many deep canyons in the continental slopes around the United States and elsewhere prompted a debate over whether they were excavated by turbidity currents. Daly (1936, p. 26) suggested that negatively buoyant flows of suspended sediment could be produced by wave agitation during sea level lowerings. He wrote, “The water…was specially loaded with suspended sediment and therefore had effective density exceeding that of cleaner seawater elsewhere. The loaded water tended to slide down continental slopes, along the bottom. The question arises as to whether the velocities of the density currents sufficed to have eroded the actual trenches. A definite answer cannot be given, but the general hypothesis should be tested by appropriate observations.”

Shepard (1937) was initially skeptical of the efficacy of muddy turbidity currents in contributing significantly to canyon development by erosion and that submarine turbidity currents produced by wave agitation or river mouth outflows could also be effective. As further evidence became available, however, Shepard (1981) suggested that canyons probably form over long periods from multiple causes, including debris flows, landsliding, faults and bioerosion. Such processes are apparently capable of excavating indurated sediment and rock, which are observed exposed in the walls of some submarine canyons (Shepard, 1981).

Normark and Carlson (2003) considered the general issue of canyon development, noting that canyon size is oddly completely unrelated (perhaps even inversely related) to the amount of sediment that has passed through them, suggested by the volume of the associated submarine fan. If erosion rules of the kind used in terrestrial bedrock stream modeling were to represent the dominating process of development of submarine canyons by sedimentary flows (e.g., erosion rate proportional to flow power or shear stress; Howard, 1994), the relationship between canyon depth and fan volume, though perhaps not simple, would certainly not be nonexistent, as Normark and Carlson (2003) found. This suggests that other factors are at play in setting how canyons develop (Shepard, 1981). One of the largest canyons, for example, the Zhemchug Canyon in the Bering Sea, may have developed by a series of massive landslides (Carlson and Karl, 1988), and landsliding may more generally be important for canyon development (Normark and Carlson, 2003). Nevertheless, the submersible observations outlined in the following suggest that some bedrock was exhumed locally by sedimentary flows. The landslide explanation leaves open some questions concerning the abundance of bedrock exposures in canyons, questions discussed herein.

Addressing the bed shear stresses required for bedrock erosion potentially provides a further constraint on flow properties, supplementing the information on canyon and channel sedimentary flows currently available from a variety of sources. That other information includes the speeds with which telephone cables have been broken by historical flows (Heezen and Ewing, 1952; Hsu et al., 2008), variations in the texture (grain size) of deposits (Abd El-Gawad et al., 2012; Bowen et al., 1984; Komar, 1985; Pirmez and Imran, 2003; Sadler, 1982; Stow and Bowen, 1980; Talling, 2001), the geometries of channel levees (Skene et al., 2002) and channel deposits (Peakall et al., 2000), evidence for supercritical to subcritical flow transitions (Covault et al., 2014; Fildani et al., 2006), the differing elevations of levees caused by Coriolis effects (Komar, 1969), constraints on kinetic energy from superelevation (Lamb et al., 2008; Muck and Underwood, 1990), bedload movements from monuments installed in canyon floors (Paull et al., 2010) and from dune tracking (Smith et al., 2007), and acoustic imaging of turbidity currents of mine tailings (Hay, 1987). Measurements of turbidity current velocity structures have been made with acoustic Doppler current profilers (ADCP) suspended over channels (Xu, 2011), and laboratory experiments have been carried out to simulate various aspects of flow behavior (e.g., as reviewed by Meiburg and Kneller, 2010; Kneller and Buckee, 2000).

Inferences on flow behavior from many of these methods individually are limited; for example, the ADCP measurements have so far only recorded small flows, not those responsible for transporting the major sediment volumes forming the turbidites found on fans (Paull et al., 2005a). In the laboratory experiments, it is difficult to scale all the similarity numbers simultaneously to the natural flows (prototypes), making the accuracy to which the results correspond to the prototypes uncertain (Meiburg and Kneller, 2010; Mitchell, 2012; Parsons and Garcia, 1998; Sequeiros et al., 2010). Despite the undeniable progress in this field, there is therefore still a need to continue uncovering further information on the properties of subaqueous sedimentary flows. In particular, while there has been recent progress in developing understanding of how sedimentary flows initiate channels by erosion through weakly consolidated sediments on continental slopes (Covault et al., 2014; Fildani et al., 2013; Normark et al., 2009; Paull et al., 2013; Rowland et al., 2010; Stewart and Long, 2014), Traer et al. (2012) demonstrated how uncertainties in parameters such as those representing the picking up of bed material make forward modeling of flow behavior still impractical. Furthermore, these efforts also leave open the issue of how indurated sediment or bedrock is eroded in submarine slopes.

Erosion within rivers involves abrasion by bedload and suspended particles and quarrying or plucking of blocks from the bed (Hancock et al., 1998; Whipple et al., 2000). The terms quarrying and plucking appear to be used interchangeably in the geomorphological literature (Miller, 1991; Tinkler and Wohl, 1998); in the following, quarrying is used synonymously for both terms. Quarrying is more effective than abrasion where the rock is closely jointed, involving sliding of in situ rock bodies along joints from the shear stress directly imposed on them by the flow and/or uplift of particles from bedrock by the Bernoulli effect (i.e., reduced pressure in the fast-moving flow compared with more static fluid beneath the particles). Abrasion is generally considered less effective, but is the remaining important erosion mechanism where the rock is not jointed (massive) or can be important where the rock is weak and impacted by hard clasts. According to Snyder et al. (2003), a rare flood in Fall Creek, New York, visibly moved blocks to 0.3 m in thickness. The shear stresses within the reach studied were estimated to have been 100–200 Pa, which Snyder et al. (2003) suggested were just sufficient to cause bedrock quarrying (a reach can be any arbitrarily chosen segment of a stream, though reach scale for evaluating shear stress appears to refer to hundreds to thousands of meters; Hancock et al., 1998). A compilation of reach-scale stresses of rivers in flood in a variety of tectonic settings by Mitchell et al. (2013) suggested that this 100 Pa level was a generally important threshold for bedrock erosion by rivers. Given that entrenchment in submarine channels also involves flowing fluids, the Bernoulli effect and direct coupling of shear stress to bed particles should also operate, causing removal of those particles from the bed. This prompts the question of whether the 100 Pa shear stress threshold might also apply to submarine canyons where indurated sediment and rock now crop out.

