Hybrid submarine flows comprising turbidity current and cohesive debris flow: Deposits, theoretical and experimental analyses, and generalized models Open Access

Submarine flows of sediment driven by their excess density can run out for tens to hundreds (and on occasions thousands) of kilometers, sometimes across remarkably low seafloor gradients of 0.1°–0.01° (Talling et al., 2012a). They dominate sediment transport into many parts of the deep ocean, and produce some of the most extensive and voluminous sediment accumulations on Earth. Understanding these flows is challenging because they are remarkably difficult to monitor directly. The speeds of flows that run out beyond the continental slope have been measured accurately in only a few locations (Heezen and Ewing, 1952, 1955; Piper et al., 1999; Piper and Savoye, 1993; Mulder et al., 1997; Khripounoff et al., 2003, 2009; Vangriesheim et al., 2009; Hsu et al., 2008; Carter et al., 2012), and the vertical profile of sediment concentration has never been measured directly within such long run out flows. This paucity of direct observations provides a stark contrast to other major sediment transport processes such as rivers that have been closely monitored in action. Understanding submarine flows remains a major challenge, as the only record we have of most flows is the sediment deposit that they leave behind.

Previous work has illustrated how long run out distances can be achieved by turbulent sediment suspensions, called turbidity currents, that incrementally deposit layers of clean sand and mud. This work includes seminal contributions by Bill Normark and colleagues that elegantly combined field observations from the modern seafloor and ancient rock with quantitative modeling (Bowen et al., 1984; Normark, 1989; Normark et al., 1993, 2002, 2006). Here I address a second type of long run out flow that reaches the distal parts of submarine fans. These hybrid flows include both turbidity current and mud-rich (cohesive) debris flow (Talling et al., 2012a), and their deposits comprise mud-rich debrite sand encased within turbidite clean sand and mud. This type of deposit was described first by Wood and Smith (1958), and was noted subsequently by Van Vliet (1978), Hiscott and Middleton (1979, 1980), and Ricci-Lucchi and Valmori (1980). However, only recently has it become apparent that this type of “linked” debrite-turbidite bed is common in many locations worldwide (Fig. 1; Haughton et al., 2003; Talling et al., 2004; Amy et al., 2009; Haughton et al., 2009). Hybrid flow deposits are the norm rather than the exception in the distal parts of some submarine fans (Haughton et al., 2003, 2009; Talling et al., 2004, 2012a, 2012b), and they can involve very large amounts of sediment. One of the hybrid submarine flows that is described here transported 10 times the annual sediment flux of all of the world’s rivers combined (Talling et al., 2007a). Understanding hybrid submarine flows is therefore important for determining how sediment is transported globally. Hybrid deposits often contain large amounts of organic carbon, and may be a significant process for burying and sequestering that organic carbon in deep water (Galy et al., 2007; Saller et al., 2008). Hybrid debrite-turbidite deposits are also important because they are a significant departure from widely cited models for submarine flow deposits, such as those of Bouma (1962), Lowe (1982), Mutti (1992), and Mulder and Alexander (2001). Such generalized models capture how we think submarine flows most commonly behave at a single location (Bouma, 1962) or evolve spatially along the flow path (Lowe, 1982, fig. 10 therein). Generalized models are therefore needed that include hybrid flow deposits, and that predict how hybrid flows originate and evolve.

Hybrid debrite-turbidite beds are also important because they occur in subsurface petroleum reservoirs (Haughton et al., 2003, 2009), some of which hold large amounts of oil and gas. Mud-rich debrite sandstone has much lower permeability, and hence lower reservoir quality, than clean turbidite sandstone. Muddy debrite sandstone layers act as baffles to fluid flow within the reservoir (Amy et al., 2009). It is therefore important to predict more accurately the location, extent, and shape of debrite and turbidite sandstone layers in reservoirs containing hybrid beds, in order to extract efficiently oil and gas reserves.

This contribution stresses how quantitative insights from theory and laboratory experiments play an important role in testing hypotheses for the origin of hybrid beds, and their varied features. Large-scale field data sets from modern systems are also important for documenting relationships between hybrid beds and seafloor gradient, and provide direct information in source areas, which are typically inferred with less certainty from ancient core or outcrop data. In a few locations, individual ancient hybrid flow deposits or packages of beds can be mapped for long distances, providing important insights into flow transformation and evolution, which goes beyond the vertical sequence of deposits seen at one location. This contribution therefore assembles key information from theory, laboratory experiments, modern systems, ancient outcrops, and subsurface cores.

The first aim of this contribution is to summarize the field observations from modern submarine fans, ancient rock outcrops, and subsurface oil and gas reservoirs (Figs. 1 and 2). What features are observed consistently in hybrid beds from these disparate locations? Which features differ, thereby defining different types of hybrid bed?

The second aim is to provide a quantitative framework for understanding hybrid flows that combines observations from laboratory experiments and theoretical analysis. When will sand grains or mud clasts of different sizes be supported by cohesive debris flow? What determines the mud content in the matrix of the debris flow deposit? Small amounts of cohesive mud can dampen turbulence, leading to flow transformations from fully turbulent condition to laminar plug flow. Under what conditions will such flow transformation occur, and how are they recorded by deposits? A wide-ranging theoretical analysis is presented that predicts generally how flows with increasing mud content and cohesive strength will behave. This continuum is illustrated using the rheology and consolidation behavior of a suspension with increasing kaolin clay concentration. Observations from laboratory experiments involving sediment flows with increasing cohesive strength are then summarized, together with how these experimental flows might scale to much larger and often faster submarine flows. This theoretical and experimental evidence is then used to provide a series of general models for how hybrid flows with increasing cohesive strength will tend to evolve and deposit sediment.

The final aim is to answer a series of key questions about hybrid beds. In which depositional settings do hybrid beds occur, and where are they most common? How do cohesive debris flows originate within hybrid submarine flows, and why are the cohesive debris flow deposits often absent in proximal locations? Why do cohesive debrites occur consistently at the same level within most hybrid beds? How can lateral changes sometimes occur from turbidite-dominated to debrite-dominated hybrid beds, without changes in the total bed thickness? What determines the location, extent, and planform shape of cohesive debrites within hybrid beds? What factors control the presence (or lack) of stratigraphic clustering of hybrid beds? What are the implications of this work for predicting the distribution and geometry of linked debrite-turbidite beds in subsurface hydrocarbon reservoirs? A comparison is made to the previous classification scheme for hybrid flows of Haughton et al. (2009), and it is shown how this work complements, amends, and extends that classification scheme. I conclude with suggestions for further work to clarify the origin and behavior of hybrid submarine flows.

The terminology used herein was described in more detail by Talling et al. (2012a). The term turbidity current is used to denote flow that is fully turbulent and that deposits clean turbidite sand incrementally in a layer by layer fashion. Laboratory experiments suggest that incremental deposition of clean sand tends to occur during turbulent (rather than laminar) flow conditions (Sumner et al., 2009). Differential settling and spatial or temporal changes in flow speed lead to graded clean-sand turbidite layers unless flow is steady and uniform (Kneller and Branney, 1995) or suppressed at high concentrations. If outsize mud clasts are present, they occur along discrete horizons in the deposit.

High-density turbidity currents are characterized by high near-bed sediment concentrations that produce hindered settling and dampen turbulence. Sediment deposition from high-density turbidity currents occurs from high-concentration near-bed layers that may be laminar or weakly turbulent (traction carpets or laminar shear layers; Hiscott, 1994; Sumner et al., 2008). These near-bed layers differ from debris flows because they are driven by the overlying flow (and not their own downslope weight), and they produce planar laminated (T_B) or massive (T_A) deposits, the thicknesses of which are unrelated to the overall thickness of the flow (Kuenen, 1966a; Bannerjee, 1977; Arnott and Hand, 1989; Sumner et al., 2008). A nondepositional (bypassing) flow that has a high near-bed sediment concentration, but that is turbulent, is also termed a high-density turbidity current.

If sediment fallout rates are sufficiently slow, low-density turbidity currents produce ripple-scale cross-lamination (T_C), overlain by fine-scale fine-grained planar laminae (T_D), or underlain by planar lamination produced by low-amplitude bedwaves (T_B-2).

The term debris flow is used for flows that are laminar (or almost laminar) and deposit sand and mud in an en masse fashion. Experiments suggest that en masse settling, better termed consolidation, tends to occur from laminar plug flows (Sumner et al., 2009). Sediment can be supported in debris flows due to matrix strength, buoyancy (reduced density contrast between clasts and surrounding matrix), grain to grain interactions, and excess pore fluid pressures. The term cohesive debris flow is used here to denote debris flows in which the amount of fine mud, and resulting cohesive strength of muddy fluid, is sufficient to support at least sand grains. This does not preclude the simultaneous operation of other support processes on occasions, such as excess pore pressure. Debris flows in which cohesive strength of muddy fluid is insufficient to support sand are termed poorly cohesive, or entirely noncohesive if they contain no cohesive mud (as in Lowe, 1976). The term liquefied flow denotes that excess pore pressures either fully or mostly support sediment, such as in the experiments of Iverson et al. (2010) and Breien et al. (2010). The term hybrid flow denotes a single flow event that comprises both turbidity current and debris flow (Haughton, et al., 2009).

Larger and smaller grains tend not to segregate from a debris flow, unless it has very low strength, thereby producing predominantly ungraded deposits (debrites) (Sumner et al., 2009). Exceptions are large outsize clasts that may preferentially settle, and the uppermost few centimeters of the debrite that may be graded due to mixing and dilution by overlying seawater. Debrites lack sedimentary structures formed by bedload reworking. Chaotically distributed mud clasts occur in some debrites, and a sharp grain size break always occurs at the top of the debrite interval. Debris flows produce a deposit having a thickness more closely related to the flow thickness, and debris flows can come to a halt (freeze) such that their deposits pinch out abruptly in areas of low relief (Amy et al., 2005; Amy and Talling, 2006; Talling et al., 2007b). Nondepositional (bypassing) flows that are laminar are termed debris flow. The term fluid mud (McAnally et al., 2007) is used to denote laminar flows that contain only mud and that lack sand or larger mud clasts.

Transitional flow denotes the turbulence structures that occur as flow transforms between fully turbulent and fully laminar (plug flow) states (Baas and Best, 2008; Baas et al., 2009; Haughton et al., 2009; Sumner et al., 2009; Baas et al., 2011).

The muddy sandstone intervals that are included in this study have features indicative of deposition via cohesive debris flow (Figs. 1 and 2). Higgs (2010) noted that postdepositional (in situ) liquefaction of turbidite sand and overlying background (hemipelagic) mud can generate muddy sandstones, and that these muddy sands could contain clasts of overlying background mudstone. Higgs (2010) proposed that such liquefaction could be generated by seismic shaking. Such a process may generate muddy sandstones in other locations; however, I concur with Haughton et al. (2010), and do not think that this process generated the muddy sandstone intervals described here.

Field observations from a series of deposits have been summarized and synthesized from more than 20 widespread locations. This breadth of observational data is needed to illustrate the considerable variability in cohesive debrite and hybrid bed character (Figs. 1 and 2).

Cohesive debris flows can produce very thick (tens of meters) deposits that contain abundant clasts chaotically distributed within a homogenized muddy sand matrix. These very thick debrites have a lobate or blocky morphology, often with sufficient relief to be visible in bathymetric mapping of the seafloor, and can often be mapped back to the vicinity of initial slope failure on the modern seafloor. Examples of very thick cohesive debrites include those produced by the 1929 Grand Banks slope failure off Newfoundland (Piper et al., 1999; Mosher and Piper, 2007), the Storegga Slide offshore Norway (Haflidason et al., 2005), the BIG’95 slide on the Ebro continental margin (Lastras et al., 2005), and those resulting large-scale slope failures off the northwest African margin (Masson et al., 1993, 1997, 2010).