In this study the implications of bedrock erosion for flow properties are investigated, assuming that the 100 Pa and 1000 Pa shear stress levels apply to submarine canyons. Data were compiled from two contrasting canyons where bedrock has been observed from submersibles in the channel walls or floor (Hendrickson Canyon in the New Jersey slope; McHugh et al., 1993; Monterey Canyon, California; McHugh et al., 1998). Requiring a minimum bed shear stress of 100 Pa allows the flow weight to be estimated, and from that weight, the minimum flow density and/or thickness. For Monterey Canyon, where turbidity currents dominate sediment transfer to the fan, current density was calculated. For Hendrickson Canyon, the flow thickness was estimated and found to correspond with debris flows of high yield strength (Talling et al., 2012), in keeping with preservation of debris flow structures in cores of debrites recovered from the slope and upper rise. The apparently modest values estimated in both localities imply that shear stresses generated by sedimentary flows are adequate for erosion of bedrock by quarrying.


Bed shear stresses (τ0) opposing river flow were calculated from 
where ρ is the flood water density, g is the acceleration due to gravity (9.8 m s–2), R is the hydraulic radius (approximately the flow depth for wide rivers), and θ is the bed slope angle, usually calculated over a reach scale of 100–1000 m (Hancock et al., 1998). Equation 1 represents the stresses required to balance the weight of the flow, neglecting stream-wise pressure gradients and acceleration, which are usually considered minor if the stress estimate (an average) is made over the scale of a river’s reach.

The riverbed shear stresses for extreme flood stages compiled by Mitchell et al. (2013) are reproduced in Figure 1 (from the data sources given in the figure caption). Stresses were computed using Equation 1 with R generally estimated from trim lines of debris left by floods or extents of removed vegetation. Flood return intervals or flood ages represented by these data were 1–350 yr; multiple decades were typical. The rivers are in a variety of climatic and tectonic environments; some are in regions of convergent tectonics (Torto, Torrente L’Apa, and South Fork Hoh) and some are in extensional environments (Xerias and Nahal Paran), whereas Sandy Creep gorge is within a declining orogen.

Despite the wide range of environment and return period, Figure 1 shows that bed shear stresses were mostly restricted to 100–1000 Pa. The lower 100 Pa level suggests that the 100–200 Pa threshold mentioned by Snyder et al. (2003) is a common threshold stress involved in producing geomorphologically meaningful erosion (i.e., erosion capable of opposing the long-term steepening of the channel by tectonic uplift). In this interpretation, in areas of tectonic uplift where channels undergo bed shear stresses during floods of <100 Pa, the channels progressively steepen tectonically until flood stresses reach 100 Pa and bedrock begins to erode, moderating further steepening. The upper limit of the histograms in Figure 1 at 1000 Pa, in contrast, may be associated with rapid erosion; i.e., if a stream repeatedly undergoes stresses much above 1000 Pa, it is rapidly entrenched, reducing its overall gradient until flood stresses decline below 1000 Pa and erosion becomes less effective. The upper 1000 Pa of the stress distribution may also reflect a finite range of tectonic steepening rates in terrestrial environments.


Riverbed studies suggest that erosion of bedrock occurs more rapidly by quarrying rather than by abrasion when the bed is strongly jointed, whereas slower abrasion dominates when the rock is more massive (Hancock et al., 1998); therefore, the following estimates of minimum flow weights were calculated effectively assuming that the dominant erosion mechanism is quarrying. In practice, however, the stress levels in Figure 1 involve riverbeds eroding by both quarrying and abrasion mechanisms, so it is difficult to assign different stress levels to the different mechanisms and stress thresholds are likely to vary with joint spacings, particle size, and weathering. The following calculations exploit only an apparent average tendency.

Applying these stress conditions to submarine channels requires that some process is available to fracture or open joints in rocks to allow erosion to continue with progressive exhumation. In rivers, stress release accompanying reduction in confining stress as rock is exhumed is considered important for creating joints (Haxby and Turcotte, 1976; Selby, 1993). The beds of mountain streams are commonly affected by frost action and other weathering processes during the winter low-discharge season. These processes may be important for weakening and developing joints (Whipple et al., 2000).

A shear stress of 100 Pa also coincides roughly with the threshold of motion of boulder-sized clasts (Carling, 1983), consistent with bedload movements also being important to erosion (Cowie et al., 2008). A mobile bedload may enhance long-term erosion by quarrying mechanisms by promoting the growth of fractures and opening of joints in rock by percussion effects (Whipple et al., 2000).

Within submarine channels, it is less clear that percussion effects would be so important where the flows are muddy. Weathering along joints will also be different. However, stress release mechanisms should operate and there is some evidence of burrowing activity by organisms causing weakening of rock (Dillon and Zimmerman, 1970; Malahoff et al., 1982; Paull et al., 2005b; Shepard, 1981; Valentine et al., 1980; Warme et al., 1978). A further mechanism of joint formation for Hendrickson Canyon, from mineralogical changes in the rock (McHugh et al., 1993), is outlined in the following.

Ambient ocean hydrostatic pressure is assumed to have no effect on joint formation in the deep sea, as laboratory experiments on rock samples typically show pore pressure is acquired over experimental time scales of days to weeks (Hackston, 2013). For extraction of particles from the bed to occur rapidly, water must quickly enter the fracture space underlying the particles in order to restore hydrostatic pressure during extraction, otherwise a negative pressure relative to hydrostatic pressure will work to keep fractures closed, as has been proposed to explain how blocks in submarine debris avalanche deposits remain intact despite pervasive jigsaw cracks (Mitchell et al., 2002). As with the similar mechanism involved in breaching of submarine sands (Mastbergen and Van den Berg, 2003), this probably slows submarine bedrock erosion.

Although the Bernoulli equation is derived assuming that the flow has negligible viscosity (Munson et al., 1990), laboratory experiments with particle-laden non-Newtonian fluids with Bingham-plastic rheologies reveal pressure variations not unlike those produced with the Bernoulli equation (López and Srivastava, 2002). Pressures measured in experimental turbidity currents with mean sediment concentrations to 32.5 kg m–3 also showed pressure reductions as predicted with the Bernoulli equation (Eggenhuisen and McCaffrey, 2012). The extent of pressure reduction with velocity is unclear where the fluid has very high effective viscosity, as suggested to be the case for flows within Hendrickson Canyon. Nevertheless, inhomogeneities in rheology seem likely to produce more varied flow velocity and hence pressures, leading to quarrying.

Research reports on the erosion by subaerial debris flows have also described quarrying of particles from jointed rock, but emphasized the importance of inertial impacts from large clasts within the flows in causing erosion (Stock and Dietrich, 2006). Normal stresses recorded during passage of rocky debris flows can exceed 50 times those due to their static weights (McCoy et al., 2013). However, the clasts in the debris flows suggested here as being likely to have traversed Hendrickson Canyon were muddy and deformed plastically (McHugh et al., 2002), so it seems unlikely they would have produced the same impact normal stresses. Rather, shear stresses are suggested to be the more important stress components for erosion by these flows, as with rivers.


Figure 2 shows a bathymetric map of the canyon with the path of DSV (deep submergence vehicle) Alvin dive A2131 reported by Eittreim et al. (1989). The dive transcripts (by H.G. Greene and C. Baxter, in Eittreim et al., 1989) mention poor visibility, but suggest that a granular fractured bedrock likely to be granite was observed briefly in the lower slope and a siltstone in the upper slope, whereas rock talus was observed at the base of the slope. The two bedrock lithologies are marked in Figure 2, derived from McHugh et al. (1998). High backscattering in deep-tow sonar data interpreted by McHugh et al. (1998) (area of the green polygon, Fig. 2) suggests that this bedrock extends for several kilometers along the canyon wall; high backscattering probably was produced by large rock particles (Mitchell, 1993a), which were reported to cover much of the bed (Eittreim et al., 1989). Paull et al. (2005a) mentioned that the southern wall of the canyon is composed of granodiorite between 1400 and 1700 m depth farther south of Figure 2. Granite, granodiorite, siltstone, sandstone, and other indurated sediment samples have been recovered elsewhere in the bay by dredging, many of which were likely extracted from rock outcrops (Greene, 1977; Martin, 1964). A lithologic map of Greene et al. (1989) also suggests that rock outcrops are widespread.

Greene et al. (2002) reviewed tectonic structures within Monterey Bay. The spur comprising the rock outcrops shown in Figure 2 is described as the Monterey Meander. The straight canyon trend on the southwest of this feature follows the trend of the right-lateral strike-slip Navy fault. Greene et al. (2002) suggested that rock weakness associated with fracturing along this fault and with the parallel Seaside fault on the other side of the meander has been exploited by turbidity currents to create this morphology; they observed a steep inner canyon slope on the northern outer bend of the meander, which may indicate undermining by turbidity current erosion. A steep inner slope also occurs in the canyon segment to the southwest of the spur, as the ridges between slope gullies are truncated and ∼100 m of the lower slope elevation is unusually steep (illustrated by the two profiles in the inset, Fig. 2). Although faults may have controlled the canyon trends and helped to expose rock in the canyon wall, some process is still needed to explain how fractured rock has been removed.

Away from debrites associated with the Sur Slide (a feature resulting from failure of the lower continental slope; Normark and Gutmacher, 1988), 3.5 kHz sediment profiler records and seismic reflection images show stratified reflectivity typical of turbidites (Fildani and Normark, 2004; Hess and Normark, 1976; Normark, 1970, 1999; Normark et al., 1983). Alternating turbidite-like sand and mud deposits have been recovered in sediment cores (Gardner et al., 1991; Normark et al., 2002) and observed from a submersible in channel walls exposed by erosion (Normark, 1998). Sand within the modern upper canyon likely forms much of the particulate material of sedimentary flows traversing the canyon (Paull et al., 2005a; Smith et al., 2005). Turbidity currents rather than denser debris flows probably passed down Monterey Canyon and supplied sediment to the Monterey Fan.

Talling et al. (2013) made a further distinction between dense but thin sandy flows and more dilute flows imaged in the ADCP measurements (Xu, 2010, 2011; Xu et al., 2004). Talling et al. (2013) suggested that the dense but thin sandy flows likely are only 4–10 m thick because coring from a remotely operated vehicle (ROV) within the inner canyon recovered coarse sand as much as 2–6 m above the canyon floor, and thin sand layers as much as 10 m above the canyon floor (Paull et al., 2010, 2005b). Crescentic bedforms within the sandy canyon floor migrate upcanyon (Paull et al., 2010; Smith et al., 2007; Xu et al., 2008) and were interpreted by Talling et al. (2013) as having been produced by a combination of breaching of sand (Mastbergen and Van den Berg, 2003) on the bedform lee slopes and alternations between supercritical and subcritical Froude number, as with cyclic steps (Fildani et al., 2006). Although thin, these flows were powerful enough to displace 45–60 kg concrete monuments (Paull et al., 2010). In the ADCP measurements of more dilute but still supercritical flows by Xu (2010, 2011) and Xu et al. (2004), peak current speeds were measured 5–12 m above the canyon floor. Whether the dense flows of Talling et al. (2013) were truly thin flows or merely the dense components of compound flows (Cartigny et al., 2013) is not evaluated here. However, the flow density required to cause bedrock erosion was also estimated assuming a 10-m-thick flow to see if such a thin flow density would be within ranges of acceptable values.