Glacigenic debris flows on trough-mouth fans also produce very thick (10–50 m) cohesive debrites, comprising chaotically arranged clasts in a generally ungraded finer grained matrix (King et al., 1996; Laberg and Vorren, 2000; Kilfeather et al., 2010). Individual debris flow lobes can extend for 100–200 km and be 2–10 km wide, and contain 0.1–50 km³ of sediment. The debris flows initiate on gradients of ∼1°–3°, and terminate on gradients of ∼0.2°–0.5°, with multiple lobes stacked and offset (King et al., 1998; Laberg and Vorren, 2000; Elverhoi et al., 2007). The final thickness of the debrite deposits suggests that the debris flows had high yield strengths of 600 to ∼13,000 Pa, assuming (1) that their density was the same as at present (∼1800 kg m^–3) and (2) seafloor gradients of ∼0.5°–2°. These very thick debris flows tend to lack any overlying turbidites, either due to limited mixing with seawater or because the turbidity current bypassed on steeper slopes and only deposited sediment beyond the termination of the debris flows. The termination of the Canary debris flow in the Madeira Abyssal Plain is encased within graded turbidite sand (Weaver et al., 1994), suggesting that the debris flow and turbidity current reached the basin at a similar time, and therefore had broadly similar speeds.

Tripsanas et al. (2008) and Tripsanas and Piper (2008) described a range of debris flow and other mass flow deposits from the continental slope offshore eastern Canada, and within minibasins in the continental slope or at the Sigsbee Escarpment in the Gulf of Mexico. These include debris avalanche, coherent but deformed strata within slide or slumps, and cohesive debrites containing clasts. The debris flow deposits are typically no more than ∼3 m thick. A striking feature of this comprehensive analysis of shallow (<20 m) cores is that turbidite clean sand is rare, and that most debris flow deposits (or other types of mass flow deposit) are not associated with encasing turbidite sands. My experience of logging more than 30 (as much as 20 m long) sediment cores from the continental slope offshore from the Nile delta is that 1–4-m-thick debrite intervals are common, but they are not associated with encasing turbidite sand (also see Ducassou et al., 2013). This suggests that hybrid beds are relatively rare or absent on continental slopes. The reason for this is most likely that any turbidity current formed by dilution of debris flow material is highly mobile, and flows down the relatively steep continental slope without depositing turbidite sand. Therefore, debrites and turbidite sand layers are not deposited in the same geographic location.

Outsize hybrid megabeds comprise a clast-rich debrite interval that is tens of meters (to 250 m) thick (Fig. 1I; Johns et al., 1981; Labaume et al., 1987; Kleverlaan, 1987; Payros et al., 1999). Some megabeds have been mapped for 100 km, and contain 6–60 km³ or more of sediment (Johns et al., 1981; Kleverlaan, 1987). Clasts are supported in a muddy matrix, and can be as long as 100 m. They may comprise extrabasinal or intrabasinal strata. Clast size is often ungraded vertically, but there may be grading of the largest clasts (Kleverlaan, 1987). The debrite interval is overlain sharply by a graded turbidite comprising conglomerate, sandstone, or mudstone that may be 6–20 m thick (Labaume et al., 1987; Kleverlaan, 1987). This turbidite layer can sometimes extend further down the basin than the debris flow deposits (Payros et al., 1999). Unlike most examples of hybrid beds with thinner debrites, the debrite interval within the megabed is not typically underlain by turbidite sand, although there may be local areas of clast-supported conglomerate (Payros et al., 1999). This may be because these particularly thick debris flows were relatively fast moving and outran the turbidity current, although this is not consistent with the lack of motion within the plug recorded by intact pectin shells in some megabeds (Kleverlaan, 1987). Alternatively, these debris flows may have completely eroded any initial turbidite deposits.

Relatively thin (<2 m) cohesive debrites are described in three informative large-scale data sets from modern submarine fans, showing how cohesive debrite location and extent are related to seafloor morphology.

A hybrid flow deposit has been mapped for more than 1500 km between the Agadir Basin and the Seine and Madeira Abyssal Plains offshore northwest Africa, providing an unusually complete view of hybrid flow evolution (Fig. 3; Wynn et al., 2002; Talling et al., 2007a; Frenz et al., 2008; Wynn et al., 2010; Sumner et al., 2012). This flow contained ∼130 km³ of sediment, and was triggered by a landslide near the upper Agadir Canyon. The flow was initially very powerful. It eroded ∼1 m of sediment in locations ∼300 m above the canyon floor (Talling et al., 2007a; Wynn et al., 2010), and cut a spectacular field of ∼1-km-wide by ∼10-m-deep scours at the canyon mouth (Huvenne et al., 2009; Macdonald et al., 2011). Canyon-mouth erosion provided as much as 30 km³ of mud (Huvenne et al., 2009), thereby elevating substantially the mud content of the flow. The initial fraction of mud in the flow may already have been considerable, as the initial slope failure involved muddy continental margin sediment.

The flow deposited very little sediment beyond the canyon mouth, across 150 km of seafloor (Fig. 3). A remarkably subtle (but fourfold) slope break from 0.05° to 0.02° eventually triggered deposition from both the turbidity current and debris flow (Fig. 3; Talling et al., 2007a; Sumner et al., 2012). Similar ratios of benthic foraminifera in turbidite and debrite divisions show that sand-sized particles were well mixed within the flow as it moved across the area of sediment bypass (Talling et al., 2007a). The hybrid flow was therefore at least weakly turbulent during its initial stages.

The debris flow deposit is <150 cm thick and comprises mainly ungraded sandstone with high mud matrix content (Fig. 3). It extends for 250 km and is as wide as 80 km in places (Fig. 3; Talling et al., 2007a). The debrite is found in two separate low-gradient areas (<0.02°), and is absent across an area of slightly steeper (0.05°) slope in the central Agadir Basin (Fig. 3; Wynn et al., 2010). There are no grain-size discontinuities or other evidence of debris flow bypass within the turbidite deposit in this area of steeper slope (Fig. 3B). Faster velocity therefore most likely caused local transformation of debris flow back into turbidity current in this area (Wynn et al., 2010).

This data set is important because it illustrates the behavior of clast-rich cohesive debris flows in a well-studied submarine fan system. Debris flow deposits with a remarkable frond-like planform shape occur at the distal fringe of the most recently active lobe of the Mississippi fan (Fig. 4A; Nelson et al., 1992; Twichell et al., 1992, 1995, 2009; Talling et al., 2010). This frond-like pattern is most likely due to local breaching of levees, such that debris flow material flowed down the dip of the levees (Talling et al., 2010). Nelson et al. (1992), Twichell et al. (1992), and Schwab et al. (1996) concluded that these deposits comprise a complex arrangement of thin interbedded turbidity current and debris flow material. In Talling et al. (2010), we suggested that some intervals previously described as in-place turbidites are actually clasts within a single debrite interval that is ∼1–2 m thick (Fig. 4C). These clasts include boulders with diameters >50 cm, which are wider than the core barrel (Talling et al., 2010). This difference in interpretation highlights the issue of distinguishing intervals of intact strata from large clasts within a debrite using relatively narrow core (Haughton et al., 2009; Jackson et al., 2009), especially because elongate bedded clasts can be flat lying.

The debrite is underlain by a thin (18–30 cm) layer of clean massive sand (Fig. 4C). Emplacement of this clean sand layer appears to be closely linked to debris flow deposition because the clean sand layer pinches out in a location similar to that of the overlying debrite. In Talling et al. (2010), we inferred that the basal clean sand most likely settled out from the overlying debris flows, as seen in laboratory experiments involving debris flows with low strength (∼5–10 Pa; Marr et al., 2001; Sumner et al., 2009). The debrite interval is overlain by hemipelagic mud, and lacks an overlying interval of turbidite sand or mud that is seen in many hybrid beds (Fig. 1). These debris flows failed to produce a significant turbidity current, as cores located just beyond the debrite termination contain only very thin silty laminae (Fig. 4B); this suggests that the debris flows were sufficiently slow moving to not undergo significant mixing with seawater. It is unlikely that a large turbidity current produced by the debris flow would bypass this part of the study area, as the seafloor has a gradient of only 0.06°.

The abundance of organic carbon in the debrite matrix indicates that the debris flows did not originate from failures on the distal fan lobes, or failures of the main channel levee, as they comprise deposits with lower organic carbon content. The debris flow matrix came from a source more than 600 km away near the Mississippi Canyon (Twichell et al., 1992; Nelson et al., 1992; Talling et al., 2010). Organic-rich debris flow deposits are also found near the junction of the lobe’s feeder channel and main fan channel, and plug the upper parts of the main fan channel (Nelson et al., 1992; Twichell et al., 1996). It is therefore plausible that the debris flows that reached the distal fan left a trail of debrite deposits that extends back to the vicinity of the Mississippi Canyon. This, together with the large clasts within the distal fan debrites, shows that the debris flows did not originate through flow transformation from initially more dilute flows in which clasts were supported primarily by turbulence. The largest clasts must have been carried initially by dense flow in which the clasts were buoyant or supported by matrix strength (Talling et al., 2010).

A third key data set comes from sediment lobes at the mouth of the most recently active channel-levee system in the Nile deep-sea fan (Ducassou et al., 2008; Migeon et al., 2010; Ducassou et al., 2013). This data set is important because it shows how channel-levee topography affects debrite distribution. A unit with high acoustic backscatter covers an area of ∼100 × 60 km, and has an abrupt fingered termination that resembles the shape of the Mississippi fan deposits (Migeon et al., 2010). Sediment cores confirm that this high-backscatter unit comprises a series of 3–6 thin (0.3–1.2 m) debris flow deposits, as well as intervening thin turbidite beds and hemipelagic mud (Migeon et al., 2010). Debrite intervals contain numerous mud clasts that are chaotically distributed in a mud-rich matrix (Ducassou et al., 2009; Migeon et al., 2010; Ducassou et al., 2013). The low seafloor gradients (<0.2°) and debrite thickness suggest that these debris flows had low yield strength. Some of the clast-rich debrites are underlain directly by relatively thin (5–10 cm) clean sand or silt intervals. Some of these basal clean sand or silt intervals have well-developed planar or cross-lamination, indicating deposition by turbidity current, most likely in the same event as the overlying muddy debrite (Migeon et al., 2010).

The fingered high-backscatter area associated with debrites surrounds the elevated channel-levee ridge (Migeon et al., 2010); this suggests that the flows responsible for the debrites reached the distal fan mainly through the channel. However, debrites are absent in two cores from low-backscatter areas immediately adjacent to the channel (Migeon et al., 2010). I propose that debris flow deposition was initiated by decreasing gradients as flows moved radially down the flanks of the channel-levee ridge. Flow adjacent to the channel was either turbidity current, or a debris flow that bypassed sediment across slopes >∼0.2°.

Studies of rock outcrops are now described starting with three of the most informative locations, where individual hybrid beds, or packages of beds containing hybrid beds, were mapped over long distances. Some of the more detailed outcrop studies include detailed analyses of grain size and mud content analyses for hybrid beds, or information from clast and matrix composition that helps to constrain the origin of sediment within the flows.

This location provides especially detailed information on hybrid flow evolution and the planform shape of debrites within hybrid beds (described in more detail in Talling et al., 2012b). It is the only ancient sequence where individual hybrid beds have been mapped out for more than 100 km (Amy and Talling, 2006; Talling et al., 2007b, 2007c, 2012b, 2012c). The beds were deposited in a relatively flat basin plain that lacked channels. Cohesive debrites are present locally in almost all larger volume flow deposits, the volume of which indicates that they were generated by slope failure. The cohesive debrite intervals are always underlain by clean sand, and overlain by turbidite sand or mud. Debrites only occur in thick beds that also contain T_A and T_B divisions, and are always absent in thin beds only comprising T_C,T_D, and T_E intervals. Debrite intervals can extend for 40–80 km down the basin axis, and at least 10–15 km across the basin. The debrite intervals are subdivided into two types with different downflow geometries.

Clast-rich debrites are underlain by massive and relatively thin clean sandstone (Fig. 5B), and pinch out abruptly over a few kilometers in a downflow direction (Fig. 6). The massive basal sand interval pinches out at the same location as the debrite, suggesting that the two intervals are formed by flows that are closely linked. The sand may originate from late-stage settling of sand from the debris flow plug, or result from a forerunning turbidity current that is very closely linked to the debris flow. The massive sand lacks evidence of dewatering structures, indicating that it did not lubricate the overlying debris flow by dewatering. The composition of clasts and abundance of organic material in the matrix indicates that these clast-rich debris flows originated outside the basin plain outcrops, but had sufficiently low strength to travel for tens of kilometers across the basin plain (Talling et al., 2012b).

Clast-poor debrites contain either no clasts, or millimeter-scale mud chips. They are typically underlain by relatively thick intervals of clean sand that contain clear evidence of deposition via turbidity current. In some locations, clast-poor debrite intervals infill the relief above dune crests (Fig. 5A), suggesting rather gentle emplacement. These clast-poor debrites gradually taper and fine downbasin, grading into graded muddy siltstone in distal sections. They are absent in the most proximal outcrops, where beds comprise only turbidite sand and mud. It is most likely that the clast-poor debris flows formed via flow transformation from initial turbidity currents that deposited the proximal turbidites (Talling et al., 2012b). This flow transformation may have been due to the development of colloidal bonds between mud particles as the flow decelerated (Sumner et al., 2009; Talling et al., 2012b, 2012c). Clast-poor debris flows appear to have had even lower strength than clast-rich debris flows (Fig. 5A), and they may have mixed more readily with surrounding seawater, causing a downflow dilution and the lateral change to deposition of graded silts in the distal basin (Fig. 6). A single bed can contain both clast-rich debrite and clast-poor debrite at different locations across the basin. This may result from lateral changes in debris flow strength within a single flow event (Talling et al., 2012b).

Ito (2008) analyzed hybrid beds within a 30-m-thick interval that could be correlated for 20 km from mid-fan channel to distal lobe settings using marker (ash band) horizons. His study included detailed analyses of grain size, and is noteworthy because the debrite matrix has a relatively low cohesive mud content, the importance of which is discussed herein (also see Talling et al., 2012c). The channel and proximal fan outcrops lack cohesive debrites, which are found only in the intermediate sections. Debrites are absent in the most distal locations that comprise only thin-bedded turbidites (Ito, 2008).

Debrite intervals are 0.2–1.6 m thick, have irregular grading patterns (Fig. 1J), and contain abundant clasts that are 2–60 cm long. The mud content in debrite intervals finer than 20 μm is 8%–14%, which exceeds the mud content (4%–5% finer than 20 μm) seen in basal cleaner sand intervals. In some cases the debrite comprises the base of the bed, with a subtle planar laminated fabric at its base (Ito, 2008, figs. 5B and 8A therein). In other examples, the debrite is underlain by massive clean turbidite sandstone that is normally graded and coarser.

Hodgson (2009) presented a detailed analysis of thin (0.1–1 m) hybrid beds based on outcrops in which sandstone packages (as opposed to individual beds) have been mapped out over tens of kilometers, thereby providing important information on debrite distribution. The correlated sandstone packages show lateral changes from relatively fixed feeder channels to lobes characterized by bed amalgamation, the depth of amalgamation in the lobe settings decreasing toward the fan fringe. The cohesive debrites only occur in distal lobe settings and on the eastern side of the fan fringe, suggesting that local topography may influence their final location (Hodgson, 2009). The debrites are first seen ∼15 km downfan from the feeder channel exposures. The lack of debrites in the proximal lobe and feeder channels suggests that debris flows either bypassed the areas without depositing debrite, or that the debris flows formed by a relatively late transformation on the distal lobe from initial turbidity currents.

Two types of cohesive debrites are seen (Fig. 1F). The first type of debrite contains abundant carbonaceous organic matter dispersed within a muddy sand or muddy silt matrix. The organic material is chaotically distributed, and the debrite intervals are 1–10 cm thick. Hodgson (2009) suggested that the organic-rich sediment originated from shelf sequences. The organic-rich debrite is underlain by subtly normally graded clean sand that lacks structures. The second type of cohesive debrites contains abundant mud clasts dispersed in a muddy sand matrix. The clast-rich cohesive debrites are typically somewhat (10–50 cm) thicker, and the clasts are not rich in organics. Hodgson (2009) inferred that the clasts most likely came from a source different from that of the organic-rich sediment on the continental slope or proximal fan. Underlying basal clean sand is typically massive, but can show lamination in its upper part. The transition between both types of debrite and the underlying clean sand is most commonly abrupt, but can be banded or loaded.

Individual cohesive debrites layers can be walked out for ∼500 m, and sometimes 1 km, but they are not laterally extensive marker horizons, in part due to pervasive bed amalgamation (Hodgson, 2009). Lateral changes occur in the same layer between the two types of organic-rich and clast-rich debrite, suggesting that they have the same overall origin. The same bed can display a lateral change from being mainly clean turbidite sand to being mainly cohesive debrite (Fig. 1F). The overall bed thickness does not change markedly across such lateral transitions, and the debrite thickness is locally compensated by the turbidite sand thickness. This geometry could suggest that the depositional processes for the cohesive debrite and massive turbidite were not dissimilar, such that a transition from one process to another could occur over a relatively short lateral distance.

Sylvester and Lowe (2004) provided an analysis of two outcrops that contain two types of muddy sandstone interval that are 10–50 cm thick, which is important because of the detailed textural data that they present (Fig. 1G). This work illustrates that distinct types of debrite can occur in the same sequence. The first type of muddy sandstone is relatively coarse, poorly graded and sorted, and has a matrix mud content of 15%–35% (Fig. 1G; Sylvester and Lowe, 2004). It can contain relatively large clasts that tend to be grouped in the upper part of the debrite interval. This type of coarser muddy debrite is underlain by massive or planar laminated, graded, clean sandstone deposited by turbidity current. The upper part of this clean sandstone can show banding and dewatering, although these features are more poorly developed than in the Britannia Formation (Lowe and Guy, 2000), as discussed subsequently. The debrite interval is overlain by a grain size break that separates it from dilute turbidity current deposits comprising ripple cross laminated turbidite sand.

The second type of muddy debrite sandstone is distinctly finer grained, better sorted, and has 27%–37% mud matrix (Sylvester and Lowe, 2004). It is the only debrite deposit described in this contribution that occurs within ripple cross-laminated sandstone, as all of the other examples are not underlain by such T_C divisions. The debrite is generally clast poor, but can contain occasional small mud clasts. Its grain size distribution is similar to that of the surrounding T_C divisions, and it is not separated from these adjacent intervals by a grain size break.

Hybrid beds have been described in a series of other ancient rock sequences representing lobe or more distal depositional settings. Haughton et al. (2009) and Pyles and Jennette (2009) illustrate the occurrence of hybrid beds containing cohesive debrites in the Ballybunion and Inishcorker outcrops of the Carboniferous Ross Sandstone in western Ireland, which have a sheet-like geometry and represent a lobe setting. Haughton et al. (2009) attributed the hybrid beds in the basal Ross Formation to a period of fan initiation associated with incision up slope, due to a slope that was ‘out of grade,’ although the coeval upslope outcrops are not described. Fan fringe or lobe deposits in the Aberystwyth Grits include beds that contain cohesive debrite intervals that are 0.5–1.5 m thick with particularly large clasts, sometimes to several meters in length (Wood and Smith, 1959; Talling et al., 2004). Hiscott and Middleton (1979) describe ∼40-cm-thick intervals of sand in the Ordovician Tourelle Formation in Canada with abundant mud matrix that contain numerous clasts, which may be as much as several meters long. These cohesive debrites lack underlying clean sand layers, and have a flat base. Jackson et al. (2009) describe hybrid beds in outcrops with inferred to be from a mid-lobe setting. Debrites were very rare in channel outcrops that comprised amalgamated thick-bedded turbidites, and absent in outcrops representing more distal fan fringes that were characterized only by sheet-like thin-bedded turbidites. Gonzales-Bonorino and Middleton (1976) inferred late-stage transitions to debris flow for beds in a nonchannelized distal setting in Devonian rocks in western Argentina, although they do not present detailed logs describing the bed character. Van Vliet (1978) documents that clast-rich muddy debrite occur within a fan fringe setting in Tertiary strata outcropping along the Atlantic Coast of northern Spain. The clast-rich debrites occur in thicker beds with a basal clean sandstone interval that is massive or contains dish structures. Clasts can be as long as 1 m and comprise intraformational strata. Van Vliet (1978) attributed such clast-rich debrites to erosion by initial turbidity currents within the fan fringe. Lowe (1982, fig. 13C therein) illustrates a hybrid bed with a relatively thin interval of graded clean sand, overlain by a cohesive debrite, in which clasts are more common near the top of the debrite. Haughton et al. (2009) notes that hybrid beds with clast-rich muddy debrite intervals occur in the Namurian Mam Tor Sandstone in England and the Macigno Formation and Cilento flysch of the Italian Apennines. Mud-rich debrites within hybrid beds are also well developed in the Eocene Tyree Formation in Oregon (Haughton, 2010).

Haughton et al. (2003) and Haughton et al. (2009) provided particularly detailed descriptions of hybrid beds within cored sequences from Jurassic (e.g., Magnus and Miller) and Paleocene (e.g., Forties Sandstone) subsurface reservoir units in the North Sea (Fig. 1E). Hybrid beds are common in lateral and frontal fan fringe settings in which bed amalgamation is rare. In some cores, >80% of the fan fringe beds contain a muddy debrite interval (Haughton et al., 2009). Debrite intervals are rare or absent in mid-fan sequences located further upflow that show greater bed amalgamation, and where beds comprise mainly or exclusively clean sand.

Beds with debrites range from a few centimeters to >1.5 m in thickness, with an average of ∼20 cm. Debrite intervals typically have abundant mud clasts (to 1 m in length but more commonly a few centimeters long) and sheared sand patches, and their muddy matrix can contain dispersed larger sand grains or granules. Carbonaceous organic material is typically abundant within the debrite, and both clasts and organic material can be fractionated toward the top or base of the debrite. Some of the dispersed sand grains can be larger than the sand grains in underlying clean sand intervals. The composition of clasts within some beds shows that the clasts are exotic, and they cannot be matched to local fan fringe deposits. The debrite interval is typically underlain by a clean sandstone interval that is either massive or less commonly laminated, that commonly shows dewatering structures such as pipes and dishes. In some cases, one or more stages of sand injections penetrate the overlying debrite interval. In fewer cases, the debrite is underlain by a very thin interval of clean sandstone or siltstone, or the debrite lacks a basal sand. The boundary between the basal clean sand and overlying muddy debrite is commonly sharp, but it can be transitional over a few centimeters, or display a series of centimeter- to decimeter-spaced laminations. The debrite interval is commonly overlain by a laminated sand or silt interval that grades into turbidite mud, both of which are undeformed.

The exotic mud clasts and abundant organic material suggest that some of these flows originated a significant distance updip from the fan fringe (Haughton et al., 2003, 2009), and did not result from local or basin margin failures on the distal fan. The absence of debrites in more proximal locations is unlikely to be due only to bed amalgamation, and suggests that debris flows either bypassed through such locations, or formed through flow transformation from turbidity currents. In Haughton et al. (2009), a model was favored in which the debris flows form via flow transformation due to erosion of a muddy substrate; they attributed some debrite-prone packages to periods in which the submarine fan was out of grade, and erosion was more common in updip locations. The basal sandstones described in Haughton et al. (2003, 2009) often contain abundant evidence of soft sediment deformation; they therefore suggested that dewatering of the basal sand plays an important part in debris flow run out, and this hypothesis is discussed more fully herein.

Several publications have provided detailed descriptions of slurry beds within the Aptian Britannia Sandstone Member in subsurface cores from the North Sea, where the unit is a major gas reservoir. This example is important because it is distinctly different from most other types of hybrid beds or cohesive debrites, in which repeated banding is less well developed. The slurry beds can be unusually thick (to several tens of meters) and are dominated by repeated banding at a wide variety of scales (Fig. 1H; Lowe and Guy, 2000; Lowe et al., 2003; Barker et al., 2008). Slurry beds differ significantly from debrites formed of a single interval of muddy sandstone matrix with chaotic clasts, and such clast-rich debrites also occur in the Britannia Member (Lowe and Guy, 2000, fig. 8 therein). Individual couplets in slurry beds typically comprise a pair of light and dark bands. Water escape features in the lighter bands are common, but truncate against the sharp base of darker bands. Lighter bands can also be seen to founder into underlying darker bands. Lowe and Guy (2000) argued that the darker bands contained higher amounts of detrital mud matrix, although this was challenged by Blackbourn and Thomson (2000). Banding can occur on a very wide range of scales with individual light-dark couplets ranging from 50 cm to a few millimeters in thickness. Slurry beds comprise intervals with numerous repeated couplets, which can be several meters thick. It is important to distinguish between such unusually thick intervals formed of numerous repeated couplets and debrites comprising a single muddy sandstone interval (Haughton et al., 2009).