Figure 3 shows the bathymetry of part of the New Jersey lower slope, including Hendrickson Canyon, produced from multibeam echosounder data. It reveals many channels with broad floors crossed by semicircular embayments, which may be the result of slope failure (McAdoo et al., 2000; McHugh et al., 1993). Immediately upslope from the steep embayment walls, narrow linear depressions run upslope for 1–2 km (e.g., sites 1 and 2 in Fig. 3, within Hendrickson Canyon). These were suggested (Mitchell, 2005b, 2006) to have been created by acceleration of sedimentary flows as they approach the steep escarpments of the embayments, much like how water accelerates in drawdown reaches above waterfalls, leading to greater shear stress and hence erosion. Toward the upper slope and shallower water depths, channels are more V shaped and dendritic in plan view (around the 500 m contour in Fig. 4A), a morphology interpreted generally as the result of erosion by sedimentary flows and failure of deposits in canyon walls and heads (Mountain et al., 1996; Pratson and Ryan, 1996; Pratson et al., 1994).

Geotechnical measurements were made on surficial sediment cores recovered extensively from the U.S. Atlantic slope in the 1970s. The recovered sediments are generally silty muds within the slope, but sand forms isolated layers or in places dominates near the shelf edge (Doyle et al., 1979). Atterberg laboratory tests are used to quantify sediment behavior, with the liquid limit representing the water content above which samples develop fluid slurry consistencies when sheared (Terzaghi and Peck, 1967). In the upper 1–2 m, water contents were found to be 60%–80% of dry sediment weight, compared with liquid limits that had a mean value of 78% (Keller et al., 1979). Measurements on some individual cores from the slope recorded water contents above the liquid limit (Bennett et al., 1980). Shear vane measurements on these shallow sediments revealed undrained shear strengths to ∼20 kPa, but strength was reduced to <5 kPa and commonly <1 kPa on remolding (Bennett et al., 1980; Keller et al., 1979). Analyses on 2 m cores collected 200 km farther southwest of the area in Figure 2 by McGregor et al. (1979) revealed remolded sediment shear strengths of 0.8–5.5 kPa that were typically 2–4 times smaller than natural shear strengths in the cores.

The geotechnical behavior of more deeply buried sediments is less certain and could be different from consolidation effects, because sediments of different mineralogies may have been deposited from melting ice in the cooler climatic conditions expected for this margin (Duncan and Goff, 2001; Knebel, 1984). Five shear vane measurements on cores from the upper 30 m at Ocean Drilling Program (ODP) Site 1073 (Fig. 4) recorded natural shear strengths of 9–21 kPa (Christie-Blick et al., 1998). Penetrometer readings from ODP Sites 902–904 show increasing strength with depth and consolidation, with local reversals associated with layers of differing sediment texture (Mountain et al., 1994). If remolding leads to a similar collapse of shear strength by a factor of 2–4 as with surficial sediments (McGregor et al., 1979), debris flows with shear strengths of <1 kPa probably form when only the uppermost ∼10 m of sediment fails, whereas deeper failures lead to sediment slides. Some wet bulk density values representative of the uppermost slope were obtained from ODP Sites 1073 and 903. The upper 10 m typically has an average wet bulk density of 1700 kg m–3 (higher values at Site 903 possibly reflect an earlier phase of compaction of the sediment and subsequent exhumation, which is common on this margin; Blum et al., 1996; Mountain et al., 1996).

Observations of bedrock exposures made along the two DSV Alvin submersible dives (Fig. 3) were described by McHugh et al. (1993); two cross sections based on their observations are shown in the lower inset in Figure 3, with lithologies projected updip to the channel axis. The lithology labeled chalk in Figure 3 was described as an impure chalk (50%–67% carbonate, remainder siliceous) that is only semi-indurated, soft and friable. The lithology labeled porcellanite, in contrast, is a hard and indurated siliceous porcellanite chalk (42%–56% carbonate). In outcrops, this rock type is jointed parallel and perpendicular to bedding, probably a result of contraction associated with an opal-A to opal-CT transformation. A talus of this material comprised tabular blocks, the shapes of which are evidently controlled by the joints. Blocks are 10–30 cm thick, but individually can be only 1–2 mm thick. Some in situ porcellanite was described by McHugh et al. (1993) as being exposed in the channel floor or appearing to be covered by only a thin veneer of sediment. Some grooves a few centimeters deep occurred in the channel floor porcellanite, but the tabular blocks showed no evidence for abrasion.