Thinner intervals (<10 cm) comprising a smaller number (typically <5) of couplets can occur in hybrid beds, with banding occurring between the basal clean sandstone interval and the overlying debrite interval. This type of thinner banded interval was observed in subsurface cores from the Jurassic and Paleocene sequences in the North Sea, and assigned to the H3 interval by Haughton et al. (2009) (Fig. 2A). Lowe and Sylvester (2004, SB5 bed) also showed a <10-cm-thick interval comprising a small number of bands below a debrite interval. However, in many hybrid beds the boundary between basal clean sand and overlying muddy debrite is relatively sharp, and in some other cases mud content increases in a series of 2–4 steps spaced over centimeters to decimeters (Haughton et al., 2009; my own core logs from the North Sea and Gulf of Mexico). The thickness and number of repeated banded couplets seen in the Britannia Formation are therefore unusual, and can exceed greatly the thickness and number of bands seen in the hybrid beds described here from other locations.

Lowe and Guy (2000) and Lowe et al. (2003) proposed that the light and dark couplets in slurry beds formed through cyclic building up of cohesive mud particles in the near-bed boundary layer. Initially, noncohesive sand and silt particles settled out from this boundary layer to form lighter bands. The volume concentration of mud in the boundary layer subsequently increased due to continued settling, and mud floccules were broken up by turbulence, leading to the suppression of turbulence in the boundary layer due to cohesive mud bonds, and deposition of a dark band. The cyclic process bears some resemblance to the model of Stow and Bowen (1980) for laminated turbidite mud intervals. This type of cyclic depositional process may occur, but it has yet to be reproduced in flume experiments.

I provide here a quantitative framework for understanding cohesive debris flow and hybrid flows that combines observations from laboratory experiments and theoretical analysis. Experimental and theoretical analyses play an important role in the study of submarine flows due to the almost complete lack of direct measurements from the flows in the deep ocean. This discussion forms the basis for a series of generalized models for submarine flows with increasing cohesive sediment strength (outlined later herein).

Ungraded mud-rich matrix in a cohesive debrite provides evidence that sand grains were supported by the strength of the surrounding muddy fluid, such that preferential settling and segregation of the larger sand grains did not occur. Such segregation settling typifies fully turbulent turbidity currents. If the sand grains were supported by excess pore pressures, as was the case in the laminar dense liquefied flows described by Breien et al. (2010), then differential settling and incremental deposition would occur after those excess pore pressures started to dissipate.

This relationship assumes spherical grains of uniform density, and has been broadly validated by laboratory experiments (Hampton, 1975; Amy et al., 2006; Sumner et al., 2009). It shows that sand grains with diameters of ∼250–500 μm and the density of quartz (∼2600 kg m^–3) can be supported in muddy fluid having a yield strength of only 0.1–3 Pa (Fig. 7). Many of the sand-sized grains in these flows are either quartz or other minerals that have a density similar to quartz (e.g., feldspars). Platy-shaped mica grains, or carbonaceous organic material with much lower density, would only settle through muddy fluids with significantly lower strengths than those shown in Figure 7.

Annular flume experiments (in which circular flow is driven by paddles) have shown how a critical mud concentration in the flow is necessary to support sand (Fig. 8; Hampton, 1975; Sumner et al., 2009; Baas et al., 2011), and how shearing of muddy fluid can reduce its strength substantially (Hampton, 1975; Coussot, 1995). The sand is supported by a network of bonds formed by surface charges between the colloidal mud particles, and the strength of these bonds is dependent on the mud mineralogy and water chemistry, as well as mud concentration, and shear rate. Hampton (1975) originally concluded that just 0.6%–1.5% volume of kaolin would keep fine sand aloft, although these figures were extrapolated from experiments that had contained >4.5%–7% volume kaolin. Hampton (1975) noted that even smaller amounts of stronger clays such as bentonite will support sand, and that shearing weakens the muddy suspension significantly, such that the size of sand supported after shearing was typically one-half to one-third of that supported before shearing. Sumner et al. (2009) found higher concentrations of kaolin (>∼14% volume) were needed to support fine to medium (63–250 mm) sand in shearing flows (Fig. 8). This result was consistent with the kaolin mud contents observed to support sand in the static settling tube experiments of Amy et al. (2006), and consistent with theoretical predictions based on the equation proposed by Johnson (1970).

Sumner et al. (2009) observed that sand could sometimes settle out from the flow at a late stage, sometimes even after the initial flow had stopped moving, for kaolin concentrations of between ∼10.25% and 14.25% volume (Fig. 8). A similar late stage of larger sand grains was observed by Marr et al. (2001) from debris flow with low strength. Initial settling of a few larger sand particles appears to break cohesive bonds between mud particles locally within the plug, leading to further settling of sand through the weakened pipe area, with upward expulsion of water through the same pipes (Sumner et al., 2009). Some sand remained within the plug. This process can produce a distinctly cleaner sand layer at the base of the deposit, overlain by a thicker interval of muddy debrite sand that contains pipe structures. The clean basal sand layer can be either graded or ungraded, and need not be strongly normally graded (Sumner et al., 2009). The basal sand in the Sumner et al. (2009, their web fig. 1) experiments comprised ∼50% sediment coarser than ∼125 μm, while the overlying muddy sand had ∼10% sediment coarser than ∼125 μm. So, not all of the larger sand grains settled into the basal sand. The thickness of the basal sand can depend on the thickness of the overlying debris flow, and the duration of deposition if it forms incrementally beneath the late stages of flow.

This is an important question because mud content determines the reservoir quality of the debrite sandstone, and because it has been suggested that laminar debris flows can also deposit clean sand having a mud content that resembles that of turbidites (Shanmugam and Moiola, 1995; Talling et al., 2012c). An approximate minimum volume concentration of cohesive (<20–30 μm) fine mud can be calculated for a cohesive debrite plug in which the sand grains are supported by the strength of the mud (Kuenen, 1966b). The analysis neglects other support mechanisms that may help to support the sand grains. It is assumed that the sand grains are closely packed, such that the muddy pore fluid comprises 25%–50% of the deposit volume at the time of deposition (Allen, 1985, and references therein). The volume of muddy pore fluid will be closer to 25%–35% in more poorly sorted sand-silt mixtures (Bandini and Sathiskumar, 2009). A threshold mud concentration of 14% by volume within the muddy fluid is used to illustrate the approach, based on experiments using kaolin (Fig. 8; Amy et al., 2006; Sumner et al., 2009). This results in a minimum cohesive mud volume concentration in the entire deposit of 7% (when there is 50% pore space between sand grains) to 3.5% (when there is 25% pore space between sand grains). Kaolin is a relatively weak clay mineral, and it is likely to be somewhat weaker than mixtures of clay minerals found in most turbidity currents. This means that the volume of clay concentrations in the deposit could be lower for other clay minerals, yet still support sand (Hampton, 1975). If sand grains are not closely packed, and more widely dispersed in the plug, then the mud concentration can be much higher within the debrite. Burial will generally result in a reduction in porosity, although porosity loss may be offset by cementation. If all of the initial porosity is lost, then the minimum volume kaolin mud concentration will rise to between 14% and 7%, assuming that compaction only results in the loss of water from pore spaces.

In Talling et al. (2012c), we described layers of clean sand that pinch out abruptly in the Marnoso-arenacea Formation beds. These clean sand layers contain <14% (and typically <10%) mud finer than 20 μm, as measured in scanning electron microscope images (Talling et al., 2012c, fig. 7 therein). This mud content is similar to that seen in the turbidite sandstone intervals within these beds. In contrast, mud-rich debrites in these beds contain 18%–60% mud finer than 20 μm, as described in the preceding discussion of the Miocene Marnoso-arenacea Formation. The clean sand debris flows often have a distinctive swirly texture, most likely recording pervasive liquefaction, and their abrupt pinchout within this low-gradient basin plain provides evidence of debris flow deposition (Talling et al., 2012c).

Further work is needed to determine whether the strength of marine clay is typically only slightly weaker than that of kaolin, but assuming that this is the case, cohesive debrite sandstone intervals in cores and outcrop will tend to have minimum cohesive fine mud volume concentrations of ∼5% to ∼10% (Kuenen, 1966b). Amy et al. (2006, fig. 2 therein) summarized field data suggesting that cohesive debris flows tend to have cohesive (<20–30 μm) fine mud contents in excess of 10%–15% volume. Cohesive debrite mud fraction should generally exceed that of turbidity currents (and fluidized layers; Breien et al., 2010) in which the cohesive fine mud concentration is insufficient to support sand grains.

Outsize mud clasts are found within many debris flow deposits; their density at the time of deposition differs significantly from that of quartz, and therefore larger mud clasts can perhaps be supported (Fig. 9B). Mud-clast density is a function of the depth to which the mud has been buried and compacted, before being eroded and incorporated into the debris flow. Mud density in the upper 10 m of sediment below the seafloor can range from 1350 to 1600 kg m^–3 (Flemings et al., 2006; Tripsanas et al., 2008; Expedition 333 Scientists, 2011). Lower density mud can occur in the uppermost tens of centimeters below the seafloor, but such lower density mud tends to act as fluid mud (McAnally et al., 2007). Mud densities that are >>1600 kg m^–3 can occur if the mud is buried more deeply before being exhumed (Flemings et al., 2006).

Equation 1 was used to estimate the largest diameter of mud clast with variable density that could be supported in kaolin suspensions of varying yield strength and density (Fig. 9A). It is apparent that once the mud clast has a density that is less than that of the surrounding muddy fluid, then the fluid will be able to support very large mud clasts (Fig. 9A). Such positively buoyant mud clasts could in some cases rise to the top of the debris flow. Mud clasts will tend to be buoyant in the denser and therefore stronger and more coherent debris flows, but large yield strengths may prevent buoyant clasts rising to the top of the debris flow.

Flume experiments show that relatively small amounts of cohesive fine mud (<4% volume) can dampen turbulence very effectively, especially at lower flow velocities when shear is reduced (Fig. 8; Baas and Best, 2002; Baas et al., 2009, 2011; Sumner et al., 2009). This process can cause transformation from initially fully turbulent flow (turbidity current) to laminar plug flow (debris flow) as the speed of the flow decreases. Flow transformation initiates in areas of lower shear in a decelerating flow, as decreasing turbulence intensity allows bonds to form between mud particles. Baas and Best (2008) illustrated how a laminar plug can progressively thicken, such that turbulence is restricted to a progressively narrower zone at the bed. It is also likely that the transformation will begin in the slower moving tail of a flow. This will favor development of laminar debris flow toward the rear of the event, and debrite deposition toward the top of the resulting deposit. The experiments of Baas et al. (2009) and Sumner et al. (2009) suggest that such flow transform may be commonplace in submarine flows containing even a small cohesive mud fraction, as every flow will at some point decelerate to a standstill. In the experiments, flow transformation occurred at ∼1.2 m/s for suspensions with ∼12% volume kaolin, and at speeds of ∼0.2 m/s for kaolin suspensions of ∼2% volume (Fig. 8; Baas et al., 2009; Sumner et al., 2009). Flow transformation will tend to increase the viscosity of the fluid through stronger mud bonds, and this may cause further deceleration, producing a positive feedback.

Sumner et al. (2009) showed how deposition of turbidite or cohesive debrite sand is linked to flow state (turbulent or laminar) during sand deposition. Turbidite sand tended to be deposited under turbulent flow conditions before transformation to laminar flow occurred as flow decelerated (Fig. 8). Debrite sand was deposited from laminar flow, if the transformation from turbulent to laminar flow occurred at higher flow velocities before the sand started to settle out. This resulted in the generation of a laminar plug flow from which muddy debrite sand was eventually deposited en masse. At intermediate concentrations, sand settled out from the laminar plug at a late stage, sometimes even after flow stopped (Sumner et al., 2009; see also discussion of Experimental Observations). This association between the type of sand deposited and the flow state results from near coincidence between the cohesive mud content necessary to support sand and the cohesive mud content needed to suppress turbulence in the experiments. It remains to be seen whether such a coincidence characterizes a wider range of marine mud compositions.

Laboratory experiments have shown how a layer of water may be injected under the head of a submarine debris flow, if the dynamic pressure developed at the nose of the flow exceeds the downward directed weight of debris (Mohrig et al., 1998; Harbitz et al. 2003; De Blasio et al., 2004). Hydroplaning will therefore characterize faster debris flows, or thinner debris flows that have lower density. A further condition for hydroplaning is that the debris is sufficiently impermeable to prevent rapid dissipation of the overridden water. Water overridden at the debris flow head will tend to lubricate the debris flow. It has been proposed that hydroplaning is one explanation for the long run out of submarine debris flows across low-gradient seafloor (Mohrig et al., 1998; De Blasio et al., 2004). The injection of water during hydroplaning will only penetrate for a limited distance under the head of a hydroplaning debris flow, where it will tend to be shear mixed back into the body of the flow. In the laboratory experiments of Mohrig et al. (1998), this distance is a few tens of centimeters. Hydroplaning therefore only lubricates the very front of the flow, and may not lubricate the trailing body of the flow. Greater lubrication of the head can lead to detachment of faster moving outrunner blocks from the main debris flow. Hydroplaning may protect underlying deposits from erosion, but only at the front of the flow.

Experimental and theoretical work by Ilstad et al. (2004) suggests that water overridden at the head will often be mixed back into the body of the debris flow, due to high shear rates at the bottom of the flow. Ilstad et al. (2004) termed this process shear wetting, and it would tend to produce a basal sediment layer with high water content, rather than a basal layer comprising only seawater. Such a basal water-rich sediment mixture could still act to lubricate a debris flow. It may be even more important in submarine debris flows that tend to be faster moving than their laboratory counterparts.

The uppermost ∼20 cm of the modern seafloor can often have very low strength (soupy when cored) mud. This is why box corers (rather than piston or gravity) corers are typically used to sample the sediment-water interface. It is possible that this type of low-strength mud layer can also lubricate a debris flow that passes across it, in a manner similar to shear wetting.

Outrunner blocks have been observed that most likely result from basal lubrication of debris flows (Harbitz et al., 2003; De Blasio et al., 2004). However, in many cases outrunner blocks are not observed, and the lobate submarine debris flow deposits resemble the deposits of terrestrial debris flows that did not hydroplane (e.g., Laberg and Vorren, 2000; Talling et al., 2010). Shear wetting would generate homogenized layers of flow with lower sediment concentrations than the overlying debris flow. Erosion of seafloor sediment can also generate basal layers in the flow, such as the muddy sand layer described by Gee et al. (1999) beneath the Canary debris flow. Layers of massive clean sand are observed commonly below debrites, but it is difficult to envisage how such clean sand layers could record basal shear wetting, or seafloor erosion. Shear mixing with either the muddy seafloor or the muddy debris flow would be expected to produce a muddy sand layer, not a clean sand layer.

Rates of mixing between debris flow and the surrounding seawater or turbidity current are possibly very important controls on submarine flows (Kuenen, 1951; Hampton, 1972; Mohrig et al., 1998; Marr et al., 2001; Mohrig and Marr, 2003; Ilstad et al., 2004; Felix and Peakall, 2006). Mixing and dilution can decrease the strength and viscosity of the debris flow in a nonlinear fashion, and determine the rate at which associated turbidity currents are produced. The rate of mixing can determine the duration and run out distance of a debris flow, if the debris flow is transformed completely into turbidity current.

Shear stresses between the debris flow and surrounding water, and erosion rates, are higher at the head than above the body of the debris flow. The head can be eroded by detachment of single grains at low shear stresses that exceed a critical value. This critical value of the shear stress (typically 0.5–2 Pa) can be orders of magnitude lower than the sediment mixture’s yield strength, so erosion and mixing can initiate well before the material is internally deformed. As shear stresses increase at the head, intact chunks of material, and then a discrete layer of sediment, may be flung backward toward the body.

Laboratory studies have tended to focus on mixing at the head, as the head dominates smaller volume experiments. The body will comprise a far greater proportion of most submarine debris flows, and may play a greater role in mixing and dilution (Talling et al., 2002). Erosion of the body will also begin through detachment of individual grains, and then proceed to erosion of chunks as shear stresses increase. As with the head, experiments suggest that mixing becomes much more efficient once the body becomes turbulent, and mixing can also occur through breaking waves along the upper surface of the body (Felix and Peakall, 2006).

A reduction in sediment density, especially through turbulent mixing of the flow interior, will decrease the strength and viscosity of a sediment mixture in a strongly nonlinear fashion. This can promote more rapid mixing, and the reduced viscosity will tend to increase the speed of the flow. This positive feedback may produce a bifurcation in flow behavior, such that some flows undergo much more effective mixing and dilution than other flows. Flow behavior will be heavily dependent on whether the head and especially the body of a flow become turbulent.

A theoretical analysis explores how submarine flows behave as strength of the sediment mixture forming the flow increases; this analysis extends on work by Hampton (1975), Hiscott and Middleton (1979), Locat et al., (1996), Schwab et al. (1996), and Talling et al. (2002, 2010). Flow evolution is illustrated by a series of plots based on the rheology of a kaolin suspension (Fig. 9; rheology from Coussot, 1995).

A plot of flow speed against volume concentration of kaolin shows that increasing viscosity can become important at relatively modest mud concentrations. For example, increasing viscosity causes a decrease in flow speeds once the volume concentration of kaolin in the suspension exceeds ∼6% for a 1-m-thick flow on a gradient of 0.1° (Fig. 9E). Direct measurements of hyperconcentrated flow speeds in the Yellow River in China similarly suggest that viscous forces become important at relatively low sediment concentrations, such that maximum flow speeds are reached at ∼5% sediment volume concentration (Van Maren et al., 2009).

The calculations suggest that a thin (∼2 m) flow with yield strength of <∼10 Pa, on a gradient of <0.05°, is likely to be at least weakly turbulent. Thicker debris flows (>5 m) with greater yield strengths (>100 Pa) are also likely to be turbulent on steeper gradients of >∼0.1°. This analysis suggests that the (often clast rich) sediment mixtures responsible for depositing cohesive debrites distally may often be at least weakly turbulent on steeper slopes closer to the source.

The analysis based on kaolin rheology suggests that hydroplaning could be common, and that a sufficiently thick high-strength mixture of kaolin will nearly always hydroplane, if it is able to move (Fig. 10). A flow (or block) that is hydroplaning may travel beyond the position at which a non-hydroplaning flow will stop. Faster moving flows that hydroplane may also be prone to becoming turbulent, and this may be accentuated by decreases in flow viscosity at higher shear rates (Fig. 9E). Laboratory experiments have mainly focused on hydroplaning of relatively thin and laminar flows. Turbulence may cause mixing of overridden seawater back into the main body of the flow in many situations. The analysis predicts that moving flows with uniform thickness and composition will undergo a transition from turbulent to laminar flow as seafloor gradients decline. A more enigmatic prediction is that thicker flows that are turbulent sometimes come to a halt without a transition to laminar flow; this prediction appears to be unrealistic, unless there is an abrupt transition from turbulent flow to no motion.

Here I explore whether mixing and dilution by surrounding seawater prevents long run out of low-strength debris flows. Laboratory experiments illustrate how relatively small changes in debris flow sediment concentration (e.g., 2%–3% volume concentration of kaolin; Hampton, 1972) and yield strength can lead to disproportionately large changes in the rate of mixing and the rate of turbidity current generation. Dilution of the debris flow can reduce its viscosity substantially, potentially leading to flow acceleration and therefore even higher mixing rates.

Flume experiments illustrate in how low-strength (<5 Pa) muddy suspensions mix rather easily and erode rapidly. Winterwerp and Kranenburg (1997) found that a suspension containing 2.5% kaolin was eroded at a rate of ∼5 mm/s by an overlying flow traveling at ∼50 cm/s. A 1 m thickness of this material would be rapidly eroded in ∼3 min. A suspension containing 7.5% volume concentration kaolin was eroded at a rate of 0.2 mm by similar flow speeds, such that 1 m of erosion would take ∼1.5 h. Winterwerp and Kranenburg (1997) allowed the kaolin suspension to consolidate for several hours, and estimated that it had a yield strength of ∼4 Pa; the rheology of their kaolin suspensions may span the boundary from fluid mud to very low strength debris flows that can support sand (flow types A and B in Fig. 9). The results of Winterwerp and Kranenburg (1997) support qualitative observations from other annular flume experiments in which mud suspensions with yield strengths of <10 Pa mixed rapidly with overlying water at flow speeds of a few tens of centimeters per second (Esther Sumner, 2010, personal commun.). This observation is important because it suggests that thin (∼1 m), low-strength (<1 Pa) debris flows could not travel for long distances at speeds of even a few tens of centimeters per second without mixing almost completely with the surrounding seawater. Calculations of flow speeds for such low volume concentration kaolin mixtures suggest that speeds of 1 m/s would often be exceeded, even on gradients of just 0.1° (Fig. 9E). This suggests that very low strength (clast poor) cohesive debris flows tend not to flow for long distances, but form at a late stage from flow transformation.

Changes in excess pore pressure can profoundly change cohesive debris flow motion (Iverson and Vallence, 2001; Iverson et al., 2010), and indeed in flows in which the excess pore fluid pressure rather than cohesive matrix strength supports sand grains. Compaction of the debris during motion would cause increases in the debris flow’s strength and viscosity. Dissipation of excess pore pressures after the debris flow has stopped can lead to foundering of overlying layers, and in situ soft sediment deformation.

The coefficient of consolidation depends upon the permeability and stiffness of the material, and can therefore vary many orders of magnitude. For example, the addition of ∼5% volume mud finer than 64 mm to an initial mixture of sand and gravel reduced the coefficient of consolidation from 10^–4 to 10^–7 in the experiments of Iverson et al. (2010). As the time taken for pore pressure to dissipate is proportional to the coefficient of consolidation, small increases in mud content profoundly change the period over which excess pore pressures remain (Fig. 11).

Mud contents necessary to support sand are likely to result in very low coefficients of consolidation (<10^–7); this means that little excess pore pressure is dissipated during cohesive debris flow motion over periods of as much as several hours (Fig. 11). It is therefore likely that the debris will still be partly or wholly liquefied for significant periods after it is deposited. It is therefore unsurprising that overlying ripple cross-laminated sands often founder into underlying cohesive debrite (Butler and McCaffrey, 2010).

Small amounts of cohesive mud can dampen turbulence, especially at slower flow velocities (Fig. 8; Baas et al., 2009; Sumner et al., 2009). Baas et al. (2009) reported that just 0.75% volume kaolin (a rather weak clay mineral) was sufficient to dampen turbulence locally in a flow moving at <50 cm/s (Fig. 8); however, for flow traveling at 1 m/s, turbulence was damped when kaolin volume concentrations exceeded ∼6% (Fig. 8). Many submarine flows carry significant amounts of cohesive mud (Talling et al., 2012a). Small volume fractions (<<1%) of noncohesive grains can also dampen turbulence (Wright and Parker, 2004; Cantero et al., 2012). This suggests that transitions from turbulent to laminar flow may be relatively common as submarine flows decelerate.

A bifurcation in flow behavior may result from the shear thinning properties of many marine muds (Fig. 9E; Coussot et al., 2002; Jeong, 2010; Jeong et al., 2010) and due to positive feedbacks associated with mixing and dilution (see discussion of Potential Bifurcation in Mixing and Flow Behavior). Deceleration may lead to increased viscosity, which in turn causes further deceleration of the flow, and therefore even higher viscosity (Coussot et al., 2002). Increased velocity may reduce viscosity, thereby promoting even faster flow. This may favor late-stage transformation from turbulent to laminar flow as flows decelerate.