Interpretations of cores recovered by scientific drilling and sediment profiler records suggest the types of sediment flows or movements that have been responsible for excavating the slope channels. The drill sites are located relative to the slope channels in Figure 4A. McHugh et al. (2002) characterized the sediments recovered at ODP Sites 902–906; at Site 905, in the continental rise, they described muddy debrites and slump deposits forming intervals of 30 m in the middle Miocene and 215 m in the early Pleistocene stratigraphy. According to McHugh et al. (1996), clasts within the Pleistocene mass transport deposits are of varied age (middle and late Eocene, Pliocene, and Pleistocene) and contain benthic foraminifera assemblages that could have originated from water depths of <200 m to 2000 m. Clast lithologies are varied but include Eocene biosiliceous chalk. McHugh et al. (2002) interpreted these deposits as having originated from slope failure in the adjacent continental slope. The sediment profiler data collected through Site 905 also show that the seabed has small-scale relief, and coincident watergun seismic data show a chaotic facies typical of debrites (Mountain et al., 1994). Deep-tow sidescan sonar data collected southwest of Deep Sea Drilling Project Site 613 revealed blocks (olistoliths) in the upper continental rise (Farre and Ryan, 1987). In the upper slope, many reflectors in high-resolution seismic data (Pratson et al., 1994; Robb et al., 1981) are rounded and dip in to the channels (not abruptly truncated at channel walls), suggesting that the interchannel ridges there have grown by aggradation (Mitchell and Huthnance, 2007), possibly from the dilute upper parts of sedimentary flows.

Although ODP Site 905 was not located downslope of Hendrickson Canyon, the GLORIA (geological long range inclined asdic) sidescan sonar data in Figure 4B reveal that it is in a similar band of high acoustic backscattering emanating from slope canyons. High backscattering implies significant relief of features at the seabed or within a few meters of the bed, given the GLORIA system’s ability to penetrate shallow muds (Mitchell, 1993b). Lee et al. (2002) interpreted small depressions in channels of the New Jersey margin at the slope-rise transition as plunge pools caused by flow hydraulic jumps and/or excavations by impacts of sedimentary flows. A small closed-contour depression occurring at the base of Hendrickson Canyon (Fig. 4A) may have such an origin. Furthermore, 3.5 kHz sediment profiler records more widely showing abundant diffraction hyperbolae, typical of a rugged seabed, with deeper reflectors obscured (Pratson and Laine, 1989), also suggest widespread mass transport deposits such as debrites.


If the bed shear stress during flow is assumed to have been at least 100 Pa in order to be capable of excavating bedrock along the channels, and possibly 1000 Pa if rapid erosion was required to produce deep exposure, estimates of the flow density can be made, i.e., equating the component of flow weight resolved parallel to the bed to these levels of bed failure stress (Mitchell, 2005a). The following calculations are based on the bodies of the currents rather than their heads, because turbidity currents in submarine channels are considered long-lived events (Damuth et al., 1988) and the few monitoring time-series data available show that peak current velocities occurred after the front had passed (Talling et al., 2013). The importance of the body to erosion is supported by evidence of migrating knickpoints in submarine canyons elsewhere and in the data shown in Figure 3 that suggest that erosion increases where flows accelerate on increasing gradients (Mitchell, 2006); such erosional behavior is less likely caused by turbidity current heads, which vary in velocity less with gradient (Middleton, 1966a).

In earlier work on turbidity currents, a modified form of the Chezy formula for flow speed, U, was proposed for the bodies of nonaccelerating currents (Middleton, 1966b): 
where h is flow depth, S is the bed gradient and ρ′ is the fractional excess density of the flow relative to its surroundings, i.e., (ρf – ρw)/ρw, where ρf is flow density and ρw is ambient water density. The product f(1 + α) is a modified Darcy-Weisbach factor in which f represents friction with the bed and fα represents the effect of momentum transfer from entrainment of ambient water into the flow. For Froude supercritical flow, Komar (1971, 1975) suggested the value α = 0.9, which would imply that approximately half of the flow weight is supported by basal friction and half by entrainment. Estimates of α compiled by Sequeiros (2012) vary by an order of magnitude for supercritical flows, with one extreme value of 8, although the average is visually estimated to be <0.9.
To evaluate the basal flow stress, a modified form of Equation 1 was developed as follows. The flow was assumed to be in local equilibrium (an assumption returned to later). If the flow is not accelerating, the specific weight of the flow in ambient water (Δρgh, where Δρ is ρf – ρw) must be supported, implying that the sum of the shear stresses above and below the flow must balance the component of that weight resolved parallel with the bed, Δρghsinθ (using h in place of R, i.e., the flows are much wider than they are deep). Thus, if there were no friction above the flow, the basal shear stress would then be simply Δρghsinθ. To evaluate the effect of upper stresses, note first that boundary stress τ0 in turbulent flows is typically proportional to U2; thus, from Equation 2, τ0 is inversely proportional to (1 + α) (i.e., not inversely with the square root of [1 + α], as appears in Equation 2). To modify Equation 1 for submarine flows, we replace R with h, ρ with Δρ and divide by (1 + α): 

Note also that with α = 0 and air for the ambient fluid (Δρ = ρɸ), Equation 3 reduces to Equation 1. With α = 0.9, the ratio 1/(1 + α) equals 0.53. Using the extreme α = 8 and a lower bound of α = 0, the full potential range for the ratio 1/(1 + α) is 1–0.11, though in practice it is likely to be modestly larger than 0.53 given the average α < 0.9.

Equation 3 can in principle be inverted to estimate minimum Δρ, assuming that there is a set of processes involved in excavating bedrock channels by quarrying similar to those occurring in rivers and that they require similar minimum basal flow shear stress. For the submarine channels, bed gradients were measured from the multibeam sonar data over kilometer length scales, reflecting the greater dimensions of the flows compared with rivers (an order of magnitude or more in height) and because the gradients may have been somewhat different during bedrock erosion compared with the modern channel gradients, so short-scale gradients would not in any case be representative. Because quarrying likely occurs during surges in fluid velocity, it is not entirely clear that a 100 Pa threshold is appropriate for submarine channels, and properties derived using Equation 3 may be somewhat biased; velocity fluctuations associated with Kelvin-Helmholtz instabilities in turbidity currents (Cartigny et al., 2013; Kneller and Buckee, 2000) probably lead to different in situ velocity and stress fluctuations (Best et al., 2005) compared with rivers. Furthermore, form drag supports some of the flow weight (Garcia and Parker, 1993), but cannot be evaluated where submarine bedrock channel floors are not observable. Nevertheless, the results are worth considering if the rough nature of these calculations is accepted and we are mindful that these issues may bias the results.