Experiments in which mixtures of mud and sand are released into flume tanks illustrate the dynamics of cohesive debris flows, and how debris flows generate turbidity currents through mixing with surrounding water (Hampton, 1972; Mohrig et al., 1998; Marr et al., 2001; Mohrig and Marr, 2003; Ilstadt et al., 2004; Breien et al., 2010). It can be useful to group experiments according to the coherency of the debris flow (Marr et al., 2001). Coherency expresses the relative magnitudes of the sediment’s yield strength and dynamic pressures it undergoes due to shear with surrounding water and seafloor. Strongly coherent flows tend to have higher yield strengths and/or travel at lower speeds, as the shear forces tend to scale with the square of flow speed.

Strongly coherent debris flows are prone to hydroplaning at their head, and undergo relatively little mixing with surrounding seawater. Mixing occurs mainly at the head of the debris through erosion of single grains at lower speeds, and chunks or a sheared layer of material at higher speeds. This material travels backward, and may partly settle back into the body. The finer grained material forms a dilute turbidity current above the body. Strongly coherent debris flows tend to produce rather small volume and dilute turbidity currents, which initially trail behind the debris flow. However, the dilute turbidity currents tend to run out beyond the location where the debris flow comes to a halt. The highly coherent debris flow tends to extend back continuously to the point at which it initiates, although it can locally display tensional cracks (Mohrig et al., 1998). Sand is trapped within the laminar plug composing the body of the debris flow, and does not segregate.

There is a gradual continuum between strongly coherent and weakly coherent debris flows, with the degree of mixing with surrounding seawater increasing progressively, so that progressively more voluminous and denser turbidity currents are produced. The head of the debris flow becomes turbulent as the relative magnitude of shear stresses increase.

The morphology, dynamics, and deposits of weakly coherent debris flows differ significantly from those of strongly coherent debris flows. The head and frontal part of the body become increasingly turbulent, and this turbulence promotes much more effective mixing with surrounding seawater. The trailing part of the body develops a pronounced interface separating a basal high-density flow from an upper, much more dilute, turbidity current (Marr et al., 2001; Mohrig and Marr, 2003; Ilstad et al., 2004; Breien et al., 2010). The experiments of Ilstad et al. (2004) and Breien et al. (2010) showed that excess pore pressures in the high-density layer are sufficient to support the sediment. Breien et al. (2010) showed how the frontal part of the high-density layer comprises a liquefied layer from which sand settles out, incrementally depositing a layer of clean massive sand. Their experiments show how the rear of the dense layer comprises the same muddy sediment that was introduced into the tank, as it is relatively protected from mixing. Sediment in the rear part of the dense flow layer is supported by muddy matrix strength rather than by liquefaction, and deposits muddy sand en masse when it comes to a halt (Breien et al., 2010). The frontal part of the dense layer that deposits clean sand is continuously being fed from the rear part of the dense layer, such that the two parts of the layer are closely coupled.

The laboratory-scale flows are relatively thin (<50 cm) and slow moving (40–100 cm/s), such that the dynamic pressures on the experimental flows may be much less than in faster moving submarine debris flows. Sediment strength and viscosity tend to be more important in the relatively slow experimental flows. For example, the slow speed of the experimental flows tends to reduce the importance of shear wetting at their base. This means that sediment mixtures with weaker strength in the laboratory flows will tend to reproduce the behavior of faster flowing mixtures with much greater strength in full-scale submarine examples.

This scaling can be expressed by the ratio of the sediment yield strength and the dynamic pressure at the flow front. However, other key properties will not scale up in the same way as this ratio. This can be illustrated by considering an experimental debris flow with yield strength of 30 Pa and speed of 50 cm/s, and a submarine debris flow with yield strength of 1080 Pa traveling at 3 m/s; both have similar ratios of yield strength to frontal dynamic pressure. The much stronger submarine flow will be able to support much larger grains or clasts. The critical shear stress needed to erode the surface of a sediment mixture (typically 0.5–2 Pa) can be orders of magnitude lower than its yield strength (Talling et al., 2002), and much higher surface erosion rates may characterize the faster submarine debris flow. Laboratory studies of mixing processes have tended to concentrate on processes that occur near the head, rather than on the upper surface of the debris flow body. This is partly because the head of a small-volume experimental debris flows forms a disproportionately large fraction of the debris flow, and mixing along the body may be more important for much larger submarine flows (Talling et al., 2002).

Differences in shear rates between laboratory experiments and full-scale submarine flows may also be important because of the shear thinning rheology that characterizes many marine clays (Jeong, 2010; Jeong et al., 2010). This shear thinning rheology can cause substantial (orders of magnitude) decreases in the viscosity of the debris flow at high shear rates, or increases in viscosity as shear rates decrease (Fig. 9E; Coussot et al., 2002; Jeong, 2010; Jeong et al., 2010). A decrease in viscosity can lead to even faster flow, or an increase in viscosity can cause further flow deceleration. These positive feedbacks can therefore potentially lead to a bifurcation in flow velocity and flow behavior that will not be seen clearly in short-lived laboratory experiments. Other scaling issues include the relatively short distance that experimental flows travel that, together with differential speeds in different parts of the event, will determine how a flow event organizes during longer run out in the ocean. A final important point is that submarine debris flows will initially accelerate and eventually decelerate to a standstill, such that the dynamic pressures undergone by a parcel of sediment may vary substantially through time. The coherency of a debris flow may vary therefore substantially through its history, with stronger coherency characterizing the final stages of the flow.

The importance of these scaling issues has only been partially addressed in previous publications, even those that consider geometric scaling and Froude or flow Reynolds numbers. An even clearer understanding of how debris flow dynamics scale up from laboratory to full-scale submarine examples is a priority for future experimental studies.

Field observations combined with the experimental and theoretical analysis produce a generalized model for cohesive debris flows and hybrid flows (Fig. 12). The classification is based on increasing strength of the sediment mixture that tends to produce more strongly coherent debris flows (Marr et al., 2001). The yield strengths quoted for each type of debris flow are approximate, as flow behavior will also depend on other factors such as flow speed and the dynamic pressures exerted on the sediment mixture. Decreasing shear will tend to decrease, sometimes substantially, the viscosity and strength of shear thinning muddy fluids.

Very high strength (>100 Pa) debris flows produce debrites that are often tens of meters thick (Fig. 12A) and that support large clasts (Fig. 9). Such thick cohesive debrites tend to be restricted to the continental slope (e.g., Laberg and Vorren, 2000). However, a few very thick debris flows have run out onto low-gradient (<∼0.2°) distal fans and basin plains to produce megabeds (Labaume et al., 1987; Kleverlaan, 1987). High-strength (10–100 Pa) cohesive debris flow will generally produce similar but thinner (0.5–3 m) deposits on continental slopes (Tripsanas et al., 2008).

Very high and high-strength debris flows are unlikely to become turbulent, even on steeper gradients (Fig. 9), and their deposits tend to extend back to the vicinity of the original slope failure. Mixing with the surrounding seawater is limited, and any dilute turbidity currents that are produced by mixing are of limited volume. The dilute turbidity current tends to lag behind the head of the debris flow, but can run ahead of the debris flow, after the debris flow comes to halt. Cohesive debrite intervals are overlain by the trailing dilute turbidity current deposits, which form a relatively small fraction of the overall deposit. In some cases, the dilute turbidity current may bypass across steeper gradient seafloor, and deposit in other locations further down the fan.

The frontal part of faster moving higher strength debris flows is prone to hydroplaning, if it is sufficiently thick to move (Fig. 10). However, water overridden at the head may often be mixed with the overlying debris flow, to form a basal layer of low sediment concentration. This basal layer may lubricate the flow, allowing increased speeds and run out.

Intermediate-strength (∼5–100 Pa) debris flows produce thinner (1–2 m) cohesive debrites in distal locations on submarine fans, beyond the steeper continental slope (Fig. 12B). The debrites are commonly clast rich, and clasts can be >1 m in length if they are of lower density than the matrix. The sediment-charged flow is laminar during the later depositional stages of flow, but it may be initially weakly turbulent on steeper gradients. Hydroplaning may occur at the head, but the overridden water may be thoroughly mixed into the body. Relatively long run out distances can result only from the lower strength of the debris flow, without the need for hydroplaning. In some cases, these intermediate-strength debris flows produce deposits with intricate digitate planform shapes (Fig. 12B).

Limited shear mixing may lead to the formation of low-volume dilute turbidity currents, as was the case for higher strength debris flows, such that fine-grained turbidite composes a small part of the overall deposit. For example, the Mississippi fan debrites pass abruptly into very thin muddy turbidites. However, as mixing increases progressively larger volume dilute turbidity currents may be generated, and these turbidity currents may run out well beyond the debris flow (as seen in Bed 2.5 of the Marnoso-arenacea Formation; Talling et al., 2012b). Debrites may be underlain by rather thin massive clean sand intervals, which tend to pinch out at the same place as the overlying debrite. The basal sand may result from initial deposition from high-density turbidity current, generated by mixing at the head of the flow that is undergoing relatively high dynamic pressures. Alternatively, the thin basal clean sand may be formed by late-stage settling of sand from the debris flow plug, possibly at times even after the debris flow has stopped moving (Sumner et al., 2009), or due to shear shinning effects during the flow.

Although their cohesive strength is strong enough to carry sand, these lower strength (0.1 to ∼5 Pa) debris flows generally produce finer grained muddy sandstones with few, if any, small clasts (Fig. 12C). Their deposits tend to be relatively thin. Mixing with surrounding seawater is more effective, such that much of the deposit comprises high-density and low-density turbidite. The basal turbidite clean sand may at times be relatively thick, and the debrite can transition downbasin into muddy silt. The lowest strength clast-poor debris flows may generate deposits that form bulls-eye patterns in basin lows, rather than having digitate (fingered) planform shapes seen in intermediate-strength debris flows (Fig. 12C). Any water overridden at the turbulent flow front is rapidly assimilated into the flow, and these flows therefore do not hydroplane.

The debrite is absent in proximal deposits, and this may be a combination of two distinct processes that are not mutually exclusive. First, their low strength ensures that these flows may often be initially fully turbulent on steeper slopes, such that the debris flow forms via flow transformation from turbidity current (Figs. 6 and 12C; Sumner et al., 2009). This transformation will begin in slower moving parts of the flow, such as at the rear. The resulting weak cohesive debris flow may travel only for a short distance, at a slow speed, because such low-strength debris flow mixes easily with surrounding seawater (Fig. 12). Second, the low coherency of these flows may ensure that mixing is especially vigorous at the front of the flow (Marr et al., 2001), generating a high-density turbidity current (or a dense fluidized sand layer) in front of the trailing cohesive debris flow (Breien et al., 2010). This, together with an overlying dilute turbidity current, may protect the debris flow composing the rear of the event from mixing. The forerunning turbidity current may be generated by this progressive mixing of the initial debris flow, explaining why the cohesive debrite is almost always found at the boundary between high-density and low-density turbidite (Fig. 2).

A further decrease in cohesive strength (<0.1 Pa) produces a transition into fluid mud layers (Fig. 12D), the strength of which is insufficient to support sand. Low-strength debris flows (that carry sand) may share many aspects of fluid mud behavior, such as ponding in basinal lows and late-stage transformation from turbulent to laminar flow. The behavior of very low strength debris flows will also tend toward that of turbidite mud deposition (McCave and Jones, 1988; Talling et al., 2012a).

Previous studies suggested that hybrid beds are common in distal depositional settings, and that they are typically absent in more proximal settings (Haughton et al., 2003; Talling et al., 2004). This much wider review of hybrid bed occurrence in 3 modern systems, 14 ancient outcrop sequences, and more than 5 ancient subsurface sequences strongly supports both conclusions. Almost all of the hybrid beds described here are from distal lobe (fan fringe) settings with relatively infrequent or shallow bed amalgamation, or basin-plain settings with almost no bed amalgamation. The only exceptions are one of the two types of hybrid beds (finer grained muddy sandstones) described by Sylvester and Lowe (2004) in Carpathian outcrops inferred to represent a channel fill. Haughton et al. (2009) also stated that hybrid beds occur in the upper part of channel-fill deposits in a core from the Paleocene Schiehallion oil field, located west of the Shetland Islands, although a detailed description of these hybrid beds was not provided.