The values used in the calculation are listed in Table 1. The maximum heights (h) of the flows were measured from the channel reliefs in cross sections drawn where shown in Figures 2 and 3 (difference in elevation between the channel thalweg and the mean elevation of the two adjacent peak elevations of the interchannel ridges). Flows are expected to avulse (spread out) where their thicknesses exceed these heights, so h values provide maximum flow thicknesses. For Monterey Canyon, because the lower 100 m slopes in the cross sections suggest intense scouring by the dense, high-velocity part of the flows, an alternative flow density for h = 100 m was also calculated. A further calculation with h = 10 m is also provided based on the observations suggesting dense sandy flows (Talling et al., 2013). The channel gradients were measured by regressing the elevation data collected along the down-channel lines shown in Figures 2 and 3. For Hendrickson Canyon these closely follow the bedrock surface, but the channel floor in the Monterey Canyon is buried by recent sediment so the gradient value there is less certain.

Table 1 shows values of Δρ computed by inverting Equation 3 using the values of h and sinθ given, assuming that α = 0.9. The equivalent volumetric concentrations of solids are also given, based on a quartz grain density of 2650 kg m–3. These values were computed assuming a minimum τ0 = 100 Pa; if instead τ0 = 1000 Pa were assumed, the density and concentration values would be 10 times larger.

Profiles of excess density with altitude above the bed for supercritical flows vary from concave upward to convex upward, with the greatest densities near the flow bases reflecting the effect of fallout of particles (Sequeiros et al., 2010). If the density profile is assumed to be triangular (decreasing linearly from a maximum at the flow base) and we allow the full height of the flow to be double the 100 m height of dense flowage in Monterey Canyon, the peak excess density at the base of the flow would be 11–110 kg m–3 (0.6%–6% solids concentration) with 100–1000 Pa shear stress constraints. These values would be 10 times larger if the 10 m height were used.

In the sites above the escarpment lips in Hendrickson Canyon, the flows are likely to have been accelerating, as occurs in the drawdown reaches of rivers above waterfalls. The flow would then not be in equilibrium, as was assumed here. The accelerations in drawdown reaches should lead to greater bed shear stress than more nearly equilibrium flow upstream away from the escarpment (this is suggested by the narrow slots appearing close to the lips; Mitchell, 2006). The flow also thins vertically to maintain discharge relative to the flow father upstream. The calculations based on 100–1000 Pa bed shear stress and equilibrium flow therefore lead to overestimates of the flow weight and density. This opposes the bias suggested earlier for the values in Table 1 for Hendrickson Canyon and creates greater uncertainty for those values. We include these values only for general comparison with the Monterey values, in practice concentrating on reconstructing debris flow properties, as such flows are more likely in Hendrickson Canyon.


As the sedimentary flows occupying the New Jersey slope channels appear to have been dominated by debris flows (McHugh et al., 1996, 2002), the following calculations address their properties. As with the turbidity currents, the constraints assumed are that flow basal shear stress must have been >100 Pa, and perhaps may have reached 1000 Pa, to cause rapid erosion. The debris flow weight per unit area was computed from τ0/sinθ and expressed in terms of equivalent thickness of pre-failure sediment of wet bulk density 1700 kg m–3 by dividing the computed weight by Δρg. This calculation assumes that the flow weight is fully supported by bed friction, ignoring downflow accelerations and pressure gradients. In some laboratory experiments of debris flows, turbulence developed above them (Felix and Peakall, 2006; Hampton, 1972) implies that some of the flow weight is supported above as well. Numerical modeling by Norem et al. (1990) of a submarine flow of clay-rich material suggested that ∼30% of the flow weight was supported by resistance above the flow. Although stresses above the flow are unlikely to approach those at the base of the flow, the calculation of flow thickness was repeated assuming that only half the weight was supported at the base, conservatively representing this uncertainty. The resulting values are plotted in Figures 5A and 5B for 100 and 1000 Pa bed shear stresses, respectively (bold vertical bars represent the range assuming fully or only half flow weight supported at base). The thicknesses estimated assuming that the flows are fully supported at their base are also suspected to be minima because of the effects of flow heterogeneity.

The other lines shown in Figures 5A and 5B represent various conditions for submarine debris flow computed by Talling (2013) and Talling et al. (2012) based on the rheological properties of kaolin-water mixtures (Coussot, 1995). The heavy dotted lines show threshold thicknesses for debris flow motion based on the sediment yield strength τy, computed from τy/Δρgsinθ. This calculation assumes that the flow base is coupled to the bed (note that in Fig. 5A the thickness of Talling et al. [2012] for threshold of motion of a 1000 Pa yield strength debris flow coincides, as would be expected, with the thickness computed assuming a 1000 Pa basal shear stress and similarly for the 100 Pa values in Fig. 5B). However, motion of thinner flows is possible if the base decouples from the bed by hydroplaning. This smaller flow thickness was computed by Talling et al. (2012) using information from Mohrig et al. (1998). The remaining (fine dotted) lines (Fig. 5) represent the boundaries between turbulent and nonturbulent flow based on a criterion of Hampton (1972) and with flow speed computed using a simple analytical expression for laminar flows based on the mixture viscosity. The graphs in Figures 5A and 5B were computed with kaolin-water mixtures of yield strengths 100 and 1000 Pa, respectively.

Locat and Lee (2002) showed graphs of the dependence of flow shear strength on liquidity index. Using the average liquidity index of 9 cores analyzed by McGregor et al. (1979) of 1.41 and Equation 8 of Locat and Lee (2002), a flow shear strength of 823 Pa is predicted, suggesting that the graph for a stiff flow (Fig. 5A) is the more appropriate. This would be in keeping with the preservation of coherent mud structures in cores (McHugh et al., 1996, 2002) that suggests that turbulence was modest.

According to Talling (2013) kaolin is a weak clay mineral, and in experiments of Hampton (1975), kaolin-water mixtures had smaller yield strengths than mixtures with montmorillonite. As the New Jersey slope sediments contain a variety of other clay minerals (chlorite, smectite, and illite; McHugh et al., 1996), debris flows on the slope potentially have greater yield strengths for given densities than modeled with the kaolin. Thickness thresholds may therefore in practice be greater than shown in Figures 5A and 5B.


The flow density estimates for Monterey Canyon can be compared in a general way with density estimates for turbidity currents in the literature to judge if they are realistic. According to Sequeiros (2012), turbidity currents that have been measured in the field usually have <1% volume concentrations of solids (a 16.5 kg m–3 excess density for solids density 2650 kg m–3) and only rarely exceed 5% (82.5 kg m–3 excess density). These values largely represent dilute marine flows or underflows in reservoirs; particle concentrations within large natural turbidity currents capable of transporting sediment to below the continental slope have not been measured in the field (Talling et al., 2013) and instrumentation is often damaged by larger flows (Paull et al., 2003). Komar (1975) suggested that a 200 kg m–3 excess density might be possible for such large turbidity currents. Van Tassell (1981) reconstructed properties of a major turbidity current based on grain-size characteristics of a turbidite on the Silver Abyssal Plain that suggested a 70 kg m–3 density excess. Bowen et al. (1984) estimated a more modest 0.2% (4 kg m–3 excess density) for Holocene turbidites on the Navy submarine fan off California. Although some of these values may be extreme, even the dilute measured flows (Sequeiros, 2012) have densities that exceed the minimum density required for bedrock erosion, according to the results in Table 1 for 100-m-thick Monterey Canyon flows, so turbidity currents seem generally capable of eroding bedrock by quarrying. Even the 53 kg m–3 excess density for the 10 m flow thickness is within the range of above values. As a 100 m flow depth is smaller than the reliefs of many other canyons or slope channels (Goff, 2001) and bed gradients in Monterey Canyon are typical of slope canyons globally (Harris and Whiteway, 2011), bed shear stresses are likely to reach levels capable of eroding bedrock more generally. The outcrops of indurated sediment or rock in lower continental slope canyons are therefore explainable by erosion by sedimentary flows.

Further corroboration is provided by observations of coarse particles; Komar (1970) interpreted the presence of cobbles in some submarine channels as implying a minimum shear stress for bedload movement. Particles as large as boulder size were apparently mobilized by the 1929 Grand Banks turbidity current (Piper et al., 1988). Vibro-coring in the Monterey Canyon floor has recovered gravel particles and clasts of dimensions larger than the 7.65 cm diameter of the corer and therefore coarser than gravel (Paull et al., 2010). Shear stresses required for bedload movements of cobbles would need to have exceeded ∼30–100 Pa based on measurements in streams (Carling, 1983), lending further support to the view that the minimum 100 Pa shear stress is occasionally exceeded in Monterey Canyon.

Where the calculations were based on the full height of each canyon, the derived excess density values in Table 1 for Hendrickson Canyon (2.4 and 2.0 kg m–3) are larger than the 0.88 kg m–3 value for the Monterey Canyon, reflecting the smaller relief of Hendrickson Canyon (although this difference may also be an artifact of acceleration). Although extreme, these are within the range of acceptable values for turbidity currents (Sequeiros, 2012), so erosion by such currents cannot be ruled out on this basis. However, erosion by debris flows is suspected to be more likely given the predominance of debrites and other mass transport deposits (McHugh et al., 2002), so the estimated thickness values plotted in Figure 5 are discussed here. The shear stress conditions for bedrock erosion suggest that bedrock-eroding flows must be at least 1–10 m thick in terms of 1700 kg m–3 sediment density. These thicknesses are minima also given potential flow dilution with ambient water and dilation of the failed material (for such thin flows, with the gradients measured ∼1 km from the escarpments, accelerations are not considered to have biased these values greatly). The head of Hendrickson Canyon in Figure 4A and cross sections in Figure 6 from this and other canyon walls show that embayments of many tens of meters of relief are common. Flow weights corresponding to 1–10 m sediment thicknesses therefore seem reasonable.

As the Hendrickson Canyon debris flows are suspected to have been stiff (based on remold strengths, shear strength predicted from liquid limits, and preservation of coherent structures in deposits), flows of ∼1000 Pa yield strength are suspected. The predicted thicknesses should therefore be compared first with the predictions of Talling (2013) and Talling et al. (2012) in Figure 5A rather than Figure 5B. If 1000 Pa basal shear stresses were involved in erosion, the flow would have likely been laminar. For flowage to have occurred at the low bedrock erosion stresses (nearer to 100 Pa) but still involving a stiff 1000 Pa yield strength flow, the flow front would need to have been hydroplaning. In that case, the thinner flow (and hence smaller flow weight) would have been supported via a rapidly shearing thin basal water layer with a 100 Pa basal shear stress. Alternatively, shear stresses of 100 Pa could be achieved without hydroplaning if the flow had a yield strength of 100 Pa as in Figure 5B and only moderately turbulent. Reconstructions in which flow was laminar or only modestly turbulent are preferred in order to explain the sediment structures observed in the deposits. Thus, stiff debris flows ∼10 m thick and moderately turbulent, or perhaps as little as 1 m thick if hydroplaning, or less stiff debris flows ∼1 m thick, would be compatible with the rheological arguments of Talling et al. (2012) based on these equilibrium flow weight arguments.