Generally thicker (>3 m to tens of meters) cohesive debris flow deposits are relatively common on proximal continental slopes, but turbidites encasing these deposits are poorly developed or absent (Tripsanas et al., 2008). This may be due to the higher strength of such debris flows that reduces mixing with seawater and turbidity current generation. However, it is more likely that turbidity currents that are generated bypass the steeper continental slope and deposit farther from the source. Examples of megabeds with very thick cohesive debrite and turbidite sand intervals are also found in basin plain or lobe settings, rather than on continental slopes.

A potentially important observation is the relatively consistent level at which cohesive debrite intervals occur within hybrid beds (Fig. 2). The cohesive debrite interval is almost always underlain by massive, or occasionally planar laminated, clean sandstone (broadly equivalent to T_A or T_B intervals; see later discussion about its origin). The only examples of cohesive debrites underlain by ripple cross-laminated sand (T_C) are the finer grained cohesive debrites of Sylvester and Lowe (2004). Clast-poor (lower strength) debrite intervals in the Marnoso-arenacea Formation are sometimes directly underlain by dune-scale (20–90 cm wavelength) cross-bedding. Such large-scale bedforms are often inferred to originate from relatively dilute flow (Southard, 1991), but Baas et al. (2011) showed how they may also originate below more concentrated mud-rich flows. Cohesive debrite intervals are not observed within thinly (<30 cm) bedded turbidites characterized by T_C, T_D, or T_E intervals in any of these examples described here. However, cohesive debrites are commonly overlain by T_C, T_D, or T_E intervals deposited from relatively dilute turbidity currents, from which they are typically separated by a grain size break.

It therefore appears that cohesive debrites consistently mark the boundary between deposition from high-density and low-density turbidity currents (Figs. 1 and 2). As a bed is built up progressively, the vertical bed structure records the longitudinal flow structure at that site. The position of the debrite may be due to the speed of the cohesive debris flow being consistently intermediate between that of high-density and low-density turbidity currents, across the same flow path. The debris flow therefore arrives at the site after the high-density turbidity current, but before the arrival of low-density turbidity current. Such a relationship might also be consistent with lubrication of the debris flow by dewatering of previously deposited high-density turbidite sand, and a lack of lubrication by underlying low-density turbidite sand. However, this explanation is not favored because some basal high-density turbidite intervals have undeformed laminations, and because strongly contorted laminations show that low-density (T_C) turbidite sand can also undergo pervasive syndepositional dewatering.

The debris flows only occur in flow events that also contain high-density turbidity currents, as well as trailing low-density turbidity currents. This relationship could be due mainly to the longer run out of low-density turbidity currents than either debris flow or high-density turbidity current. Confinement of cohesive debris flows in channels will also explain their paucity in levee sequences, which are dominated by thin low-density turbidites. However, bed correlation in the Marnoso-arenacea Formation shows that clast-poor cohesive debrite can sometimes replace thick intervals of ripple cross-laminated low-density turbidite over a few kilometers in a downflow direction (Talling et al., 2012b). This transition suggests that the presence of cohesive debris flow can also sometimes suppress ripple development in the low-density tail of the flow event (cf. Sylvester and Lowe, 2004).

Some field observations suggest that deposition by cohesive debris flows and high-density turbidity currents can on occasions be closely linked. Hybrid beds containing cohesive debrites often have thickness distributions very similar to those of beds comprising only turbidite sandstone in the same sequence. Hodgson (2009, fig. 3C therein) illustrated how the relative thickness of high-density turbidite and clast-rich muddy debrite sand can change rapidly, despite the overall bed thickness remaining near constant. This does not appear to result from local nests of clasts within muddy debrite sandstone, as clast-free areas comprise clean turbidite sand. A similar geometrical relationship occurs over somewhat longer distances for hybrid beds with clast-rich cleaner sand debrites in the Marnoso-arenacea Formation (Talling et al., 2012c). This type of bed geometry could suggest that clast-rich debris flows tend to erode out a similar thickness of previously deposited high-density turbidite, but it is surprising that the thickness of eroded and deposited sediment is so similar. Alternatively, this geometry could indicate that there is a close link between cohesive debris flow deposition and high-density turbidite deposition, such that the two processes deposit similar sediment thicknesses in adjacent locations.

Correlation of hybrid beds containing clast-poor debrites shows that the debrite can sometimes be on top of a relatively constant thickness of high-density turbidite, such that the overall bed thickness changes significantly. This different (noncompensating) internal bed geometry suggests that the clast-poor debris flows are less erosional, or that deposition from the clast-poor debrite and high-density turbidity current are not closely linked.

Four general types of model can be proposed for the origin of cohesive debris flows within hybrid flow events.

The debris flow can originate from the same initial source as the rest of the hybrid flow, for example from an initial slope failure. The debris flow component of the flow needs to bypass through more proximal parts of the fan, where only turbidite sand is deposited. Bypass could occur if the matrix strength, and other support mechanisms such as excess pore pressure, is sufficient to keep the sand suspended. Unlike turbidity currents, debris flows deposit en masse, and this could mean that they are more likely to bypass sediment initially before deposition finally occurs. Larger clasts may segregate from a bypassing debris flow, potentially leaving a coarse lag of clasts in proximal areas of bypass (Talling et al., 2007a). Hydroplaning could aid bypassing, although this process will only affect the very front of the flow, and turbulence can lead to mixing of overridden water back into the flow.

Subaerial debris flows can leave behind relatively thin coarse-grained deposits near their margins on steep (>1°) slopes (Pierson et al., 1990; Revellino et al., 2004). Bypass by subaerial debris flows typically occurs for <1 or 2 km (Revellino et al., 2004), although subaerial debris flows have traveled as much as 55 km with limited deposition (Pierson et al., 1990). Debris flow deposits in areas of subaerial bypass can contain boulders, and are typically tens of centimeters thick (Pierson et al., 1990; Revellino et al., 2004). To explain the location of cohesive debrites in submarine flow deposits, bypass of much larger sediment volumes would need to occur on much lower gradients. Such bypass may be recorded by grain size breaks between turbidite sand and mud in proximal deposits, but there are typically no other signs of large-volume bypass in proximal turbidites. This might be consistent with the generally larger volumes and lower yield strengths of submarine debris flows, which are generally able to travel across lower seafloor gradients (<0.1°).

This model would be consistent with a debrite matrix, which is often rich in organic matter, coming primarily from shallow water. However, the model needs to explain why the clasts within that matrix typically lack abundant organic material, and therefore often have a composition different from the organic-rich matrix. A potential explanation is that the original slope failure was of weaker strata rich in organics and interbedded more resistant intervals. Such interbedded sequences of weaker organic-rich turbidites and stronger hemipelagic mud can occur on continental slopes offshore major river deltas, due to cyclic changes in sea level that result in episodic deposition of turbidite packages (e.g., Ducassou et al., 2008, 2013).

A second hypothesis is that the cohesive debris flows originated through local erosion of muddy seafloor sediment along the flow path. The eroded muddy material causes turbulence to be damped and local transformation from turbidity current to debris flow. In this scenario, the debris flow contains additional sediment that comes locally from the seafloor. Experiments have shown how small increases in cohesive mud content can lead to the support of sand grains (Sumner et al., 2009), and how increased mud content can dampen flow turbulence (Baas et al., 2009). It is possible that a small amount of erosion could increase the mud content above a threshold value, leading to flow transformation.

Clasts in the debris flow should come from deeper water (see Ito, 2008). However, other debrites contain exotic clasts that could not come from local erosion of the seafloor (Talling et al., 2007b; Haughton et al., 2009). Erosion of muddy seafloor sediment should generally decrease the relative abundance of organic carbonaceous material, as seafloor sediment contains less organic material in deeper water locations. However, in many cases, the debrite intervals have more abundant organic material than adjacent turbidite mud and sand or hemipelagic mud. Cohesive debrite is typically underlain by a continuous layer of clean sand, which forms a barrier between the debrite and areas of eroded seafloor. Unless this clean sand forms through late-stage settling of sand from the debris flow, it is not clear as yet how such a continuous barrier would form if the debris flow was generated by seafloor erosion.

A third model is that the cohesive debris flow results from flow transformation from an initially turbulent part of the flow (turbidity current, according to our definitions), but without the need for local erosion of muddy seafloor. In this model, flow deceleration and reduced shearing of the muddy fluid lead to flow transformation as the existing mud within the flow increases flow viscosity and dampens turbulence at slower flow speeds. Experiments suggest that such flow transformation may be a general characteristic of muddy flows, the amount of mud controlling the speed at which transformation occurs. If transformation occurs before all the sand has settled out of the flow, then muddy debrite sand will result.

Debrites containing large clasts are less likely to have been deposited in this way, as the large clasts would need to be carried within the initially turbulent flow. This might only occur if turbulence was very strong (in which case one might expect the clasts to rapidly break up), or if the turbulent sediment mixture was sufficiently dense that clasts were nearly buoyant. This process is more likely to produce clast-poor debrites. This transformation process might occur first in slower moving parts of the flow, ensuring that the transformation may tend to start at the back of the flow. This could explain why clast-poor (low strength) debrites are often underlain by thick turbidite sand, deposited from the still turbulent front of the flow event.

Two variants of this model can be proposed. First, the cohesive debris flow may run out for long distances after formation. Second, the debris flow tends to form close to the site of debrite deposition. Very weak debris flows will tend to mix efficiently with surrounding seawater, even at low speeds, favoring formation locally near the site of deposition. However, the theoretical analysis presented here shows how somewhat stronger debris flows could run out for long distances on low gradients, if not diluted by shear mixing. The observation that cohesive debrites consistently underlie rippled intervals suggests that the debris flows were faster moving than the dilute turbidity currents that deposited the rippled interval. This in turn suggests that the debris flows were relatively fast moving, and could be formed from initially fast moving and turbulent flow.

It is unlikely that the cohesive debris flows were generated by loading and failure of basin margin slopes by an initial turbidity current (McCaffrey and Kneller, 2001). This process could not easily explain the organic-rich matrix of many cohesive debrites, or the exotic mud clasts seen in certain debrites (Talling et al., 2004; Haughton et al., 2009), and in some cases it can be shown that the debrites are not located near basin margins or topographic highs (Talling et al., 2007b).

The difference between two of these models is really a matter of the terminology used to describe an initial sediment laden dense flow. If an initial dense turbulent suspension is called a turbidity current, then the debris flow forms via flow transformation in model 3. If the initial sediment-charged but turbulent initial suspension was called a debris flow, then the debris flow forms from the initial failure (model 1).

Dune cross-bedding, strong normal grading, and planar laminations provide clear evidence of deposition from a forerunning turbidity current in some cases (Talling et al., 2007b). However, it is less easy to determine unambiguously whether massive clean sandstone is deposited incrementally from a laminar dense fluidized flow of the type produced in the experiments of Breien et al. (2010), or from a turbulent high-density turbidity current. Clean sand layers formed from dense fluidized layers might pinch out more abruptly once the debris flow stops, while the run out of high-density turbidity current may be less connected to debris flow motion.

Relatively thin massive graded or ungraded clean sand layers could potentially be formed by sand settling from the debris flow plug at a late stage, or after the debris flow has stopped (Sumner et al., 2009). Such a process will tend to occur only in relatively weak debris flows, and will form basal sand that terminates where the debris flow stops.

Rather thin cohesive debrites within hybrid beds, and debrites lacking surrounding turbidite (Migeon et al., 2010), can be found as much as several hundred kilometers from their source in areas of very low gradient (0.05°) seafloor (Talling et al., 2004). The mobility of thin submarine debris flows therefore tends to be much greater than most of their subaerial counterparts.

This contribution has shown how thin and low-strength submarine debris flows can support sand, but still run out across very low-gradient seafloor (Fig. 9B). Therefore, debris flows need not necessarily be lubricated at their base to achieve these run out distances. In some cases, submarine debris flows may carry very large clasts that are positively buoyant to the fringes of submarine fans (Talling et al., 2010; Fig. 5B). The conditions for hydroplaning may occur commonly (Fig. 10), but water overridden by such hydroplaning flows may often be mixed back into the body of the debris flow.

Haughton et al. (2003, 2009) proposed that dewatering of the clean basal sand often plays an important role in the long run out of overlying debris flow. In some of the examples described here this process did not occur, as cross-lamination or planar lamination in the underlying clean sand is not deformed (Fig. 1; Talling et al., 2004, 2007b). Haughton et al. (2003, 2009) described a series of hybrid beds in which the underlying clean sand is injected and preserves evidence of dewatering. It is difficult to determine whether this evidence means that dewatering played a major role in debris flow motion, or whether the beds result from rapid loading, because it is unknown whether such dewatered clean sands terminate at the same location as the debrite. However, basal dewatering of the clean sand is apparent in the Haughton et al. (2003, 2009) examples.

Mud-rich cohesive debrite sandstone intervals have much lower permeability than clean sand deposited by turbidity currents (Amy et al., 2009). This means that the debrite sandstone intervals have unfavorable reservoir quality and baffles to subsurface flow, and it is important to predict their geometry to extract oil and gas effectively. Hybrid beds are common in systems worldwide, and most beds in some subsurface cores (e.g., Haughton et al., 2009) can be hybrid beds.

There is a strong tendency for hybrid beds to occur in the intermediate parts of larger systems, as cohesive debrites are absent in more proximal and the most distal locations. The more proximal locations comprise only thick-bedded clean sandstones formed by high-density turbidity currents, while the distal areas are typically dominated thin turbidites deposited by dilute turbidity currents.

Cohesive debris flow deposition is more strongly affected by small changes in seafloor gradient than turbidity current deposition (Fig. 3B; Talling et al., 2007a; Wynn et al., 2010). This means that the location of debrite intervals is controlled by (often subtle) changes in basin morphology. For example, basin morphology can determine whether debrites form the fringes or core of a sequence. Debrites on the Nile fan form a fringe around the end of the upstanding ridge formed by the channel system (Migeon et al., 2010); this is because flows that exited the channel underwent a radial break in slope, and therefore formed debrite in a fringe. In the Agadir Basin, cohesive debrite is found with a bulls-eye pattern in the basin’s two flattest areas (Fig. 3). The Agadir Basin lacks an upstanding channel-levee ridge, and debrite therefore occurs beyond subtle (0.05°–0.02°) breaks in slope at the basin center.

The location and shape of cohesive debrites will also depend on the relative strength of a debris flow (Fig. 12). Very high strength debris flows tend to produce thick clast-rich deposits in a single lobe, typically on the continental slope (Laberg and Vorren, 2000). Intermediate-strength debris flows can produce thinner clast-rich debris flows with a digitate frond-like planform shape, which may lack thick encasing turbidite sand, such as on the Mississippi and Nile fans (Fig. 4). The digitate planform shape is diagnostic of debris flow deposition, and can be visible in attribute maps of high-resolution three-dimensional seismic reflection data. As debris flow strength decreases, debris flow deposits will tend to run out further from source, although the run out is also dependent on factors such as debris flow volume. Very low strength (clast poor) debris flows may lack a digitate planform shape and may pond in a bulls-eye shape in the flattest parts of a basin (similar to fluid mud) (Figs. 3 and 12).

Cohesive debrites tend to occur in the distal (but not most distal parts) of submarine fans. Patterns of fan progradation may therefore be an important control on the overall distribution of hybrid beds in a stratigraphic interval (Hodgson, 2009; Pyles and Jennette, 2009). Channel-lobe progradation may lead to intervals rich in hybrid beds, overlain by more proximal lobes of channel-levee facies that lack hybrid beds, and underlain by very thin bedded turbidites from the most distal fan locations. Retrogression of the channel lobe may lead to the inverse succession, although this may be rarer due to abrupt lobe abandonment because of channel avulsion. Larger scale progradation of the entire fan system may also control hybrid bed frequency over longer time scales. Hybrid beds can show very little if any clustering in situations where the depositional setting did not fluctuate strongly, such as the thick basin plain sequence of the (inner part of the) Marnoso-arenacea Formation (Talling et al., 2012b).

Hybrid beds are associated with larger volume flows that include a high-density turbidity current component, and tend to be lacking in small dilute turbidity currents. This means that clustering of thick high-density turbidites will tend to be associated with clustering of hybrid beds within those thicker bedded packages; this can result from either channel-lobe migration or fluctuations in sea level.

Haughton et al. (2009) suggested that clustering of hybrid beds may result commonly from periods in which systems are out of grade. Such a model infers that hybrid beds are formed primarily by erosion of the muddy seafloor along parts of the flow path (model 2) that are periodically out of grade. The field observations reviewed here suggest that a significant number of cohesive debris flows, rich in organic material that most likely came from initial slope failure, are not formed by local erosion along the flow path. Laboratory experiments show that any flow that contains a significant amount of cohesive mud content can undergo late-stage transformation from turbulent to laminar flow. This cohesive mud fraction can be present from the start of the event, or be picked up en route via erosion. It is not clear whether “out of grade” refers to submarine canyons cut into the continental slope, or local topographic anomalies further down the fan. Submarine canyons tend to be areas of net erosion over long periods, and might be said to always be out of grade. The available evidence therefore does not provide strong support for a model in which local erosion on the fan determines clustering of hybrid beds, although such a process could potentially occur in some locations.

Haughton et al. (2009, fig. 3 therein) provided a graphic log showing an idealized five-part hybrid event bed (H1–H5); their model provides an excellent starting point for describing the basic structure of hybrid beds. Figure 2A provides a summary of hybrid bed character seen in this wider collection of examples, and includes variations in hybrid bed character and examples of downflow bed geometries (Fig. 2B). I emphasize here the variability in hybrid bed types as cohesive debris flow strength increases, and how flow evolution leads to different bed geometries.

Haughton et al. (2009) ascribed the basal clean sand interval (H1; Fig. 2) to deposition by a high-density turbidity current. This is the depositional mechanism in most cases (see discussion of Origins of Basal Clean Sand), but basal clean sand may be formed in a smaller number of cases by late-stage settling of sand from the debris flow plug (Sumner et al., 2009), or incrementally by a forerunning laminar dense liquefied flow of the type observed in the experiments of Breien et al. (2010). Haughton et al. (2009) emphasized dewatering of the dense basal sand as a mechanism for long run out. Here I show that some basal sand intervals have not dewatered and contain undeformed planar or dune-scale lamination, and that low-strength cohesive debris flows can potentially run out for very long distances without the need for basal lubrication.

An interval with repeated banding separated the basal high-density turbidite from the overlying cohesive debrite in the Haughton et al. (2009) model. This transition can be abrupt and occur gradationally over a few centimeters, or more rarely occur in a series of steps with progressively increasing mud content. Dune-scale cross-bedding is also rarely observed at this boundary (Figs. 1 and 2). Repeated banding of the type depicted in the Haughton et al. (2009) idealized model is rare, although seen in other locations (Sylvester and Lowe, 2004). As yet, it has only been shown to form very thick intervals in the Britannia Formation in the North Sea (Fig. 1; Lowe and Guy, 2000). The Haughton et al. (2009) model is therefore amended to show the variability of the transitional H2 interval, and to show that repeated banding is not the norm (Fig. 2).

The cohesive debrite interval contains clasts of variable types, or no clasts, as described by Haughton et al. (2009). Sand injection does not occur in many examples reviewed here, although it is common in the hybrid beds described by Haughton et al. (2009) in the North Sea (see discussion of Hybrid Beds in Subsurface Cores). The Haughton et al. (2009) model is extended herein by showing how increasing cohesive debrite strength influences that character of the hybrid bed (Fig. 12).

As noted by Haughton et al. (2009), the uppermost part of the hybrid bed comprises fine-grained low-density turbidity current deposits (Fig. 2). In some cases the debrite is directly overlain by turbidite mud. In other cases, the debrite is overlain by ripple cross-laminated sand, which have sometimes partly foundered into the underlying debrite. Planar laminated intervals tend to overlie ripple cross-laminated sand in the caps of hybrid beds reviewed here, rather than occurring below ripple cross-laminated intervals (as in the Haughton et al., 2009, model). The planar laminated intervals in the H4 interval therefore correspond to the T_D division rather than the T_B division (Fig. 2).

Hybrid beds are a major departure from previous widely cited models for submarine flow deposits that capture our understanding of the flows (e.g., Bouma, 1962; Lowe, 1982; Mutti, 1992; Mulder and Alexander, 2001). Cohesive debris flows are best classified in terms of a continuum of decreasing cohesive strength (Talling et al., 2012a; Fig. 12). High-strength (∼100 to >1000 Pa) debris flows tend to produce clast-rich debrites that are relatively thick and extend back to near the site of original slope failure. They are mostly restricted to steeper continental slopes, but sometimes form megabeds on basin plains, and tend to lack well-developed encasing turbidite sand. Intermediate cohesive strength (∼5 to ∼100 Pa) debris flows often contain clasts, and may commonly reach the distal low-gradient parts of submarine fans, to produce debrites that are less than a couple of meters thick. Clast composition shows that debris flow can be very far-traveled, and meter-sized clasts can be rafted for long distances on low gradients if the clasts are less dense than surrounding flow. Low-strength (0.1–5 Pa) cohesive debris flows generally lack larger (greater than millimeter) clasts. Intermediate- and low-strength debris flows may evolve from initially turbulent and sediment-charged flows on steeper slopes in proximal areas, and their deposits are commonly restricted to hybrid beds in the distal parts of submarine fans. Intermediate-strength clast-rich debris flows may run out for long distances on low gradients (as shown by clast compositions) without hydroplaning. Very low strength cohesive debris flows most likely form via local flow transformation near the final site of debrite deposition, and emplacement must be gentle if they are not to mix with surrounding seawater. As cohesive strength is further reduced there is a transition into fluid mud layers that lack sand, and the processes that deposit turbidite mud. The location and shape of cohesive debrites are controlled strongly by subtle changes in seafloor gradient. Cohesive debrites in hybrid beds may form a fringe around upstanding channel-levee ridges, or occur in the central lowest part of basin plains lacking such uplifted ridges. This allows cohesive debrite geometry to be predicted for subsurface oil and gas reservoirs, if basin floor morphology is well understood. Small fractional changes in mud content lead to changes in cohesive strength, viscosity, permeability, and rates of excess pore pressure dissipation that span several orders of magnitude. Debris flows may be strongly shear thinning; this, together with mixing processes, can lead to bifurcation in flow behavior. Small amounts of mud can dampen turbulence effectively, especially as flow decelerates, and flow transformation may be common. This ensures that hybrid flows and cohesive debris flows display a wide range of flow behavior and deposit geometries.

This synthesis includes insights from a wide range of previous work, for which I am grateful. Lawrence Amy (Tullow Oil) and Esther Sumner (University of Southampton) played major roles in completing laboratory experiments and field work in the Marnoso-arenacea Formation, the latter in collaboration with Giuseppe Malgesini (National Oceanography Centre, NOC) and Fabrizio Felletti (University of Milan). Field data collection in the Moroccan turbidite system benefitted from the considerable efforts of Russell Wynn, Michael Frenz, Doug Masson, James Hunt, Christopher Stevenson (all at NOC), and others over many years. Sebastian Migeon, Jean Mascle, and Emmanuelle Ducassou (Geoscience Azur) provided recently published information on the Nile system. David Twichell (U.S. Geological Survey) helped greatly in the analysis of Mississippi fan cores. This work was part of a UK-TAPS (U.K. Turbidite Architecture and Process Studies) project, funded by NERC (Natural Environment Research Council) Ocean Margins LINK grants NER/T/S/2000/0106 and NER/M/S/2000/00264, and cosponsored by ConocoPhillips, BHP Billiton, and Shell. The reviews of Joris Eggenhuisen, David Piper, and an anonymous reviewer were very much appreciated.