Computer-generated images of the geomorphologies of submarine canyons can appear remarkably similar to those of river networks (Belderson and Kenyon, 1976; Belderson and Stride, 1969; McGregor et al., 1982). Measurements on multibeam bathymetry data show that the canyons can also be similar in a quantitative geometrical sense. For example, the channels in many networks obey Playfair’s law (tributaries join main stems at the same elevation without hanging; Mitchell, 2004), channel long profiles are commonly concave upward with concavity indices similar to those of rivers (Mitchell, 2005b), and the morphologies of knickpoints (anomalously steep reaches) also vary from smooth to sharp-lipped, suggesting varied erosion mechanisms (as in rivers; Mitchell, 2006). Although how fluvial-like geomorphology develops in submarine slopes is still unclear, if muddy flows can erode indurated sediment, as suggested here, resistance to erosion would not impede that development.

Given that turbidity currents and debris flows are apparently capable of eroding bedrock, the question arises as to why rock is not exposed more commonly in canyon walls; the impression obtained from the literature is that, while consolidated sediment is commonly exposed, rock or indurated sediment is only exposed in a minority of the continental slope canyons globally. At large scales and over long periods, continental margins tend to prograde seaward from sediment deposition on the slope (e.g., Fulthorpe and Austin, 1998). These landscapes are therefore aggrading, behavior opposite to that of steep regions on the continents, which are commonly denuding. Part of the reason may be that most submarine sedimentary flows, even though they are supercritical, deposit sediment on the continental slopes (Gerber et al., 2009), and in many cases channel relief may be maintained by aggradation on the interchannel ridges or terraces (Mitchell and Huthnance, 2007; Paull et al., 2013) rather than by entrenchment of the channels. Furthermore, the wide range of turbidite thicknesses in sedimentary basins, in places spanning at least two orders of magnitude (Beattie and Dade, 1996; Rothman et al., 1994), combined with observations of variably coarser bases of the thicker turbidites (Sadler, 1982; Talling, 2001), suggests that slope channels probably undergo a wide range of bed shear stress from turbidity currents (Mitchell, 2006). We therefore need to consider the effect of variability of shear stress in submarine channels, much like in studies of bedrock river channels (Tucker and Bras, 2000).

If a bed shear stress closer to 1000 Pa is needed for rapid erosion, the lack of extensive bedrock exposure may be a constraint on the upper limit of the frequency distribution of typical flow stress at continental margins. For channel gradients with typical θ = 2° and with α = 0.9, if the shear stress rarely exceeds 1000 Pa, the flow mass excess Δρh per unit area may then only rarely exceed ∼5000 kg m–2, or the mean bed normal stress Δρgh only rarely exceed ∼50 kPa. Alternatively, the sediments accumulated within canyons by smaller flows (Paull et al., 2005a) may not be fully mobilized during the extreme flows, leaving the bedrock protected and less rapidly eroded, as has been suggested for bedrock river channels with thick alluvial cover (Sklar and Dietrich, 2001). This question may ultimately be answered because of the increasing interest by the oil and gas industry in seismically imaging and drilling of deep-water channel and fan deposits.


To explain bedrock erosion by quarrying in Monterey Canyon, a minimum flow excess density of 5–50 kg m–3 is predicted if the flows are assumed to be 100 m thick and a minimum 100–1000 Pa bed shear stress is required for erosion by quarrying. If the flows were only 10 m thick, a minimum 50 kg m–3 excess density is predicted for a minimum 100 Pa bed shear stress. Given that these values are generally within the range of predicted flow densities, erosion by quarrying provides a viable explanation for exposure of rock along the canyon walls. Furthermore, as the geometrical properties of Monterey Canyon are typical of other canyons, these results suggest that bedrock exposures within canyons more generally are explainable by erosion by sedimentary flows.

If exposure of porcellanite in Hendrickson Canyon is attributed to quarrying by debris flows, the equivalent minimum thicknesses of the flows also predicted with the 100–1000 Pa basal shear stress constraint are 1–10 m. These are comparable with depths of embayments of the upper slope. Based on liquid limits in geotechnical measurements and other observations, the flows were likely to have been stiff, with yield strengths of ∼1000 Pa. Such flows would have been either laminar if hydroplaning or modestly turbulent if toward the 10 m end of the thickness estimates.

Because sedimentary flows produce sufficient basal shear stresses to erode, these results also prompt the question of why rock is not more commonly exposed in the canyons of the world’s continental slopes. In contrast with denudation on the continents above sea level, continental slopes prograde with time, so they are essentially aggrading landscapes. Addressing this issue may involve reconstructing the effective upper limit of the frequency distribution of basal shear stress produced by sedimentary flows on continental slopes. Assuming that bed shear stress rarely exceeds 1000 Pa and that bed slope angles in canyons are typically ∼2°, the mass of flow solids (product of excess density and thickness) only rarely exceeds 5000 kg m–2 and bed normal stress rarely exceeds 50 kPa. Alternatively, in these commonly aggrading landscapes, deposits of smaller flows filling canyons are not completely mobilized by the largest flows, causing an armoring of canyon beds.

The multibeam sonar data contributing to this study in Figure 2 were generously provided by the Monterey Bay Aquarium Research Institute. Figures in this article were prepared with the GMT (Generic Mapping Tools) free software system (Wessel and Smith, 1991). Pete Talling is acknowledged for useful advice on the assumptions and Monterey Canyon work. Thorough and insightful reviews by Joris Eggenhuisen, Andrea Fildani, and an anonymous reviewer led to significant improvements of this article.